1 2018-07-08 Paul Koning <ni1d@arrl.net>
3 * doc/md.texi (define_subst): Document how multiple occurrences of
4 the same argument in the replacement pattern are handled.
6 2018-07-11 Paul Koning <ni1d@arrl.net>
8 * doc/extend.texi (Common Variable Attributes): Move "mode" into
10 (Common Type Attributes): Add "mode" attribute.
12 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
14 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
15 stream DECL_ORIGINAL_TYPE.
16 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
17 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
18 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
19 Do not walk original type.
20 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
22 (write_ts_decl_non_common_tree_pointers): Do not stream
24 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
25 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
27 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
29 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
30 thread twice from the same starting edge.
32 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
34 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
35 * gimple.c (gimple_call_nonnull_result_p): ...here...
36 (gimple_call_nonnull_arg): ...and here.
37 * gimple.h (gimple_call_nonnull_result_p): New.
38 (gimple_call_nonnull_arg): New.
40 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
42 * config/arm/arm-cpus.in: Move information from fpu field of each
43 cpu definition to the isa field.
44 * config/arm/parsecpu.awk (fpu): Delete match rule.
45 (gen_comm_data): Don't add bits from the CPU's FPU entry.
47 2018-07-11 Richard Biener <rguenther@suse.de>
50 * dwarf2out.c (init_sections_and_labels): Use
51 output_asm_line_debug_info consistently.
52 (dwarf2out_early_finish): Likewise.
53 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
56 2018-07-11 Richard Biener <rguenther@suse.de>
58 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
59 Rework father_bb setting in a way to avoid propagating constants
60 multiple times on a loop body.
62 2018-07-10 Mark Wielaard <mark@klomp.org>
65 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
67 2018-07-10 Richard Biener <rguenther@suse.de>
69 * hash-map.h (hash_map::iterator::operator*): Return
70 references to key and value.
72 2018-07-10 Jakub Jelinek <jakub@redhat.com>
75 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
76 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
77 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
78 (gimplify_omp_for): For composite loops, move outer
79 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
80 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
81 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
82 TREE_LIST for both the original class iterator and the "last" helper
83 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
84 loop, remember has_decl_expr from outer composite loops for the
85 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
87 2018-07-09 Martin Sebor <msebor@redhat.com>
91 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
92 accessing implicitly initialized array elements.
93 * expr.c (string_constant): Handle string initializers of
94 character arrays within aggregates.
95 * gimple-fold.c (fold_array_ctor_reference): Add argument.
96 Store element offset. As a special case, handle zero size.
97 (fold_nonarray_ctor_reference): Same.
98 (fold_ctor_reference): Add argument. Store subobject offset.
99 * gimple-fold.h (fold_ctor_reference): Add argument.
101 2018-07-09 Paul Koning <ni1d@arrl.net>
103 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
104 (pdp11_insn_cost): New function.
105 (pdp11_md_asm_adjust): New function.
106 (TARGET_INVALID_WITHIN_DOLOOP): Define.
107 (pdp11_rtx_costs): Update to match machine better.
108 (output_addr_const_pdp11): Correct format mismatch warnings.
109 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
110 * config/pdp11/pdp11.md: General change to add base_cost and/or
111 length attributes for use by new pdp11_insn_cost function.
112 (MIN_BRANCH): Correct definition.
114 (doloop_end): Use standard pattern name for looping pattern.
115 (doloop_end_nocc): New.
116 (movsf): Add another constraint alternative.
117 (zero_extendqihi2): Add constraint alternatives for not in place
119 (zero_extendhisi2): Remove.
120 (shift patterns): Add CC handling variants.
124 (define_peephole2): New peephole to recognize mov that sets CC for
127 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
129 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
130 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
133 2018-07-09 Jakub Jelinek <jakub@redhat.com>
136 * real.c (real_nextafter): Return true if result is denormal.
138 2018-07-09 Martin Liska <mliska@suse.cz>
140 * common.opt: Add back wrongly removed attribute.
142 2018-07-09 Richard Biener <rguenther@suse.de>
145 * dwarf2out.c (gen_block_die): For an early generated DIE
146 always output high/low PC attributes.
148 2018-07-09 Tom de Vries <tdevries@suse.de>
150 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
151 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
152 onto VAR_DECL with abstract origin.
154 2018-07-07 Jim Wilson <jimw@sifive.com>
156 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
158 2018-07-07 Tom de Vries <tdevries@suse.de>
160 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
162 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
164 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
166 (combine_bound): Use wide-int overflow calculation instead of
168 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
170 * fold-const.c (int_const_binop_2): Same.
171 (extract_muldiv_1): Same.
172 (fold_div_compare): Same.
173 (fold_abs_const): Same.
175 * poly-int.h (add): Same.
179 * predict.c (predict_iv_comparison): Same.
180 * profile-count.c (slow_safe_scale_64bit): Same.
181 * simplify-rtx.c (simplify_const_binary_operation): Same.
182 * tree-chrec.c (tree_fold_binomial): Same.
183 * tree-data-ref.c (split_constant_offset_1): Same.
184 * tree-if-conv.c (idx_within_array_bound): Same.
185 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
186 * tree-ssa-phiopt.c (minmax_replacement): Same.
187 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
188 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
189 * vr-values.c (vr_values::adjust_range_with_scev): Same.
190 * wide-int.cc (wi::add_large): Same.
191 (wi::mul_internal): Same.
192 (wi::sub_large): Same.
193 (wi::divmod_internal): Same.
194 * wide-int.h: Change overflow type to overflow_type for neg, add,
195 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
196 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
197 mul_internal, divmod_internal.
198 (overflow_type): New enum.
199 (accumulate_overflow): New.
201 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
203 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
204 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
206 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
208 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
209 argument is checked for zero before entering loop, avoid checking again.
211 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
213 * gimplify.h (generic_expr_could_trap_p): Set as global function.
214 * gimplify.h (generic_expr_could_trap_p): Likwise.
215 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
217 2018-07-06 Jakub Jelinek <jakub@redhat.com>
219 PR tree-optimization/86401
220 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
221 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
222 (fold_bit_and_mask): ... here. New helper function for match.pd.
223 * fold-const.h (fold_bit_and_mask): Declare.
224 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
226 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
229 * target.def (translate_mode_attribute): New hook.
230 * targhooks.h (default_translate_mode_attribute): Declare.
231 * targhooks.c (default_translate_mode_attribute): New function.
232 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
233 * doc/tm.texi: Regenerate.
234 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
235 (rs6000_translate_mode_attribute): New function.
237 2018-07-06 Paul Koning <ni1d@arrl.net>
239 * doc/md.texi (define_split): Document DONE and FAIL.
240 (define_peephole2): Ditto.
242 2018-07-05 Jeff Law <law@redhat.com>
244 PR tree-optimization/86010
245 * tree-ssa-dse.c (compute_trims): More aggressively trim at
246 both the head and tail of mem* and str* calls.
248 2018-07-05 Jim Wilson <jimw@sifive.com>
250 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
252 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
254 * config/aarch64/aarch64-simd.md: correct flags text for
255 MIN_EXPR replacement.
257 2018-07-05 James Clarke <jrtc27@jrtc27.com>
259 * configure: Regenerated.
261 2018-07-05 Carl Love <cel@us.ibm.com>
263 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
264 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
265 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
266 VSX_BUILTIN_DOUBLEL_V4SF.
268 2018-07-05 Martin Sebor <msebor@redhat.com>
271 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
272 than its domain to compute its the upper bound of a char array.
274 2018-07-05 Nathan Sidwell <nathan@acm.org>
276 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
277 * doc/cpp.texi: Update comment.
278 * doc/tm.texi: Rebuilt.
279 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
280 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
281 * doc/extend.texi (Backwards Compatibility): Clarify it is system
282 headers affected by extern "C".
283 * system.h: Poison NO_IMPLICIT_EXTERN_C.
284 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
285 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
286 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
287 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
288 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
289 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
290 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
291 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
292 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
293 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
294 config/sparc/sp64-elf.h, config/spu/spu.h,
295 config/stormy16/stormy16.h, config/v850/v850.h,
296 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
297 define NO_IMPLICIT_EXTERN_C.
298 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
300 2018-07-05 Tamar Christina <tamar.christina@arm.com>
303 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
304 instead of GET_MODE_SIZE when comparing Units.
306 2018-07-05 Tamar Christina <tamar.christina@arm.com>
309 * rtlanal.c (set_noop_p): Constrain on mode change,
310 include hard-reg-set.h
312 2018-07-05 Tamar Christina <tamar.christina@arm.com>
314 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
316 2018-07-05 Jakub Jelinek <jakub@redhat.com>
319 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
322 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
324 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
326 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
329 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
331 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
333 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
336 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
338 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
340 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
343 2018-07-04 Martin Liska <mliska@suse.cz>
345 * tree-switch-conversion.c: Define
346 max_ratio_for_speed and max_ratio_for_size constants.
348 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
349 Martin Liska <mliska@suse.cz>
354 * common.opt: Rename align options with 'str_' prefix.
355 * common/config/i386/i386-common.c (set_malign_value): New
357 (ix86_handle_option): Use it to set -falign-* options/
358 * config/aarch64/aarch64-protos.h (struct tune_params): Change
359 type from int to string.
360 * config/aarch64/aarch64.c: Update default values from int
362 * config/alpha/alpha.c (alpha_override_options_after_change):
364 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
365 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
366 max skip conditionally.
367 * config/i386/freebsd.h (SUBALIGN_LOG): New.
368 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
369 max skip conditionally.
370 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
371 max skip conditionally.
372 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
373 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
374 max skip conditionally.
375 * config/i386/i386.c (struct ptt): Change type from int to
377 (ix86_default_align): Set default values.
378 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
379 max skip conditionally.
380 * config/i386/iamcu.h (SUBALIGN_LOG): New.
381 (ASM_OUTPUT_MAX_SKIP_ALIGN):
382 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
383 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
384 max skip conditionally.
385 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
386 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
387 * config/i386/x86-64.h (SUBALIGN_LOG): New.
388 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
389 max skip conditionally.
390 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
391 * config/ia64/ia64.c (ia64_option_override): Set default values
392 for alignment options.
393 * config/m68k/m68k.c: Handle new str_align_* options.
394 * config/mips/mips.c (mips_set_compression_mode): Change
396 (mips_option_override): Set default values for options.
397 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
399 * config/rs6000/rs6000.c (rs6000_option_override_internal):
401 * config/rx/rx.c (rx_option_override): Likewise.
402 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
403 (LABEL_ALIGN): Use align_labels_log.
404 (LOOP_ALIGN): Use align_loops_align.
405 * config/s390/s390.c (s390_asm_output_function_label): Use new
407 * config/sh/sh.c (sh_override_options_after_change):
408 Change type of constants.
409 * config/spu/spu.c (spu_sched_init): Likewise.
410 * config/sparc/sparc.c (sparc_option_override): Set default
412 * config/visium/visium.c (visium_option_override): Likewise.
413 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
414 emit p2align format with last argument if it's not needed.
415 * doc/invoke.texi: Document extended format of -falign-*.
416 * final.c: Use align_labels alignment.
417 * flags.h (struct target_flag_state): Change type to use
419 (struct align_flags_tuple): New.
420 (struct align_flags): Likewise.
421 (align_loops_log): Redefine macro to use new types.
422 (align_loops_max_skip): Redefine macro to use new types.
423 (align_jumps_log): Redefine macro to use new types.
424 (align_jumps_max_skip): Redefine macro to use new types.
425 (align_labels_log): Redefine macro to use new types.
426 (align_labels_max_skip): Redefine macro to use new types.
427 (align_functions_log): Redefine macro to use new types.
428 (align_loops): Redefine macro to use new types.
429 (align_jumps): Redefine macro to use new types.
430 (align_labels): Redefine macro to use new types.
431 (align_functions): Redefine macro to use new types.
432 (align_functions_max_skip): Redefine macro to use new types.
433 (align_loops_value): New macro.
434 (align_jumps_value): New macro.
435 (align_labels_value): New macro.
436 (align_functions_value): New macro.
437 * function.c (invoke_set_current_function_hook): Propagate
438 alignment values from flags to global variables default in
440 * ipa-icf.c (sem_function::equals_wpa): Use
441 cl_optimization_option_eq instead of memcmp.
442 * lto-streamer.h (cl_optimization_stream_out): Support streaming
444 (cl_optimization_stream_in): Likewise.
445 * optc-save-gen.awk: Support strings in cl_optimization.
446 * opth-gen.awk: Likewise.
447 * opts.c (finish_options): Remove error checking of invalid
449 (MAX_CODE_ALIGN): Remove.
450 (MAX_CODE_ALIGN_VALUE): Likewise.
451 (parse_and_check_align_values): New function.
452 (check_alignment_argument): Likewise.
453 (common_handle_option): Use check_alignment_argument.
454 * opts.h (parse_and_check_align_values): Declare.
455 * toplev.c (init_alignments): Remove.
456 (read_log_maxskip): New.
457 (parse_N_M): Likewise.
458 (parse_alignment_opts): Likewise.
459 (backend_init_target): Remove usage of init_alignments.
460 * toplev.h (parse_alignment_opts): Declare.
461 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
463 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
464 * tree.c (cl_option_hasher::equal): New.
465 * varasm.c: Use new global macros.
467 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
469 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
470 Use a simpler align directive also if MAXSKIP = ALIGN-1.
471 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
472 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
473 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
474 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
475 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
476 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
477 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
478 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
479 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
480 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
481 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
483 2018-07-04 Martin Liska <mliska@suse.cz>
484 Jonathan Wakely <jwakely@redhat.com>
486 * coverage.c: Use correct type.
487 * doc/invoke.texi: Language correction.
489 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
492 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
493 ENDBRANCH for non-tail call which may return via indirect branch.
494 * doc/extend.texi: Document indirect_return attribute.
496 2018-07-03 Martin Sebor <msebor@redhat.com>
498 PR tree-optimization/86274
499 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
501 (format_floating): Correct handling of infinities and NaNs.
503 2018-07-03 Martin Sebor <msebor@redhat.com>
505 * print-tree.c (print_real_cst): New function.
506 (print_node_brief): Call it.
509 2018-07-03 Jeff Law <law@redhat.com>
511 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
512 into a single pattern.
514 * config/h8300/h8300.md (ors code_iterator): New.
515 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
516 a single pattern and single splitter.
517 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
518 (iorqi3_1, xorqi3_1): Likewise.
519 (iorqi3, xorqi3 expanders): Similarly.
521 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
522 (movmd_internal) into a single pattern using the P mode iterator.
523 (movmd splitters): Similarly.
524 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
525 (movsd splitters): Similarly.
527 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
528 ADDB, ADDW and ADDL into a single ADD attribute which selects the
529 right table based on the size of the operand.
530 * config/h8300/h8300.md (length_table): Corresponding changes. All
531 references to "addb", "addw" and "addl" changed to "add".
532 (btst patterns): Merge two variants into a single pattern.
533 (tstqi, tsthi): Likewise.
534 (addhi3_incdec, addsi3_incdec): Likewise.
535 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
536 (mulhi3, mulsi3): Likewise.
537 (udivhi3, udivsi3): Likewise.
538 (divhi3, divsi3): Likewise.
539 (andorqi3, andorhi3, andorsi3): Likewise.
541 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
544 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
545 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
547 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
549 PR tree-optimization/85694
550 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
551 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
553 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
554 (<u>avg<mode>3_ceil): New patterns.
556 2018-07-03 David Malcolm <dmalcolm@redhat.com>
558 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
559 scan-tree-dump directive.
560 * gcc.dg/vect/slp-perm-2.c: Likewise.
561 * gcc.dg/vect/slp-perm-3.c: Likewise.
562 * gcc.dg/vect/slp-perm-5.c: Likewise.
563 * gcc.dg/vect/slp-perm-6.c: Likewise.
564 * gcc.dg/vect/slp-perm-7.c: Likewise.
565 * gcc.dg/vect/slp-perm-8.c: Likewise.
567 2018-07-03 Marek Polacek <polacek@redhat.com>
570 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
572 2018-07-03 Richard Biener <rguenther@suse.de>
575 * tree-ssa-structalias.c (find_func_clobbers): Properly
576 handle indirect calls.
578 2018-07-03 Jeff Law <law@redhat.com>
580 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
581 (shifts): New code iterator.
582 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
583 expander. Fix HImode handling on H8/SX.
584 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
585 (subqi3, subhi3, subsi3 expanders): Likewise.
586 (andqi3, andhi3, andsi3 expanders): Likewise.
587 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
588 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
589 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
590 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
591 (zero_extendqihi2, zero_extendqisi2): Likewise.
592 (extendqihi2, extendqisi2): Likewise.
593 (rotlqi3, rotlhi3, rotlsi3): Likewise.
594 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
595 (rotlqi3_1, rotlhi3_1): Likewise.
596 (logicalhi3_sn, logicalsi3_sn): Likewise.
597 (logicalhi3, logicalsi3): Likewise.
599 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
601 * tree-vect-patterns.c (vect_recog_rotate_pattern)
602 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
603 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
604 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
605 type to append_pattern_def_seq instead of creating a stmt_vec_info
607 (build_mask_conversion): Likewise. Remove vinfo argument.
608 (vect_add_conversion_to_patterm): Likewise, renaming to...
609 (vect_add_conversion_to_pattern): ...this.
610 (vect_recog_mask_conversion_pattern): Update call to
611 build_mask_conversion. Pass the vector type to
612 append_pattern_def_seq here too.
613 (vect_recog_gather_scatter_pattern): Update call to
614 vect_add_conversion_to_pattern.
616 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
618 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
619 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
620 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
621 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
622 STMT_VINFO_PATTERN_DEF_SEQ to null here.
623 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
624 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
625 append_pattern_def_seq instead of new_pattern_def_seq.
626 (vect_recog_divmod_pattern): Do both of the above.
627 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
630 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
632 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
633 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
634 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
635 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
636 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
637 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
638 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
639 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
640 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
641 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
642 parameter with a single stmt_vec_info.
643 (vect_recog_func_ptr): Likewise.
644 (vect_recog_gather_scatter_pattern): Likewise, folding in...
645 (vect_try_gather_scatter_pattern): ...this.
646 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
647 the stmt_vec_info of the statement to be matched. Don't clear
648 STMT_VINFO_RELATED_STMT.
649 (vect_pattern_recog): Update call accordingly.
651 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
653 PR tree-optimization/85694
654 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
655 (uavgM3_ceil): Document new optabs.
656 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
657 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
659 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
660 (savg_ceil_optab): New optabs.
661 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
662 (vect_vect_recog_func_ptrs): Add it.
663 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
664 constant directly from the associated lhs.
666 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
668 * tree-vect-patterns.c (vect_split_statement): New function.
669 (vect_convert_input): Use it to try to split an existing cast.
671 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
673 * poly-int.h (print_hex): New function.
674 * dumpfile.h (dump_dec, dump_hex): Declare.
675 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
676 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
677 min_input_precision, operation_precision and operation_sign.
678 * tree-vect-patterns.c (vect_get_range_info): New function.
679 (vect_same_loop_or_bb_p, vect_single_imm_use)
680 (vect_operation_fits_smaller_type): Delete.
681 (vect_look_through_possible_promotion): Add an optional
682 single_use_p parameter.
683 (vect_recog_over_widening_pattern): Rewrite to use new
684 stmt_vec_info infomration. Handle one operation at a time.
685 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
686 (vect_truncatable_operation_p, vect_set_operation_type)
687 (vect_set_min_input_precision): New functions.
688 (vect_determine_min_output_precision_1): Likewise.
689 (vect_determine_min_output_precision): Likewise.
690 (vect_determine_precisions_from_range): Likewise.
691 (vect_determine_precisions_from_users): Likewise.
692 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
693 (vect_vect_recog_func_ptrs): Put over_widening first.
695 (vect_pattern_recog): Call vect_determine_precisions.
697 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
699 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
700 statements that have been replaced by further pattern statements.
701 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
703 2018-07-03 Richard Biener <rguenther@suse.de>
705 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
706 always set *dt. Dump vectype in vectype overload.
707 * dumpfile.h (dump_gimple_expr): New function.
708 (dump_gimple_expr_loc): Likewise.
709 * dumpfile.c (dump_gimple_expr): New function.
710 (dump_gimple_expr_loc): Likewise.
712 2018-07-02 Jeff Law <law@redhat.com>
714 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
715 the H8/300, H8/300H and H8/S variants into a single pattern.
716 (movhi_h8300, movqi_h8300hs): Similarly.
717 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
718 (QHI mode iterator): New.
720 * config/h8300/h8300.md: Remove trailing whitespace.
722 2018-07-02 Jim Wilson <jimw@sifive.com>
724 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
725 instead of emit_insn for interrupt returns.
726 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
727 (riscv_sret, riscv_uret): Likewise.
729 2018-07-02 David Malcolm <dmalcolm@redhat.com>
731 * pretty-print.c (selftest::test_pp_format): Move save and restore
732 of quotes to class auto_fix_quotes, and add an instance.
733 * selftest.c: Include "intl.h".
734 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
735 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
736 * selftest.h (selftest::auto_fix_quotes): New class.
738 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
740 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
741 (aarch64_sve_prepare_conditional_op): Remove.
742 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
743 Allow aarch64_simd_reg_or_zero as select operand; remove
744 the aarch64_sve_prepare_conditional_op call.
745 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
746 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
747 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
748 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
749 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
750 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
751 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
752 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
753 and a splitters to match all of the *_any patterns.
754 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
756 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
757 (SVE_COND_FP_BINARY_REV): Remove.
758 (sve_int_op_rev, sve_fp_op_rev): New.
759 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
760 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
761 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
762 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
763 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
764 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
765 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
766 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
767 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
769 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
770 Remove match_dup 1 from the inner unspec.
771 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
773 * config/aarch64/aarch64.md (movprfx): New attr.
774 (length): Default movprfx to 8.
775 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
776 (*madd<SVE_I>, *msub<SVE_I): Likewise.
777 (*<su>mul<SVE_I>3_highpart): Likewise.
778 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
779 (*v<ASHIFT><SVE_I>3): Likewise.
780 (*<su><MAXMIN><SVE_I>3): Likewise.
781 (*<su><MAXMIN><SVE_F>3): Likewise.
782 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
783 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
784 (*div<SVE_F>4): Likewise.
786 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
788 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
791 2018-07-02 Richard Biener <rguenther@suse.de>
793 PR tree-optimization/86363
794 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
795 memset argument refers to a non-variable address.
797 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
799 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
800 {PLUS,MINUS}_EXPR code to...
801 (adjust_symbolic_bound): ...here,
802 (combine_bound): ...here,
803 (set_value_range_with_overflow): ...and here.
805 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
807 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
809 (extract_range_from_abs_expr): ...here.
811 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
813 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
814 -fno-omit-frame-pointer when not optimizing.
816 2018-07-02 Martin Liska <mliska@suse.cz>
819 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
820 (propagate_nothrow): Likewise.
822 2018-07-02 Martin Liska <mliska@suse.cz>
825 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
827 2018-07-02 David Malcolm <dmalcolm@redhat.com>
829 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
830 function in r262149, changing "loc" param from source_location to
831 const dump_location_t &.
832 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
833 declaration, as above.
835 2018-07-01 Paul Koning <ni1d@arrl.net>
837 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
838 -munit-asm, -mgnu-asm, -mdec-asm.
839 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
840 (pdp11_output_labelref): New.
841 (pdp11_output_def): New.
842 (pdp11_output_addr_vec_elt): New.
843 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
844 %# and %@ format codes.
845 (pdp11_option_override): New.
846 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
847 (pdp11_output_ident): New.
848 (pdp11_asm_named_section): New.
849 (pdp11_asm_init_sections): New.
850 (pdp11_file_start): New.
851 (pdp11_file_end): New.
852 (output_ascii): Use .ascii/.asciz for -mdec-asm.
853 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
854 %o, like %c but octal.
855 (pdp11_option_override): New.
856 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
858 (DATA_SECTION_ASM_OP): Ditto.
859 (READONLY_DATA_SECTION_ASM_OP): New.
860 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
861 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
862 (ASM_OUTPUT_LABELREF): Ditto.
863 (ASM_OUTPUT_DEF): Ditto.
864 (ASM_OUTPUT_EXTERNAL): New.
865 (ASM_OUTPUT_SOURCE_FILENAME): New.
866 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
867 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
868 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
869 %# and %@ format codes.
870 * config/pdp11/pdp11.opt (mgnu-asm): New.
871 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
872 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
873 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
875 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
877 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
878 dereferencing path[] beyond its length.
880 (debug_all_paths): New.
881 (rewire_first_differing_edge): New.
882 (adjust_paths_after_duplication): New.
883 (duplicate_thread_path): Call adjust_paths_after_duplication.
885 (thread_through_all_blocks): Add new argument to
886 duplicate_thread_path.
888 2018-06-30 Jim Wilson <jimw@sifive.com>
890 * config/riscv/predicates.md (p2m1_shift_operand): New.
891 (high_mask_shift_operand): New.
892 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
893 pattern using p2m1_shift_operand.
894 (lshsi3_zero_extend_3+2): New combiner pattern using
895 high_mask_shift_operand.
897 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
899 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
901 (vect_recog_rotate_pattern): ...here.
902 (vect_convert_input): Try to insert casts of invariants in the
904 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
905 preheader to be empty.
907 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
909 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
910 vector type. If given, install it in the new statement's
912 (vect_element_precision): New function.
913 (vect_unpromoted_value): New struct.
914 (vect_unpromoted_value::vect_unpromoted_value): New function.
915 (vect_unpromoted_value::set_op): Likewise.
916 (vect_look_through_possible_promotion): Likewise.
917 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
918 (vect_widened_op_tree, vect_convert_input): Likewise.
919 (vect_convert_inputs, vect_convert_output): Likewise.
920 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
921 to handle the optional cast of the multiplication result and
922 vect_widened_op_tree to detect the widened multiplication itself.
923 Do not require the input and output of promotion casts to have
924 the same sign, but base the signedness of the operation on the
925 input rather than the result. If the pattern includes two
926 promotions, check that those promotions have the same sign.
927 Do not restrict the MULT_EXPR handling to a double-width result;
928 handle quadruple-width results and wider. Use vect_convert_inputs
929 to convert the inputs to the common type.
930 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
931 to handle the optional cast of the ABS result. Also allow a sign
932 change or a sign extension between the ABS and MINUS.
933 Use vect_widened_op_tree to detect the widened subtraction and use
934 vect_convert_inputs to convert the inputs to the common type.
935 (vect_handle_widen_op_by_const): Delete.
936 (vect_recog_widen_op_pattern): New function.
937 (vect_recog_widen_mult_pattern): Use it.
938 (vect_recog_widen_shift_pattern): Likewise.
939 (vect_recog_widen_sum_pattern): Use
940 vect_look_through_possible_promotion to handle the promoted
943 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
945 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
946 the containing gimple_seq *.
947 * gimple-iterator.h (gsi_for_stmt): Declare it.
948 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
949 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
950 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
951 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
952 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
954 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
956 (vect_mark_pattern_stmts): ...here. Handle cases in which the
957 statement being replaced is part of an existing pattern
958 definition sequence, inserting the new pattern statements before
960 (vect_pattern_recog_1): Don't return a bool. If the statement
961 is already part of a pattern, instead apply pattern matching
962 to the pattern definition statements. Don't clear the
963 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
964 (vect_pattern_recog): Don't break after the first match;
965 continue processing the pattern definition statements instead.
966 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
968 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
970 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
971 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
972 (vect_recog_widen_sum_pattern): Use it.
974 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
976 * tree-vect-loop.c (vectorizable_reduction): Assert that the
977 phi is not a pattern statement and has not been replaced by
979 * tree-vect-patterns.c (type_conversion_p): Don't check
980 STMT_VINFO_IN_PATTERN_P.
981 (vect_recog_vector_vector_shift_pattern): Likewise.
982 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
983 the pattern statement rather than the original statement; check
984 directly for a WIDEN_MULT_EXPR here.
985 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
986 vect_is_simple_use to return the pattern statement rather
987 than the original statement; use is_pattern_stmt_p to check
988 for such a pattern statement.
989 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
990 to return the pattern statement rather than the original statement;
991 don't do the same transformation here.
992 (vect_is_simple_use): If the defining statement has been replaced
993 by a pattern statement, return the pattern statement instead.
994 Remove the corresponding (local) transformation from the vectype
997 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
999 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1000 end and default to null.
1001 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1002 (vectorizable_reduction): Update calls accordingly, dropping the
1003 gimple ** argument if the passed-back statement isn't needed.
1004 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1005 (vect_recog_rotate_pattern): Likewise.
1006 (vect_recog_mask_conversion_pattern): Likewise.
1007 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1008 (vect_mask_constant_operand_p): Likewise.
1009 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1010 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1011 (get_group_load_store_type, get_load_store_type): Likewise.
1012 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1013 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1014 (vectorizable_conversion, vectorizable_assignment): Likewise.
1015 (vectorizable_shift, vectorizable_operation): Likewise.
1016 (vectorizable_store, vect_is_simple_cond): Likewise.
1017 (vectorizable_condition, vectorizable_comparison): Likewise.
1018 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1019 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1020 and move it to the end. Cope with null def_stmt_outs.
1022 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1024 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1026 2018-06-29 Jeff Law <law@redhat.com>
1028 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1029 displacements for TARGET_V850E2V3 and newer.
1030 (TARGET_LRA_P): Remove. Defaults to LRA now.
1031 * config/v850/v850.md (sign23byte_load): Remove.
1032 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1033 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1035 2018-06-29 Martin Liska <mliska@suse.cz>
1038 * coverage.c (coverage_init): Mangle full path name.
1039 * doc/invoke.texi: Document the change.
1040 * gcov-io.c (mangle_path): New.
1041 * gcov-io.h (mangle_path): Likewise.
1042 * gcov.c (mangle_name): Use mangle_path for path mangling.
1044 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1046 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1047 if starting source register is not even.
1049 2018-06-29 Martin Liska <mliska@suse.cz>
1051 PR tree-optimization/86263
1052 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1053 Make edge redirection.
1055 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1057 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1058 (dump_scope_depth): New variable.
1059 (get_dump_scope_depth): New function.
1060 (dump_begin_scope): New function.
1061 (dump_end_scope): New function.
1062 * dumpfile.h (get_dump_scope_depth): New declaration.
1063 (dump_begin_scope): New declaration.
1064 (dump_end_scope): New declaration.
1065 (class auto_dump_scope): New class.
1066 (AUTO_DUMP_SCOPE): New macro.
1067 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1070 2018-06-29 Richard Biener <rguenther@suse.de>
1072 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1073 compute_all_dependences succeeds.
1074 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1075 exceed --param loop-max-datarefs-for-datadeps.
1077 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1079 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1081 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1084 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1085 alternative 0 in preferred_for_speed attribute.
1087 2018-06-28 Paul Koning <ni1d@arrl.net>
1089 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1090 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1091 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1092 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1093 for shift insn patterns.
1095 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1097 * cgraph.c (cgraph_node::get_body): Replace assignments to
1098 "dump_file" with calls to set_dump_file.
1099 * dumpfile.c (alt_dump_file): Make static, and group with...
1100 (alt_flags): ...this definition.
1101 (dumps_are_enabled): New variable.
1102 (refresh_dumps_are_enabled): New function.
1103 (set_dump_file): New function.
1104 (set_alt_dump_file): New function.
1105 (gcc::dump_manager::dump_start): Replace assignments to
1106 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1108 (gcc::dump_manager::dump_finish): Likewise.
1109 * dumpfile.h (alt_dump_file): Delete decl.
1110 (dumps_are_enabled): New variable decl.
1111 (set_dump_file): New function decl.
1112 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1114 * tree-nested.c (lower_nested_functions): Replace assignments to
1115 "dump_file" with calls to set_dump_file.
1117 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1119 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1120 goto_locus of each outgoing edge of each basic block.
1122 2018-06-28 Richard Biener <rguenther@suse.de>
1124 * dwarf2out.c (decl_scope_table): Remove.
1125 (push_decl_scope): Likewise.
1126 (pop_decl_scope): Likewise.
1127 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1128 (gen_struct_or_union_type_die): Likewise.
1129 (gen_tagged_type_die): Likewise.
1130 (dwarf2out_init): Do not initialize decl_scope_table.
1131 (dwarf2out_c_finalize): Do not free it.
1133 2018-06-28 Richard Biener <rguenther@suse.de>
1135 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1136 deciding whether to not re-use a DIE.
1138 2018-06-28 Richard Biener <rguenther@suse.de>
1140 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1141 DW_AT_abstract_origin attribute.
1143 2018-06-28 Martin Liska <mliska@suse.cz>
1145 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1146 Use newly introduced constants.
1147 * tree-switch-conversion.h (struct jump_table_cluster):
1148 Define max_ratio_for_size and max_ratio_for_speed.
1150 2018-06-28 Martin Liska <mliska@suse.cz>
1152 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1153 Add new checking assert to catch invalid state.
1154 (jump_table_cluster::can_be_handled): Handle single case
1156 (jump_table_cluster::is_beneficial): Bail out for such case.
1157 (bit_test_cluster::find_bit_tests):
1158 Add new checking assert to catch invalid state.
1159 (bit_test_cluster::can_be_handled): Handle single case
1161 (bit_test_cluster::is_beneficial): Bail out for such case.
1162 (switch_decision_tree::analyze_switch_statement):
1165 2018-06-28 Martin Liska <mliska@suse.cz>
1167 * common.opt: Introduce -completion option.
1168 * gcc.c (driver_handle_option): Handle it.
1169 (driver::main): Print completions if completion
1171 * opt-suggestions.c (option_proposer::get_completions):
1173 (option_proposer::suggest_completion): Likewise.
1174 (option_proposer::find_param_completions): Likewise.
1175 (verify_autocompletions): Likewise.
1176 (test_completion_valid_options): Likewise.
1177 (test_completion_valid_params): Likewise.
1178 (in_completion_p): Likewise.
1179 (empty_completion_p): Likewise.
1180 (test_completion_partial_match): Likewise.
1181 (test_completion_garbage): Likewise.
1182 (opt_proposer_c_tests): Likewise.
1183 * opt-suggestions.h: Declare new functions.
1184 * opts.c (common_handle_option): Handle OPT__completion_.
1185 * selftest-run-tests.c (selftest::run_tests): Add
1186 opt_proposer_c_tests.
1187 * selftest.c (assert_str_startswith): New.
1188 * selftest.h (assert_str_startswith): Likewise.
1189 (opt_proposer_c_tests): New.
1190 (ASSERT_STR_STARTSWITH): Likewise.
1192 2018-06-28 Martin Liska <mliska@suse.cz>
1194 * Makefile.in: Add opt-suggestions.o.
1195 * gcc-main.c: Include opt-suggestions.h.
1196 * gcc.c (driver::driver): Likewise.
1197 (driver::~driver): Remove m_option_suggestions.
1198 (driver::build_option_suggestions): Moved to option_proposer.
1199 (driver::suggest_option): Likewise.
1200 (driver::handle_unrecognized_options): Use option_proposer.
1201 * gcc.h (class driver): Add new memver m_option_proposer.
1202 * opt-suggestions.c: New file.
1203 * opt-suggestions.h: New file.
1205 2018-06-28 Martin Liska <mliska@suse.cz>
1207 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1208 (auto_string_vec::~auto_string_vec): Likewise.
1210 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1212 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1213 prevent_decl_creation_for_types fields up and add reset_location field.
1214 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1215 statement if id->reset_location is true.
1216 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1217 id->reset_location is true.
1218 (copy_phis_for_bb): Force input_location on the arguments if
1219 id->reset_location is true.
1220 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1221 is set on the function to be inlined.
1223 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1225 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1227 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1229 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1230 registers for Pmode.
1231 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1232 hard registers for the clobbered pseudo.
1234 2018-06-27 Paul Koning <ni1d@arrl.net>
1236 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1237 mutually exclusive options.
1238 * config/pdp11/constraints.md (h): New constraint.
1239 (O): Update definition to match shift code generation.
1240 (D): New constraint.
1241 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1243 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1245 (output_jump): Change arguments.
1246 (pdp11_fixed_cc_regs): New function.
1247 (pdp11_cc_mode): Ditto.
1248 (pdp11_expand_shift): Ditto.
1249 (pdp11_assemble_shift): Ditto.
1250 (pdp11_small_shift): Ditto.
1251 (pdp11_branch_cost): Remove.
1252 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1254 (pdp11_register_move_cost): Update for CC registers.
1255 (pdp11_rtx_costs): Add case for LSHIFTRT.
1256 (pdp11_output_jump): Add CCNZ mode conditional branches.
1257 (notice_update_cc_on_set): Remove.
1258 (pdp11_cc_mode): New function.
1259 (simple_memory_operand): Correct pre/post decrement case.
1260 (no_side_effect_operand): New function.
1261 (pdp11_regno_reg_class): Add CC_REGS class.
1262 (pdp11_fixed_cc_regs): New function.
1263 (pdp11_small_shift): New function.
1264 (pdp11_expand_shift): New function to expand shift insns.
1265 (pdp11_assemble_shift): New function to output shifts.
1266 (pdp11_branch_cost): Remove.
1267 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1268 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1269 (WCHAR_TYPE): Ditto.
1270 (PTRDIFF_TYPE): Ditto.
1271 (ADJUST_INSN_LENGTH): New macro.
1272 (FIXED_REGISTERS): Add CC registers.
1273 (CALL_USED_REGISTERS): Ditto.
1275 (REG_CLASS_NAMES): Ditto.
1276 (REG_CLASS_CONTENTS): Ditto.
1277 (SELECT_CC_MODE): Use new function.
1278 (TARGET_FLAGS_REGNUM): New macro.
1279 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1280 (cc0_reg_rtx): Remove.
1281 (CC_STATUS_MDEP): Remove.
1282 (CC_STATUS_MDEFP_INIT): Remove.
1283 (CC_IN_FPU): Remove.
1284 (NOTICE_UPDATE_CC): Remove.
1285 (REGISTER_NAMES): Add CC registers.
1286 (BRANCH_COST): Change to constant 1.
1287 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1289 * config/pdp11/pdp11.opt (mbcopy): Remove.
1290 (mbcopy-builtin): Remove.
1291 (mbranch-cheap): Remove.
1292 (mbranch-expensive): Remove.
1293 * config/pdp11/predicates.md (expand_shift_operand): Update to
1294 match shift code generation.
1295 (ccnz_operator): New predicate.
1296 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1297 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1298 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1300 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1301 description of O constraint.
1303 2018-06-27 Jeff Law <law@redhat.com>
1304 Austin Law <austinklaw@gmail.com>
1306 * config/v850/v850.md (addsi3_set_flags): New pattern.
1307 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1308 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1309 (zero_extendhisi2_v850_set_flags): Likewise.
1310 (zero_extendqisi2_v850_set_flags): Likewise.
1311 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1312 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1313 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1315 * config/v850/v850-protos.h (notice_update_cc): Remove.
1316 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1317 (v850_print_operand): Handle 'D' and "d".
1318 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1319 Add handling of arithmetic/logical operations compared against zero.
1320 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1321 Do not look at v850_compare_op, instead get mode from last argument.
1322 (v850_gen_compare): Remove
1323 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1324 after reload for prologue insns.
1325 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1327 (construct_save_jarl): Likewise.
1328 (TARGET_FLAGS_REGNUM): Define.
1329 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1330 (NOTICE_UPDATE_CC): Remove.
1331 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1332 than cc0. Conditionalize on reload_completed.
1333 (cmpsi_insn, setfcc_insn): Likewise.
1334 (tst1 splitter): Turn into define_and_split which sets the flags
1336 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1337 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1338 (cstoresf4, cstoredf4): Clobber the flags.
1339 (cmpsi, cmpsf, cmpdf): Remove expanders.
1340 (setf_insn): Remove pattern.
1341 (addsi3): Turn into define_and_split which clobbers the flags after
1342 reload and a suitable pattern (addsi3_clobber_flags) for use after
1344 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1345 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1346 (ashrsi3, ashrsi3_v850e2): Likewise.
1347 (bins): Clobber the flags.
1348 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1349 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1350 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1351 (call_value_internal_short, call_value_internal_long): Likewise.
1352 (callt_save_interrupt, callt_return_interrupt): Likewise.
1353 (save_interrupt, return_interrupt): Likewise.
1354 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1355 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1356 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1357 (All FP comparisons): Only allow after reload has completed.
1359 (divh, divhu): Tweak output template.
1360 (branch_z_normal, branch_z_invert): Remove
1361 (branch_nz_normal, branch_nz_invert): Likewise.
1362 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1364 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1365 * config/v850/v850.c (notice_update_cc): Remove.
1366 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1367 (CC_NO_CARRY): Likewise.
1368 (NOTICE_UPDATE_CC): Define to nothing.
1369 * config/v850/v850.md: Remove block comment on cc0 handling
1370 Remove "cc" attribute from all patterns. Remove cc_status handling
1371 from all patterns. Minor formatting fixes.
1373 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1375 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1376 (cortex-a76.cortex-a55): Likewise.
1377 * config/aarch64/aarch64-tune.md: Regenerate.
1378 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1379 cortex-a76.cortex-a55.
1381 2018-06-27 Jeff Law <law@redhat.com>
1383 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1384 (MULTILIB_DIRNAMES): Similarly.
1386 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1388 * gimple.h (gimple_return_retbnd): Delete.
1389 (gimple_return_set_retbnd): Likewise.
1390 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1391 gimple_return_set_retbnd.
1392 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1393 gimple_return_retbnd and adjust.
1394 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1395 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1396 Explicitly return NULL in a couple more cases. Move assertion
1397 on debug statements and remove unreachable code.
1398 (reset_debug_binding): Do not test id->retbnd.
1399 (expand_call_inline): Do not set it.
1401 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1403 * configure.ac: Add --disable-gcov option.
1404 * configure: Regenerate.
1405 * Makefile.in: Honour @enable_gcov@.
1406 * doc/install.texi: Document --disable-gcov.
1408 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1410 * config/arm/arm-cpus.in (cortex-a76): New entry.
1411 (cortex-a76.cortex-a55): Likewise.
1412 * config/arm/arm-tables.opt: Regenerate.
1413 * config/arm/arm-tune.md: Likewise.
1414 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1415 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1416 cortex-a76.cortex-a55.
1418 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1421 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1423 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1425 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1427 (EPILOGUE_USES): Likewise.
1429 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1431 * tree-inline.c (remap_location): New function extracted from...
1432 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1433 (copy_phis_for_bb): ...here. Call remap_location.
1434 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1436 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1438 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1439 unaligned vsx for 16B memset.
1441 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1444 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1445 ieee128_float_type_node to long_double_type_node unless
1446 TARGET_LONG_DOUBLE_128 is set.
1448 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1450 * cfgloop.c (get_loop_location): Convert return type from
1451 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1452 by implicit construction from rtx_insn *, and using
1453 dump_user_location_t::from_function_decl for the fallback case.
1454 * cfgloop.h (get_loop_location): Convert return type from
1455 location_t to dump_user_location_t.
1456 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1457 dump_printf_loc to pass in a dump_location_t rather than a
1458 location_t, via the gimple stmt.
1459 * coverage.c (get_coverage_counts): Update calls to
1460 dump_printf_loc to pass in dump_location_t rather than a
1462 * doc/optinfo.texi (Dump types): Convert example of
1463 dump_printf_loc from taking "locus" to taking "insn". Update
1464 description of the "_loc" calls to cover dump_location_t.
1465 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1467 (dump_user_location_t::dump_user_location_t): New constructors,
1468 from gimple *stmt and rtx_insn *.
1469 (dump_user_location_t::from_function_decl): New function.
1470 (dump_loc): Make static.
1471 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1472 const dump_location_t &.
1473 (dump_generic_expr_loc): Delete.
1474 (dump_printf_loc): Convert param "loc" from location_t to
1475 const dump_location_t &.
1476 (selftest::test_impl_location): New function.
1477 (selftest::dumpfile_c_tests): New function.
1478 * dumpfile.h: Include "profile-count.h".
1479 (class dump_user_location_t): New class.
1480 (struct dump_impl_location_t): New struct.
1481 (class dump_location_t): New class.
1482 (dump_printf_loc): Convert 2nd param from source_location to
1483 const dump_location_t &.
1484 (dump_generic_expr_loc): Delete.
1485 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1486 const dump_location_t &.
1487 * gimple-fold.c (fold_gimple_assign): Update call to
1488 dump_printf_loc to pass in a dump_location_t rather than a
1489 location_t, via the gimple stmt.
1490 (gimple_fold_call): Likewise.
1491 * gimple-loop-interchange.cc
1492 (loop_cand::analyze_iloop_reduction_var): Update for change to
1493 check_reduction_path.
1494 (tree_loop_interchange::interchange): Update for change to
1496 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1497 change in return-type of find_loop_location.
1498 (graphite_regenerate_ast_isl): Likewise.
1499 * graphite-optimize-isl.c (optimize_isl): Likewise.
1500 * graphite.c (graphite_transform_loops): Likewise.
1501 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1502 pass in a dump_location_t rather than a location_t, via the
1504 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1505 * ipa.c (walk_polymorphic_call_targets): Likewise.
1506 * loop-unroll.c (report_unroll): Convert "locus" param from
1507 location_t to dump_location_t.
1508 (decide_unrolling): Update for change to get_loop_location's
1510 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1511 location_t to dump_user_location_t.
1512 (grid_find_single_omp_among_assignments_1): Updates calls to
1513 dump_printf_loc to pass in a dump_location_t rather than a
1514 location_t, via the gimple stmt.
1515 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1516 location_t to dump_location_t. Updates calls to dump_printf_loc
1517 to pass in a dump_location_t rather than a location_t, via the
1519 (grid_inner_loop_gridifiable_p): Likewise.
1520 (grid_dist_follows_simple_pattern): Likewise.
1521 (grid_gfor_follows_tiling_pattern): Likewise.
1522 (grid_target_follows_gridifiable_pattern): Likewise.
1523 (grid_attempt_target_gridification): Convert initialization
1524 of local "grid" from memset to zero-initialization; FIXME: does
1525 this require C++11? Update call to dump_printf_loc to pass in a
1526 optinfo_location rather than a location_t, via the gimple stmt.
1527 * profile.c (read_profile_edge_counts): Updates call to
1528 dump_printf_loc to pass in a dump_location_t rather than a
1530 (compute_branch_probabilities): Likewise.
1531 * selftest-run-tests.c (selftest::run_tests): Call
1533 * selftest.h (dumpfile_c_tests): New decl.
1534 * tree-loop-distribution.c (pass_loop_distribution::execute):
1535 Update for change in return type of find_loop_location.
1536 * tree-parloops.c (parallelize_loops): Likewise.
1537 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1538 "locus" from location_t to dump_user_location_t.
1539 (canonicalize_loop_induction_variables): Likewise.
1540 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1541 for change in return type of find_loop_location.
1542 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1543 to dump_printf_loc to pass in a dump_location_t rather than a
1544 location_t, via the stmt.
1545 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1547 * tree-vect-loop-manip.c (find_loop_location): Convert return
1548 type from source_location to dump_user_location_t.
1549 (vect_do_peeling): Update for above change.
1550 (vect_loop_versioning): Update for change in type of
1552 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1553 from location_t to dump_user_location_t.
1554 (vect_estimate_min_profitable_iters): Update for change in type
1556 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1557 location_t to dump_location_t.
1558 (vect_slp_bb): Update for change in type of vect_location.
1559 * tree-vectorizer.c (vect_location): Convert from source_location
1560 to dump_user_location_t.
1561 (try_vectorize_loop_1): Update for change in vect_location's type.
1562 (vectorize_loops): Likewise.
1563 (increase_alignment): Likewise.
1564 * tree-vectorizer.h (vect_location): Convert from source_location
1565 to dump_user_location_t.
1566 (find_loop_location): Convert return type from source_location to
1567 dump_user_location_t.
1568 (check_reduction_path): Convert 1st param from location_t to
1569 dump_user_location_t.
1570 * value-prof.c (check_counter): Update call to dump_printf_loc to
1571 pass in a dump_user_location_t rather than a location_t; update
1572 call to error_at for change in type of "locus".
1573 (check_ic_target): Update call to dump_printf_loc to
1574 pass in a dump_user_location_t rather than a location_t, via the
1577 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1579 * config/s390/s390.h (enum processor_flags): Do not use
1580 default tune parameter when -march was specified.
1582 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1585 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1586 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1589 2018-06-26 Richard Biener <rguenther@suse.de>
1591 PR tree-optimization/86287
1593 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1594 (vect_analyze_loop): Initialize n_stmts.
1596 2018-06-26 Richard Biener <rguenther@suse.de>
1599 * fold-const.c (fold_convertible_p): Pointer extension
1602 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1605 * dwarf2out.c (loc_list_has_views): Adjust comments.
1606 (dw_loc_list): Split single cross-partition range with
1609 2018-06-25 Jeff Law <law@redhat.com>
1611 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1612 on -mbig-switch by default.
1614 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1616 (const_float_0_operand): Remove unused predicate.
1617 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1618 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1619 (recipsf2): New expander. Original pattern now called
1621 (recipdf2, recipdf2_insn): Similarly.
1622 (rsqrtsf2, rsqrtsf2_insn): Similarly
1623 (rsqrtdf2, rsqrtdf2_insn): Similarly
1625 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1627 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1628 Simplify logic for FreeBSD (twice).
1630 2018-06-25 Martin Sebor <msebor@redhat.com>
1632 PR tree-optimization/86204
1633 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1634 a strnlen result if it's less than the length of the string.
1636 2018-06-25 Martin Sebor <msebor@redhat.com>
1638 PR tree-optimization/85700
1639 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1640 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1641 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1643 2018-06-25 Martin Sebor <msebor@redhat.com>
1645 * doc/extend.texi (Zero-length arrays): Update and clarify.
1647 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1649 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1650 added IEEE/IBM long double multilib support on PowerPC little
1651 endian Linux systems.
1652 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1653 (MULTILIB_DEFAULTS): Likewise.
1654 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1656 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1657 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1658 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1660 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1663 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1664 (REORDER_45): Likewise.
1666 2018-06-25 Jeff Law <law@redhat.com>
1668 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1669 dividend to 32 bits. Adjust length.
1670 (udivmodhi4): Cleanup output template. Fix length.
1672 2018-06-25 Carl Love <cel@us.ibm.com>
1674 * config/rs6000/vsx.md: Change word selector to prefered location.
1676 2018-06-25 Richard Biener <rguenther@suse.de>
1678 PR tree-optimization/86304
1679 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1680 epilogue-if-converted loops as well.
1682 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1684 * lto-section-out.c (lto_begin_section): Do not print section
1685 name for noaddr and unnumbered dumps.
1687 2018-06-25 Richard Biener <rguenther@suse.de>
1689 * tree-vectorizer.h (struct vec_info_shared): New structure
1690 with parts split out from struct vec_info and loop_nest from
1691 struct _loop_vec_info.
1692 (struct vec_info): Adjust accordingly.
1693 (struct _loop_vec_info): Likewise.
1694 (LOOP_VINFO_LOOP_NEST): Adjust.
1695 (LOOP_VINFO_DATAREFS): Likewise.
1696 (LOOP_VINFO_DDRS): Likewise.
1697 (struct _bb_vec_info): Likewise.
1698 (BB_VINFO_DATAREFS): Likewise.
1699 (BB_VINFO_DDRS): Likewise.
1700 (struct _stmt_vec_info): Add dr_aux member.
1701 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1702 (DR_MISALIGNMENT_UNINITIALIZED): New.
1703 (set_dr_misalignment): Adjust.
1704 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1705 (vect_analyze_loop): Adjust prototype.
1706 (vect_analyze_loop_form): Likewise.
1707 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1708 Compute dependences lazily.
1709 (vect_record_base_alignments): Use shared datarefs/ddrs.
1710 (vect_verify_datarefs_alignment): Likewise.
1711 (vect_analyze_data_refs_alignment): Likewise.
1712 (vect_analyze_data_ref_accesses): Likewise.
1713 (vect_analyze_data_refs): Likewise.
1714 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1715 constructor parameter for shared part.
1716 (vect_analyze_loop_form): Pass in shared part and adjust.
1717 (vect_analyze_loop_2): Pass in storage for the number of
1718 stmts. Move loop nest finding to the caller. Compute
1720 (vect_analyze_loop): Pass in shared part.
1721 (vect_transform_loop): Verify shared datarefs are unchanged.
1722 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1723 constructor parameter for shared part.
1724 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1725 (vect_slp_bb): Verify shared datarefs are unchanged before
1727 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1729 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1730 DR_MISALIGNMENT_UNINITIALIZED.
1731 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1732 parameter for shared part.
1733 (vec_info::~vec_info): Adjust.
1734 (vec_info_shared::vec_info_shared): New.
1735 (vec_info_shared::~vec_info_shared): Likewise.
1736 (vec_info_shared::save_datarefs): Likewise.
1737 (vec_info_shared::check_datarefs): Likewise.
1738 (try_vectorize_loop_1): Construct shared part live for analyses
1739 of a single loop for multiple vector sizes.
1740 * tree-parloops.c (gather_scalar_reductions): Adjust.
1742 2018-06-25 Richard Biener <rguenther@suse.de>
1744 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1745 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1746 (vect_analyze_data_refs): Remove similar code from here and
1747 simplify accordingly.
1749 2018-06-25 Richard Biener <rguenther@suse.de>
1751 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1752 for reverse storage order accesses rather than asserting
1753 they cannot happen here.
1755 2018-06-25 Tom de Vries <tdevries@suse.de>
1758 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1759 Use data16 instead of .byte for insn prefix.
1761 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1764 * parser.c (make_char_string_pack): Pass this literal chars
1765 through cpp_interpret_string.
1766 (cp_parser_userdef_numeric_literal): Check the result of
1767 make_char_string_pack.
1769 2018-06-24 Maya Rashish <coypu@sdf.org>
1771 * ginclude/stddef.h: Simplify conditions around avoiding
1772 re-definition of __size_t.
1774 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1776 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1777 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1779 2018-06-22 Maya Rashish <coypu@sdf.org>
1781 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1782 made to OpenBSD and NetBSD through the years.
1784 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1786 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1787 behavior of vec_pack (vector double, vector double) to match
1788 behavior of vec_float2 (vector double, vector double).
1790 2018-06-22 Olivier Hainque <hainque@adacore.com>
1792 * gimplify.c (gimplify_function_tree): Prevent creation
1793 of a trampoline for the address of the current function
1794 passed to entry/exit instrumentation hooks.
1796 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1799 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1802 2018-06-22 Martin Liska <mliska@suse.cz>
1804 PR tree-optimization/86263
1805 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1806 Bail out if is_enabled is false.
1807 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1809 (jump_table_cluster::is_enabled): New function.
1811 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1813 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1814 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1815 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1816 (lto_input_ts_binfo_tree_pointers): Likewise.
1817 * tree-streamer-out.c (streamer_write_tree_bitfields,
1818 write_ts_binfo_tree_pointers): Likewise.
1819 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1821 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1823 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1825 2018-06-22 Martin Liska <mliska@suse.cz>
1827 * symbol-summary.h (get): Make it pure and inline move
1828 functionality from ::get function.
1829 (get): Remove and inline into ::get and ::get_create.
1830 (get_create): Move code from ::get function.
1832 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1835 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1836 -x assembler-with-cpp.
1838 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1840 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1841 _FILE_OFFSET_BITS=64 for C++.
1843 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1845 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1846 conversion insn that shows up when pr85657-3.c is compiled using
1847 IEEE 128-bit long double.
1848 (neg<mode>2_internal): Use the correct mode to check whether the
1849 mode is IBM extended.
1850 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1851 multiply and divide external functions from being created more
1854 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1856 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1858 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1859 the edge can be forwarded.
1860 (cfg_layout_merge_blocks): Likewise.
1862 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1864 * except.c (finish_eh_generation): Commit edge insertions only after
1865 the EH edges have been redirected from post-landing to landing pads.
1867 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1869 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1870 create_tmp_var_for to create the FRAME decl.
1871 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1873 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1875 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1876 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1877 debug statement when resetting its value.
1878 (expand_call_inline): Copy the locus of the call onto the assignment
1879 of the return value, if any. Use local variable in more cases.
1881 2018-06-21 Martin Liska <mliska@suse.cz>
1883 * ipa-pure-const.c (propagate_nothrow): Use
1884 funct_state_summaries->get.
1885 (dump_malloc_lattice): Likewise.
1886 (propagate_malloc): Likewise.
1888 2018-06-21 Richard Biener <rguenther@suse.de>
1890 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1891 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1892 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1894 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1895 BLOCK_ABSTRACT_ORIGIN unconditionally.
1897 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1899 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1901 * ipa-reference.c (ipa_reference_c_finalize): Delete
1902 ipa_ref_opt_sum_summaries and set it to NULL.
1904 2018-06-21 Tom de Vries <tdevries@suse.de>
1906 PR tree-optimization/85859
1907 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1908 test with comment from bb_no_side_effects_p.
1910 2018-06-21 Richard Biener <rguenther@suse.de>
1912 PR tree-optimization/86232
1913 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1914 max for constant niter.
1916 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1918 * config/aarch64/aarch64-simd.md
1919 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1921 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1923 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1924 Make opernads of the unspec commutative.
1926 2018-06-21 Richard Biener <rguenther@suse.de>
1928 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1929 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1930 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1931 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1933 (vect_record_base_alignments): Do not record base alignment
1934 for gathers or scatters.
1935 (vect_compute_data_ref_alignment): Drop return value that is always
1936 true. Bail out early for gathers or scatters.
1937 (vect_enhance_data_refs_alignment): Bail out early for gathers
1939 (vect_find_same_alignment_drs): Likewise.
1940 (vect_analyze_data_refs_alignment): Remove dead code.
1941 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1942 (vect_analyze_data_refs): For possible gathers or scatters do
1943 not create an alternate DR, just check their possible validity
1944 and mark them. Adjust DECL_NONALIASED handling to not rely
1946 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1947 update inits of gathers or scatters.
1948 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1949 Also copy gather/scatter flag to pattern vinfo.
1951 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1953 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1954 behavior of vec_packsu (vector unsigned long long, vector unsigned
1955 long long) to match behavior of vec_packs with same signature.
1957 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1958 Thomas Schwinge <thomas@codesourcery.com>
1959 Cesar Philippidis <cesar@codesourcery.com>
1961 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1962 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1963 (gimplify_adjust_omp_clauses): Likewise.
1964 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1965 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1966 (gimplify_omp_target_update): Update handling of acc update and
1968 * omp-low.c (install_var_field): Remove unused parameter
1969 base_pointers_restrict.
1970 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1971 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1973 (omp_target_base_pointers_restrict_p): Delete.
1974 (scan_omp_target): Update call to scan_sharing_clauses.
1975 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1977 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1978 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1979 (convert_local_omp_clauses): Likewise.
1980 * tree-pretty-print.c (dump_omp_clause): Likewise.
1981 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1983 (omp_clause_code_name): Likewise.
1985 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1988 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1991 PR tree-optimization/86231
1992 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1993 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1995 2018-06-20 Tom de Vries <tdevries@suse.de>
1997 PR tree-optimization/86097
1998 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1999 iv type if signedness of iv type is not the same as that of *nit.
2001 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2003 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2004 EDGE_EH edges, verify they are all EDGE_EH.
2006 2018-06-20 Maya Rashish <coypu@sdf.org>
2008 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2010 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2012 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2013 * config/aarch64/aarch64.c (xgene1_tunings): Add
2014 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2015 (aarch64_mode_valid_for_sched_fusion_p):
2016 Allow 16-byte modes.
2017 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2018 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2020 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2022 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2023 * config/aarch64/iterators.md (VQ2): New mode iterator.
2025 2018-06-20 Martin Liska <mliska@suse.cz>
2027 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2028 Change default ratio from 10 to 8.
2030 2018-06-20 Martin Liska <mliska@suse.cz>
2032 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2034 (bit_test_cluster::find_bit_tests): Likewise.
2035 (switch_decision_tree::analyze_switch_statement): Find clusters.
2036 * tree-switch-conversion.h (struct jump_table_cluster): Document
2039 2018-06-20 Martin Liska <mliska@suse.cz>
2041 * tree-switch-conversion.c (switch_conversion::collect):
2042 Record m_uniq property.
2043 (switch_conversion::expand): Bail out for special conditions.
2044 (group_cluster::~group_cluster): New.
2045 (group_cluster::group_cluster): Likewise.
2046 (group_cluster::dump): Likewise.
2047 (jump_table_cluster::emit): New.
2048 (switch_decision_tree::fix_phi_operands_for_edges): New.
2049 (struct case_node): Remove struct.
2050 (jump_table_cluster::can_be_handled): New.
2051 (case_values_threshold): Moved to header.
2052 (reset_out_edges_aux): Likewise.
2053 (jump_table_cluster::is_beneficial): New.
2054 (bit_test_cluster::can_be_handled): Likewise.
2055 (add_case_node): Remove.
2056 (bit_test_cluster::is_beneficial): New.
2057 (case_bit_test::cmp): New.
2058 (bit_test_cluster::emit): New.
2059 (expand_switch_as_decision_tree_p): Remove.
2060 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2061 (fix_phi_operands_for_edge): Likewise.
2062 (switch_decision_tree::analyze_switch_statement): New.
2063 (compute_cases_per_edge): Move ...
2064 (switch_decision_tree::compute_cases_per_edge): ... here.
2065 (try_switch_expansion): Likewise.
2066 (switch_decision_tree::try_switch_expansion): Likewise.
2067 (record_phi_operand_mapping): Likewise.
2068 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2069 (emit_case_decision_tree): Likewise.
2070 (switch_decision_tree::emit): Likewise.
2071 (balance_case_nodes): Likewise.
2072 (switch_decision_tree::balance_case_nodes): Likewise.
2073 (dump_case_nodes): Likewise.
2074 (switch_decision_tree::dump_case_nodes): Likewise.
2075 (emit_jump): Likewise.
2076 (switch_decision_tree::emit_jump): Likewise.
2077 (emit_cmp_and_jump_insns): Likewise.
2078 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2079 (emit_case_nodes): Likewise.
2080 (switch_decision_tree::emit_case_nodes): Likewise.
2081 (conditional_probability): Remove.
2082 * tree-switch-conversion.h (enum cluster_type): New.
2084 (struct cluster): Likewise.
2085 (cluster::cluster): Likewise.
2086 (struct simple_cluster): Likewise.
2087 (simple_cluster::simple_cluster): Likewise.
2088 (struct group_cluster): Likewise.
2089 (struct jump_table_cluster): Likewise.
2090 (struct bit_test_cluster): Likewise.
2091 (struct min_cluster_item): Likewise.
2092 (struct case_tree_node): Likewise.
2093 (case_tree_node::case_tree_node): Likewise.
2094 (jump_table_cluster::case_values_threshold): Likewise.
2095 (struct case_bit_test): Likewise.
2096 (struct switch_decision_tree): Likewise.
2097 (struct switch_conversion): Likewise.
2098 (switch_decision_tree::reset_out_edges_aux): Likewise.
2100 2018-06-20 Martin Liska <mliska@suse.cz>
2102 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2103 (hoist_edge_and_branch_if_true): Likewise.
2104 (expand_switch_using_bit_tests_p): Likewise.
2105 (struct case_bit_test): Likewise.
2106 (case_bit_test_cmp): Likewise.
2107 (emit_case_bit_tests): Likewise.
2108 (switch_conversion::switch_conversion): New class.
2109 (struct switch_conv_info): Remove old struct.
2110 (collect_switch_conv_info): More to ...
2111 (switch_conversion::collect): ... this.
2112 (check_range): Likewise.
2113 (switch_conversion::check_range): Likewise.
2114 (check_all_empty_except_final): Likewise.
2115 (switch_conversion::check_all_empty_except_final): Likewise.
2116 (check_final_bb): Likewise.
2117 (switch_conversion::check_final_bb): Likewise.
2118 (create_temp_arrays): Likewise.
2119 (switch_conversion::create_temp_arrays): Likewise.
2120 (free_temp_arrays): Likewise.
2121 (gather_default_values): Likewise.
2122 (switch_conversion::gather_default_values): Likewise.
2123 (build_constructors): Likewise.
2124 (switch_conversion::build_constructors): Likewise.
2125 (constructor_contains_same_values_p): Likewise.
2126 (switch_conversion::contains_same_values_p): Likewise.
2127 (array_value_type): Likewise.
2128 (switch_conversion::array_value_type): Likewise.
2129 (build_one_array): Likewise.
2130 (switch_conversion::build_one_array): Likewise.
2131 (build_arrays): Likewise.
2132 (switch_conversion::build_arrays): Likewise.
2133 (gen_def_assigns): Likewise.
2134 (switch_conversion::gen_def_assigns): Likewise.
2135 (prune_bbs): Likewise.
2136 (switch_conversion::prune_bbs): Likewise.
2137 (fix_phi_nodes): Likewise.
2138 (switch_conversion::fix_phi_nodes): Likewise.
2139 (gen_inbound_check): Likewise.
2140 (switch_conversion::gen_inbound_check): Likewise.
2141 (process_switch): Use the newly created class.
2142 (switch_conversion::expand): New.
2143 (switch_conversion::~switch_conversion): New.
2144 * tree-switch-conversion.h: New file.
2146 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2148 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2149 tree-vect-patterns.c.
2150 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2151 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2152 (vect_recog_sad_pattern): Likewise.
2153 (vect_recog_widen_sum_pattern): Likewise.
2154 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2155 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2156 (vect_recog_rotate_pattern): Likewise.
2157 (vect_recog_mult_pattern): Likewise.
2158 (vect_recog_vector_vector_shift_pattern): Likewise.
2159 (vect_recog_divmod_pattern): Likewise.
2160 (vect_recog_mixed_size_cond_pattern): Likewise.
2161 (vect_recog_bool_pattern): Likewise.
2162 (vect_recog_mask_conversion_pattern): Likewise.
2163 (vect_try_gather_scatter_pattern): Likewise.
2164 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2165 (vect_recog_over_widening_pattern): Likewise.
2166 (vect_recog_gather_scatter_pattern): Likewise.
2167 (vect_recog_func_ptr): Move from tree-vectorizer.h
2168 (vect_vect_recog_func_ptrs): Move further down the file.
2169 (vect_recog_func): Likewise. Remove the third argument.
2170 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2171 (vect_pattern_recog_1): Expect the pattern function to do any
2172 necessary target tests. Also expect it to provide a vector type.
2173 Remove the type_in handling.
2175 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2177 * tree-vect-patterns.c (vect_pattern_detected): New function.
2178 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2179 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2180 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2181 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2182 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2183 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2184 (vect_recog_mask_conversion_pattern)
2185 (vect_try_gather_scatter_pattern): Likewise.
2187 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2189 * tree-vect-patterns.c (vect_get_internal_def): New function.
2190 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2191 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2192 (search_type_for_mask_1): Use it.
2194 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2196 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2197 redundant WIDEN_SUM_EXPR handling.
2198 (vect_recog_sad_pattern): Likewise.
2200 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2202 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2203 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2204 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2205 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2206 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2208 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2210 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2211 use the stmt_vec_info of the original bb statement for the
2212 new zero assignment, even if the call is part of a pattern.
2214 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2216 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2217 that the sequence is attached to the original statement rather
2218 than the pattern statement.
2219 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2220 PATTERN_DEF_SEQ from the original statement rather than
2221 the main pattern statement.
2222 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2223 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2224 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2226 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2228 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2229 definition statements before the early exit for statements that aren't
2231 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2233 (vect_transform_loop): ...here. Process pattern definition
2234 statements without first checking whether the main pattern
2235 statement is live or relevant.
2237 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2239 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2240 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2242 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2244 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2245 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2246 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2247 (expand_block_compare): Change select_block_compare_mode call.
2248 (expand_strncmp_align_check): Use new functions, fix comment.
2249 (emit_final_str_compare_gpr): New function.
2250 (expand_strn_compare): Refactor and clean up code.
2251 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2253 2018-06-19 Tony Reix <tony.reix@atos.com>
2254 Damien Bergamini <damien.bergamini@atos.com>
2255 David Edelsohn <dje.gcc@gmail.com>
2257 * collect2.c (static_obj): New variable.
2258 (static_libs): New variable.
2259 (is_in_list): Uncomment declaration.
2260 (main): Track AIX libraries linked statically.
2261 (is_in_list): Uncomment definition.
2262 (scan_prog_file): Don't add AIX shared libraries initializer
2263 to constructor list if linking statically.
2265 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2267 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2269 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2271 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2273 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2276 2018-06-19 Martin Liska <mliska@suse.cz>
2278 * config/i386/i386.c (ix86_can_inline_p): Do not use
2279 ipa_fn_summaries::get_create.
2280 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2282 (devirtualization_time_bonus): Likewise.
2283 (ipcp_propagate_stage): Likewise.
2284 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2285 (edge_set_predicate): Likewise.
2286 (evaluate_conditions_for_known_args): Likewise.
2287 (evaluate_properties_for_edge): Likewise.
2288 (ipa_call_summary::reset): Tranform to ...
2289 (ipa_call_summary::~ipa_call_summary): ... this.
2290 (ipa_fn_summary::reset): Transform to ...
2291 (ipa_fn_summary::~ipa_fn_summary): ... this.
2292 (ipa_fn_summary_t::remove): Rename to ...
2293 (ipa_fn_summary_t::remove_callees): ... this.
2294 (ipa_fn_summary_t::duplicate): Use placement new
2295 instead of memory copy.
2296 (ipa_call_summary_t::duplicate): Likewise.
2297 (ipa_call_summary_t::remove): Remove.
2298 (dump_ipa_call_summary): Change get_create to get.
2299 (ipa_dump_fn_summary): Dump only when summary exists.
2300 (analyze_function_body): Use symbol_summary::get instead
2302 (compute_fn_summary): Likewise.
2303 (estimate_edge_devirt_benefit): Likewise.
2304 (estimate_edge_size_and_time): Likewise.
2305 (inline_update_callee_summaries): Likewise.
2306 (remap_edge_change_prob): Likewise.
2307 (remap_edge_summaries): Likewise.
2308 (ipa_merge_fn_summary_after_inlining): Likewise.
2309 (write_ipa_call_summary): Likewise.
2310 (ipa_fn_summary_write): Likewise.
2311 (ipa_free_fn_summary): Likewise.
2312 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2313 (struct ipa_call_summary): Likewise.
2314 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2316 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2317 (estimate_size_after_inlining): Likewise.
2318 (estimate_growth): Likewise.
2319 (growth_likely_positive): Likewise.
2320 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2321 (inline_call): Likewise.
2322 * ipa-inline.c (caller_growth_limits): Likewise.
2323 (can_inline_edge_p): Likewise.
2324 (can_inline_edge_by_limits_p): Likewise.
2325 (compute_uninlined_call_time): Likewise.
2326 (compute_inlined_call_time): Likewise.
2327 (want_inline_small_function_p): Likewise.
2328 (edge_badness): Likewise.
2329 (update_caller_keys): Likewise.
2330 (update_callee_keys): Likewise.
2331 (inline_small_functions): Likewise.
2332 (inline_to_all_callers_1): Likewise.
2333 (dump_overall_stats): Likewise.
2334 (early_inline_small_functions): Likewise.
2335 (early_inliner): Likewise.
2336 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2337 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2338 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2339 * ipa-split.c (execute_split_functions): Likewise.
2340 * symbol-summary.h: Likewise.
2341 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2343 2018-06-19 Richard Biener <rguenther@suse.de>
2345 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2346 (vectorize_loops): ... here. Fix dbgcnt handling.
2347 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2349 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2352 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2353 ieee128 argument takes up only one (vector) register, not two (floating
2356 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2358 * gimplify.c (gimplify_init_constructor): Really never clear for an
2359 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2361 2018-06-19 Richard Biener <rguenther@suse.de>
2363 PR tree-optimization/86179
2364 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2365 after failed recognition.
2367 2018-06-18 Martin Sebor <msebor@redhat.com>
2370 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2371 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2372 Handle integer subtraction.
2373 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2374 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2376 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2378 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2379 param from rtx to rtx_insn *.
2380 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2382 (frv_ifcvt_modify_insn): Likwise.
2383 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2384 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2385 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2386 as_a <rtx_insn *> cast to local "unprotected_region" once
2387 it's been established that it's not NULL or pc_rtx.
2388 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2389 param "sethi" from rtx to rtx_insn *.
2390 (nds32_group_float_insns): Likewise for param "insn".
2391 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2393 (vax_output_int_subtract): Likewise.
2394 * config/vax/vax.c (vax_output_int_add): Likewise for param
2396 (vax_output_int_subtract): Likewise.
2397 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2398 (emit_pattern_after): Likewise for param "after".
2399 (emit_insn_after): Likewise.
2400 (emit_jump_insn_after): Likewise.
2401 (emit_call_insn_after): Likewise.
2402 (emit_debug_insn_after): Likewise.
2403 (emit_pattern_before): Likewise for param "before".
2404 (emit_insn_before): Likewise.
2405 (emit_jump_insn_before): Likewise.
2406 * final.c (get_insn_template): Likewise for param "insn", removing
2408 * output.h (get_insn_template): Likewise for 2nd param.
2409 * rtl.h (emit_insn_before): Likewise.
2410 (emit_jump_insn_before): Likewise.
2411 (emit_debug_insn_before_noloc): Likewise.
2412 (emit_insn_after): Likewise.
2413 (emit_jump_insn_after): Likewise.
2414 (emit_call_insn_after): Likewise.
2415 (emit_debug_insn_after): Likewise.
2416 (set_insn_deleted): Likewise for param.
2418 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2421 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2422 floating point modes, so that IFmode is numerically greater than
2423 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2424 to declare the ordering. This prevents IFmode from being
2425 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2426 machine. Include rs6000-modes.h to share the fractional values
2427 between genmodes* and the rest of the compiler.
2431 * config/rs6000/rs6000-modes.h: New file.
2432 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2433 meaning of rs6000_long_double_size so that 126..128 selects an
2434 appropriate 128-bit floating point type.
2435 (rs6000_option_override_internal): Likewise.
2436 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2437 (TARGET_LONG_DOUBLE_128): Change the meaning of
2438 rs6000_long_double_size so that 126..128 selects an appropriate
2439 128-bit floating point type.
2440 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2441 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2442 source and destination to match the standard usage.
2443 (truncifkf2): Likewise.
2444 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2445 ISA 2.07 to use an explicit clobber, instead of passing in a
2447 (copysign<mode>3_soft): Likewise.
2449 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2451 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2452 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2453 (vect_slp_analyze_instance_dependence): Likewise.
2454 (vect_enhance_data_refs_alignment): Likewise.
2455 (vect_analyze_data_refs_alignment): Likewise.
2456 (vect_slp_analyze_and_verify_instance_alignment
2457 (vect_analyze_data_ref_accesses): Likewise.
2458 (vect_prune_runtime_alias_test_list): Likewise.
2459 (vect_analyze_data_refs): Likewise.
2460 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2461 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2462 (vect_analyze_scalar_cycles_1): Likewise.
2463 (vect_get_loop_niters): Likewise.
2464 (vect_analyze_loop_form_1): Likewise.
2465 (vect_update_vf_for_slp): Likewise.
2466 (vect_analyze_loop_operations): Likewise.
2467 (vect_analyze_loop): Likewise.
2468 (vectorizable_induction): Likewise.
2469 (vect_transform_loop): Likewise.
2470 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2471 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2472 (vect_make_slp_decision): Likewise.
2473 (vect_detect_hybrid_slp): Likewise.
2474 (vect_slp_analyze_operations): Likewise.
2475 (vect_slp_bb): Likewise.
2476 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2477 (vectorizable_bswap): Likewise.
2478 (vectorizable_call): Likewise.
2479 (vectorizable_simd_clone_call): Likewise.
2480 (vectorizable_conversion): Likewise.
2481 (vectorizable_assignment): Likewise.
2482 (vectorizable_shift): Likewise.
2483 (vectorizable_operation): Likewise.
2484 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2486 2018-06-18 Martin Sebor <msebor@redhat.com>
2488 PR tree-optimization/81384
2489 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2490 * builtins.c (expand_builtin_strnlen): New function.
2491 (expand_builtin): Call it.
2492 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2493 * builtins.def (BUILT_IN_STRNLEN): New.
2494 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2495 Warn for bounds in excess of maximum object size.
2496 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2497 single-value ranges. Handle strnlen.
2498 (handle_builtin_strlen): Handle strnlen.
2499 (strlen_check_and_optimize_stmt): Same.
2500 * doc/extend.texi (Other Builtins): Document strnlen.
2502 2018-06-18 Maya Rashish <coypu@sdf.org>
2504 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2505 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2506 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2508 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2510 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2512 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2514 * tree.c (escaped_string::escape): Replace cast to char * by
2515 const_cast<char *> (unescaped).
2517 2018-06-18 Nick Clifton <nickc@redhat.com>
2520 * tree.c (escaped_string): New class. Converts an unescaped
2521 string into its escaped equivalent.
2522 (warn_deprecated_use): Use the new class to convert the
2523 deprecation message, if present.
2524 (test_escaped_strings): New self test.
2525 (test_c_tests): Add test_escaped_strings.
2526 * doc/extend.texi (deprecated): Add a note that the
2527 deprecation message is affected by the -fmessage-length
2528 option, and that control characters will be escaped.
2529 (#pragma GCC error): Document this pragma.
2530 (#pragma GCC warning): Likewise.
2531 * doc/invoke.texi (-fmessage-length): Document this option's
2532 effect on the #warning and #error preprocessor directives and
2533 the deprecated attribute.
2535 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2537 * tree.c (decl_value_expr_lookup): Revert latest change.
2538 (decl_value_expr_insert): Likewise.
2540 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2542 * gimplify.c (nonlocal_vlas): Delete.
2543 (nonlocal_vla_vars): Likewise.
2544 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2546 (gimplify_body): Do not create and destroy nonlocal_vlas.
2547 * tree-nested.c: Include diagnostic.h.
2548 (use_pointer_in_frame): Tweak.
2549 (lookup_field_for_decl): Add assertion and declare the transformation.
2550 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2551 internal error when the reference is in a wrong context. Do not
2552 create a debug decl by default.
2553 (note_nonlocal_block_vlas): Delete.
2554 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2555 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2556 create a debug decl by default.
2557 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2558 call is in a wrong context.
2559 (fixup_vla_decls): New function.
2560 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2561 debug variables were created.
2562 * tree.c (decl_value_expr_lookup): Add checking assertion.
2563 (decl_value_expr_insert): Likewise.
2565 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2568 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2569 * tree-scalar-evolution.c (interpret_expr): Likewise.
2570 (expression_expensive_p): Likewise.
2571 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2572 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2573 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2574 (ssa_defined_by_minus_one_stmt_p): New.
2576 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2579 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2580 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2581 * dojump.c (do_jump): Likewise.
2582 * expr.c (expand_expr_real_2): Check operand type's sign.
2583 * fold-const.c (const_unop): Handle ABSU_EXPR.
2584 (fold_abs_const): Likewise.
2585 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2586 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2587 (strip_sign_op_1): Likesise.
2588 * match.pd: Add new pattern to generate ABSU_EXPR.
2589 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2590 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2591 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2592 * tree-inline.c (estimate_operator_cost): Likewise.
2593 * tree-pretty-print.c (dump_generic_node): Likewise.
2594 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2595 * tree.def (ABSU_EXPR): New.
2597 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2600 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2601 documented as preserved for backward compatibility only.
2602 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2604 PR rtl-optimization/86108
2605 * bb-reorder.c (create_forwarder_block): Renamed to ...
2606 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2607 jump from new landing pad to the second part.
2608 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2611 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2614 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2615 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2616 Only call store_expr for halves if the mode is the same.
2619 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2620 Fix up comment formatting.
2622 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2624 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2625 (typed_splay_tree::closure,
2626 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2627 (typed_splay_tree::typed_splay_tree,
2628 typed_splay_tree::operator =): Declared private.
2629 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2630 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2631 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2632 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2633 typed_splay_tree::splay_tree_splay,
2634 typed_splay_tree::splay_tree_foreach_helper,
2635 typed_splay_tree::splay_tree_insert,
2636 typed_splay_tree::splay_tree_remove,
2637 typed_splay_tree::splay_tree_lookup,
2638 typed_splay_tree::splay_tree_predecessor,
2639 typed_splay_tree::splay_tree_successor,
2640 typed_splay_tree::splay_tree_min,
2641 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2642 (typed_splay_tree::root, typed_splay_tree::comp,
2643 typed_splay_tree::delete_key,
2644 typed_splay_tree::delete_value): New data members.
2645 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2646 typed_splay_tree::remove.
2648 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2650 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2651 -mginv and -mno-ginv to the assembler.
2652 * config/mips/mips.opt (-mcrc): New option.
2654 * doc/invoke.text (-mcrc): Document.
2657 2018-06-15 Nick Clifton <nickc@redhat.com>
2660 * tree.c (escaped_string): New class. Converts an unescaped
2661 string into its escaped equivalent.
2662 (warn_deprecated_use): Use the new class to convert the
2663 deprecation message, if present.
2664 (test_escaped_strings): New self test.
2665 (test_c_tests): Add test_escaped_strings.
2666 * doc/extend.texi (deprecated): Add a note that the
2667 deprecation message is affected by the -fmessage-length
2668 option, and that control characters will be escaped.
2669 (#pragma GCC error): Document this pragma.
2670 (#pragma GCC warning): Likewise.
2671 * doc/invoke.texi (-fmessage-length): Document this option's
2672 effect on the #warning and #error preprocessor directives and
2673 the deprecated attribute.
2675 2018-06-15 Richard Biener <rguenther@suse.de>
2677 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2678 here, also noting vector size used.
2679 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2680 size used in MSG_OPTIMIZED_LOCATIONS dump.
2681 (pass_slp_vectorize::execute): Adjust.
2683 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2686 * config/arc/arc.c (arc_return_address_register): Fix
2689 2018-06-15 Richard Biener <rguenther@suse.de>
2692 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2693 leave useless conversion stripping to force_gimple_operand_gsi.
2694 (gimplify_build2): Likewise.
2695 (gimplify_build1): Likewise.
2697 2018-06-15 Richard Biener <rguenther@suse.de>
2700 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2701 before adjusting their block.
2703 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2705 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2706 multilibs for *-*-rtems*.
2707 * config/riscv/t-rtems: New file.
2709 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2712 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2713 unsigned_type_for returns NULL.
2716 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2717 subregs of multi-word pseudos unless the float mode has word size.
2719 2018-06-14 Richard Biener <rguenther@suse.de>
2722 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2723 duplicate and harmful type_hash_canon.
2724 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2726 2018-06-14 Richard Biener <rguenther@suse.de>
2729 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2732 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2734 * config/rtems.h (STDINT_LONG32): Define.
2736 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2737 Prachi Godbole <prachi.godbole@imgtec.com>
2739 * config/mips/mips-cpus.def: Define P6600.
2740 * config/mips/mips-tables.opt: Regenerate.
2741 * config/mips/mips.c (mips_ucbranch_type): New enum.
2742 (mips_rtx_cost_data): Add support for P6600.
2743 (mips_issue_rate): Likewise.
2744 (mips_multipass_dfa_lookahead): Likewise.
2745 (mips_avoid_hazard): Likewise.
2746 (mips_reorg_process_insns): Likewise.
2747 (mips_classify_branch_p6600): New function.
2748 * config/mips/mips.h (TUNE_P6600): New define.
2749 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2750 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2751 * config/mips/mips.md: Include p6600.md.
2752 (processor): Add p6600.
2753 * config/mips/p6600.md: New file.
2754 * doc/invoke.texi: Add p6600 to supported architectures.
2756 2018-06-13 Martin Sebor <msebor@redhat.com>
2758 PR tree-optimization/86114
2759 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2761 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2763 2018-06-13 Richard Biener <rguenther@suse.de>
2765 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2766 Properly set vector type of the intermediate stmt.
2767 * tree-vect-stmts.c (vectorizable_operation): The destination
2768 var always has vectype_out type.
2770 2018-06-13 Jeff Law <law@redhat.com>
2772 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2773 integer 0 for argument to print_rtl_with_bb.
2774 (rl78_reorg): Likewise.
2776 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2778 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2779 from rtx to rtx_insn *.
2780 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2782 (add_sched_insns_for_speculation): Likewise for local "target",
2783 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2784 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2785 from rtx_insn ** to rtx_code_label **.
2786 (reorg_emit_nops): Likewise.
2787 (c6x_reorg): Likewise for local "call_labels".
2788 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2790 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2791 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2792 the loops over LABEL_REFS.
2793 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2795 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2796 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2797 (split_branches): Strengthen local "olabel" from rtx to
2798 rtx_insn *, adding a safe_as_a cast.
2799 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2801 (add_insn_after): Likewise for first two params.
2802 (add_insn_before): Likewise.
2803 (remove_insn): Likewise for param.
2804 (emit_pattern_before_noloc): Likewise for second and third params.
2805 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2806 (emit_call_insn_before_noloc): Likewise.
2807 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2809 (emit_barrier_before): Likewise.
2810 (emit_label_before): Strengthen "label" param from "rtx" to
2811 "rtx_code_label *". Strengthen "before" param from "rtx" to
2813 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2815 (emit_pattern_after_noloc): Likewise.
2816 (emit_insn_after_noloc): Likewise.
2817 (emit_jump_insn_after_noloc): Likewise.
2818 (emit_call_insn_after_noloc): Likewise.
2819 (emit_debug_insn_after_noloc): Likewise.
2820 (emit_barrier_after): Likewise.
2821 (emit_label_after): Likewise for both params.
2822 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2823 "loc" param from "int" to "location_t".
2824 (emit_insn_after_setloc): Likewise.
2825 (emit_jump_insn_after_setloc): Likewise.
2826 (emit_call_insn_after_setloc): Likewise.
2827 (emit_debug_insn_after_setloc): Likewise.
2828 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2829 "loc" param from "int" to "location_t".
2830 (emit_pattern_before): Convert NULL_RTX to NULL.
2831 (emit_insn_before_setloc): Convert "loc" param from "int" to
2833 (emit_jump_insn_before_setloc): Likewise.
2834 (emit_call_insn_before_setloc): Likewise.
2835 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2836 rtx_insn *. Convert "loc" param from "int" to "location_t".
2837 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2838 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2839 Convert 3rd param from "int" to "location_t".
2840 (emit_barrier_before, emit_barrier_after, next_real_insn):
2841 Strengthen param from rtx to rtx_insn *.
2842 (emit_label_before): Strengthen 1st param from "rtx" to
2843 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2845 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2846 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2847 Strengthen 2nd param from "rtx" to "rtx_insn *".
2848 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2849 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2850 Likewise. Convert 3rd param from "int" to "location_t".
2851 (emit_label_after): Strengthen 1st param from "rtx" to
2853 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2855 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2856 from "rtx" to "rtx_insn *".
2858 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2860 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2861 bodies streamed in with -Q.
2862 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2863 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2864 * gimple-streamer-out.c: Include gimple-pretty-print.h
2865 (output_bb): Dump stmts streamed.
2866 * lto-section-out.c: Include print-tree.h
2867 (lto_begin_section): Dump sections created.
2868 (lto_output_decl_index): Dump decl encoded.
2869 * lto-streamer-out.c: Include print-tree.h
2870 (create_output_block): Dump output block created.
2871 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2872 (output_function): Dump function output.
2873 (output_constructor): Dump constructor streamed.
2874 (write_global_stream): Output indexes encoded.
2875 (produce_asm_for_decls): Dump streams encoded.
2876 * lto-streamer.c (streamer_dump_file): New global var.
2877 * lto-streamer.h (streamer_dump_file): Declare.
2878 * passes.c (ipa_write_summaries): Initialize streamer dump.
2879 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2882 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2885 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2886 offsets for register save directives. Emit a second batch of save
2887 directives, if need be, when the function accesses prior frames.
2889 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2891 * config/arc/fpu.md (fmasf4): Force operand to register.
2892 (fnmasf4): Likewise.
2894 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2896 * config/arc/arc-protos.h (arc_pad_return): Remove.
2897 * config/arc/arc.c (machine_function): Remove force_short_suffix
2899 (arc_print_operand): Adjust printing of '&'.
2900 (arc_verify_short): Remove conditional printing of short suffix.
2901 (arc_final_prescan_insn): Remove reference to size_reason.
2902 (pad_return): New function.
2903 (arc_reorg): Call pad_return.
2904 (arc_pad_return): Remove.
2905 (arc_init_machine_status): Remove reference to force_short_suffix.
2906 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2907 (attr length): When attribute iscompact is true force to 2
2908 regardless; in the case of maybe check if we want to force the
2909 instruction to have 4 bytes length.
2910 (nopv): Change it to generate 4 byte long nop as well.
2911 (blockage): New pattern.
2912 (simple_return): Remove call to arc_pad_return.
2913 (p_return_i): Likewise.
2915 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2917 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2919 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2921 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2924 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2926 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2927 for ARC700 and ARCv2.
2929 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2932 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2933 operands[2] instead of operands[1].
2936 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2938 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2939 case, check whether the outer register overlaps an unallocatable
2940 register, not just whether it fits the required class.
2942 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2944 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2946 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2947 (memrefs_conflict_p): Likewise.
2948 (init_alias_analysis): Likewise.
2949 * cfgexpand.c (expand_debug_expr): Likewise.
2950 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2951 * cse.c (fold_rtx): Likewise.
2952 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2953 * expr.c (emit_block_move_hints): Likewise.
2954 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2955 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2956 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2957 (emit_group_store): Likewise.
2958 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2959 to read the PRE/POST_MODIFY increment.
2960 * calls.c (store_one_arg): Use strip_offset.
2961 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2963 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2965 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2966 (simplify_binary_operation_1): Extend CONST_INT handling to
2968 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2969 than a HOST_WIDE_INT.
2970 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2972 (adjust_mems, add_stores): Update accodingly.
2973 (vt_canonicalize_addr): Track polynomial offsets.
2974 (emit_note_insn_var_location): Likewise.
2975 (vt_add_function_parameter): Likewise.
2976 (vt_initialize): Likewise.
2978 2018-06-12 Jeff Law <law@redhat.com>
2980 * config.gcc (alpha*-*-freebsd*): Remove.
2981 * config/alpha/freebsd.h: Remove.
2983 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2986 * spellcheck-tree.c (levenshtein_distance): Rename to...
2987 (get_edit_distance): ...this, and update for underlying renaming.
2988 * spellcheck-tree.h (levenshtein_distance): Rename to...
2989 (get_edit_distance): ...this.
2990 * spellcheck.c (levenshtein_distance): Rename to...
2991 (get_edit_distance): ...this. Convert from Levenshtein distance
2992 to Damerau-Levenshtein distance by supporting transpositions of
2993 adjacent characters. Rename "v1" to "v_next" and "v0" to
2995 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2996 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2997 update for underlying renaming.
2998 (selftest::levenshtein_distance_unit_test): Rename to...
2999 (selftest::test_get_edit_distance_unit): ...this, and update for
3000 underlying renaming.
3001 (selftest::test_find_closest_string): Add example from PR 69968
3002 where transposition helps
3003 (selftest::test_metric_conditions): Update for renaming.
3004 (selftest::test_metric_conditions): Likewise.
3005 (selftest::spellcheck_c_tests): Likewise.
3006 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3007 (get_edit_distance): ...this.
3008 (best_match::consider): Update for renaming.
3010 2018-06-12 Martin Sebor <msebor@redhat.com>
3012 PR tree-optimization/85259
3013 * builtins.c (compute_objsize): Handle constant offsets.
3014 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3015 true iff a warning has been issued.
3016 * gimple.h (gimple_nonartificial_location): New function.
3017 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3018 gimple_nonartificial_location and handle -Wno-system-headers.
3019 (handle_builtin_stxncpy): Same.
3021 2018-06-12 Martin Sebor <msebor@redhat.com>
3024 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3026 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3028 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3029 BUILTIN_VEC_XST entries for pointer to double and long long.
3031 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3034 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3036 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3039 2018-06-12 Martin Liska <mliska@suse.cz>
3041 * doc/options.texi: Document IntegerRange.
3043 2018-06-12 Martin Liska <mliska@suse.cz>
3045 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3046 * opt-functions.awk: Handle Deprecated flag.
3047 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3049 (read_cmdline_option): Report warning for a deprecated option.
3050 * opts.h (struct cl_option): Add new field cl_deprecated.
3051 (CL_ERR_DEPRECATED): New.
3053 2018-06-12 Martin Liska <mliska@suse.cz>
3055 * doc/options.texi: Document Deprecated option flag.
3057 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3059 * config/arc/arc-arch.h (arc_extras): New enum.
3060 (arc_cpu_t):Add field extra.
3061 (arc_cpu_types): Consider the extras.
3062 * config/arc/arc-cpus.def: Add extras info.
3063 * config/arc/arc-opts.h (processor_type): Consider extra field.
3064 * config/arc/arc.c (arc_override_options): Handle extra field.
3066 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3068 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3069 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3071 * config/arc/arc.c (arc_init): Update pic warning.
3072 (irq_range): Update irq range parsing warnings.
3073 (arc_override_options): Update various warning messages.
3074 (arc_handle_aux_attribute): Likewise.
3076 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3078 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3080 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3082 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3083 compiler messages directives.
3085 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3087 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3088 * config/mips/mips-tables.opt: Regenerate.
3089 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3091 * doc/invoke.texi: Document -march=i6500.
3093 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3095 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3096 (i6400_gpmul): Add cpu_unit.
3097 (i6400_gpdiv): Likewise.
3098 (i6400_msa_add_d): Update reservations.
3099 (i6400_msa_int_add) Likewise.
3100 (i6400_msa_short_logic3) Likewise.
3101 (i6400_msa_short_logic2) Likewise.
3102 (i6400_msa_short_logic) Likewise.
3103 (i6400_msa_move) Likewise.
3104 (i6400_msa_cmp) Likewise.
3105 (i6400_msa_short_float2) Likewise.
3106 (i6400_msa_div_d) Likewise.
3107 (i6400_msa_long_logic1) Likewise.
3108 (i6400_msa_long_logic2) Likewise.
3109 (i6400_msa_mult) Likewise.
3110 (i6400_msa_long_float2) Likewise.
3111 (i6400_msa_long_float4) Likewise.
3112 (i6400_msa_long_float5) Likewise.
3113 (i6400_msa_long_float8) Likewise.
3114 (i6400_fpu_fadd): Include frint type.
3115 (i6400_fpu_store): New define_insn_reservation.
3116 (i6400_fpu_load): Likewise.
3117 (i6400_fpu_move): Likewise.
3118 (i6400_fpu_fcmp): Likewise.
3119 (i6400_fpu_fmadd): Likewise.
3120 (i6400_int_mult): Include imul3nc type and update reservation.
3121 (i6400_int_div): Include idiv3 type and update reservation.
3122 (i6400_int_load): Update to check type not move_type.
3123 (i6400_int_store): Likewise.
3124 (i6400_int_prefetch): Set zero latency.
3126 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3128 * gcc.c: Document new %@{...} sequence.
3129 (LINK_COMMAND_SPEC): Use it for the -L switches.
3130 (cpp_unique_options): Use it for the -I switches.
3131 (at_file_argbuf): New global variable.
3132 (in_at_file): Likewise.
3133 (alloc_args): Create at_file_argbuf.
3134 (clear_args): Truncate at_file_argbuf.
3135 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3136 (open_at_file): New function.
3137 (close_at_file): Likewise.
3138 (create_at_file): Delete.
3139 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3142 (validate_switches_from_spec): Deal with %@{...} sequence.
3143 (validate_switches): Likewise.
3144 (driver::finalize): Call clear_args.
3146 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3148 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3150 2018-06-11 Martin Sebor <msebor@redhat.com>
3152 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3153 (Wno-class-memaccess): Add @opindex.
3154 (Wno-templates, Wno-multiple-inheritance): Same.
3155 (Wno-virtual-inheritance, Wno-namespaces): Same.
3156 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3157 (Wno-format-overflow, Wno-format-truncation): Same.
3158 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3159 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3160 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3161 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3162 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3163 (Wno-misspelled-isr): Same.
3165 2018-06-11 Martin Sebor <msebor@redhat.com>
3167 * PR tree-optimization/86083
3168 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3170 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3172 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3174 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3177 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3178 on the correct operand.
3179 (*movdi_internal64): Ditto.
3181 2018-06-11 Martin Liska <mliska@suse.cz>
3183 PR tree-optimization/86089
3184 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3186 2018-06-11 Julia Koval <julia.koval@intel.com>
3188 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3189 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3190 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3192 2018-06-11 Olivier Hainque <hainque@adacore.com>
3194 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3195 for Ada with strict dwarf2.
3197 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3200 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3203 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3205 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3207 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3209 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3212 2018-06-08 Martin Liska <mliska@suse.cz>
3214 * tree-cfg.h (debug_function): Fix argument type to match
3217 2018-06-08 Martin Liska <mliska@suse.cz>
3219 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3220 Remove usage of MPX-related (and removed) fields.
3221 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3223 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3225 * cfg.c (debug): Use TDF_NONE rather than 0.
3226 * cfghooks.c (debug): Likewise.
3227 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3228 (struct dump_option_value_info): Convert to...
3229 (struct kv_pair): ...this template type.
3230 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3232 (optinfo_verbosity_options): Likewise.
3233 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3235 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3236 than int for "optgroup_flags" param.
3237 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3239 (dump_dec): Likewise.
3240 (dump_finish): Use TDF_NONE rather than 0.
3241 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3242 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3243 than 0. Update for change to option_ptr.
3244 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3245 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3246 0. Update for changes to optinfo_verbosity_options and
3248 (opt_info_switch_p): Convert optgroup_flags from int to
3250 (dump_basic_block): Use dump_flags_t rather than int
3251 for "dump_kind" param.
3252 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3253 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3254 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3255 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3256 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3257 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3258 TDF_NONE): Convert from macros to...
3259 (enum dump_flag): ...this new enum.
3260 (dump_flags_t): Update to use enum.
3261 (operator|, operator&, operator~, operator|=, operator&=):
3262 Implement for dump_flags_t.
3263 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3264 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3265 Convert from macros to...
3266 (enum optgroup_flag): ...this new enum.
3267 (optgroup_flags_t): New typedef.
3268 (operator|, operator|=): Implement for optgroup_flags_t.
3269 (struct dump_file_info): Convert field "alt_flags" to
3270 dump_flags_t. Convert field "optgroup_flags" to
3272 (dump_basic_block): Use dump_flags_t rather than int for param.
3273 (dump_generic_expr_loc): Likewise.
3274 (dump_dec): Likewise.
3275 (dump_register): Convert param "optgroup_flags" to
3277 (opt_info_enable_passes): Likewise.
3278 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3280 * gimple-pretty-print.c (debug): Likewise.
3281 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3282 (merged_store_group::apply_stores): Likewise.
3283 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3284 * gimple.c (verify_gimple_pp): Likewise.
3285 * graphite-poly.c (print_pbb_body): Likewise.
3286 * passes.c (pass_manager::register_one_dump_file): Convert
3287 local "optgroup_flags" to optgroup_flags_t.
3288 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3290 (debug_body): Likewise.
3291 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3292 to optgroup_flags_t.
3293 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3295 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3296 (convert_mult_to_fma): Likewise.
3297 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3298 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3299 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3300 "dump_kind" to dump_flags_t.
3302 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3304 * config/rs6000/rs6000.c (min, max): Delete.
3306 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3308 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3309 -mabi=spe and -mabi=no-spe.
3311 2018-06-08 Martin Liska <mliska@suse.cz>
3313 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3314 where we expect an existing summary.
3316 2018-06-08 Martin Liska <mliska@suse.cz>
3318 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3319 * ipa-inline.h (estimate_edge_growth): Likewise.
3321 2018-06-08 Martin Liska <mliska@suse.cz>
3323 * cgraph.c (function_version_hasher::hash): Use
3324 cgraph_node::get_uid ().
3325 (function_version_hasher::equal):
3326 * cgraph.h (cgraph_node::get_uid): New method.
3327 * ipa-inline.c (update_caller_keys): Use
3328 cgraph_node::get_uid ().
3329 (update_callee_keys): Likewise.
3330 * ipa-utils.c (searchc): Likewise.
3331 (ipa_reduced_postorder): Likewise.
3332 * lto-cgraph.c (input_node): Likewise.
3333 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3334 * symbol-summary.h (symtab_insertion): Likewise.
3335 (symtab_removal): Likewise.
3336 (symtab_duplication): Likewise.
3337 * tree-pretty-print.c (dump_function_header): Likewise.
3338 * tree-sra.c (convert_callers_for_node): Likewise.
3340 2018-06-08 Martin Liska <mliska@suse.cz>
3342 * cgraph.c (symbol_table::create_edge): Always assign a new
3344 (symbol_table::free_edge): Do not recycle numbers.
3345 * cgraph.h (cgraph_edge::get): New method.
3346 * symbol-summary.h (symtab_removal): Use it.
3347 (symtab_duplication): Likewise.
3348 (call_summary::hashable_uid): Remove.
3350 2018-06-08 Martin Liska <mliska@suse.cz>
3352 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3353 (initialize_growth_caches): Remove.
3354 (free_growth_caches): Likewise.
3355 (do_estimate_edge_time): Use edge_growth_cache.
3356 (do_estimate_edge_size): Likewise.
3357 (do_estimate_edge_hints): Likewise.
3358 * ipa-inline.c (reset_edge_caches): Likewise.
3359 (recursive_inlining): Likewise.
3360 (inline_small_functions): Likewise.
3361 * ipa-inline.h (initialize_growth_caches): Remove.
3362 (estimate_edge_size): Likewise.
3363 (estimate_edge_time): Likewise.
3364 (estimate_edge_hints): Likewise.
3365 (reset_edge_growth_cache): Likewise.
3366 * symbol-summary.h (call_summary::remove): New method.
3368 2018-06-08 Martin Liska <mliska@suse.cz>
3370 * ipa-cp.c (class edge_clone_summary): New summary.
3371 (grow_edge_clone_vectors): Remove.
3372 (ipcp_edge_duplication_hook): Remove.
3373 (class edge_clone_summary_t): New call_summary class.
3374 (ipcp_edge_removal_hook): Remove.
3375 (edge_clone_summary_t::duplicate): New function.
3376 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3377 (create_specialized_node): Likewise.
3378 (ipcp_driver): Initialize edge_clone_summaries and do not
3381 2018-06-08 Martin Liska <mliska@suse.cz>
3383 * symbol-summary.h (get): New function.
3384 (call_summary::m_initialize_when_cloning): New class member.
3386 2018-06-08 Martin Liska <mliska@suse.cz>
3388 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3389 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3390 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3391 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3392 (remove_cgraph_node_from_order): Use the removed_nodes set.
3393 (do_per_function_toporder): Likwise.
3394 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3395 instead of summary_uid.
3396 (symtab_removal): Likewise.
3397 (symtab_duplication): Likewise.
3399 2018-06-08 Martin Liska <mliska@suse.cz>
3401 * ipa-cp.c (ipcp_store_bits_results): Use
3402 ipcp_transformation_sum.
3403 (ipcp_store_vr_results): Likewise.
3404 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3406 (ipcp_transformation_initialize): ... this.
3407 (ipa_set_node_agg_value_chain):
3408 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3409 (write_ipcp_transformation_info): Likewise.
3410 (read_ipcp_transformation_info): Likewise.
3411 (ipcp_update_bits): Likewise.
3412 (ipcp_update_vr): Likewise.
3413 (ipcp_transform_function): Likewise.
3414 * ipa-prop.h: Rename ipcp_transformation_summary to
3415 ipcp_transformation.
3416 (class ipcp_transformation_t): New function summary.
3417 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3418 (ipa_get_agg_replacements_for_node): Likewise.
3420 2018-06-08 Martin Liska <mliska@suse.cz>
3422 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3424 (class funct_state_summary_t): New function_summary class.
3425 (has_function_state): Remove.
3426 (get_function_state): Likewise.
3427 (set_function_state): Likewise.
3428 (add_new_function): Likewise.
3429 (funct_state_summary_t::insert): New function.
3430 (duplicate_node_data): Remove.
3431 (remove_node_data): Remove.
3432 (funct_state_summary_t::duplicate): New function.
3433 (register_hooks): Create new funct_state_summaries.
3434 (pure_const_generate_summary): Use it.
3435 (pure_const_write_summary): Likewise.
3436 (pure_const_read_summary): Likewise.
3437 (propagate_pure_const): Likewise.
3438 (propagate_nothrow): Likewise.
3439 (dump_malloc_lattice): Likewise.
3440 (propagate_malloc): Likewise.
3441 (execute): Do not register hooks, just remove summary
3443 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3446 2018-06-08 Martin Liska <mliska@suse.cz>
3448 * ipa-reference.c (remove_node_data): Remove.
3449 (duplicate_node_data): Likewise.
3450 (class ipa_ref_var_info_summary_t): New class.
3451 (class ipa_ref_opt_summary_t): Likewise.
3452 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3453 (get_reference_optimization_summary): Use
3454 ipa_ref_opt_sum_summaries.
3455 (set_reference_vars_info): Remove.
3456 (set_reference_optimization_summary): Likewise.
3457 (ipa_init): Create summaries.
3458 (init_function_info): Use function summary.
3459 (ipa_ref_opt_summary_t::duplicate): New function.
3460 (ipa_ref_opt_summary_t::remove): New function.
3461 (get_read_write_all_from_node): Fix GNU coding style.
3462 (propagate): Use function summary.
3463 (write_node_summary_p): Fix GNU coding style.
3464 (stream_out_bitmap): Likewise.
3465 (ipa_reference_read_optimization_summary): Use function summary.
3466 (ipa_reference_c_finalize): Do not release hooks.
3468 2018-06-08 Martin Liska <mliska@suse.cz>
3470 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3471 (analyze_function_body): Extract multiple calls of get_create.
3472 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3473 * ipa-inline.c (recursive_inlining): Use ::get method.
3474 * ipa-inline.h (estimate_edge_growth): Likewise.
3476 2018-06-08 Martin Liska <mliska@suse.cz>
3478 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3480 (hsa_function_summary::hsa_function_summary): Use the new enum
3482 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3483 * hsa-gen.c (hsa_get_host_function): Likewise.
3484 (get_brig_function_name): Likewise.
3485 * ipa-hsa.c (process_hsa_functions): Likewise.
3486 (ipa_hsa_write_summary): Likewise.
3487 * symbol-summary.h (symtab_duplication): Use ::get function/
3488 (get): New function.
3490 2018-06-08 Martin Liska <mliska@suse.cz>
3492 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3494 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3495 (hsa_register_kernel): Likewise.
3496 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3497 * hsa-gen.c (hsa_get_host_function): Likewise.
3498 (get_brig_function_name): Likewise.
3499 (generate_hsa): Likewise.
3500 (pass_gen_hsail::execute): Likewise.
3501 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3502 (devirtualization_time_bonus): Likewise.
3503 (ipcp_propagate_stage): Likewise.
3504 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3505 (edge_set_predicate): Likewise.
3506 (evaluate_conditions_for_known_args): Likewise.
3507 (evaluate_properties_for_edge): Likewise.
3508 (ipa_fn_summary::reset): Likewise.
3509 (ipa_fn_summary_t::duplicate): Likewise.
3510 (dump_ipa_call_summary): Likewise.
3511 (ipa_dump_fn_summary): Likewise.
3512 (analyze_function_body): Likewise.
3513 (compute_fn_summary): Likewise.
3514 (estimate_edge_devirt_benefit): Likewise.
3515 (estimate_edge_size_and_time): Likewise.
3516 (estimate_calls_size_and_time): Likewise.
3517 (estimate_node_size_and_time): Likewise.
3518 (inline_update_callee_summaries): Likewise.
3519 (remap_edge_change_prob): Likewise.
3520 (remap_edge_summaries): Likewise.
3521 (ipa_merge_fn_summary_after_inlining): Likewise.
3522 (ipa_update_overall_fn_summary): Likewise.
3523 (read_ipa_call_summary): Likewise.
3524 (inline_read_section): Likewise.
3525 (write_ipa_call_summary): Likewise.
3526 (ipa_fn_summary_write): Likewise.
3527 (ipa_free_fn_summary): Likewise.
3528 * ipa-hsa.c (process_hsa_functions): Likewise.
3529 (ipa_hsa_write_summary): Likewise.
3530 (ipa_hsa_read_section): Likewise.
3531 * ipa-icf.c (sem_function::merge): Likewise.
3532 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3533 (do_estimate_edge_time): Likewise.
3534 (estimate_size_after_inlining): Likewise.
3535 (estimate_growth): Likewise.
3536 (growth_likely_positive): Likewise.
3537 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3538 (inline_call): Likewise.
3539 * ipa-inline.c (caller_growth_limits): Likewise.
3540 (can_inline_edge_p): Likewise.
3541 (can_inline_edge_by_limits_p): Likewise.
3542 (compute_uninlined_call_time): Likewise.
3543 (compute_inlined_call_time): Likewise.
3544 (want_inline_small_function_p): Likewise.
3545 (edge_badness): Likewise.
3546 (update_caller_keys): Likewise.
3547 (update_callee_keys): Likewise.
3548 (recursive_inlining): Likewise.
3549 (inline_small_functions): Likewise.
3550 (inline_to_all_callers_1): Likewise.
3551 (dump_overall_stats): Likewise.
3552 (early_inline_small_functions): Likewise.
3553 (early_inliner): Likewise.
3554 * ipa-inline.h (estimate_edge_growth): Likewise.
3555 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3556 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3557 * ipa-prop.h (IPA_NODE_REF): Likewise.
3558 (IPA_EDGE_REF): Likewise.
3559 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3560 (propagate_malloc): Likewise.
3561 * ipa-split.c (execute_split_functions): Likewise.
3562 * symbol-summary.h: Rename get to get_create.
3564 (get_create): Likewise.
3565 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3567 2018-06-08 Martin Liska <mliska@suse.cz>
3569 * symbol-summary.h (release): Move definition out of class
3571 (symtab_removal): Likewise.
3572 (symtab_duplication): Likewise.
3574 2018-06-08 Martin Liska <mliska@suse.cz>
3576 * symbol-summary.h (function_summary): Move constructor
3577 implementation out of class declaration.
3578 (release): Likewise.
3579 (symtab_insertion): Likewise.
3580 (symtab_removal): Likewise.
3581 (symtab_duplication): Likewise.
3584 2018-06-08 Martin Liska <mliska@suse.cz>
3586 * Makefile.in: Remove support for MPX (macros, related functions,
3587 fields in cgraph_node, ...).
3588 * builtin-types.def (BT_BND): Likewise.
3589 (BT_FN_BND_CONST_PTR): Likewise.
3590 (BT_FN_CONST_PTR_BND): Likewise.
3591 (BT_FN_VOID_PTR_BND): Likewise.
3592 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3593 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3594 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3595 (expand_builtin_mempcpy_with_bounds): Likewise.
3596 (expand_builtin_memset_with_bounds): Likewise.
3597 (expand_builtin_memset_args): Likewise.
3598 (std_expand_builtin_va_start): Likewise.
3599 (expand_builtin): Likewise.
3600 (expand_builtin_with_bounds): Likewise.
3601 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3602 (DEF_LIB_BUILTIN_CHKP): Likewise.
3603 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3604 (DEF_CHKP_BUILTIN): Likewise.
3605 (BUILT_IN_MEMCPY): Likewise.
3606 (BUILT_IN_MEMMOVE): Likewise.
3607 (BUILT_IN_MEMPCPY): Likewise.
3608 (BUILT_IN_MEMSET): Likewise.
3609 (BUILT_IN_STPCPY): Likewise.
3610 (BUILT_IN_STRCAT): Likewise.
3611 (BUILT_IN_STRCHR): Likewise.
3612 (BUILT_IN_STRCPY): Likewise.
3613 (BUILT_IN_STRLEN): Likewise.
3614 (BUILT_IN_MEMCPY_CHK): Likewise.
3615 (BUILT_IN_MEMMOVE_CHK): Likewise.
3616 (BUILT_IN_MEMPCPY_CHK): Likewise.
3617 (BUILT_IN_MEMSET_CHK): Likewise.
3618 (BUILT_IN_STPCPY_CHK): Likewise.
3619 (BUILT_IN_STRCAT_CHK): Likewise.
3620 (BUILT_IN_STRCPY_CHK): Likewise.
3621 * calls.c (store_bounds): Likewise.
3622 (emit_call_1): Likewise.
3623 (special_function_p): Likewise.
3624 (maybe_warn_nonstring_arg): Likewise.
3625 (initialize_argument_information): Likewise.
3626 (finalize_must_preallocate): Likewise.
3627 (compute_argument_addresses): Likewise.
3628 (expand_call): Likewise.
3629 * cfgexpand.c (expand_call_stmt): Likewise.
3630 (expand_return): Likewise.
3631 (expand_gimple_stmt_1): Likewise.
3632 (pass_expand::execute): Likewise.
3633 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3634 (cgraph_node::remove): Likewise.
3635 (cgraph_node::dump): Likewise.
3636 (cgraph_node::verify_node): Likewise.
3637 * cgraph.h (chkp_function_instrumented_p): Likewise.
3638 (symtab_node::get_alias_target): Likewise.
3639 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3640 (cgraph_local_p): Likewise.
3641 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3642 (cgraph_edge::rebuild_references): Likewise.
3643 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3644 (walk_polymorphic_call_targets): Likewise.
3645 (cgraph_node::expand_thunk): Likewise.
3646 (symbol_table::output_weakrefs): Likewise.
3647 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3648 (ix86_handle_option): Likewise.
3649 * config/i386/constraints.md: Likewise.
3650 * config/i386/i386-builtin-types.def (BND): Likewise.
3654 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3655 (BDESC_FIRST): Likewise.
3657 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3658 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3659 * config/i386/i386.c (enum reg_class): Likewise.
3660 (ix86_target_string): Likewise.
3661 (ix86_option_override_internal): Likewise.
3662 (ix86_conditional_register_usage): Likewise.
3663 (ix86_valid_target_attribute_inner_p): Likewise.
3664 (ix86_set_indirect_branch_type): Likewise.
3665 (ix86_set_current_function): Likewise.
3666 (ix86_function_arg_regno_p): Likewise.
3667 (init_cumulative_args): Likewise.
3668 (ix86_function_arg_advance): Likewise.
3669 (ix86_function_arg): Likewise.
3670 (ix86_pass_by_reference): Likewise.
3671 (ix86_function_value_regno_p): Likewise.
3672 (ix86_function_value_1): Likewise.
3673 (ix86_function_value_bounds): Likewise.
3674 (ix86_return_in_memory): Likewise.
3675 (ix86_setup_incoming_vararg_bounds): Likewise.
3676 (ix86_va_start): Likewise.
3677 (indirect_thunk_need_prefix): Likewise.
3678 (print_reg): Likewise.
3679 (ix86_print_operand): Likewise.
3680 (ix86_expand_call): Likewise.
3681 (ix86_output_function_return): Likewise.
3682 (reg_encoded_number): Likewise.
3683 (BDESC_VERIFYS): Likewise.
3684 (ix86_init_mpx_builtins): Likewise.
3685 (ix86_init_builtins): Likewise.
3686 (ix86_emit_cmove): Likewise.
3687 (ix86_emit_move_max): Likewise.
3688 (ix86_expand_builtin): Likewise.
3689 (ix86_builtin_mpx_function): Likewise.
3690 (ix86_get_arg_address_for_bt): Likewise.
3691 (ix86_load_bounds): Likewise.
3692 (ix86_store_bounds): Likewise.
3693 (ix86_load_returned_bounds): Likewise.
3694 (ix86_store_returned_bounds): Likewise.
3695 (ix86_class_likely_spilled_p): Likewise.
3696 (ix86_hard_regno_mode_ok): Likewise.
3697 (x86_order_regs_for_local_alloc): Likewise.
3698 (ix86_mitigate_rop): Likewise.
3699 (ix86_bnd_prefixed_insn_p): Likewise.
3700 (ix86_mpx_bound_mode): Likewise.
3701 (ix86_make_bounds_constant): Likewise.
3702 (ix86_initialize_bounds): Likewise.
3703 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3704 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3705 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3706 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3707 (TARGET_CHKP_BOUND_MODE): Likewise.
3708 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3709 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3710 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3711 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3712 * config/i386/i386.h (TARGET_MPX): Likewise.
3713 (TARGET_MPX_P): Likewise.
3714 (VALID_BND_REG_MODE): Likewise.
3715 (FIRST_BND_REG): Likewise.
3716 (LAST_BND_REG): Likewise.
3717 (enum reg_class): Likewise.
3718 (BND_REG_P): Likewise.
3719 (BND_REGNO_P): Likewise.
3720 (BNDmode): Likewise.
3721 (ADJUST_INSN_LENGTH): Likewise.
3722 * config/i386/i386.md: Likewise.
3723 * config/i386/i386.opt: Likewise.
3724 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3725 (defined): Likewise.
3726 (LINK_MPX): Likewise.
3727 (MPX_SPEC): Likewise.
3728 (LIBMPX_SPEC): Likewise.
3729 (LIBMPXWRAPPERS_SPEC): Likewise.
3730 (CHKP_SPEC): Likewise.
3731 * config/i386/predicates.md: Likewise.
3732 * dbxout.c (dbxout_type): Likewise.
3733 * doc/extend.texi: Likewise.
3734 * doc/invoke.texi: Likewise.
3735 * doc/md.texi: Likewise.
3736 * doc/tm.texi: Likewise.
3737 * doc/tm.texi.in: Likewise.
3738 * dwarf2out.c (is_base_type): Likewise.
3739 (gen_formal_types_die): Likewise.
3740 (gen_subprogram_die): Likewise.
3741 (gen_type_die_with_usage): Likewise.
3742 (gen_decl_die): Likewise.
3743 (dwarf2out_late_global_decl): Likewise.
3744 * expr.c (expand_assignment): Likewise.
3745 (emit_storent_insn): Likewise.
3746 (store_expr_with_bounds): Likewise.
3747 (store_expr): Likewise.
3748 (expand_expr_real_1): Likewise.
3749 * expr.h (store_expr_with_bounds): Likewise.
3750 * function.c (use_register_for_decl): Likewise.
3751 (struct bounds_parm_data): Likewise.
3752 (assign_parms_augmented_arg_list): Likewise.
3753 (assign_parm_find_entry_rtl): Likewise.
3754 (assign_parm_is_stack_parm): Likewise.
3755 (assign_parm_load_bounds): Likewise.
3756 (assign_bounds): Likewise.
3757 (assign_parms): Likewise.
3758 (expand_function_start): Likewise.
3759 * gcc.c (CHKP_SPEC): Likewise.
3760 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3761 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3762 (wrestrict_dom_walker::check_call): Likewise.
3763 * gimple.c (gimple_build_call_from_tree): Likewise.
3764 * gimple.h (enum gf_mask): Likewise.
3765 (gimple_call_with_bounds_p): Likewise.
3766 (gimple_call_set_with_bounds): Likewise.
3767 * gimplify.c (gimplify_init_constructor): Likewise.
3768 * ipa-cp.c (initialize_node_lattices): Likewise.
3769 (propagate_constants_across_call): Likewise.
3770 (find_more_scalar_values_for_callers_subset): Likewise.
3771 * ipa-hsa.c (process_hsa_functions): Likewise.
3772 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3773 * ipa-icf.c (sem_function::merge): Likewise.
3774 * ipa-inline.c (early_inliner): Likewise.
3775 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3776 (warn_function_cold): Likewise.
3777 (propagate_pure_const): Likewise.
3778 * ipa-ref.h (enum GTY): Likewise.
3779 * ipa-split.c (find_retbnd): Likewise.
3780 (consider_split): Likewise.
3781 (split_function): Likewise.
3782 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3783 * ipa.c (walk_polymorphic_call_targets): Likewise.
3784 (symbol_table::remove_unreachable_nodes): Likewise.
3785 (process_references): Likewise.
3786 (cgraph_build_static_cdtor_1): Likewise.
3787 * lto-cgraph.c (lto_output_node): Likewise.
3788 (output_refs): Likewise.
3789 (compute_ltrans_boundary): Likewise.
3790 (input_overwrite_node): Likewise.
3791 (input_node): Likewise.
3792 (input_cgraph_1): Likewise.
3793 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3794 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3795 (class pass_chkp_instrumentation_passes): Likewise.
3796 (make_pass_chkp_instrumentation_passes): Likewise.
3797 * passes.def: Likewise.
3798 * rtl.h (struct GTY): Likewise.
3799 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3800 * stor-layout.c (layout_type): Likewise.
3801 * symtab.c: Likewise.
3802 * target.def: Likewise.
3803 * targhooks.c (default_chkp_bound_type): Likewise.
3804 (default_chkp_bound_mode): Likewise.
3805 (default_builtin_chkp_function): Likewise.
3806 (default_chkp_function_value_bounds): Likewise.
3807 (default_chkp_make_bounds_constant): Likewise.
3808 (default_chkp_initialize_bounds): Likewise.
3809 * targhooks.h (default_chkp_bound_type): Likewise.
3810 (default_chkp_bound_mode): Likewise.
3811 (default_builtin_chkp_function): Likewise.
3812 (default_chkp_function_value_bounds): Likewise.
3813 (default_chkp_make_bounds_constant): Likewise.
3814 (default_chkp_initialize_bounds): Likewise.
3815 * toplev.c (compile_file): Likewise.
3816 (process_options): Likewise.
3817 * tree-core.h (DEF_BUILTIN): Likewise.
3818 (DEF_BUILTIN_CHKP): Likewise.
3819 * tree-inline.c (declare_return_variable): Likewise.
3820 (remap_gimple_stmt): Likewise.
3821 (copy_bb): Likewise.
3822 (initialize_inlined_parameters): Likewise.
3823 (expand_call_inline): Likewise.
3824 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3825 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3826 (make_pass_ipa_chkp_produce_thunks): Likewise.
3827 (make_pass_chkp): Likewise.
3828 (make_pass_chkp_opt): Likewise.
3829 (make_pass_chkp_instrumentation_passes): Likewise.
3830 * tree-pretty-print.c (dump_generic_node): Likewise.
3831 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3832 * tree-ssa-dce.c (propagate_necessity): Likewise.
3833 (eliminate_unnecessary_stmts): Likewise.
3834 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3835 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3836 * tree-ssa-sccvn.h: Likewise.
3837 * tree-ssa-strlen.c (get_string_length): Likewise.
3838 (valid_builtin_call): Likewise.
3839 (adjust_last_stmt): Likewise.
3840 (handle_builtin_strchr): Likewise.
3841 (handle_builtin_strcpy): Likewise.
3842 (handle_builtin_stxncpy): Likewise.
3843 (handle_builtin_memcpy): Likewise.
3844 (handle_builtin_strcat): Likewise.
3845 (strlen_check_and_optimize_stmt): Likewise.
3846 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3847 * tree-streamer-in.c: Likewise.
3848 * tree-streamer.c (record_common_node): Likewise.
3849 * tree.c (tree_code_size): Likewise.
3850 (wide_int_to_tree_1): Likewise.
3851 (type_contains_placeholder_1): Likewise.
3852 (build_common_tree_nodes): Likewise.
3853 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3854 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3855 (POINTER_BOUNDS_P): Likewise.
3856 (BOUNDED_TYPE_P): Likewise.
3857 (BOUNDED_P): Likewise.
3858 (CALL_WITH_BOUNDS_P): Likewise.
3859 (pointer_bounds_type_node): Likewise.
3860 * value-prof.c (gimple_ic): Likewise.
3861 * var-tracking.c (vt_add_function_parameters): Likewise.
3862 * varasm.c (make_decl_rtl): Likewise.
3863 (assemble_start_function): Likewise.
3864 (output_constant): Likewise.
3865 (maybe_assemble_visibility): Likewise.
3866 * varpool.c (ctor_for_folding): Likewise.
3867 * chkp-builtins.def: Remove.
3868 * ipa-chkp.c: Remove.
3869 * ipa-chkp.h: Remove.
3870 * rtl-chkp.c: Remove.
3871 * rtl-chkp.h: Remove.
3872 * tree-chkp-opt.c: Remove.
3873 * tree-chkp.c: Remove.
3874 * tree-chkp.h: Remove.
3876 2018-06-07 Carl Love <cel@us.ibm.com>
3878 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3879 vextract_fp_from_shortl): Add BE support.
3881 2018-06-07 Paul Koning <ni1d@arrl.net>
3883 * compare-elim.c (try_merge_compare): Don't merge compare if
3884 address contains a side effect.
3885 (try_eliminate_compare): Likewise.
3887 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3889 * config.gcc: Support "tremont".
3890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3891 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3893 * config/i386/i386.c (m_TREMONT): Define.
3894 (processor_target_table): Add "tremont".
3895 (PTA_TREMONT): Define.
3896 (ix86_lea_outperforms): Add TARGET_TREMONT.
3897 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3898 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3899 and M_INTEL_GOLDMONT_PLUS.
3900 (fold_builtin_cpu): Add "tremont".
3901 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3902 (ix86_option_override_internal): Add "tremont".
3903 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3904 (processor_type): Add PROCESSOR_TREMONT.
3905 * config/i386/x86-tune.def: Add m_TREMONT.
3906 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3908 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3910 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3911 symbol defined for msp430i* devices to be lower case.
3913 gcc/testsuite/gcc.target/msp430/
3914 * msp430i-device-symbol.c: New test.
3915 * msp430f-device-symbol.c: New test.
3916 * msp430.h: New test header file.
3918 2018-06-07 Richard Biener <rguenther@suse.de>
3920 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3921 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3922 Properly wrap signed arithmetic if overflow wraps.
3924 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3926 PR tree-optimization/69615
3927 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3928 of a cast from a same precision integral SSA_NAME in a bb dominated
3929 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3930 cast to utype if rhs2 has already a compatible type.
3932 2018-06-07 Richard Biener <rguenther@suse.de>
3934 PR tree-optimization/85935
3935 * graphite-scop-detection.c (find_params_in_bb): Analyze
3936 condition operands with respect to the correct loop. Assert
3937 the analysis doesn't fail.
3939 2018-06-04 Carl Love <cel@us.ibm.com>
3941 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3942 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3943 as it is slightly cheaper.
3944 (first_match_or_eos_index_<mode>):
3945 Calculate index using natural element order.
3946 (first_match_index_<mode>):
3947 Calculate index using natural element order.
3948 (first_match_or_eos_index_<mode>):
3949 Calculate index using natural order.
3950 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3951 for BE and LE modes.
3952 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3953 P9V_BUILTIN_VCLZLSBB_V16QI.
3954 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3957 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3959 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3960 indentation and line wrap for many prototypes. Add missing
3961 @smallexample directives around block of prototypes for vec_xl and
3964 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3966 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3967 track if we pass or return IEEE 128-bit floating point.
3968 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3969 C++ mangling that is compatible with GCC 8.1.
3970 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3971 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3972 floating point types.
3973 (rs6000_function_arg_advance_1): Likewise.
3974 (rs6000_mangle_type): Optionally generate mangled names that match
3975 what GCC 8.1 generated for IEEE 128-bit floating point types.
3976 (rs6000_globalize_decl_name): If we have an external function that
3977 passes or returns IEEE 128-bit types, generate a weak reference
3978 from the mangled name used in GCC 8.1 to the current mangled
3980 (rs6000_init_builtins): Make __ibm128 use the long double type if
3981 long double is IBM extended double. Make __float128 use the long
3982 double type if long double is IEEE 128-bit.
3985 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3986 macro for __ibm128 built-in functions.
3987 (PACK_IF): Add __ibm128 pack/unpack functions.
3988 (UNPACK_IF): Likewise.
3989 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3990 enable long double built-in functions if long double is IEEE
3991 128-bit floating point.
3992 (rs6000_invalid_builtin): Update long double built-in function
3994 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3995 functions, adjust the built-in function to use the long double
3996 built-in function if __ibm128 and long double are the same type.
3997 * doc/extend.texi (PowerPC builtins): Update documention for
3998 __builtin_{,un}pack_longdouble. Add documentation for
3999 __builtin_{,un}pack_ibm128.
4001 2018-06-06 Jim Wilson <jimw@sifive.com>
4003 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4004 (struct machine_function): New field interrupt_mode.
4005 (riscv_handle_type_attribute): New function. Add forward declaration.
4006 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4007 (riscv_expand_epilogue): Check interrupt_mode field.
4008 (riscv_set_current_function): Check interrupt attribute args and
4009 set interrupt_mode field.
4010 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4011 (riscv_sret, riscv_uret): New.
4012 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4013 new arguments to interrupt attribute.
4015 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4018 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4019 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4021 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4023 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4024 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4025 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4026 several redundant entries.
4028 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4030 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4031 type from "rtx" to "rtx_insn *".
4032 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4033 for local "call_insn", removing cast.
4034 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4036 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4038 PR tree-optimization/86066
4039 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4040 for BIT_INSERT_EXPR stores.
4042 2018-06-06 Richard Biener <rguenther@suse.de>
4044 PR tree-optimization/86062
4045 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4046 component refs ontop
4047 of to be offsetted base.
4049 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4051 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4052 to be static and remove check on interrupt attribute name.
4054 gcc/testsuite/gcc.target/msp430/
4055 * function-attributes-4.c: New test.
4056 * static-interrupts.c: New test.
4058 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4060 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4061 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4063 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4066 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4068 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4069 Remove second argument, change how error is called.
4070 (aarch64_layout_arg): Remove second argument from
4071 aarch64_err_no_fpadvsimd call.
4072 (aarch64_init_cumulative_args): Ditto.
4073 (aarch64_gimplify_va_arg_expr): Ditto.
4074 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4076 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4078 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4079 (*simple_return_indirect_internal<mode>): Rename from
4080 simple_return_indirect_internal. Use W mode iterator.
4081 (rstorssp): New expander.
4082 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4083 (clrssbsy): New expander.
4084 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4086 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4088 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4090 (cmse_check_pointed_object): Likewise.
4092 2018-06-05 Martin Liska <mliska@suse.cz>
4094 PR gcov-profile/47618
4095 * doc/invoke.texi: Document how -fprofile-dir format
4098 2018-06-05 Richard Biener <rguenther@suse.de>
4100 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4101 removal pretend DOM info isn't available so we do not update
4102 it and only remove edges, not dominated blocks. Actually free
4103 DOM info in case we removed something. Remove unreachable blocks.
4104 (mfb_keep_latches): Work with either DOM info or marked backedges.
4105 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4106 first. Mark backedges if DOM info isn't available.
4107 (Re-)compute DOM info after cleanup_control_flow_pre.
4109 2018-06-05 Richard Biener <rguenther@suse.de>
4111 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4112 (locus_discrim_hasher::hash): Adjust.
4113 (locus_discrim_hasher::equal): Likewise.
4114 (next_discriminator_for_locus): Work on line directly.
4115 (same_line_p): Pass in expanded locus1 as well.
4116 (assign_discriminators): Avoid redundant location expansions.
4118 2018-06-05 Richard Biener <rguenther@suse.de>
4120 PR tree-optimization/86046
4121 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4122 if required after clearing TREE_ADDRESSABLE.
4124 2018-06-05 Richard Biener <rguenther@suse.de>
4126 PR tree-optimization/86047
4127 * tree-ssa-loop.c (for_each_index): Glob handling of all
4128 decls and constants and really handle all of them.
4130 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4133 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4134 qualifier_void_pointer and qualifier_const_void_pointer.
4135 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4136 (arm_init_builtins): Handle the above.
4137 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4138 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4139 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4142 2018-06-05 Martin Liska <mliska@suse.cz>
4144 * auto-profile.c (read_autofdo_file): Do not use
4145 gcov_ctr_summary struct.
4146 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4147 * coverage.c (struct counts_entry): Likewise.
4148 (read_counts_file): Read just single summary entry.
4149 (get_coverage_counts): Use gcov_summary struct.
4150 * coverage.h (get_coverage_counts): Likewise.
4151 * gcov-dump.c (dump_working_sets): Likewise.
4152 (tag_summary): Dump just single summary.
4153 * gcov-io.c (gcov_write_summary): Write just histogram
4155 (gcov_read_summary): Read just single summary.
4156 (compute_working_sets): Use gcov_summary struct.
4157 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4158 of GCOV_COUNTERS_SUMMABLE.
4159 (GCOV_COUNTERS_SUMMABLE): Remove.
4160 (GCOV_FIRST_VALUE_COUNTER): Replace with
4161 GCOV_COUNTER_V_INTERVAL.
4162 (struct gcov_ctr_summary): Remove.
4163 (struct gcov_summary): Directly use fields of former
4165 (compute_working_sets): Use gcov_summary struct.
4166 * gcov.c (read_count_file): Do not use ctrs fields.
4167 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4169 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4171 * profile.c: Likewise.
4172 * profile.h: Likewise.
4174 2018-06-05 Martin Liska <mliska@suse.cz>
4176 PR gcov-profile/84846
4177 * gcov.c (output_lines): Print working directory only
4178 in intermediate format.
4180 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4182 * config/s390/s390-builtin-types.def: Add void function type.
4183 * config/s390/s390-builtins.def: Use the function type for the
4186 2018-06-04 Jim Wilson <jimw@sifive.com>
4188 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4190 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4191 and maybe_eh_return. Change regno to unsigned int. Use new args to
4192 handle EH_RETURN_DATA_REGNO registers properly.
4193 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4194 (riscv_expand_epilogue): Update comment. Change argument name and
4195 type. Update code to use new name and type. Pass new args to
4196 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4198 * config/riscv/riscv.md (NORMAL_RETURN): New.
4199 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4200 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4201 (eh_return): Call gen_eh_return_internal and emit barrier.
4202 (eh_return_internal): Call riscv_expand_epilogue.
4204 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4206 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4207 bit_insertion field and declare can_be_merged_into method.
4208 (merged_store_group::can_be_merged_into): New method.
4209 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4210 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4211 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4213 2018-06-04 Richard Biener <rguenther@suse.de>
4215 PR tree-optimization/85955
4216 * builtins.c (fold_builtin_sincos): Convert pointers to
4217 destination to appropriate type before dereferencing.
4219 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4221 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4223 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4225 * expr.c (expand_expr_real_1): Force the operand into memory if
4226 its TYPE_MODE is BLKmode and if there is no integer mode for
4227 the number of bits being extracted.
4229 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4233 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4234 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4236 2018-06-04 Richard Biener <rguenther@suse.de>
4238 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4239 (cleanup_tree_cfg_noloop): ... single caller. Do
4240 start_recording_case_labels later.
4242 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4244 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4245 to _IMMINTRIN_H_INCLUDED.
4246 * config/i386/pconfigintrin.h: Ditto.
4247 * config/i386/waitpkgintrin.h: Ditto.
4248 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4249 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4250 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4251 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4252 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4253 waitpkgintrin.h and cldemoteintrin.h.
4255 2018-06-04 Richard Biener <rguenther@suse.de>
4257 PR tree-optimization/86038
4258 * tracer.c (find_best_successor): Check probability for
4259 being initialized, bail out if not.
4261 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4264 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4265 of bits to ignore when comparing architectures.
4267 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4269 PR tree-optimization/69615
4270 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4271 maximum or minimum of the type, try to merge it also as if
4272 range1 is + [-, x - 1] or + [x + 1, -].
4275 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4277 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4279 PR tree-optimization/86034
4280 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4281 the unsigned bitfield type in a bit insertion sequence if it does not
4282 have a larger precision than the bitfield size.
4283 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4285 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4287 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4289 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4291 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4292 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4293 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4294 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4296 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4298 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4299 Disable -fdelete-null-pointer-checks for ELF toolchain.
4301 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4302 Kito Cheng <kito.cheng@gmail.com>
4304 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4305 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4306 (nds32*-*-*): Add glibc and uclibc conditions.
4307 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4308 (TARGET_EXCEPT_UNWIND_INFO): Define.
4309 * config/nds32/elf.h: New file.
4310 * config/nds32/linux.h: New file.
4311 * config/nds32/nds32-elf.opt: New file.
4312 * config/nds32/nds32-linux.opt: New file.
4313 * config/nds32/nds32-fp-as-gp.c
4314 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4315 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4317 (nds32_asm_file_end): Ditto.
4318 (nds32_print_operand): Ditto.
4319 (nds32_insert_attributes): Ditto.
4320 (nds32_init_libfuncs): New function.
4321 (TARGET_HAVE_TLS): Define.
4322 (TARGET_INIT_LIBFUNCS): Define.
4323 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4325 (TARGET_ELF): Apply different mcmodel setting.
4326 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4327 been migrated into elf.h and linux.h files.
4328 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4329 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4330 (mcmodel): The content has been migrated into nds32-elf.opt and
4331 nds32-linux.opt files.
4332 * config/nds32/t-elf: New file.
4333 * config/nds32/t-linux: New file.
4335 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4336 Shiva Chen <shiva0217@gmail.com>
4338 * config/nds32/constants.md (unspec_volatile_element): Add
4339 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4340 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4342 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4343 (make_pass_nds32_fp_as_gp): Declare.
4344 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4346 (nds32_asm_function_end_prologue): Remove unused asm output.
4347 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4348 (nds32_asm_file_start): Output necessary fp_as_gp information.
4349 (nds32_option_override): Adjust register usage.
4350 (nds32_expand_prologue): Consider fp_as_gp situation.
4351 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4352 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4355 (simple_return): Ditto.
4356 (omit_fp_begin): Output special directive for fp_as_gp.
4357 (omit_fp_end): Output special directive for fp_as_gp.
4358 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4359 mforbid-fp-as-gp): New options.
4361 2018-06-01 Mark Wielaard <mark@klomp.org>
4363 * dwarf2out.c (dwarf2out_finish): Remove generation of
4364 DW_AT_loclists_base.
4366 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4368 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4369 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4370 (struct merged_store_group): Add bit_insertion field.
4371 (dump_char_array): Use standard hexadecimal format.
4372 (merged_store_group::merged_store_group): Set bit_insertion to false.
4373 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4374 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4375 also print the mask in the dump file.
4376 (pass_store_merging::gate): Minor tweak.
4377 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4378 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4379 stores with INTEGER_CST stores.
4380 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4381 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4382 and use it throughout. Generate bit insertion sequences if need be.
4383 (pass_store_merging::process_store): Remove redundant condition.
4384 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4386 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4388 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4389 the 128-bit floating point types. Fix function comment.
4391 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4393 * config/aarch64/aarch64-simd.md
4394 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4396 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4399 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4401 PR tree-optimization/85989
4402 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4404 (backprop::intersect_uses): Check it when deciding whether this
4405 is a backedge reference.
4406 (backprop::process_block): Add each phi to m_visited_phis
4407 after visiting it, then clear it at the end.
4409 2018-06-01 Richard Biener <rguenther@suse.de>
4411 * tree-vectorizer.h (vect_dr_stmt): New function.
4412 (vect_get_load_cost): Adjust.
4413 (vect_get_store_cost): Likewise.
4414 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4415 Use vect_dr_stmt instead of DR_SMTT.
4416 (vect_record_base_alignments): Likewise.
4417 (vect_calculate_target_alignment): Likewise.
4418 (vect_compute_data_ref_alignment): Likewise and make static.
4419 (vect_update_misalignment_for_peel): Likewise.
4420 (vect_verify_datarefs_alignment): Likewise.
4421 (vector_alignment_reachable_p): Likewise.
4422 (vect_get_data_access_cost): Likewise. Pass down
4423 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4424 (vect_get_peeling_costs_all_drs): Likewise.
4425 (vect_peeling_hash_get_lowest_cost): Likewise.
4426 (vect_enhance_data_refs_alignment): Likewise.
4427 (vect_find_same_alignment_drs): Likewise.
4428 (vect_analyze_data_refs_alignment): Likewise.
4429 (vect_analyze_group_access_1): Likewise.
4430 (vect_analyze_group_access): Likewise.
4431 (vect_analyze_data_ref_access): Likewise.
4432 (vect_analyze_data_ref_accesses): Likewise.
4433 (vect_vfa_segment_size): Likewise.
4434 (vect_small_gap_p): Likewise.
4435 (vectorizable_with_step_bound_p): Likewise.
4436 (vect_prune_runtime_alias_test_list): Likewise.
4437 (vect_analyze_data_refs): Likewise.
4438 (vect_supportable_dr_alignment): Likewise.
4439 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4440 (vect_gen_prolog_loop_niters): Likewise.
4441 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4442 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4444 (vect_recog_mask_conversion_pattern): Likewise.
4445 (vect_try_gather_scatter_pattern): Likewise.
4446 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4447 to vect_get_store_cost.
4448 (vect_get_store_cost): Get stmt_info instead of DR.
4449 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4450 (vect_get_load_cost): Get stmt_info instead of DR.
4452 2018-06-01 Richard Biener <rguenther@suse.de>
4455 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4456 vars wrapped in MEM_REFs.
4458 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4460 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4461 Fix subreg tests so that we only return a choice between
4462 GENERAL_REGS and FP_REGS if the original classes included both.
4464 2018-06-01 Richard Biener <rguenther@suse.de>
4467 * tree-ssa-structalias.c (get_function_part_constraint):
4468 Handle NULL fi->decl.
4469 (find_func_aliases_for_call): Properly handle indirect
4470 fi from direct call.
4471 (find_func_clobbers): Likewise.
4472 (ipa_pta_execute): Likewise.
4473 (create_variable_info_for): For functions that are ifunc_resolver
4474 resolve to a varinfo that contains the result of the resolver call.
4475 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4478 2018-05-31 Michael Collison <michael.collison@arm.com>
4480 * config/aarch64/aarch64.md:
4481 (*fix_to_zero_extenddfdi2): New pattern.
4482 * gcc.target/aarch64/fix_extend1.c: New testcase.
4484 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4488 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4489 and BUILT_IN_STRNCMP_EQ.
4490 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4491 BUILT_IN_STRNCMP_EQ.
4492 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4493 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4494 (gimple_fold_builtin): Likewise.
4495 * tree-ssa-strlen.c (compute_string_length): New function.
4496 (determine_min_obsize): New function.
4497 (handle_builtin_string_cmp): New function to handle calls to
4498 string compare functions.
4499 (strlen_optimize_stmt): Add handling to builtin string compare
4501 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4502 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4503 * tree.c (build_common_builtin_nodes): Add new defines of
4504 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4506 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4509 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4510 functions with naked attribute.
4512 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4514 * config/i386/sse.md (avx_vec_concat<mode>):
4515 Substitute concat_tg_mode mode attribute with xtg_mode.
4516 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4517 (concat_tg_mode): Remove mode attribute.
4519 2018-05-31 Martin Sebor <msebor@redhat.com>
4522 * calls.c (alloc_max_size): Correct a logic error/typo.
4523 Treat excessive arguments as infinite. Warn for invalid arguments.
4524 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4526 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4529 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4530 and movx for Haswell.
4532 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4533 Cesar Philippidis <cesar@codesourcery.com>
4536 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4537 when emitting error on private/firstprivate reductions.
4538 * omp-low.c (lower_omp_target): Avoid reference-type processing
4539 on pointers for firstprivate clause.
4541 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4543 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4546 * config/aarch64/aarch64-simd.md
4547 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4548 (aarch64_ld1_x3_<mode>): Likewise
4549 (aarch64_st1x2<VALLDIF:mode>): Likewise
4550 (aarch64_st1_x2_<mode>): Likewise
4551 (aarch64_st1x3<VALLDIF:mode>): Likewise
4552 (aarch64_st1_x3_<mode>): Likewise
4553 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4554 (vld1_s8_x3): Likewise.
4555 (vld1_u16_x3): Likewise.
4556 (vld1_s16_x3): Likewise.
4557 (vld1_u32_x3): Likewise.
4558 (vld1_s32_x3): Likewise.
4559 (vld1_u64_x3): Likewise.
4560 (vld1_s64_x3): Likewise.
4561 (vld1_f16_x3): Likewise.
4562 (vld1_f32_x3): Likewise.
4563 (vld1_f64_x3): Likewise.
4564 (vld1_p8_x3): Likewise.
4565 (vld1_p16_x3): Likewise.
4566 (vld1_p64_x3): Likewise.
4567 (vld1q_u8_x3): Likewise.
4568 (vld1q_s8_x3): Likewise.
4569 (vld1q_u16_x3): Likewise.
4570 (vld1q_s16_x3): Likewise.
4571 (vld1q_u32_x3): Likewise.
4572 (vld1q_s32_x3): Likewise.
4573 (vld1q_u64_x3): Likewise.
4574 (vld1q_s64_x3): Likewise.
4575 (vld1q_f16_x3): Likewise.
4576 (vld1q_f32_x3): Likewise.
4577 (vld1q_f64_x3): Likewise.
4578 (vld1q_p8_x3): Likewise.
4579 (vld1q_p16_x3): Likewise.
4580 (vld1q_p64_x3): Likewise.
4581 (vst1_s64_x2): Likewise.
4582 (vst1_u64_x2): Likewise.
4583 (vst1_f64_x2): Likewise.
4584 (vst1_s8_x2): Likewise.
4585 (vst1_p8_x2): Likewise.
4586 (vst1_s16_x2): Likewise.
4587 (vst1_p16_x2): Likewise.
4588 (vst1_s32_x2): Likewise.
4589 (vst1_u8_x2): Likewise.
4590 (vst1_u16_x2): Likewise.
4591 (vst1_u32_x2): Likewise.
4592 (vst1_f16_x2): Likewise.
4593 (vst1_f32_x2): Likewise.
4594 (vst1_p64_x2): Likewise.
4595 (vst1q_s8_x2): Likewise.
4596 (vst1q_p8_x2): Likewise.
4597 (vst1q_s16_x2): Likewise.
4598 (vst1q_p16_x2): Likewise.
4599 (vst1q_s32_x2): Likewise.
4600 (vst1q_s64_x2): Likewise.
4601 (vst1q_u8_x2): Likewise.
4602 (vst1q_u16_x2): Likewise.
4603 (vst1q_u32_x2): Likewise.
4604 (vst1q_u64_x2): Likewise.
4605 (vst1q_f16_x2): Likewise.
4606 (vst1q_f32_x2): Likewise.
4607 (vst1q_f64_x2): Likewise.
4608 (vst1q_p64_x2): Likewise.
4609 (vst1_s64_x3): Likewise.
4610 (vst1_u64_x3): Likewise.
4611 (vst1_f64_x3): Likewise.
4612 (vst1_s8_x3): Likewise.
4613 (vst1_p8_x3): Likewise.
4614 (vst1_s16_x3): Likewise.
4615 (vst1_p16_x3): Likewise.
4616 (vst1_s32_x3): Likewise.
4617 (vst1_u8_x3): Likewise.
4618 (vst1_u16_x3): Likewise.
4619 (vst1_u32_x3): Likewise.
4620 (vst1_f16_x3): Likewise.
4621 (vst1_f32_x3): Likewise.
4622 (vst1_p64_x3): Likewise.
4623 (vst1q_s8_x3): Likewise.
4624 (vst1q_p8_x3): Likewise.
4625 (vst1q_s16_x3): Likewise.
4626 (vst1q_p16_x3): Likewise.
4627 (vst1q_s32_x3): Likewise.
4628 (vst1q_s64_x3): Likewise.
4629 (vst1q_u8_x3): Likewise.
4630 (vst1q_u16_x3): Likewise.
4631 (vst1q_u32_x3): Likewise.
4632 (vst1q_u64_x3): Likewise.
4633 (vst1q_f16_x3): Likewise.
4634 (vst1q_f32_x3): Likewise.
4635 (vst1q_f64_x3): Likewise.
4636 (vst1q_p64_x3): Likewise.
4638 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4640 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4641 user_label_prefix to name.
4643 * tree-core.h: Update comment about the format of NAME string
4644 passed to handler in attribute_spec.
4646 * config/msp430/msp430.md: Remove erroneous subreg expression from
4647 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4648 zero_extend{q,h}isi2.
4650 2018-05-30 Borislav Petkov <bp@suse.de>
4652 * doc/extend.texi: Document some architecture specific
4653 constraints and sort entries.
4655 2018-05-30 Martin Sebor <msebor@redhat.com>
4658 * builtins.c (expand_builtin_stpcpy_1): New function.
4659 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4660 only if the former succeeds.
4662 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4664 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4667 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4669 * doc/invoke.texi (-flinker-output): Document
4671 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4673 * passes.c (ipa_write_summaries): Only modify statements if body
4675 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4676 incrementally linking.
4677 (ipa_passes): Likewise.
4678 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4679 pass down resolution info.
4680 * common.opt (flag_incremental_link): Update info.
4681 * gcc.c (plugin specs): Turn flinker-output=* to
4682 -plugin-opt=-linker-output-known
4683 * toplev.c (compile_file): Also cut compilation when doing incremental
4685 * flag-types. (enum lto_partition_model): Add
4686 LTO_LINKER_OUTPUT_NOLTOREL.
4687 (invoke.texi): Add -flinker-output docs.
4688 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4689 link same way as WPA; do not stream in dead initializers.
4691 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4692 darf2out_register_external_decl): Support incremental link.
4694 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4696 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4698 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4700 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4701 it down to simple_object_copy_lto_debug_sections.
4702 (run_gcc): Determine incremental LTO link time and configure
4703 lto1 into non-wpa mode, disable renaming of debug sections.
4705 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4707 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4708 descriptions of various incorrectly documented functions.
4710 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4713 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4714 address check not strict.
4716 2018-05-30 Richard Biener <rguenther@suse.de>
4718 PR tree-optimization/85964
4719 * tracer.c (better_p): Drop initialized count check, we only
4720 call the function with initialized counts now.
4721 (find_best_successor): Do find a best edge if one
4722 has uninitialized count.
4723 (find_best_predecessor): Likewise. Do BB frequency check only
4724 if count is initialized.
4726 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4728 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4729 (aarch64_ldrstr_offset_compare): New.
4730 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4731 load/store orderings.
4732 (aarch64_gen_adjusted_ldpstp): Likewise.
4734 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4736 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4737 Check for subset of GENERAL_REGS and FP_REGS.
4738 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4741 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4743 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4744 and wi::to_poly_offset. Add the current offset and then check
4745 whether the sum fits, rather than using an unchecked addition of
4746 a checked term. Check for a shwi rather than a uhwi.
4747 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4748 (store_constructor): Use poly_int_tree_p.
4749 (expand_expr_real_1): Likewise.
4750 * function.c (assign_temp): Likewise.
4751 * fold-const.c (const_binop): Use poly_int_tree_p and
4753 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4755 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4756 to_poly_offset for MEM offsets.
4757 * ipa-icf.c (sem_variable::equals): Likewise.
4758 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4759 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4760 wi::to_poly_offset for BIT_FIELD_REF offsets.
4761 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4763 * var-tracking.c (emit_note_insn_var_location): Use
4764 tree_to_poly_uint64.
4766 2018-05-29 Jim Wilson <jimw@sifive.com>
4768 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4770 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4773 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4774 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4776 (sse4_1_round<mode>2): Use nonimmediate_operand
4777 for operand 1 predicate.
4779 2018-05-29 Martin Sebor <msebor@redhat.com>
4780 Richard Biener <rguenther@suse.de>
4783 * calls.c (get_size_range): Call determine_value_range instead
4784 of get_value_range..
4785 * tree-vrp.h (determine_value_range): Declared new function.
4786 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4788 2018-05-29 Richard Biener <rguenther@suse.de>
4790 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4791 sure to use non-pattern stmts for get_earlier_stmt arguments.
4792 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4793 called on pattern stmts.
4794 (get_later_stmt): Likewise.
4796 2018-05-29 Martin Liska <mliska@suse.cz>
4798 PR gcov-profile/85759
4799 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4802 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4804 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4806 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4810 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4811 VEC_PACK_FLOAT_EXPR): New tree codes.
4812 * tree-pretty-print.c (op_code_prio): Handle
4813 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4814 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4815 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4816 * tree-inline.c (estimate_operator_cost): Likewise.
4817 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4818 * fold-const.c (const_binop): Likewise.
4819 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4820 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4821 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4822 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4823 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4824 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4825 * expr.c (expand_expr_real_2): Likewise.
4826 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4827 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4828 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4830 * optabs.c (expand_widen_pattern_expr): For
4831 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4832 sign from result type rather than operand's type.
4833 (expand_binop_directly): For vec_packu_float_optab and
4834 vec_packs_float_optab allow result type to be different from operand's
4836 * optabs-tree.c (optab_for_tree_code): Handle
4837 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4838 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4839 * tree-vect-generic.c (expand_vector_operations_1): Handle
4840 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4841 VEC_PACK_FLOAT_EXPR.
4842 * tree-vect-stmts.c (supportable_widening_operation): Handle
4844 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4845 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4846 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4847 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4848 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4850 (vec_pack<floatprefix>_float_<mode>): New expander.
4851 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4853 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4854 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4855 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4856 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4857 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4859 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4860 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4861 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4862 VEC_PACK_FLOAT_EXPR): Document.
4864 2018-05-29 Richard Biener <rguenther@suse.de>
4866 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4868 (stmt_vec_info_vec): Make pointer.
4869 (init_stmt_vec_info_vec): Remove.
4870 (free_stmt_vec_info_vec): Likewise.
4871 (set_stmt_vec_info_vec): New function.
4872 (free_stmt_vec_infos): Likewise.
4873 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4874 (set_vinfo_for_stmt): Likewise.
4875 (get_earlier_stmt): Likewise.
4876 (get_later_stmt): Likewise.
4877 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4878 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4879 (vec_info::~vec_info): Free stmt_vec_infos.
4880 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4881 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4882 (pass_slp_vectorize::execute): Likewise.
4883 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4884 (free_stmt_vec_info_vec): Likewise.
4885 (set_stmt_vec_info_vec): New function.
4886 (free_stmt_vec_infos): Likewise.
4887 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4888 the global stmt_vec_info_vec.
4889 * tree-parloops.c (gather_scalar_reductions): Use
4890 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4893 2018-05-29 Richard Biener <rguenther@suse.de>
4895 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4897 2018-05-29 Martin Liska <mliska@suse.cz>
4898 David Malcolm <dmalcolm@redhat.com>
4900 * vec.c (test_reverse): New.
4901 (vec_c_tests): Add new test.
4902 * vec.h (vl_ptr>::reverse): New function.
4904 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4906 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4908 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4911 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4913 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4915 2018-05-28 Richard Biener <rguenther@suse.de>
4917 PR tree-optimization/85933
4918 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4919 look at stmts marked as vectorizable.
4921 2018-05-28 Richard Biener <rguenther@suse.de>
4923 PR tree-optimization/85934
4924 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4925 vector boolean check before scalar optimization.
4927 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4929 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4932 2018-05-28 Mark Wielaard <mark@klomp.org>
4934 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4935 if it is an expression containing a minus sign.
4937 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4939 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4941 2018-05-27 Paul Koning <ni1d@arrl.net>
4943 * config/pdp11/pdp11.md (truncsihi2): Remove.
4945 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4946 Chung-Ju Wu <jasonwucj@gmail.com>
4948 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4950 (unaligned_store_dw): Ditto.
4951 * config/nds32/nds32-memory-manipulation.c
4952 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4953 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4954 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4955 (emit_setmem_word_loop): Rename to ...
4956 (emit_setmem_doubleword_loop): ... this.
4957 (nds32_gen_dup_4_byte_to_word_value): New function.
4958 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4959 (nds32_expand_setmem_loop): Refine implementation.
4960 (nds32_expand_setmem_loop_v3m): Ditto.
4961 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4964 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4966 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4968 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4970 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4971 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4972 machine->attr_no_prologue_p.
4973 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4974 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4975 (nds32_expand_epilogue): Consider attr_naked_p.
4976 (nds32_expand_epilogue_v3pop): Likewise.
4977 (nds32_can_use_return_insn): Likewise.
4978 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4979 attr_no_prologue_p fields.
4980 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4982 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4985 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4987 * config/i386/sse.md
4988 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4990 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4992 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4994 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4996 (*<floatsuffix>floatv2div2sf2): Rename to ...
4997 (*float<floatunssuffix>v2div2sf2): ... this.
4998 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4999 (float<floatunssuffix>v2div2sf2_mask): ... this.
5000 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5001 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5002 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5004 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5006 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5008 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5010 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5012 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5014 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5015 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5016 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5017 gen_ufix_truncv8dfv8si2.
5018 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5019 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5020 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5021 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5022 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5023 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5024 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5025 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5027 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5031 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5033 2018-05-25 Jim Wilson <jimw@sifive.com>
5035 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5036 * config/riscv/riscv.c (struct machine_function): Add
5037 interrupt_handler_p and attribute_checked_p fields.
5038 (riscv_attribute_table): Add interrupt.
5039 (riscv_interrupt_type_p): New.
5040 (riscv_save_reg_p): Save extra regs for interrupt handler.
5041 (riscv_use_save_libcall): Return false for interrupt handler.
5042 (riscv_first_stack_step): Add forward declaration.
5043 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5044 for interrupt handler with large frame. Use it for saved reg list.
5045 (riscv_expand_prologue): Move flag_stack_usage_info support to
5046 eliminate duplication.
5047 (riscv_expand_epilogue): Generate mret for interrupt handler.
5048 (riscv_epilogue_uses): New.
5049 (riscv_can_use_return_insn): Return false for interrupt handler.
5050 (riscv_function_ok_for_sibcall): Likewise.
5051 (riscv_set_current_function): Add interrupt handler support.
5052 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5053 * config/riscv/riscv.md (UNSPECV_MRET): New.
5055 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5057 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5059 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5061 PR tree-optimization/85712
5062 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5063 this candidate has already been replaced in-situ by a copy.
5065 2018-05-25 Jason Merrill <jason@redhat.com>
5067 PR c++/80485 - inline function non-zero address.
5068 * symtab.c (nonzero_address): Check DECL_COMDAT.
5070 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5073 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5074 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5075 extension of SImode operation. Use const123_operand predicate.
5076 (*saddsi_1): Remove.
5077 (*saddl_se_1): Ditto.
5079 (*ssubl_se_1): Ditto.
5080 * config/alpha/predicates.md (const123_operand): New predicate.
5081 * config/alpha/constraints.md (P): Use IN_RANGE.
5083 2018-05-25 Richard Biener <rguenther@suse.de>
5085 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5087 (ref_maybe_used_by_stmt_p): Likewise.
5088 (stmt_may_clobber_ref_p): Likewise.
5089 (stmt_may_clobber_ref_p_1): Likewise.
5090 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5092 (ref_maybe_used_by_stmt_p): Likewise.
5093 (stmt_may_clobber_ref_p): Likewise.
5094 (stmt_may_clobber_ref_p_1): Likewise.
5095 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5096 the alias oracle to disambiguate DRs with stmts DR analysis
5098 (vect_analyze_data_refs): Do not give up on not analyzable
5099 DRs for BB vectorization. Remove code truncating the dataref
5102 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5105 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5106 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5107 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5109 2018-05-25 Richard Biener <rguenther@suse.de>
5111 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5112 function, combining stmt data ref gathering and fatal analysis
5114 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5115 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5116 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5117 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5118 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5120 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5122 PR tree-optimization/85720
5123 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5124 SCC if all partitions are builtins.
5125 (version_loop_by_alias_check): New parameter. Generate cancelable
5126 runtime alias check if all partitions are builtins.
5127 (distribute_loop): Update call to above function.
5129 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5131 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5132 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5133 (parm_default_def_partition_arg): Ditto.
5134 (set_parm_default_def_partition): Ditto.
5135 (get_parm_default_def_partitions): Ditto and make it static.
5136 (get_undefined_value_partitions): Ditto and make it static.
5137 (remove_ssa_form): Refactor call to init_var_map here.
5138 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5139 computation for loop region.
5140 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5141 (register_default_def): Delete.
5142 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5143 (parm_default_def_partition_arg): Ditto.
5144 (set_parm_default_def_partition): Ditto.
5145 (get_parm_default_def_partitions): Ditto and make it static.
5146 (get_undefined_value_partitions): Ditto and make it static.
5147 (coalesce_with_default, coalesce_with_default): Update comment.
5148 (create_coalesce_list_for_region): New func factored out from
5149 create_outofssa_var_map.
5150 (populate_coalesce_list_for_outofssa): New func factored out from
5151 create_outofssa_var_map and coalesce_ssa_name.
5152 (create_outofssa_var_map): Delete.
5153 (coalesce_ssa_name): Refactor to support live range computation.
5154 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5155 (get_parm_default_def_partitions): Delete.
5156 (get_undefined_value_partitions): Ditto.
5157 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5158 computation for loop region.
5159 (new_tree_live_info, loe_visit_block): Ditto.
5160 (live_worklist, set_var_live_on_entry): Ditto.
5161 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5162 * tree-ssa-live.h (struct _var_map): New fields.
5163 (init_var_map): Change decl.
5164 (region_contains_p): New.
5166 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5168 * tree-ssa-live.h (live_merge_and_clear): Delete.
5170 2018-05-25 Richard Biener <rguenther@suse.de>
5173 * tree-dump.c (dequeue_and_dump): Remove access to removed
5174 operand 2 of a SWITCH_EXPR.
5176 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5178 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5179 multiplication and division.
5180 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5181 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5182 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5183 (cond_udiv_optab, cond_umod_optab): New optabs.
5184 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5185 (IFN_COND_RDIV): New internal functions.
5186 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5187 TRUNC_MOD_EXPR and RDIV_EXPR.
5188 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5189 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5191 (SVE_INT_BINARY): Include mult.
5192 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5193 (optab, sve_int_op): Handle mult.
5194 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5196 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5197 for SVE_INT_BINARY_SD.
5199 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5201 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5202 (optab, sve_int_op): Handle div and udiv.
5203 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5204 for SVE_INT_BINARY_SD.
5205 (*<optab><mode>3): New insn for the same.
5207 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5209 * tree-vect-patterns.c: Include predict.h.
5210 (vect_recog_divmod_pattern): Restrict check for division support
5211 to when optimizing for size.
5213 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5215 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5216 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5217 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5218 (gimple_match_op::set_op): Likewise.
5219 (gimple_resimplify4): Declare.
5220 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5221 (expr::gen_transform): Likewise.
5222 (decision_tree::gen): Generate a simplification routine for 4 operands.
5223 * gimple-match-head.c (gimple_simplify): Add an overload for
5224 4 operands. In the top-level function, handle up to 4 call
5225 arguments and call gimple_resimplify4.
5226 (gimple_resimplify4): New function.
5227 (build_call_internal): Pass a fourth operand.
5228 (maybe_push_to_seq): Likewise.
5229 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5230 Fold VEC_COND_EXPRs of an operation and a default value into
5231 an IFN_COND_* function if possible.
5232 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5234 (SVE_COND_FP_BINARY): Include them.
5235 (optab, sve_fp_op): Handle them.
5236 (SVE_INT_BINARY_REV): New code iterator.
5237 (SVE_COND_FP_BINARY_REV): New int iterator.
5238 (commutative): New int attribute.
5239 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5241 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5243 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5244 (*cond_<optab><mode>): New patterns for reversed operands.
5246 2018-05-25 Richard Biener <rguenther@suse.de>
5248 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5249 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5250 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5251 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5252 * tree-vect-data-refs.c (everywhere): Adjust users.
5253 * tree-vect-loop.c (everywhere): Likewise.
5254 * tree-vect-slp.c (everywhere): Likewise.
5255 * tree-vect-stmts.c (everywhere): Likewise.
5256 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5258 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5260 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5262 (gcc_cv_as_section_exclude): ... this.
5263 Try Solaris as #exclude syntax.
5264 * configure: Regenerate.
5265 * config.in: Regenerate.
5266 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5268 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5269 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5271 * varasm.c (default_elf_asm_named_section): Don't check if
5272 HAVE_GAS_SECTION_EXCLUDE is defined.
5274 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5276 * doc/md.texi: Update the documentation of the cond_* optabs
5277 to mention the new final operand. Fix GET_MODE_NUNITS call.
5278 Describe the scalar case too.
5279 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5280 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5282 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5283 (get_conditional_internal_fn): Update comment.
5284 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5285 accumulator value as a final argument to conditional functions.
5286 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5287 a define_expand and add an "else" operand. Assert for now that
5288 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5289 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5290 (*cond_<optab><mode>): New patterns.
5291 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5292 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5293 (UNSPEC_COND_EOR): Delete.
5294 (optab): Remove associated mappings.
5295 (SVE_INT_BINARY): New code iterator.
5296 (sve_int_op): Remove int attribute and add "minus" to the code
5298 (SVE_COND_INT_OP): Delete.
5299 (SVE_COND_FP_OP): Rename to...
5300 (SVE_COND_FP_BINARY): ...this.
5302 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5304 * optabs.c (can_reuse_operands_p): New function.
5305 (maybe_legitimize_operands): Try to reuse the results for
5308 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5310 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5311 Add {q} suffix to insn mnemonic.
5313 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5315 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5316 (msp430_warn_func_return): New.
5318 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5320 * fold-const.c (tree_nonzero_bits): New function.
5321 * fold-const.h (tree_nonzero_bits): Likewise.
5322 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5323 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5325 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5329 * varasm.c (assemble_alias): Check ifunc_resolver only on
5332 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5335 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5336 when memory input operand is handled.
5338 2018-05-24 Luis Machado <luis.machado@linaro.org>
5340 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5342 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5344 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5346 * match.pd: Delay FMA folds until after vectorization.
5348 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5351 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5352 address check not strict.
5354 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5356 * gimple-match.h (gimple_match_op): New class.
5357 (mprts_hook): Replace parameters with a gimple_match_op *.
5358 (maybe_build_generic_op): Likewise.
5359 (gimple_simplified_result_is_gimple_val): Replace parameters with
5360 a const gimple_match_op *.
5361 (gimple_simplify): Replace code_helper * and tree * parameters with
5362 a gimple_match_op * parameter.
5363 (gimple_resimplify1): Replace code_helper *, tree and tree *
5364 parameters with a gimple_match_op * parameter.
5365 (gimple_resimplify2): Likewise.
5366 (gimple_resimplify3): Likewise.
5367 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5368 parameters with a gimple_match_op * parameter.
5369 * gimple-match-head.c (gimple_simplify): Change prototypes of
5370 auto-generated functions to take a gimple_match_op * instead of
5371 separate code_helper * and tree * parameters. Make the same
5372 change in the top-level overload and update calls to the
5373 gimple_resimplify routines. Update calls to the auto-generated
5374 functions and to maybe_push_res_to_seq in the publicly-facing
5375 operation-specific gimple_simplify overloads.
5376 (gimple_match_op::MAX_NUM_OPS): Define.
5377 (gimple_resimplify1): Replace rcode and ops with a single res_op
5378 parameter. Update call to gimple_simplify.
5379 (gimple_resimplify2): Likewise.
5380 (gimple_resimplify3): Likewise.
5381 (mprts_hook): Replace parameters with a gimple_match_op *.
5382 (maybe_build_generic_op): Likewise.
5383 (build_call_internal): Replace type, nargs and ops with
5384 a gimple_match_op *.
5385 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5386 with a single gimple_match_op *. Update calls to mprts_hook,
5387 build_call_internal and gimple_simplified_result_is_gimple_val.
5388 Factor out code that is common to the tree_code and combined_fn cases.
5389 * genmatch.c (expr::gen_transform): Replace tem_code and
5390 tem_ops with a gimple_match_op called tem_op. Update calls
5391 to the gimple_resimplify functions and maybe_push_res_to_seq.
5392 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5393 res_ops. Update call to the gimple_resimplify functions.
5394 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5395 (decision_tree::gen): Make the functions take a gimple_match_op *
5396 called res_op instead of separate res_code and res_ops parameters.
5397 Update call accordingly.
5398 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5399 and ops with a single res_op parameter. Update calls to
5400 maybe_build_generic_op and maybe_push_res_to_seq.
5401 (fold_stmt_1): Update calls to gimple_simplify and
5402 replace_stmt_with_simplification.
5403 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5404 and gimple_simplified_result_is_gimple_val.
5405 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5407 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5408 with a gimple_match_op *.
5409 (vn_nary_build_or_lookup): Likewise. Update call to
5410 vn_nary_build_or_lookup_1.
5411 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5412 gimple_match_op *. Update calls to the gimple_resimplify routines
5413 and to gimple_simplified_result_is_gimple_val.
5414 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5415 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5416 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5417 (visit_nary_op): Likewise.
5418 (visit_reference_op_load): Likewise.
5420 2018-05-23 Luis Machado <luis.machado@linaro.org>
5422 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5423 modifier for printing the step amount.
5425 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5428 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5431 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5433 * doc/sourcebuild.texi (Endianness): New subsubsection.
5435 2018-05-23 Luis Machado <luis.machado@linaro.org>
5437 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5438 <prefetch_dynamic_strides>: New const bool field.
5439 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5440 prefetch_dynamic_strides.
5441 (exynosm1_prefetch_tune): Likewise.
5442 (thunderxt88_prefetch_tune): Likewise.
5443 (thunderx_prefetch_tune): Likewise.
5444 (thunderx2t99_prefetch_tune): Likewise.
5445 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5447 (aarch64_override_options_internal): Update to set
5448 PARAM_PREFETCH_DYNAMIC_STRIDES.
5449 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5450 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5451 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5452 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5453 prefetch-dynamic-strides setting.
5455 2018-05-23 Luis Machado <luis.machado@linaro.org>
5457 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5458 <minimum_stride>: New const int field.
5459 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5460 minimum_stride field defaulting to -1.
5461 (exynosm1_prefetch_tune): Likewise.
5462 (thunderxt88_prefetch_tune): Likewise.
5463 (thunderx_prefetch_tune): Likewise.
5464 (thunderx2t99_prefetch_tune): Likewise.
5465 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5466 <default_opt_level>: Set to 3.
5467 (aarch64_override_options_internal): Update to set
5468 PARAM_PREFETCH_MINIMUM_STRIDE.
5469 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5470 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5471 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5472 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5473 stride is constant and is below the minimum stride threshold.
5475 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5477 * config/arm/arm-cpus.in (mode26): Delete.
5478 (armv4): Delete mode26 reference.
5479 * config/arm/arm.c (arm_configure_build_target): Delete use of
5482 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5484 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5486 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5487 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5489 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5490 (floatunsdidf2): Ditto.
5492 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5494 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5495 (fixuns_trunc<mode>si2_avx512f): Ditto.
5496 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5497 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5498 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5500 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5502 PR rtl-optimization/79985
5503 * df-scan.c (df_insn_refs_collect): Remove special case for
5504 global registers and asm statements.
5506 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5508 * extend.texi (Global Register Variables): Rewrite the bullet list.
5509 Note that the register is available for allocation. Note that access
5510 via inline asm must use constraints. Add note about async-signal
5511 handlers. Remove paragraph about automagic register selection.
5513 2018-05-23 Richard Biener <rguenther@suse.de>
5515 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5516 of fixed offset from memset VN.
5518 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5520 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5522 (alloc_cand_and_find_basis): Initialize first_interp field.
5523 (slsr_process_mul): Modify first_interp field.
5524 (slsr_process_add): Likewise.
5525 (slsr_process_cast): Modify first_interp field for each new
5527 (slsr_process_copy): Likewise.
5528 (dump_candidate): Dump first_interp field.
5529 (replace_mult_candidate): Process all interpretations, not just
5531 (replace_rhs_if_not_dup): Likewise.
5532 (replace_one_candidate): Likewise.
5534 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5536 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5538 (aarch64_needs_frame_chain): New function.
5539 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5541 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5544 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5545 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5546 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5547 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5548 as true for strict-align.
5549 (aarch64_can_inline_p): Perform checks even when callee has no
5550 attributes to check for strict alignment.
5551 * doc/extend.texi (AArch64 Function Attributes): Document
5553 * doc/invoke.texi: (AArch64 Options): Likewise.
5555 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5557 PR tree-optimization/85853
5558 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5559 the handling of the root of the node to...
5560 (vect_slp_analyze_node_operations_1): ...this new function,
5561 and run the whole thing with the child nodes' def types
5562 set according to their SLP node's def type.
5564 2018-05-23 Richard Biener <rguenther@suse.de>
5567 * tree-data-ref.c (create_runtime_alias_checks): Defer
5568 and ignore overflow warnings.
5570 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5572 PR tree-optimization/85822
5573 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5576 2018-05-23 Richard Biener <rguenther@suse.de>
5578 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5579 memset constants via native_interpret_expr.
5581 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5584 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5586 (cgraph_node::create_alias): Likewise.
5587 (cgraph_node::get_availability): Check ifunc_resolver instead
5588 of looking up ifunc attribute.
5589 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5590 * varasm.c (do_assemble_alias): Likewise.
5591 (assemble_alias): Likewise.
5592 (default_binds_local_p_3): Likewise.
5593 * cgraph.h (cgraph_node): Add ifunc_resolver.
5594 (cgraph_node::only_called_directly_or_aliased_p): Return false
5596 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5598 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5599 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5600 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5601 instead of looking up ifunc attribute.
5603 2018-05-22 Luis Machado <luis.machado@linaro.org>
5605 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5607 2018-05-22 Martin Sebor <msebor@redhat.com>
5610 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5611 only when expasion succeeds.
5612 (expand_builtin_strcmp): Same.
5613 (expand_builtin_strncmp): Same.
5615 2018-05-22 Martin Sebor <msebor@redhat.com>
5617 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5619 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5620 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5622 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5623 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5625 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5626 Add check for legitimate_address.
5627 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5628 (aarch64_swap_ldrstr_operands): New.
5629 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5632 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5633 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5635 * config/aarch64/aarch64.md: New patterns to generate stp
5637 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5638 single words and double words.
5639 (load_pair_sw, load_pair_dw): Likewise.
5640 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5642 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5644 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5645 for different mode ldpstp and add peephole for merged zero stores.
5647 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5649 (aarch64_gen_store_pair): Rename calls to match new patterns.
5650 (aarch64_gen_load_pair): Rename calls to match new patterns.
5651 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5652 (load_pair<DREG:mode><DREG2:mode>): ... This.
5653 (store_pair<mode>): Rename to...
5654 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5655 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5657 (V_INT_EQUIV): Handle SImode.
5658 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5661 2018-05-22 Martin Sebor <msebor@redhat.com>
5664 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5665 or ajust the presumed bound on an operation to avoid unnecessary
5668 2018-05-22 Martin Sebor <msebor@redhat.com>
5670 PR tree-optimization/85826
5671 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5672 assuming that a DECL necesarily has a constant size.
5674 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5677 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5679 2018-05-22 Richard Biener <rguenther@suse.de>
5681 PR tree-optimization/85834
5682 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5683 non-constant and non-zero memset arguments.
5685 2018-05-22 Martin Liska <mliska@suse.cz>
5688 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5690 2018-05-22 Richard Biener <rguenther@suse.de>
5692 PR tree-optimization/85863
5693 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5694 comparisons when vectype is specified.
5695 (vectorizable_condition): Do not specify vectype for
5696 vect_is_simple_cond when SLP vectorizing.
5698 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5701 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5702 define __ibm128 as long double.
5703 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5704 as a distinct type when IEEE 128-bit support is enabled.
5705 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5706 128-bit types to use the correct functions.
5707 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5708 convert between 128-bit floating point types that have different
5709 modes but the same representation, instead of using gen_lowpart to
5711 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5713 (IFKF_reg): New attributes to give the register constraints for
5715 (extend<mode>tf2_internal): New insns to mark an explicit
5716 conversion between 128-bit floating point types that have a
5717 different mode but share the same representation.
5719 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5721 PR tree-optimization/85814
5722 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5723 a null return from get_strinfo when unsharing the next
5724 strinfo in the chain.
5726 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5729 * varasm.c (weak_finish): Clean up weak_decls.
5731 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5733 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5734 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5735 UNSPEC_UADALP values.
5736 * config/aarch64/iterators.md (ABAL): New int iterator.
5739 (sur): Add mappings for the above.
5740 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5742 (aarch64_<sur>abal<mode>_4): Likewise.
5743 (aarch64_<sur>adalp<mode>_3): Likewise.
5744 (<sur>sadv16qi): New define_expand.
5746 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5748 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5749 (*movdf_internal): Ditto.
5750 (*rcpsf2_sse): Ditto.
5751 (*rsqrtsf2_sse): Ditto.
5752 (*sqrt<mode>2_sse): Ditto.
5754 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5756 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5758 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5759 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5760 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5761 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5763 * config/aarch64/arm_neon.h: Likewise.
5764 * config/aarch64/iterators.md (VQ_I): New.
5766 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5768 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5770 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5773 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5775 * config/nds32/constraints.md (S): New constraint.
5776 * config/nds32/nds32.md (call_internal): Use constraint S.
5777 (call_value_internal): Likewise.
5778 (sibcall_internal): Likewise.
5779 (sibcall_value_internal): Likewise.
5781 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5782 Chung-Ju Wu <jasonwucj@gmail.com>
5784 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5787 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5788 Chung-Ju Wu <jasonwucj@gmail.com>
5790 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5791 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5792 (nds32_rtx_costs_impl): Simplify.
5793 (nds32_address_cost_impl): Simplify.
5794 (nds32_init_rtx_costs): New function.
5795 (nds32_rtx_costs_speed_prefer): Likewise.
5796 (nds32_rtx_costs_size_prefer): Likewise.
5797 (nds32_address_cost_speed_prefer): Likewise.
5798 (nds32_address_cost_speed_fwprop): Likewise.
5799 (nds32_address_cost_size_prefer): Likewise.
5800 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5801 * config/nds32/nds32.c (nds32_option_override): Use
5802 nds32_init_rtx_costs function.
5804 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5806 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5807 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5808 (TARGET_PIPELINE_N8): Likewise.
5809 (TARGET_PIPELINE_N10): Likewise.
5810 (TARGET_PIPELINE_N13): Likewise.
5811 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5813 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5815 * config/nds32/nds32-fpu.md: Update copyright year.
5817 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5819 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5821 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5823 * config/nds32/nds32.c
5824 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5825 * config/nds32/nds32.opt (minline-asm-r15): New option.
5827 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5829 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5831 * config/nds32/nds32.md (abssi2): New pattern.
5833 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5835 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5836 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5837 Merge insn pattern from sse_cvtsi2ss<round_name> and
5838 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5839 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5840 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5841 using SWI48 mode iterator.
5842 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5843 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5844 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5845 pattern from sse_cvttss2si<round_saeonly_name>
5846 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5847 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5848 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5849 using SWI48 mode iterator.
5850 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5851 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5852 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5853 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5854 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5855 using SWI48 mode iterator.
5856 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5857 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5858 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5859 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5860 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5861 SWI48 mode iterator.
5862 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5863 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5864 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5865 pattern from sse_cvttsd2si<round_saeonly_name>
5866 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5868 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5870 * config/nds32/nds32-md-auxiliary.c
5871 (nds32_valid_smw_lwm_base_p): Refine.
5872 (nds32_output_smw_single_word): Refine.
5873 (nds32_output_smw_double_word): New.
5874 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5876 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5878 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5879 (nds32_output_stack_pop): Refine.
5880 (nds32_expand_unaligned_load): Refine.
5881 (nds32_expand_unaligned_store): Refine.
5883 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5884 Chung-Ju Wu <jasonwucj@gmail.com>
5886 * config/nds32/constants.md: Add TP_REGNUM constant.
5887 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5888 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5890 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5891 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5892 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5893 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5894 and PIC code generation.
5895 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5897 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5899 * config/nds32/nds32.md: Support TLS and PIC.
5900 * config/nds32/nds32.c: Support TLS and PIC.
5901 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5902 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5905 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5907 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5908 mode with E_ prefix.
5910 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5911 Chung-Ju Wu <jasonwucj@gmail.com>
5913 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5914 * config/nds32/nds32-md-auxiliary.c
5915 (symbolic_reference_mentioned_p): New.
5916 (nds32_legitimize_ict_address): New.
5917 (nds32_expand_ict_move): New.
5918 (nds32_indirect_call_referenced_p): New.
5919 (nds32_symbol_binds_local_p): Delete.
5920 (nds32_long_call_p): Modify.
5921 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5922 * config/nds32/nds32-protos.h
5923 (symbolic_reference_mentioned_p): Declare.
5924 (nds32_legitimize_ict_address): Declare.
5925 (nds32_expand_ict_move): Declare.
5926 (nds32_indirect_call_referenced_p): Declare.
5927 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5928 (nds32_relax_group): Use nds32_ict_const_p as condition.
5929 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5930 (nds32_asm_file_start): Output ict_model directive in asm code.
5931 (nds32_legitimate_address_p): Consider indirect call.
5932 (nds32_print_operand): Consider indirect call.
5933 (nds32_print_operand_address): Consider indirect call.
5934 (nds32_insert_attributes): Handle "indirect_call" attribute.
5935 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5936 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5937 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5938 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5939 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5940 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5941 (TARGET_ICT_MODEL_SMALL): Define.
5942 (TARGET_ICT_MODEL_LARGE): Define.
5943 * config/nds32/nds32.md (movsi): Consider ict model.
5944 (call, call_value): Consider ict model.
5945 (sibcall, sibcall_value): Consider ict model.
5946 * config/nds32/nds32.opt (mict-model): New option.
5947 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5950 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5951 Monk Chiang <sh.chiang04@gmail.com>
5952 Jim Wilson <jimw@sifive.com>
5954 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5955 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5956 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5957 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5958 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5959 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5960 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5961 compute save_libcall_adjustment properly.
5962 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5963 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5964 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5965 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5966 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5967 (ABI_SPEC): Handle mabi=ilp32e.
5968 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5969 (RVE): Add RVE mask.
5970 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5971 <-march>: Add rv32e as an example.
5973 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5976 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5977 (intra_create_variable_infos): Handle C++ constructors.
5979 2018-05-18 Martin Liska <mliska@suse.cz>
5981 * passes.def: Remove a redundant pass.
5983 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5986 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5988 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5990 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5992 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5994 (armv2, armv2a, armv3, armv3m): Delete architectures.
5995 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5996 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5997 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5999 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6000 (*mulsidi3adddi): Likewise.
6001 (mulsidi3): Likewise.
6002 (*mulsidi3_nov6): Likewise.
6003 (umulsidi3): Likewise.
6004 (umulsidi3_nov6): Likewise.
6005 (umaddsidi4): Likewise.
6006 (*umulsidi3adddi): Likewise.
6007 (smulsi3_highpart): Likewise.
6008 (*smulsi3_highpart_nov6): Likewise.
6009 (umulsi3_highpart): Likewise.
6010 (*umulsi3_highpart_nov6): Likewise.
6011 * config/arm/arm.h (arm_arch3m): Delete.
6012 * config/arm/arm.c (arm_arch3m): Delete.
6013 (arm_option_override_internal): Update armv3-related comment.
6014 (arm_configure_build_target): Delete use of isa_bit_mode32.
6015 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6016 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6017 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6019 (mulusa3): Likewise.
6020 * config/arm/arm-protos.h (arm_arch3m): Delete.
6021 * config/arm/arm-tables.opt: Regenerate.
6022 * config/arm/arm-tune.md: Likewise.
6023 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6024 deleted architectures.
6026 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6028 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6029 (armv5t, armv5te): New features.
6030 (ARMv5, ARMv5e): Delete fgroups.
6031 (ARMv5t, ARMv5te): Adjust for above changes.
6033 (armv5, armv5e): Delete arches.
6034 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6036 (*call_reg_arm): Likewise.
6037 (*call_value_reg_armv5): Likewise.
6038 (*call_value_reg_arm): Likewise.
6039 (*call_symbol): Likewise.
6040 (*call_value_symbol): Likewise.
6041 (*sibcall_insn): Likewise.
6042 (*sibcall_value_insn): Likewise.
6044 (prefetch): Likewise.
6045 (define_split and define_peephole2 dependent on arm_arch5):
6047 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6049 (TARGET_ARM_QBIT): Likewise.
6050 (TARGET_DSP_MULTIPLY): Likewise.
6051 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6052 (arm_arch5, arm_arch5e): Delete.
6053 (arm_arch5t, arm_arch5te): Declare.
6054 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6055 (arm_arch5t): Declare.
6056 (arm_option_reconfigure_globals): Update for the above.
6057 (arm_options_perform_arch_sanity_checks): Update comment, replace
6058 use of arm_arch5 with arm_arch5t.
6059 (use_return_insn): Likewise.
6060 (arm_emit_call_insn): Likewise.
6061 (output_return_instruction): Likewise.
6062 (arm_final_prescan_insn): Likewise.
6063 (arm_coproc_builtin_available): Likewise.
6064 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6065 arm_arch5e with arm_arch5t and arm_arch5te.
6066 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6067 (arm_arch5t, arm_arch5te): Declare.
6068 * config/arm/arm-tables.opt: Regenerate.
6069 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6070 * config/arm/t-multilib: Likewise.
6071 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6072 instead of arm_arch5.
6073 (*call_reg_thumb1): Likewise.
6074 (*call_value_reg_thumb1_v5): Likewise.
6075 (*call_value_reg_thumb1): Likewise.
6076 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6078 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6080 2018-05-18 Martin Liska <mliska@suse.cz>
6082 PR gcov-profile/84846
6083 * doc/gcov.texi: Document -t option of gcov tool.
6085 2018-05-18 Martin Liska <mliska@suse.cz>
6087 PR gcov-profile/84846
6088 * gcov.c (print_usage): Add new -t option.
6089 (process_args): Handle the option.
6090 (generate_results): Use stdout as output when requested by
6093 2018-05-18 Martin Liska <mliska@suse.cz>
6095 PR gcov-profile/84846
6096 * coverage.c (coverage_init): Write PWD to .gcno file.
6097 * doc/gcov.texi: Document how working directory is printed.
6098 * gcov-dump.c (dump_gcov_file): Print PWD.
6099 * gcov.c (output_intermediate_file): Likewise.
6100 (read_graph_file): Read PWD string.
6101 (output_lines): Print PWD.
6103 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6106 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6107 for retval and return false if all args to phi are zero.
6109 2018-05-18 Richard Biener <rguenther@suse.de>
6111 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6113 (evrp_dom_walker::before_dom_children): Call it.
6115 2018-05-18 Richard Biener <rguenther@suse.de>
6117 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6118 results when processing array refs with variable index.
6120 2018-05-18 Toon Moene <toon@moene.org>
6122 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6123 directly after that of -floop-interchange. Indicate that both
6124 options are enabled by default when specifying -O3.
6126 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6128 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6129 iterator. Delete separate integer-mode vec_set<mode> expander.
6130 (aarch64_simd_vec_setv2di): Delete.
6131 (vec_setv2di): Delete.
6132 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6133 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6134 the "w, r" alternative.
6136 2018-05-18 Martin Liska <mliska@suse.cz>
6138 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6139 * tree-pass.h (make_pass_lower_switch_O0): New function.
6140 * tree-switch-conversion.c (node_has_low_bound): Remove.
6141 (node_has_high_bound): Likewise.
6142 (node_is_bounded): Likewise.
6143 (class pass_lower_switch): Make it a template type and create
6145 (pass_lower_switch::execute): Add template argument.
6146 (make_pass_lower_switch): New function.
6147 (make_pass_lower_switch_O0): New function.
6148 (do_jump_if_equal): Remove.
6149 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6150 all the hard work to tree optimization passes.
6152 2018-05-18 Martin Liska <mliska@suse.cz>
6154 * dbgcnt.c (limit_low): Renamed from limit.
6155 (limit_high): New variable.
6156 (dbg_cnt_is_enabled): Check for upper limit.
6157 (dbg_cnt): Adjust dumping.
6158 (dbg_cnt_set_limit_by_index): Add new argument for high
6160 (dbg_cnt_set_limit_by_name): Likewise.
6161 (dbg_cnt_process_single_pair): Parse new format.
6162 (dbg_cnt_process_opt): Use strtok.
6163 (dbg_cnt_list_all_counters): Remove 'value' and add
6165 * doc/invoke.texi: Document changes.
6167 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6169 * doc/sourcebuild.texi (scalar_all_fma): Document.
6170 * tree.def (FMA_EXPR): Delete.
6171 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6172 * internal-fn.c (ternary_direct): New macro.
6173 (expand_ternary_optab_fn): Likewise.
6174 (direct_ternary_optab_supported_p): Likewise.
6175 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6176 * builtins.c (fold_builtin_fma): Delete.
6177 (fold_builtin_3): Don't call it.
6178 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6179 * expr.c (expand_expr_real_2): Likewise.
6180 * fold-const.c (operand_equal_p): Likewise.
6181 (fold_ternary_loc): Likewise.
6182 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6183 * gimple.c (DEFTREECODE): Likewise.
6184 * gimplify.c (gimplify_expr): Likewise.
6185 * optabs-tree.c (optab_for_tree_code): Likewise.
6186 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6187 * tree-eh.c (operation_could_trap_p): Likewise.
6188 (stmt_could_throw_1_p): Likewise.
6189 * tree-inline.c (estimate_operator_cost): Likewise.
6190 * tree-pretty-print.c (dump_generic_node): Likewise.
6191 (op_code_prio): Likewise.
6192 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6193 * tree-ssa-operands.c (get_expr_operands): Likewise.
6194 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6195 * fold-const-call.h (fold_fma): Delete.
6196 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6197 CFN_FNMA and CFN_FNMS.
6199 * genmatch.c (combined_fn): New enum.
6200 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6201 (commutative_op): New function.
6202 (commutate): Use it. Handle more than 2 operands.
6203 (dt_operand::gen_gimple_expr): Use commutative_op.
6204 (parser::parse_expr): Allow :c to be used with non-binary
6205 operators if the commutative operand is known.
6206 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6207 CFN_FMS, CFN_FNMA and CFN_FNMS.
6208 (backprop::process_assign_use): Remove FMA_EXPR handling.
6209 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6210 (gen_hsa_fma): New function.
6211 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6212 IFN_FNMA and IFN_FNMS.
6213 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6214 * gimple-fold.h (follow_all_ssa_edges): Declare.
6215 * gimple-fold.c (follow_all_ssa_edges): New function.
6216 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6217 gimple_build interface and use follow_all_ssa_edges to fold the result.
6218 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6219 instead of checking for optabs directly.
6220 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6221 rather than FMA_EXPRs.
6222 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6223 call to IFN_FMA instead of an FMA_EXPR.
6225 2018-05-17 Jim Wilson <jimw@sifive.com>
6227 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6228 sign extended promoted subreg, and the range does not have the sign bit
6229 set, then do a sign extend.
6231 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6232 test, check for sign extended subreg and/or constant operands, and
6233 do a sign extend in that case.
6235 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6237 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6238 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6240 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6241 Change logics_shift_reg to logics_shift_imm.
6242 (thunderx2t99_fp_loadpair_basic): Delete.
6243 (thunderx2t99_fp_storepair_basic): Delete.
6244 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6245 (thunderx2t99_asimd_polynomial): Delete.
6246 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6247 and neon_fp_mul_d_scalar_q.
6248 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6249 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6250 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6251 (thunderx2t99_asimd_lut): Add missing tbl types.
6252 (thunderx2t99_asimd_ext): Delete.
6253 (thunderx2t99_asimd_load1_1_mult): Delete.
6254 (thunderx2t99_asimd_load1_2_mult): Delete.
6255 (thunderx2t99_asimd_load1_ldp): New.
6256 (thunderx2t99_asimd_load1): New.
6257 (thunderx2t99_asimd_load2): Add missing *load2* types.
6258 (thunderx2t99_asimd_load3): New.
6259 (thunderx2t99_asimd_load4): New.
6260 (thunderx2t99_asimd_store1_1_mult): Delete.
6261 (thunderx2t99_asimd_store1_2_mult): Delete.
6262 (thunderx2t99_asimd_store2_mult): Delete.
6263 (thunderx2t99_asimd_store2_onelane): Delete.
6264 (thunderx2t99_asimd_store_stp): New.
6265 (thunderx2t99_asimd_store1): New.
6266 (thunderx2t99_asimd_store2): New.
6267 (thunderx2t99_asimd_store3): New.
6268 (thunderx2t99_asimd_store4): New.
6270 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6272 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6273 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6275 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6276 Segher Boessenkool <segher@kernel.crashing.org>
6279 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6282 2018-05-17 Richard Biener <rguenther@suse.de>
6284 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6285 for pruning loop and prune defs feeding only already visited PHIs.
6287 2018-05-17 Richard Biener <rguenther@suse.de>
6289 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6291 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6292 Richard Biener <rguenther@suse.de>
6294 PR tree-optimization/85793
6295 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6296 for VMAT_ELEMENTWISE.
6298 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6300 * internal-fn.h (lookup_internal_fn): Declare
6301 * internal-fn.c (lookup_internal_fn): New function.
6302 * gimple.c (gimple_build_call_from_tree): Handle calls to
6304 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6305 internal function names.
6306 * tree-pretty-print.c (dump_generic_node): Likewise.
6307 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6309 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6311 * gimple-fold.h (gimple_build): Make the function forms take
6312 combined_fn rather than built_in_function.
6313 (gimple_simplify): Likewise.
6314 * gimple-match-head.c (gimple_simplify): Likewise.
6315 * gimple-fold.c (gimple_build): Likewise.
6316 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6317 rather than gimple_build_call_internal.
6318 (get_initial_defs_for_reduction): Likewise.
6319 (vect_create_epilog_for_reduction): Likewise.
6320 (vectorizable_live_operation): Likewise.
6322 2018-05-17 Martin Liska <mliska@suse.cz>
6324 * gimple-ssa-sprintf.c (format_directive): Do not use
6325 space in between 'G_' and '('.
6327 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6330 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6331 even if the mask is not all ones.
6334 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6336 (ix86_gimple_fold_builtin): Likewise.
6339 * config/i386/i386.c: Include tree-vector-builder.h.
6340 (ix86_vector_shift_count): New function.
6341 (ix86_fold_builtin): Fold shift builtins by scalar count.
6342 (ix86_gimple_fold_builtin): Likewise.
6344 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6345 _mm512_setzero): New intrinsics.
6347 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6348 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6350 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6351 code generation for cases where splatting a value is not useful.
6352 * simplify-rtx.c (simplify_ternary_operation): Simplify
6353 vec_merge across a vec_duplicate and a paradoxical subreg forming
6354 a vector mode to a vec_concat.
6356 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6358 * config.gcc: Support "goldmont-plus".
6359 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6361 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6362 PROCESSOR_GOLDMONT_PLUS.
6363 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6364 (processor_target_table): Add "goldmont-plus".
6365 (PTA_GOLDMONT_PLUS): Define.
6366 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6367 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6368 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6369 (fold_builtin_cpu): Add "goldmont-plus".
6370 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6371 (ix86_option_override_internal): Add "goldmont-plus".
6372 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6373 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6374 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6375 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6377 2018-05-17 Richard Biener <rguenther@suse.de>
6379 PR tree-optimization/85757
6380 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6381 remove defs that only feed that PHI from further processing.
6383 2018-05-16 Jim Wilson <jimw@sifive.com>
6385 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6387 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6389 2018-05-16 Mark Wielaard <mark@klomp.org>
6391 * dwarf2out.c (count_index_strings): New function.
6392 (output_indirect_strings): Call count_index_strings and generate
6393 header for dwarf_version >= 5.
6395 2018-05-16 Mark Wielaard <mark@klomp.org>
6397 * dwarf2out.c (dwarf_FORM): New function.
6398 (set_indirect_string): Use dwarf_FORM.
6399 (reset_indirect_string): Likewise.
6400 (size_of_die): Likewise.
6401 (value_format): Likewise.
6402 (output_die): Likewise.
6403 (add_skeleton_AT_string): Likewise.
6404 (output_macinfo_op): Likewise.
6405 (index_string): Likewise.
6406 (output_index_string_offset): Likewise.
6407 (output_index_string): Likewise.
6408 (count_index_strings): Likewise.
6410 2018-05-16 Carl Love <cel@us.ibm.com>
6412 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6413 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6415 2018-05-16 Martin Jambor <mjambor@suse.cz>
6417 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6418 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6420 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6422 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6423 (fnma<mode>4): Likewise.
6424 (fms<mode>4): Likewise.
6425 (fnms<mode>4): Likewise.
6426 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6427 (aarch64_fnma<mode>4): Likewise.
6428 (aarch64_fms<mode>4): Likewise.
6429 (aarch64_fnms<mode>4): Likewise.
6430 (aarch64_fnmadd<mode>4): Likewise.
6432 2018-05-16 Jason Merrill <jason@redhat.com>
6434 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6436 2018-05-16 Richard Biener <rguenther@suse.de>
6438 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6439 (dump_stmt_cost): Declare.
6440 (add_stmt_cost): Dump cost we add.
6441 (add_stmt_costs): New function.
6442 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6444 (vect_analyze_stmt): Adjust prototype.
6445 (vectorizable_condition): Likewise.
6446 (vectorizable_live_operation): Likewise.
6447 (vectorizable_reduction): Likewise.
6448 (vectorizable_induction): Likewise.
6449 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6450 cost vector to pass to vectorizable_ and record afterwards.
6451 (vect_model_reduction_cost): Take cost vector argument and adjust.
6452 (vect_model_induction_cost): Likewise.
6453 (vectorizable_reduction): Likewise.
6454 (vectorizable_induction): Likewise.
6455 (vectorizable_live_operation): Likewise.
6456 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6457 SLP_TREE_NUMBER_OF_VEC_STMTS.
6458 (vect_analyze_slp_cost_1): Remove.
6459 (vect_analyze_slp_cost): Likewise.
6460 (vect_slp_analyze_node_operations): Take visited args and
6461 a target cost vector. Avoid processing already visited stmt sets.
6462 (vect_slp_analyze_operations): Use a local cost vector to gather
6463 costs and register those of non-discarded instances.
6464 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6465 (vect_schedule_slp_instance): Remove copying of
6466 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6468 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6469 adding cost. Record cost entry location.
6470 (vect_prologue_cost_for_slp_op): Function to compute cost of
6471 a constant or invariant generated for SLP vect in the prologue,
6472 split out from vect_analyze_slp_cost_1.
6473 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6474 (vect_model_promotion_demotion_cost): Likewise.
6475 (vect_model_store_cost): Likewise, make static.
6476 (vect_model_load_cost): Likewise.
6477 (vectorizable_bswap): Add cost vector arg and adjust.
6478 (vectorizable_call): Likewise.
6479 (vectorizable_simd_clone_call): Likewise.
6480 (vectorizable_conversion): Likewise.
6481 (vectorizable_assignment): Likewise.
6482 (vectorizable_shift): Likewise.
6483 (vectorizable_operation): Likewise.
6484 (vectorizable_store): Likewise.
6485 (vectorizable_load): Likewise.
6486 (vectorizable_condition): Likewise.
6487 (vectorizable_comparison): Likewise.
6488 (can_vectorize_live_stmts): Likewise.
6489 (vect_analyze_stmt): Likewise.
6490 (vect_transform_stmt): Adjust calls to vectorizable_*.
6491 * tree-vectorizer.c: Include gimple-pretty-print.h.
6492 (dump_stmt_cost): New function.
6494 2018-05-16 Richard Biener <rguenther@suse.de>
6496 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6497 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6498 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6499 (check_name): New callback.
6500 (dse_classify_store): Track cycles via a visited bitmap of PHI
6501 defs and simplify handling of in-loop and across loop dead stores
6502 and properly fail for loop-variant refs. Handle byte-tracking with
6503 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6506 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6508 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6509 (vect_get_mask_type_for_stmt): Likewise.
6510 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6512 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6513 to determine the statement's vector type and the vector type that
6514 should be used for calculating nunits. Deal with cases in which
6515 the type has to be deferred.
6516 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6517 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6518 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6519 (vect_determine_vf_for_stmt): New functions, split out from...
6520 (vect_determine_vectorization_factor): ...here.
6521 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6522 (vect_get_mask_type_for_stmt): New functions, split out from
6523 vect_determine_vectorization_factor.
6525 2018-05-16 Richard Biener <rguenther@suse.de>
6527 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6528 verify the [VEC_]COND_EXPR embedded comparison.
6530 2018-05-15 Martin Sebor <msebor@redhat.com>
6532 PR tree-optimization/85753
6533 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6534 RECORD_TYPE in addition to ARRAY_TYPE.
6536 2018-05-15 Martin Sebor <msebor@redhat.com>
6539 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6541 2018-05-15 Richard Biener <rguenther@suse.de>
6543 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6544 add by_clobber_p one. Change algorithm to collect all defs
6545 representing uses we need to walk and try reducing them to
6546 a single one before failing.
6547 (dse_dom_walker::dse_optimize_stmt): Adjust.
6549 2018-05-13 Mark Wielaard <mark@klomp.org>
6551 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6552 (size_of_loc_descr): Likewise.
6553 (output_loc_operands): Likewise.
6554 (output_loc_operands_raw): Likewise.
6555 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6556 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6557 (hash_loc_operands): Likewise.
6558 (compare_loc_operands): Likewise.
6560 2018-05-14 Mark Wielaard <mark@klomp.org>
6562 * dwarf2out.c (count_index_addrs): New function.
6563 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6565 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6567 PR tree-optimization/83648
6568 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6569 return value as malloc candidate.
6571 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6574 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6575 param as true in call to suggest_attribute.
6577 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6579 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6580 -mreadonly-in-sdata.
6582 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6584 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6586 (aarch64_crypto_aesd_fused): Likewise.
6588 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6590 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6591 (movsi_aarch64): Likewise.
6592 (load_pairsi): Likewise.
6593 (load_pairdi): Likewise.
6594 (store_pairsi): Likewise.
6595 (store_pairdi): Likewise.
6596 (load_pairsf): Likewise.
6597 (load_pairdf): Likewise.
6598 (store_pairsf): Likewise.
6599 (store_pairdf): Likewise.
6600 (zero_extend): Likewise.
6601 (trunc): Swap alternatives.
6602 (fcvt_target): Add '?' to prefer w over r.
6604 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6607 * config/i386/i386.md: Disallow non-commutative arithmetics in
6608 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6609 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6610 in the peephole2 before it.
6612 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6614 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6615 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6616 (ix86_handle_option): Handle -mcldemote.
6617 * config.gcc: New header.
6618 * config/i386/cldemoteintrin.h: New file.
6619 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6620 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6622 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6623 OPTION_MASK_ISA_CLDEMOTE.
6624 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6625 (ix86_valid_target_attribute_inner_p): Ditto.
6626 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6627 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6628 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6629 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6630 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6632 * config/i386/i386.opt: Add -mcldemote.
6633 * config/i386/x86intrin.h: New header.
6634 * doc/invoke.texi: Add -mcldemote.
6636 2018-05-14 Richard Biener <rguenther@suse.de>
6638 * doc/match-and-simplify.texi: Adjust :s documentation.
6640 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6642 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6643 intended memcpy size.
6644 (REORDER_45): Likewise.
6646 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6648 * sort.cc: New file.
6649 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6650 * vec.c (qsort_chk): Use gcc_qsort.
6651 * Makefile.in (OBJS-libcommon): Add sort.o.
6652 (build/sort.o): New target. Use it...
6653 (BUILD_RTL): ... here, and...
6654 (build/gencfn-macros): ... here, and...
6655 (build/genmatch): ... here.
6657 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6658 Chung-Ju Wu <jasonwucj@gmail.com>
6660 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6661 * config/nds32/nds32-graywolf.md: New file.
6662 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6663 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6665 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6666 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6667 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6668 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6669 * config/nds32/pipelines.md: Include n15 settings.
6671 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6672 Chung-Ju Wu <jasonwucj@gmail.com>
6674 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6675 * config/nds32/nds32-n13.md: New file.
6676 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6677 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6679 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6680 * config/nds32/nds32.md (pipeline_model): Add n13.
6681 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6682 * config/nds32/pipelines.md: Include n13 settings.
6684 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6685 Chung-Ju Wu <jasonwucj@gmail.com>
6687 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6688 * config/nds32/nds32-n10.md: New file.
6689 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6690 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6692 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6693 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6694 * config/nds32/nds32.md (pipeline_model): Add n10.
6695 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6696 * config/nds32/pipelines.md: Include n10 settings.
6698 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6699 Kito Cheng <kito.cheng@gmail.com>
6700 Chung-Ju Wu <jasonwucj@gmail.com>
6702 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6703 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6704 Add enum values for DSP extension instructions.
6705 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6707 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6708 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6710 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6711 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6712 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6713 * config/nds32/nds32-intrinsic.md: Likewise.
6714 * config/nds32/nds32_intrinsic.h: Likewise.
6715 * config/nds32/nds32-md-auxiliary.c: Likewise.
6716 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6717 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6718 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6719 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6720 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6721 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6722 TYPE_DMAC in switch statement.
6723 * config/nds32/nds32.c: New checking and implementation for DSP
6724 extension instructions.
6725 * config/nds32/nds32.h: Likewise.
6726 * config/nds32/nds32.md: Likewise.
6727 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6728 * config/nds32/predicates.md: Implement new predicates for DSP
6731 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6733 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6734 Reformat alternatives and attributes so it is easier to identify
6735 which constraints/attributes go with which instruction.
6736 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6737 (mov<mode>_softfloat32, FMOVE64): Likewise.
6738 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6739 (mov<mode>_softfloat64, FMOVE64): Likewise.
6741 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6743 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6745 (Basic PowerPC Built-in Functions): The new name of the
6746 subsection previously known as "PowerPC Built-in Functions".
6747 (Basic PowerPC Built-in Functions Available on all Configurations):
6749 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6750 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6751 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6752 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6754 2018-05-11 Martin Jambor <mjambor@suse.cz>
6757 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6760 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6763 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6765 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6767 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6768 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6769 (ix86_handle_option): Handle -mwaitpkg.
6770 * config.gcc: New header.
6771 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6772 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6773 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6775 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6776 OPTION_MASK_ISA_WAITPKG.
6777 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6778 (ix86_option_override_internal): Add PTA_WAITPKG.
6779 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6780 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6781 IX86_BUILTIN_TPAUSE.
6782 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6783 __builtin_ia32_umwait and __builtin_ia32_tpause.
6784 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6785 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6786 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6787 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6788 UNSPECV_TPAUSE): New.
6789 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6790 * config/i386/i386.opt: Add -mwaitpkg.
6791 * config/i386/waitpkgintrin.h: New file.
6792 * config/i386/x86intrin.h: New header.
6793 * doc/invoke.texi: Add -mwaitpkg.
6795 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6798 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6800 (cortex-m0): Use armv6s-m isa.
6801 (cortex-m0plus): Likewise.
6802 (cortex-m1): Likewise.
6803 (cortex-m0.small-multiply): Likewise.
6804 (cortex-m0plus.small-multiply): Likewise.
6805 (cortex-m1.small-multiply): Likewise.
6807 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6808 Jakub Jelinek <jakub@redhat.com>
6810 PR tree-optimization/85692
6811 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6812 source permute as well.
6814 2018-05-11 Martin Liska <mliska@suse.cz>
6817 * doc/extend.texi: Document LLVM style format for no_sanitize
6820 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6822 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6823 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6824 (mode_supports_vsx_dform_quad): Likewise.
6825 (mode_supports_vmx_dform): Move these functions to be next to the
6826 other mode_supports functions.
6827 (mode_supports_dq_form): Likewise.
6828 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6829 mode_supports_dq_form.
6830 (reg_offset_addressing_ok_p): Likewise.
6831 (offsettable_ok_by_alignment): Likewise.
6832 (rs6000_legitimate_offset_address_p): Likewise.
6833 (legitimate_lo_sum_address_p): Likewise.
6834 (rs6000_legitimize_address): Likewise.
6835 (rs6000_legitimize_reload_address): Likewise.
6836 (rs6000_secondary_reload_inner): Likewise.
6837 (rs6000_preferred_reload_class): Likewise.
6838 (rs6000_output_move_128bit): Likewise.
6840 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6842 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6843 Generate SImode target register for null target.
6844 <case IX86_BUILTIN_XGETBV>: Ditto.
6845 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6846 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6848 2018-05-10 Carl Love <cel@us.ibm.com>
6850 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6851 dcbtt and dcbtstt if operands[2] is 0.
6853 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6856 * config/i386/sse.md (usadv64qi): New expander.
6858 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6860 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6861 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6862 -maltivec=be support.
6863 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6864 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6865 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6866 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6867 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6868 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6869 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6870 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6871 altivec_vsumsws): Adjust.
6872 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6873 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6874 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6876 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6877 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6878 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6879 (altivec_lve<VI_char>x): Delete expand.
6880 (*altivec_lve<VI_char>x_internal): Rename to...
6881 (altivec_lve<VI_char>x): ... this.
6882 (altivec_lvxl_<mode>): Delete expand.
6883 (*altivec_lvxl_<mode>_internal): Rename to ...
6884 (altivec_lvxl_<mode>): ... this.
6885 (altivec_stvxl_<mode>): Delete expand.
6886 (*altivec_stvxl_<mode>_internal): Rename to ...
6887 (altivec_stvxl_<mode>): ... this.
6888 (altivec_stve<VI_char>x): Delete expand.
6889 (*altivec_stve<VI_char>x_internal): Rename to ...
6890 (altivec_stve<VI_char>x): ... this.
6891 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6892 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6893 reduc_plus_scal_<mode>): Adjust.
6894 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6896 (rs6000_cpu_cpp_builtins): Adjust.
6897 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6898 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6899 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6900 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6901 -maltivec=be support.
6902 (rs6000_split_vec_extract_var): Adjust.
6903 (rs6000_split_v4si_init): Adjust.
6904 (swap_selector_for_mode): Delete.
6905 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6906 altivec_expand_stvex_be): Delete.
6907 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6908 -maltivec=be support.
6909 (rs6000_gimple_fold_builtin): Ditto.
6910 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6912 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6913 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6914 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6915 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6916 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6917 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6918 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6919 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6920 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6921 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6922 anonymous split): Adjust.
6923 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6924 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6926 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6928 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6929 when --with-gxx-include-dir is also specified.
6930 * configure: Regenerate.
6932 2018-05-09 Jim Wilson <jimw@sifive.com>
6935 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6936 * config/riscv/t-withmultilib: New.
6937 * config/riscv/withmultilib.h: New.
6938 * doc/install.texi: Document RISC-V --with-multilib-list support.
6940 2018-05-09 Richard Biener <rguenther@suse.de>
6942 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6944 (vect_bb_vectorization_profitable_p): Adjust. Compute
6945 actual scalar cost using the cost vector and the add_stmt_cost
6948 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6950 PR rtl-optimization/85645
6951 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6952 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6954 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6956 PR rtl-optimization/85645
6957 * shrink-wrap.c (spread_components): Return a boolean saying if
6958 anything was changed.
6959 (try_shrink_wrapping_separate): Iterate spread_components until
6960 nothing changes anymore.
6962 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6964 PR rtl-optimization/85645
6965 * regrename.c (build_def_use): Also kill the chains that include the
6966 destination of a REG_CFA_REGISTER note.
6968 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6970 PR rtl-optimization/85645
6971 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6972 insn that has a REG_CFA_REGISTER note.
6974 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6976 * cfgexpand.c (expand_clobber): New function.
6977 (expand_gimple_stmt_1): Use it.
6978 * tree-vect-stmts.c (vect_clobber_variable): New function,
6980 (vectorizable_simd_clone_call): ...here.
6981 (vectorizable_store): Emit a clobber either side of an
6982 IFN_STORE_LANES sequence.
6983 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6985 2018-05-09 Tom de Vries <tom@codesourcery.com>
6988 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6989 (define_insn "trap_if_false"): Add exit after trap.
6991 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6993 PR rtl-optimization/85638
6994 * bb-reorder.c: Include common/common-target.h.
6995 (create_forwarder_block): New function extracted from...
6996 (fix_up_crossing_landing_pad): ...here. Rename into...
6997 (dw2_fix_up_crossing_landing_pad): ...this.
6998 (sjlj_fix_up_crossing_landing_pad): New function.
6999 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7000 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7001 from both partitions and exit the loop after one iteration.
7003 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7006 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7008 (Basic PowerPC Built-in Functions): The new name of the
7009 subsection previously known as "PowerPC Built-in Functions".
7010 (Basic PowerPC Built-in Functions Available on all Configurations):
7012 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7014 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7016 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7018 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7021 2018-05-08 Jim Wilson <jimw@sifive.com>
7023 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7024 (LD_EMUL_SUFFIX): New.
7025 (LINK_SPEC): Use it.
7027 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7029 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7031 (Basic PowerPC Built-in Functions): The new name of the
7032 subsection previously known as "PowerPC Built-in Functions".
7033 (Basic PowerPC Built-in Functions Available on all Configurations):
7035 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7037 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7039 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7041 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7044 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7047 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7048 after cmpelim optimization.
7050 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7052 * config.gcc: Support "goldmont".
7053 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7054 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7056 * config/i386/i386.c (m_GOLDMONT): Define.
7057 (processor_target_table): Add "goldmont".
7058 (PTA_GOLDMONT): Define.
7059 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7060 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7061 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7062 (fold_builtin_cpu): Add "goldmont".
7063 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7064 (ix86_option_override_internal): Add "goldmont".
7065 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7066 (processor_type): Add PROCESSOR_GOLDMONT.
7067 * config/i386/i386.md: Add CPU "glm".
7068 * config/i386/glm.md: New file.
7069 * config/i386/x86-tune.def: Add m_GOLDMONT.
7070 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7072 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7075 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7077 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7078 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7079 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7082 * config/i386/i386.c (ix86_fold_builtin): Handle
7083 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7086 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7087 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7089 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7092 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7093 (check_arch): Likewise.
7094 (check_fpu): Return the result rather than printing it.
7095 (end arch): Fix operator precedence.
7096 (end cpu): Likewise.
7097 (END): Print the result from check_fpu.
7099 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7100 Alan Hayward <alan.hayward@arm.com>
7101 David Sherwood <david.sherwood@arm.com>
7103 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7104 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7105 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7106 (*fcmuo<mode>_and): New patterns.
7108 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7110 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7111 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7112 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7113 (cmp_op, sve_imm_con): New code attributes.
7114 (SVE_COND_INT_CMP, imm_con): Delete.
7115 (cmp_op): Remove above unspecs from int attribute.
7116 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7118 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7119 comparison-specific unspecs.
7120 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7121 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7122 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7123 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7124 (*vec_fcm<cmp_op><mode>): Rename to...
7125 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7126 (*vec_fcmuo<mode>): Rename to...
7127 (*fcmuo<mode>): ...this and adjust likewise.
7128 (*pred_fcm<cmp_op><mode>): New pattern.
7129 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7130 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7132 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7134 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7135 (aarch64_emit_sve_predicated_cond): New function.
7136 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7137 (aarch64_emit_unspec_cond_or): Replace with...
7138 (aarch64_emit_sve_or_conds): ...this new function. Use
7139 aarch64_emit_sve_ptrue_op for the individual comparisons and
7140 aarch64_emit_binop to OR them together.
7141 (aarch64_emit_inverted_unspec_cond): Replace with...
7142 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7143 aarch64_emit_sve_ptrue_op for the comparison and
7144 aarch64_emit_unop to invert the result.
7145 (aarch64_expand_sve_vec_cmp_float): Update after the above
7146 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7148 2018-05-07 Nathan Sidwell <nathan@acm.org>
7150 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7151 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7152 (Backwards Compatibility): Likewise.
7154 2018-05-07 Luis Machado <luis.machado@linaro.org>
7158 2018-05-07 Luis Machado <luis.machado@linaro.org>
7160 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7161 <prefetch_dynamic_strides>: New const bool field.
7162 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7163 prefetch_dynamic_strides.
7164 (exynosm1_prefetch_tune): Likewise.
7165 (thunderxt88_prefetch_tune): Likewise.
7166 (thunderx_prefetch_tune): Likewise.
7167 (thunderx2t99_prefetch_tune): Likewise.
7168 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7170 (aarch64_override_options_internal): Update to set
7171 PARAM_PREFETCH_DYNAMIC_STRIDES.
7172 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7173 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7174 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7175 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7176 prefetch-dynamic-strides setting.
7178 2018-05-07 Luis Machado <luis.machado@linaro.org>
7180 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7181 <minimum_stride>: New const int field.
7182 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7183 minimum_stride field.
7184 (exynosm1_prefetch_tune): Likewise.
7185 (thunderxt88_prefetch_tune): Likewise.
7186 (thunderx_prefetch_tune): Likewise.
7187 (thunderx2t99_prefetch_tune): Likewise.
7188 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7189 (aarch64_override_options_internal): Update to set
7190 PARAM_PREFETCH_MINIMUM_STRIDE.
7191 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7192 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7193 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7194 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7195 stride is constant and is below the minimum stride threshold.
7197 2018-05-07 Luis Machado <luis.machado@linaro.org>
7199 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7202 2018-05-07 Luis Machado <luis.machado@linaro.org>
7204 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7205 <prefetch_dynamic_strides>: New const bool field.
7206 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7207 prefetch_dynamic_strides.
7208 (exynosm1_prefetch_tune): Likewise.
7209 (thunderxt88_prefetch_tune): Likewise.
7210 (thunderx_prefetch_tune): Likewise.
7211 (thunderx2t99_prefetch_tune): Likewise.
7212 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7214 (aarch64_override_options_internal): Update to set
7215 PARAM_PREFETCH_DYNAMIC_STRIDES.
7216 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7217 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7218 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7219 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7220 prefetch-dynamic-strides setting.
7222 2018-05-07 Luis Machado <luis.machado@linaro.org>
7224 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7225 <minimum_stride>: New const int field.
7226 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7227 minimum_stride field.
7228 (exynosm1_prefetch_tune): Likewise.
7229 (thunderxt88_prefetch_tune): Likewise.
7230 (thunderx_prefetch_tune): Likewise.
7231 (thunderx2t99_prefetch_tune): Likewise.
7232 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7233 (aarch64_override_options_internal): Update to set
7234 PARAM_PREFETCH_MINIMUM_STRIDE.
7235 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7236 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7237 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7238 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7239 stride is constant and is below the minimum stride threshold.
7241 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7244 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7245 the type is addressable. Don't force op into register if it has
7248 2018-05-05 Roland McGrath <mcgrathr@google.com>
7251 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7252 any section for which we don't know a specific type it should have,
7253 regardless of name. Previously this was done only for the exact
7254 names ".init_array", ".fini_array", and ".preinit_array".
7255 (default_elf_asm_named_section): Add comment about
7256 relationship with default_section_type_flags and SECTION_NOTYPE.
7257 (get_section): Don't consider it a type conflict if one side has
7258 SECTION_NOTYPE and the other doesn't, as long as neither has the
7259 SECTION_BSS et al used in the default_section_type_flags logic.
7261 2018-05-05 Tom de Vries <tom@codesourcery.com>
7264 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7265 (workaround_barsyncs): New function.
7266 (nvptx_reorg): Use workaround_barsyncs.
7267 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7268 (define_expand "nvptx_membar_cta"): New define_expand.
7269 (define_insn "*nvptx_membar_cta"): New insn.
7271 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7273 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7274 To improve optimization opportunities.
7275 * builtin-types.def: The new needed builtin types for the above.
7277 2018-05-04 Richard Biener <rguenther@suse.de>
7279 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7280 * gimple-ssa-store-merging.c
7281 (imm_store_chain_info::output_merged_store): Remove redundant create,
7282 release split_store vector contents on failure.
7283 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7284 scalar stmt vector on cache hit.
7286 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7288 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7290 * config.gcc (powerpc-xilinx-eabi*): Remove.
7291 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7293 (fusion_addis_mem_combo_load): Ditto.
7294 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7296 (rs6000_cpu_cpp_builtins): Ditto.
7297 * config/rs6000/rs6000-linux.c
7298 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7299 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7300 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7302 (rs6000_setup_reg_addr_masks): Ditto.
7303 (rs6000_init_hard_regno_mode_ok): Ditto.
7304 (rs6000_option_override_internal): Ditto.
7305 (legitimate_lo_sum_address_p): Ditto.
7306 (rs6000_legitimize_address): Ditto.
7307 (rs6000_legitimize_reload_address): Ditto.
7308 (rs6000_legitimate_address_p): Ditto.
7309 (abi_v4_pass_in_fpr): Ditto.
7310 (setup_incoming_varargs): Ditto.
7311 (rs6000_gimplify_va_arg): Ditto.
7312 (rs6000_split_multireg_move): Ditto.
7313 (rs6000_savres_strategy): Ditto.
7314 (rs6000_emit_prologue_components): Ditto.
7315 (rs6000_emit_epilogue_components): Ditto.
7316 (rs6000_emit_prologue): Ditto.
7317 (rs6000_emit_epilogue): Ditto.
7318 (rs6000_elf_file_end): Ditto.
7319 (rs6000_function_value): Ditto.
7320 (rs6000_libcall_value): Ditto.
7321 * config/rs6000/rs6000.h: Ditto.
7322 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7323 (TARGET_MINMAX): ... this. New.
7324 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7325 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7326 (*movsi_internal1_single): Delete.
7327 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7328 mfpu=, mxilinx-fpu): Delete.
7329 * config/rs6000/singlefp.h: Delete.
7330 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7331 * config/rs6000/t-rs6000: Ditto.
7332 * config/rs6000/t-xilinx: Delete.
7333 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7334 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7335 (VStype_simple): Delete.
7336 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7337 * config/rs6000/xfpu.h: Delete.
7338 * config/rs6000/xfpu.md: Delete.
7339 * config/rs6000/xilinx.h: Delete.
7340 * config/rs6000/xilinx.opt: Delete.
7341 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7342 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7344 2018-05-04 Tom de Vries <tom@codesourcery.com>
7347 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7350 2018-05-04 Richard Biener <rguenther@suse.de>
7353 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7354 (expand_complex_div_wide): Likewise.
7355 (expand_complex_operations_1): Likewise.
7356 (expand_complex_libcall): Preserve EH info of the original stmt.
7357 (tree_lower_complex): Handle removed blocks.
7358 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7359 on complex multiplication and division libcall builtins.
7361 2018-05-04 Richard Biener <rguenther@suse.de>
7364 * fold-const.c (negate_expr_p): Restrict negation of operand
7365 zero of a division to when we know that can happen without
7367 (fold_negate_expr_1): Likewise.
7369 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7372 * real.h (real_nextafter): Declare.
7373 * real.c (real_nextafter): New function.
7374 * fold-const-call.c (fold_const_nextafter): New function.
7375 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7376 CASE_CFN_NEXTTOWARD.
7377 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7378 even when arg1_mode is different from arg0_mode.
7380 2018-05-03 Nathan Sidwell <nathan@acm.org>
7382 * doc/extend.texi (Deprecated Features): Remove
7384 (Backwards Compatibility): Likewise.
7385 * doc/invoke.texi (C++ Language Options): Likewise.
7386 (C++ Dialect Options): Likewise.
7388 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7391 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7392 _mm512_mask_mullox_epi64): New intrinsics.
7394 2018-05-03 Tom de Vries <tom@codesourcery.com>
7397 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7398 dump files): Add offload-tree.
7400 2018-05-03 Richard Biener <rguenther@suse.de>
7402 PR tree-optimization/85615
7403 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7404 to loops not nested in BBs loop father to avoid creating multi-entry
7407 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7409 PR tree-optimization/70291
7410 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7411 arguments. Change return type to tree. Emit libcall as a new
7412 statement rather than replacing existing one when inplace_p is true.
7413 (expand_complex_multiplication_components): New function.
7414 (expand_complex_multiplication): Expand floating-point complex
7415 multiplication using the above.
7416 (expand_complex_division): Rename inner_type parameter to type.
7417 Update expand_complex_libcall call-site.
7418 (expand_complex_operations_1): Update expand_complex_multiplication
7419 and expand_complex_division call-sites.
7421 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7424 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7425 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7426 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7427 the highest significant bit of the shift count mask is clear. In
7428 check whether and[sq]i3 is needed verify that all significant bits
7429 of the shift count other than the highest are set.
7431 2018-05-02 Tom de Vries <tom@codesourcery.com>
7434 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7435 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7436 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7437 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7438 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7439 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7440 __builtin_goacc_parlevel_size.
7442 2018-05-02 Richard Biener <rguenther@suse.de>
7444 PR tree-optimization/85597
7445 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7446 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7449 2018-05-02 Tom de Vries <tom@codesourcery.com>
7452 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7453 dump files): Add ltrans-tree.
7455 2018-05-02 Tom de Vries <tom@codesourcery.com>
7458 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7459 dump files): Add wpa-ipa.
7461 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7463 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7464 powerpc*-*-linux*paired* target.
7465 * config/rs6000/750cl.h: Delete.
7466 * config/rs6000/paired.h: Delete.
7467 * config/rs6000/paired.md: Delete.
7468 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7470 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7471 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7472 comment. Remove paired float support.
7473 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7474 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7476 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7477 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7479 * config/rs6000/rs6000.c: Remove paired float support.
7480 (paired_expand_vector_init, paired_expand_vector_move,
7481 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7482 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7483 paired_expand_builtin, paired_expand_predicate_builtin,
7484 paired_init_builtins): Delete.
7485 * config/rs6000/rs6000.h: Remove paired float support.
7486 * config/rs6000/rs6000.md: Remove paired float support.
7487 (move_from_CR_ov_bit): Delete.
7488 * config/rs6000/rs6000.opt (mpaired): Delete.
7489 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7490 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7492 2018-05-02 Richard Biener <rguenther@suse.de>
7495 * gimplify.c (gimplify_save_expr): When in SSA form allow
7496 SAVE_EXPRs to compute to SSA vars.
7498 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7501 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7502 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7503 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7504 clobber operands[2], instead use a new pseudo. Formatting fixes.
7506 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7508 PR tree-optimization/85586
7509 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7510 exit early for statements in the same group if the accesses are
7513 2018-05-02 Tom de Vries <tom@codesourcery.com>
7516 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7519 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7521 PR tree-optimization/85143
7522 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7524 2018-05-01 Tom de Vries <tom@codesourcery.com>
7527 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7528 not found" error message.
7530 2018-05-01 Tom de Vries <tom@codesourcery.com>
7533 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7534 * vec.c (test_ordered_remove_if): New function.
7535 (vec_c_tests): Call test_ordered_remove_if.
7536 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7537 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7538 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7539 VEC_ORDERED_REMOVE_IF.
7541 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7543 PR tree-optimization/82665
7544 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7545 pointer subtraction where arguments come from a memchr call.
7547 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7549 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7550 --push-state --as-needed and --pop-state instead of --as-needed and
7551 --no-as-needed if ld supports it.
7552 * configure: Regenerated.
7555 * doc/install.texi2html: Replace _002d with - and _002a with * in
7556 generated html files using sed.
7558 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7561 * gcc-rich-location.c (blank_line_before_p): New function.
7562 (use_new_line): New function.
7563 (gcc_rich_location::add_fixit_insert_formatted): New function.
7564 * gcc-rich-location.h
7565 (gcc_rich_location::add_fixit_insert_formatted): New function.
7567 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7569 * selftest.c (assert_streq): Rename "expected" and "actual" to
7570 "val1" and "val2". Extend NULL-handling to cover both inputs
7571 symmetrically, while still requiring both to be non-NULL for a pass.
7572 * selftest.h (assert_streq): Rename "expected" and "actual" to
7574 (ASSERT_EQ): Likewise.
7575 (ASSERT_EQ_AT): Likewise.
7576 (ASSERT_KNOWN_EQ): Likewise.
7577 (ASSERT_KNOWN_EQ_AT): Likewise.
7578 (ASSERT_NE): Likewise.
7579 (ASSERT_MAYBE_NE): Likewise.
7580 (ASSERT_MAYBE_NE_AT): Likewise.
7581 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7582 the assertion to pass.
7583 (ASSERT_STREQ_AT): Likewise.
7585 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7587 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7588 interaction with -pie.
7590 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7592 * selftest.h: Fix alphabetization of per-source-file selftest
7595 2018-04-30 Jason Merrill <jason@redhat.com>
7597 PR c++/61982 - dead stores to destroyed objects.
7598 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7601 2018-04-30 Jason Merrill <jason@redhat.com>
7603 * tree.c (build_clobber): New.
7604 * tree.h: Declare it.
7605 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7607 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7609 * diagnostic-show-locus.c (layout::layout): Update for
7610 location_get_source_line returning a char_span.
7611 (struct char_span): Move to input.h.
7612 (struct correction): Update for fields in char_span becoming
7614 (struct source_line): Update for location_get_source_line
7615 returning a char_span.
7616 (layout::print_line): Likewise.
7617 * edit-context.c (edited_file::print_content): Likewise.
7618 (edited_file::print_diff_hunk): Likewise.
7619 (edited_file::print_run_of_changed_lines): Likewise.
7620 (edited_file::get_num_lines): Likewise.
7621 (edited_line::edited_line): Likewise.
7622 * final.c (asm_show_source): Likewise.
7623 * input.c (location_get_source_line): Convert return type
7624 from const char * to char_span, losing the final "line_len"
7626 (dump_location_info): Update for the above.
7627 (get_substring_ranges_for_loc): Likewise. Use a char_span
7628 when handling the literal within the line.
7629 (test_reading_source_line): Update for location_get_source_line
7630 returning a char_span.
7631 * input.h (class char_span): Move here from
7632 diagnostic-show-locus.c, converting from a struct to a class.
7633 Make data members private.
7634 (char_span::operator bool): New.
7635 (char_span::length): New.
7636 (char_span::get_buffer): New.
7637 (char_span::operator[]): New.
7638 (char_span::subspan): Make const.
7639 (char_span::xstrdup): New.
7640 (location_get_source_line): Convert return type from const char *
7641 to char_span, losing the final "line_size" param.
7643 2018-04-30 Jan Hubicka <jh@suse.cz>
7645 * lto-wrapper.c (ltrans_priorities): New static var.
7646 (cmp_priority): New.
7647 (run_gcc): Read priorities and if doing parallel build order
7648 the Makefile by them.
7650 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7652 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7654 2018-04-30 Richard Biener <rguenther@suse.de>
7656 * tree-cfg.c (verify_address): Remove base argument, add
7657 flag whether to check TREE_ADDRESSABLE and do that.
7658 (verify_expr): Remove.
7659 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7660 (verify_gimple_assign_single): Likewise.
7661 (verify_gimple_switch): Likewise.
7662 (verify_expr_location_1): Dereference tp once. Add (disabled)
7663 piece from verify_expr.
7664 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7666 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7668 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7670 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7672 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7673 (small_data_pattern): Likewise.
7674 (arc_rewrite_small_data): Likewise.
7675 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7676 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7677 (get_symbol_alignment): New function.
7678 (legitimate_small_data_address_p): Likewise.
7679 (legitimate_scaled_address): Update, call
7680 legitimate_small_data_address_p.
7681 (output_sdata): New static variable.
7682 (arc_print_operand): Update how we handle small data operands.
7683 (arc_print_operand_address): Likewise.
7684 (arc_legitimate_address_p): Update, use
7685 legitimate_small_data_address_p.
7686 (arc_rewrite_small_data_p): Remove.
7687 (arc_rewrite_small_data_1): Likewise.
7688 (arc_rewrite_small_data): Likewise.
7689 (small_data_pattern): Likewise.
7690 (compact_sda_memory_operand): Update to use
7691 legitimate_small_data_address_p and get_symbol_alignment.
7692 (prepare_move_operands): Don't rewite sdata pattern.
7693 (prepare_extend_operands): Remove.
7694 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7696 (zero_extendqisi2): Likewise.
7697 (zero_extendhisi2): Likewise.
7698 (extendqihi2): Likewise.
7699 (extendqisi2): Likewise.
7700 (extendhisi2): Likewise.
7704 * config/arc/constraints.md (Usd): Change it to memory constraint.
7706 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7708 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7709 as source of std instructions.
7710 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7711 allow 6-bit constants as source for store instructions.
7712 (movdi_insn): Update instruction pattern to allow 6-bit constants
7713 as source for store instructions.
7715 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7717 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7719 2018-04-30 Nathan Sidwell <nathan@acm.org>
7720 Sandra Loosemore <sandra@codesourcery.com>
7722 * dumpfile.c (dump_open): Allow '-' for stdout.
7723 * doc/invoke.texi (Developer Options): Document dump filename
7724 determination early. Document stdin/stdout selection.
7726 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7728 Microblaze Target: PIC data text relative
7730 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7731 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7733 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7734 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7735 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7736 New addressing mode for data-text relative position indepenedent code.
7737 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7738 'ADDRESS_SYMBOLIC_TXT_REL'.
7739 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7740 (microblaze_legitimate_pic_operand): Exclude function calls from
7741 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7742 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7744 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7745 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7746 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7747 for 'address + offset'.
7748 (microblaze_expand_prologue): Add new function prologue call for
7750 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7751 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7752 table in case of TARGET_PIC_DATA_TEXT_REL.
7753 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7754 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7755 Add new macros 'UNSPEC_TEXT',
7756 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7757 + exclude function calls from 'UNSPEC_PLT' in case of data text
7759 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7760 new target hook for generating address diff vector tables in case of
7762 * doc/tm.texi : Regenerate.
7763 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7764 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7765 of addr diff vector generation.
7766 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7767 target hook definition.
7768 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7769 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7770 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7771 option for data text relative.
7773 2018-04-30 Richard Biener <rguenther@suse.de>
7775 * tree-chrec.h (evolution_function_is_constant_p): Remove
7777 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7779 2018-04-30 Richard Biener <rguenther@suse.de>
7782 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7784 2018-04-30 Richard Biener <rguenther@suse.de>
7786 PR tree-optimization/28364
7787 PR tree-optimization/85275
7788 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7789 copying first exit test.
7791 2018-04-28 Mark Wielaard <mark@klomp.org>
7793 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7795 (dwarf_AT): Handle DW_AT_addr_base.
7796 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7798 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7801 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7802 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7803 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7804 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7806 2018-04-28 Richard Biener <rguenther@suse.de>
7808 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7809 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7810 to reflect use. Only add interesting stmts.
7812 2018-04-27 Martin Jambor <mjambor@suse.cz>
7815 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7816 the jump function allows for passing through aggregate values.
7818 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7820 * input.h (in_system_header_at): Convert from macro to inline
7822 (from_macro_expansion_at): Likewise.
7823 (from_macro_definition_at): Likewise.
7825 2018-04-27 Jeff Law <law@redhat.com>
7827 * config.gcc: Mark tile* targets as deprecated/obsolete.
7829 2018-04-27 Richard Biener <rguenther@suse.de>
7831 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7834 2018-04-27 Richard Biener <rguenther@suse.de>
7836 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7838 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7840 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7841 with Yd constraint. Set "preferred_for_speed" attribute from
7842 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7844 (*movdi_internal): Ditto.
7845 (movti_interunit splitters): Remove
7846 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7847 (movdi_interunit splitters): Ditto.
7848 * config/i386/constraints.md (Ye): Remove.
7849 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7851 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7854 * config/aarch64/constraints.md (Usg): Limit to 31.
7857 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7859 PR tree-optimization/85529
7860 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7861 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7862 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7863 zero extension or masking of the MSB bit.
7864 (optimize_range_tests): Add FIRST_BB argument, pass it through
7865 to optimize_range_tests_var_bound.
7866 (maybe_optimize_range_tests, reassociate_bb): Adjust
7867 optimize_range_tests callers.
7869 2018-04-26 Richard Biener <rguenther@suse.de>
7870 Jakub Jelinek <jakub@redhat.com>
7872 * cgraph.h (symbol_table): Just declare debug method here.
7873 * symtab.c (symbol_table::debug): Define.
7875 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7877 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7879 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7881 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7882 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7883 (*movdi_internal): Substitute Yi and Yj constraint with x
7884 and Ym and Yn constraint with y constraint. Update "isa"
7885 attribute and set "preferred_for_speed" attribute from
7886 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7887 (*movsi_internal): Ditto.
7888 (*movdf_internal): Ditto.
7889 (*movsf_internal): Ditto.
7890 (*zero_extendsidi2): Ditto.
7891 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7892 (sse2_loadld): Ditto.
7893 (*vec_extract<ssevecmodelower>_0): Ditto.
7894 (*vec_extractv4si_0_zext_sse4): Ditto.
7895 (vec_concatv2di): Ditto.
7896 (*vec_dup<mode>): Ditto.
7897 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7898 * config/i386/constraints.md (Yi): Remove.
7903 2018-04-26 Nathan Sidwell <nathan@acm.org>
7905 * dumpfile.c (dump_open): New.
7906 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7907 (dump_finish): Detect stdio/stderr by value not name.
7909 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7911 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7913 2018-04-26 Tom de Vries <tom@codesourcery.com>
7916 * config/nvptx/nvptx.c (verify_neutering_jumps)
7917 (verify_neutering_labels): New function
7918 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7920 2018-04-26 Tom de Vries <tom@codesourcery.com>
7923 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7924 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7927 2018-04-26 Richard Biener <rguenther@suse.de>
7928 Tom de Vries <tom@codesourcery.com>
7931 * lto-streamer-out.c (output_function): Fixup loops if required to match
7932 discovery done in the reader.
7934 2018-04-26 Richard Biener <rguenther@suse.de>
7936 PR tree-optimization/85116
7937 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7938 have a loop exit from the single latch predecessor. Remove
7939 case of header with just condition.
7940 (ch_base::copy_headers): Exclude infinite loops from any
7942 (pass_ch::execute): Record exits.
7944 2018-04-26 Richard Biener <rguenther@suse.de>
7946 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7947 prologue cost vector and pass it to vect_get_load_cost.
7948 (vect_get_peeling_costs_all_drs): Likewise.
7949 (vect_peeling_hash_get_lowest_cost): Likewise.
7950 (vect_enhance_data_refs_alignment): Likewise.
7952 2018-04-26 Richard Biener <rguenther@suse.de>
7955 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7956 checking of integer<->pointer conversions.
7957 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7958 sign-/zero-extending pointer types.
7959 (expand_omp_for_static_chunk): Likewise.
7961 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7962 Jean Lee <xiaoyur347@gmail.com>
7964 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7965 (TARGET_ASAN_SHADOW_OFFSET): Define.
7966 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7967 true for -fsanitize=address.
7969 2018-04-25 Mark Wielaard <mark@klomp.org>
7971 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7974 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7976 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7977 than "alu", remove explicit "memory" and "imm_disp" attributes.
7978 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7981 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7982 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7983 gen_lowpart_no_emit.
7985 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7988 * config/i386/i386.c (ix86_expand_builtin): Change memory
7989 operand to XI, extend p0 to Pmode.
7990 * config/i386/i386.md: Change unspec volatile and operand
7991 1 mode to XI, change operand 0 mode to P.
7993 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7995 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7996 GET_MODE_MASK before any checking.
7997 (nds32_can_use_bset_p): Likewise.
7998 (nds32_can_use_btgl_p): Likewise.
8000 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8002 * config/nds32/nds32-doubleword.md: New define_split pattern for
8003 illegal register number.
8005 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8007 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8009 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8011 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8013 2018-04-25 Richard Biener <rguenther@suse.de>
8015 * lto-streamer.h (LTO_major_version): Bump to 8.
8017 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8019 * BASE-VER: Set to 9.0.0.
8021 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8023 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8024 in __abskf2 and __powikf2.
8026 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8029 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8030 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8031 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8032 Use the above on operand 2. Reindent.
8033 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8035 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8038 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8040 * config/i386/i386.opt (mcet): Removed.
8041 * doc/install.texi: Remove -mcet documentation.
8042 * doc/invoke.texi: Likewise.
8044 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8047 * doc/install.texi: Remove -mcet from bootstrap-cet.
8049 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8052 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8053 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8057 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8058 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8059 containing a CONST_VECTOR.
8061 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8063 * doc/install.texi: Update newlib dependency for nvptx.
8065 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8068 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8069 instead of INTVAL when shifting x left.
8071 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8073 PR tree-optimization/85478
8074 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8075 vect_grouped_store_supported for single element vectors.
8077 2018-04-24 Richard Biener <rguenther@suse.de>
8080 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8081 load cost increase to the case of non-constant step.
8083 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8086 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8087 destination if any_malformed_asm.
8089 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8092 * expr.c (store_field): In the bitfield case, if the value comes from
8093 a function call and is returned in registers by means of a PARALLEL,
8094 do not change the mode of the temporary unless BLKmode and VOIDmode.
8096 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8098 PR rtl-optimization/85423
8099 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8100 dependencies to debug insns when the previous insn is non-debug.
8102 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8104 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8105 enums into a single definition.
8106 (fls): Fix predicates and printing.
8109 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8111 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8112 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8113 and short u6 immediate.
8114 (check_if_valid_sleep_operand): Remove.
8115 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8117 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8119 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8120 flag_always_save_lp condition.
8121 * config/nds32/nds32.opt (malways-save-lp): New option.
8123 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8125 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8126 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8127 * config/nds32/nds32.h
8128 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8129 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8131 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8133 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8134 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8136 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8137 Chung-Ju Wu <jasonwucj@gmail.com>
8139 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8141 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8142 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8144 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8146 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8148 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8150 * config/nds32/nds32-protos.h (nds32_data_alignment,
8151 nds32_local_alignment): Declare.
8152 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8153 nds32_local_alignment): New functions.
8154 (TARGET_CONSTANT_ALIGNMENT): Define.
8155 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8157 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8159 * config/nds32/nds32.c
8160 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8161 (TARGET_MODES_TIEABLE_P): Likewise.
8163 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8165 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8168 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8169 Chung-Ju Wu <jasonwucj@gmail.com>
8171 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8172 for unaligned access.
8173 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8175 * config/nds32/nds32-intrinsic.md: Likewise.
8176 * config/nds32/nds32_intrinsic.h: Likewise.
8177 * config/nds32/nds32.h (nds32_builtins): Likewise.
8178 * config/nds32/nds32.opt (munaligned-access): New option.
8179 * config/nds32/nds32.c (nds32_asm_file_start): Display
8180 flag_unaligned_access status.
8182 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8184 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8185 -mno-relax is present.
8186 * config/riscv/linux.h (LINK_SPEC): Ditto.
8188 2018-04-20 Martin Sebor <msebor@redhat.com>
8191 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8193 (gimple_fold_builtin_stxcpy_chk): Same.
8194 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8196 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8199 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8200 __powikf2 when long double is IEEE 128-bit.
8202 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8204 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8205 step to make sure stack always aligned.
8207 2018-04-20 Carl Love <cel@us.ibm.com>
8210 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8211 size check for arg0.
8213 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8214 Tom de Vries <tom@codesourcery.com>
8217 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8218 Emit insns for calls too.
8219 (nvptx_find_par): Always look for worker-level predecessor insn.
8220 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8222 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8223 (nvptx_process_pars): Propagate frames for calls.
8225 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8228 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8230 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8231 (ix86_handle_option): Don't handle OPT_mibt.
8232 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8234 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8236 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8237 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8238 (ix86_target_macros): Define __CET__ with flag_cf_protection
8239 for -fcf-protection.
8240 * config/i386/i386.c (isa2_opts): Remove -mibt.
8241 * config/i386/i386.h (TARGET_IBT): Removed.
8242 (TARGET_IBT_P): Likewise.
8243 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8244 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8245 * config/i386/i386.opt (mcet): Update help message.
8248 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8249 -mcet as an alias for -mshstk.
8251 2018-04-20 Richard Biener <rguenther@suse.de>
8254 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8255 complexity by forcing a single use of the multiply operand.
8257 2018-04-20 Martin Jambor <mjambor@suse.cz>
8260 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8261 recursion dependency to only apply to non-clones.
8263 2018-04-20 Martin Jambor <mjambor@suse.cz>
8266 * ipa-cp.c (create_specialized_node): Check that clones of
8267 self-recursive edges exist during IPA-CP.
8269 2018-04-19 Toon Moene <toon@moene.org>
8271 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8274 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8276 PR tree-optimization/85467
8277 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8278 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8279 VECTOR_CST element to type.
8281 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8284 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8285 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8286 (builtin_longjmp): Likewise.
8287 (save_stack_nonlocal): New pattern.
8288 (restore_stack_nonlocal): Likewise.
8290 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8293 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8294 Replace ASM_OUTPUT_LABEL with fprintf.
8296 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8299 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8300 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8301 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8302 define __IBT__ and __SHSTK__ for -fcf-protection.
8303 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8305 (ix86_trampoline_init): Likewise.
8306 (x86_output_mi_thunk): Likewise.
8307 (ix86_notrack_prefixed_insn_p): Likewise.
8308 (ix86_option_override_internal): Don't disallow -fcf-protection.
8309 * config/i386/i386.md (rdssp<mode>): Also enable for
8311 (incssp<mode>): Likewise.
8312 (nop_endbr): Likewise.
8313 * config/i386/i386.opt (mcet): Change help message to built-in
8317 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8318 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8319 enable CET built-in functions.
8321 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8323 * common/config/i386/i386-common.c
8324 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8325 OPTION_MASK_ISA_MOVDIRI_UNSET,
8326 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8327 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8328 * config.gcc (movdirintrin.h): New header.
8329 * config/i386/cpuid.h (bit_MOVDIRI,
8330 bit_MOVDIR64B): New bits.
8331 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8333 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8334 (VOID, PVOID, PCVOID)): New function types.
8335 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8336 __builtin_ia32_directstoreu_u64,
8337 __builtin_ia32_movdir64b): New builtins.
8338 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8339 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8341 (ix86_valid_target_attribute_inner_p): Ditto.
8342 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8343 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8344 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8345 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8346 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8347 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8348 (movdiri<mode>, movdir64b_<mode>): New.
8349 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8350 * config/i386/immintrin.h: Include movdirintrin.h.
8351 * config/i386/movdirintrin.h: New file.
8352 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8354 2018-04-19 Richard Biener <rguenther@suse.de>
8357 * cfg.c (clear_bb_flags): When loop state says we have
8358 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8360 2018-04-19 Richard Biener <rguenther@suse.de>
8362 PR tree-optimization/84737
8363 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8364 copying restrict info.
8365 (vect_setup_realignment): Use it.
8366 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8367 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8368 the first DR to all generated stores.
8369 (vectorizable_load): Likewise for loads.
8371 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8373 PR tree-optimization/85446
8374 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8375 the integral and pointer types to have the same precision.
8377 * doc/install.texi: Document --disable-cet being the default and
8380 2018-04-18 Martin Liska <mliska@suse.cz>
8382 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8385 2018-04-18 Martin Liska <mliska@suse.cz>
8388 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8391 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8392 arguments if they are comparable.
8394 2018-04-18 Martin Liska <mliska@suse.cz>
8397 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8400 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8403 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8406 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8407 ENDBR after calling __morestack.
8409 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8412 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8413 by using gcc_base_ver to generate a gcc_driver_version, and use
8414 it when generating GCC_DRIVER_NAME.
8415 * configure: Regenerate.
8417 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8420 * config.gcc: Obsolete powerpc*-*-*spe*.
8422 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8425 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8426 (stabstr_D): Change type of unum from unsigned int to
8427 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8430 2018-04-17 Jim Wilson <jimw@sifive.com>
8433 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8434 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8435 Set arg_pointer_offset after using pretend_args_size.
8437 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8439 PR rtl-optimization/85431
8440 * dse.c (record_store): Ignore zero width stores.
8443 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8444 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8445 __builtin_stack_restore rather than after it.
8446 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8447 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8448 argument instead of virtual_dynamic_stack_rtx.
8450 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8452 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8454 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8455 Add note to error message to explain internal mapping of overloaded
8456 built-in function name to non-overloaded built-in function name.
8457 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8460 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8463 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8464 where the inputs overlap with the output.
8466 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8468 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8469 (=v, v) alternative and explicit "memory" attribute.
8470 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8471 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8473 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8474 "sselog1" type instead of "sselog".
8475 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8476 "sselog". Remove explicit "memory" attribute.
8477 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8478 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8480 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8481 "sselog1" type instead of "sselog". Remove explicit "memory"
8483 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8484 use "sselog1" type instead of "sselog". Remove explicit "memory"
8486 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8487 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8489 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8490 "sselog1" type instead of "sselog". Remove explicit "memory"
8492 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8493 use "sselog1" type instead of "sselog". Remove explicit "memory"
8497 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8500 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8503 2018-04-17 Martin Jambor <mjambor@suse.cz>
8506 * ipa-cp.c (create_specialized_node): Call
8507 expand_all_artificial_thunks if necessary.
8509 2018-04-17 Martin Liska <mliska@suse.cz>
8512 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8513 in message, remote space in between '_G' and '('.
8515 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8518 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8519 avx512f_vmcmp<mode>3<round_saeonly_name>,
8520 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8521 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8522 avx512f_rndscale<mode><round_saeonly_name>,
8523 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8524 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8525 Use %<iptr>2 instead of %2 for -masm=intel.
8526 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8527 avx512f_vcvttss2usi<round_saeonly_name>,
8528 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8530 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8531 avx512f_vcvttsd2usi<round_saeonly_name>,
8532 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8533 Use %q1 instead of %1 for -masm=intel.
8534 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8535 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8536 of %3 for -masm=intel.
8537 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8539 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8541 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8543 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8544 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8546 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8548 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8549 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8550 %g1 and one with %0 and %1.
8551 (avx512er_vmrcp28<mode><round_saeonly_name>,
8552 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8554 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8555 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8556 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8557 of %0 and %{%4%} for -masm=intel.
8558 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8559 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8560 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8561 order of %0 and %{%5%}%{z%} for -masm=intel.
8563 2018-04-17 Jan Hubicka <jh@suse.cz>
8566 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8568 2018-04-17 Martin Liska <mliska@suse.cz>
8571 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8572 for target_clone error message. Make default implementation
8573 clone to be a local declaration.
8574 (separate_attrs): Add new argument and check for an empty
8576 (expand_target_clones): Handle it.
8577 (ipa_target_clone): Make redirection just for target_clones
8580 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8581 Tom de Vries <tom@codesourcery.com>
8584 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8585 tiled basic blocks without omp continue statements.
8587 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8590 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8591 vec_extract expression as having side effects to make sure it gets
8594 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8597 * config/i386/i386.c (get_builtin_code_for_version): Check
8600 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8603 * gcc/config.gcc: Support "skylake".
8604 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8606 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8607 (processor_target_table): Add "skylake".
8608 (ix86_option_override_internal): Add "skylake".
8609 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8610 PROCESSOR_CANNONLAKE.
8611 (get_builtin_code_for_version): Fix priority for
8612 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8613 PROCESSOR_SKYLAKE-AVX512.
8614 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8615 (processor_type): Add PROCESSOR_SKYLAKE.
8617 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8618 Jason Merrill <jason@redhat.com>
8621 * convert.c (convert_to_integer_1): Use direct recursion for
8622 enumeral types and types with a precision less than the number
8623 of bits in their mode.
8625 2018-04-16 Julia Koval <julia.koval@intel.com>
8628 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8629 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8631 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8634 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8635 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8636 and -mno-direct-move.
8638 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8641 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8642 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8643 Ensure negative shifts result in {0}.
8645 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8647 PR rtl-optimization/79916
8648 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8649 regs (if any) to define how to gnerate SD moves when LRA is in
8652 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8654 PR rtl-optimization/85393
8655 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8656 * except.c (expand_dw2_landing_pad_for_region): Make static.
8657 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8658 a label and unconditional jump to old_bb, rather than
8659 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8662 PR rtl-optimization/85376
8663 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8664 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8665 instead of a specific value.
8667 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8668 Bin Cheng <bin.cheng@arm.com>
8670 PR tree-optimization/82965
8671 PR tree-optimization/83991
8672 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8673 by_profile_only parameter.
8674 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8675 information if the loop was predicted to iterate too many times.
8676 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8678 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8681 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8684 2018-04-13 Martin Liska <mliska@suse.cz>
8685 Jakub Jelinek <jakub@redhat.com>
8688 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8689 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8690 * builtins.c (expand_builtin_memory_copy_args): Use
8691 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8692 handle dest_addr == pc_rtx.
8694 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8697 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8698 asked to not generate direct moves.
8699 (fix_trunc<mode>si2_stfiwx): Similar.
8700 (fix_trunc<mode>si2_internal): Similar.
8702 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8705 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8706 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8707 lookup if dest in some wider mode is known to be const0_rtx and
8708 if so, record permanent equivalence for it to be ZERO_EXTEND of
8709 the narrower mode destination.
8711 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8713 * lto-streamer-out.c (output_function): Revert 259346.
8714 * omp-expand.c (expand_oacc_for): Likewise.
8716 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8718 PR rtl-optimization/85354
8719 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8720 * sel-sched.c (sel_global_init): ... here.
8722 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8725 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8726 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8727 mode for PE-COFF targets.
8728 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8729 (i386_pe_asm_lto_end): Likewise.
8730 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8731 (TARGET_ASM_LTO_END): Likewise.
8732 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8733 (i386_pe_asm_lto_start): New function.
8734 (i386_pe_asm_lto_end): Likewise.
8736 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8737 Richard Biener <rguenther@suse.de>
8740 * lto-streamer-out.c (output_function): Fix CFG loop state before
8742 * omp-expand.c (expand_oacc_for): Handle calls to internal
8743 functions like regular functions.
8745 2018-04-12 Richard Biener <rguenther@suse.de>
8748 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8749 for the early LTO debug to properly generate references to it
8750 during DIE emission. Do not re-use that for the skeleton for
8752 (dwarf2out_early_finish): Likewise.
8754 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8757 * config/i386/sse.md
8758 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8759 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8760 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8761 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8762 and output is a reg, avoid creating invalid lowpart subreg, but
8763 instead split into a 512-bit move. Don't split if not AVX512VL,
8764 input is xmm16+ reg and output is a mem.
8765 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8766 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8767 xmm16+ reg and output is a mem.
8769 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8771 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8772 also for flag_dwarf2_cfi_asm.
8774 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8776 PR rtl-optimization/85342
8777 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8778 a bool scalar var inside of the loop instead. Don't try to update
8779 recog_data.operand after failed apply_change_group.
8781 2018-04-12 Tom de Vries <tom@codesourcery.com>
8784 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8785 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8786 array with flexible array member as array without given dimension.
8787 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8788 argument for undefined param to true.
8790 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8793 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8794 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8795 from PowerPC section.
8796 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8797 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8798 help text that is too long.
8799 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8800 help text that is too long.
8801 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8802 help text that is too long.
8804 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8806 * config/alpha/alpha.md (stack_probe_internal): Rename
8807 from "probe_stack". Update all callers.
8809 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8811 PR rtl-optimization/84566
8812 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8813 sched_macro_fuse_insns.
8815 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8818 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8819 (compute_block_dependences): ... from here.
8821 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8823 PR tree-optimization/85331
8824 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8825 from int to HOST_WIDE_INT.
8827 2018-04-11 Martin Jambor <mjambor@suse.cz>
8830 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8831 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8832 not the same as the source val.
8833 (cgraph_edge_brings_value_p): New parameter.
8834 (gather_edges_for_value): Pass destination value to
8835 cgraph_edge_brings_value_p.
8836 (perhaps_add_new_callers): Likewise.
8837 (get_info_about_necessary_edges): Likewise and exclude values brought
8838 only by self-recursive edges.
8839 (create_specialized_node): Redirect only clones of self-calling edges.
8840 (+self_recursive_pass_through_p): New function.
8841 (find_more_scalar_values_for_callers_subset): Use it.
8842 (find_aggregate_values_for_callers_subset): Likewise.
8843 (known_aggs_to_agg_replacement_list): Removed.
8844 (decide_whether_version_node): Re-calculate known constants for all
8845 remaining context clones.
8847 2018-04-11 Richard Biener <rguenther@suse.de>
8850 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8851 from early DWARF output.
8852 (dwarf2out_early_finish): Output line info unconditionally into
8853 early DWARF and add reference to it.
8855 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8858 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8859 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8860 other than V2DFmode using iptr mode attribute.
8861 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8863 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8865 PR rtl-optimization/84659
8866 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8868 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8871 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8873 (output_loc_list): Pass address of a dummy size variable even in the
8874 locview handling loop.
8875 (index_location_lists): Add comment on why skip_loc_list_entry can't
8878 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8881 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8884 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8887 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8888 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8889 and -mstring-compare-inline-limit.
8891 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8894 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8895 for stack clash protection in a register whenever we need it to be in
8898 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8900 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8901 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8903 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8906 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8908 (mlong-double-): Ditto.
8909 * config/rs6000/sysv4.opt (msdata=): Ditto.
8910 (mtls-size=): Ditto.
8912 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8914 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8915 erroneous entries for
8916 "vector int vec_ldl (int, long int *)", and
8917 "vector unsigned int vec_ldl (int, unsigned long int *)".
8918 Add comments and entries for
8919 "vector bool char vec_ldl (int, bool char *)",
8920 "vector bool short vec_ldl (int, bool short *)",
8921 "vector bool int vec_ldl (int, bool int *)",
8922 "vector bool long long vec_ldl (int, bool long long *)",
8923 "vector pixel vec_ldl (int, pixel *)",
8924 "vector long long vec_ldl (int, long long *)",
8925 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8926 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8927 type tree bool_long_long_type_node and correct definition of
8928 bool_V2DI_type_node to make reference to this new type tree.
8929 (rs6000_mangle_type): Replace erroneous reference to
8930 bool_long_type_node with bool_long_long_type_node.
8931 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8932 comments to emphasize sign distinctions for char and int types and
8933 replace RS6000_BTI_bool_long constant with
8934 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8935 use of RS6000_BTI_pixel.
8936 (bool_long_type_node): Remove this macro definition.
8937 (bool_long_long_type_node): New macro definition
8939 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8941 PR rtl-optimization/85300
8942 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8943 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8944 simplify_unary_operation fails.
8946 2018-04-10 Martin Liska <mliska@suse.cz>
8948 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8949 cgraph_edge and ipa_ref.
8951 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8955 * config/i386/sse.md
8956 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8957 computation of the VEC_MERGE selector from mask.
8958 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8959 Fix decoding of the VEC_MERGE selector into mask.
8961 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8963 PR tree-optimization/85286
8964 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8966 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8968 * final.c (final_1): Set insn_last_address as well as
8969 insn_current_address.
8971 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8974 * explow.c (emit_stack_probe): Call validize_mem on memory location
8975 before passing it to gen_probe_stack. Create address operand and
8976 legitimize it for the probe_stack_address case.
8978 2018-04-09 Jan Hubicka <jh@suse.cz>
8981 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8982 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8983 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8985 (free_lang_data): Rebuild type inheritance graph.
8987 2018-04-09 Martin Sebor <msebor@redhat.com>
8989 * invoke.texi (-finline-small-functions): Mention other optimization
8991 (-findirect-inlining, -fpartial-inlining): Same.
8992 (-finline-functions-called-once): Same.
8993 (-freorder-blocks-and-partition): Same.
8995 2018-04-09 Jan Hubicka <jh@suse.cz>
8998 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8999 jumps; choose last target that matches the criteria (i.e.
9000 no partition changes for non-crossing jumps).
9001 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9002 support for redirecting crossing jumps to non-crossing.
9004 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9006 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9007 also for naked functions.
9009 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9011 * config/arc/arc.md (add_shift): New pattern.
9012 (add_shift2): Likewise.
9013 (sub_shift): Likewise.
9014 (sub_shift_cmp0_noout): Likewise.
9015 (compare_si_ashiftsi): Likewise.
9016 (xbfu_cmp0_noout): New combine pattern.
9017 (xbfu_cmp0"): Likewise.
9018 (movsi_set_cc_insn): Place the predicable variant first.
9019 (commutative_binary_cmp0_noout): Remove clobber.
9020 (commutative_binary_cmp0): New pattern.
9021 (noncommutative_binary_cmp0): Likewise.
9022 (noncommutative_binary_cmp0_noout): Likewise.
9023 (noncommutative_binary_comparison_result_used): Removed.
9024 (rsub_cmp0): New pattern.
9025 (rsub_cmp0_noout): Likewise.
9026 (extzvsi): Changed, keep only meaningful variants.
9027 (SQH, SEZ): New iterators.
9028 (SQH_postfix): New mode attribute.
9029 (SEZ_prefix): New code attribute.
9030 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9031 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9032 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9034 (noncommutative_operator): Check the availability of barrel
9037 2018-04-09 Richard Biener <rguenther@suse.de>
9039 PR tree-optimization/85284
9040 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9041 Only use the niter constraining form of simple_iv when the exit
9044 2018-04-09 Tom de Vries <tom@codesourcery.com>
9047 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9048 (define_expand "*memory_barrier"): New define_expand.
9049 (define_insn "memory_barrier"): New insn.
9051 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9053 PR rtl-optimization/80463
9054 PR rtl-optimization/83972
9055 PR rtl-optimization/83480
9057 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9058 correct producer for the insn.
9059 (tidy_control_flow): Fixup seqnos in case of debug insns.
9061 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9063 PR rtl-optimization/83913
9065 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9066 different sched-times when merging exprs.
9068 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9070 PR rtl-optimization/83962
9072 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9073 tidy_fallthru_edge and tidy_control_flow.
9075 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9077 PR rtl-optimization/83530
9079 * sel-sched.c (force_next_insn): New global variable.
9080 (remove_insn_for_debug): When force_next_insn is true, also leave only
9081 next insn in the ready list.
9082 (sel_sched_region): When the region wasn't scheduled, make another pass
9083 over it with force_next_insn set to 1.
9085 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9087 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9089 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9090 for interrupt control.
9091 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9092 functions for interrupt control.
9093 * config/nds32/nds32-intrinsic.md: Likewise.
9094 * config/nds32/nds32_intrinsic.h: Likewise.
9095 * config/nds32/nds32.h (nds32_builtins): Likewise.
9097 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9099 * config/nds32/nds32.c (nds32_init_machine_status,
9100 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9101 strict_aligned_p field.
9102 (nds32_expand_to_rtl_hook): New function.
9103 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9104 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9106 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9107 Chung-Ju Wu <jasonwucj@gmail.com>
9109 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9110 * config/nds32/nds32-n7.md: New file.
9111 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9112 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9114 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9115 * config/nds32/nds32.md (pipeline_model): Add n7.
9116 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9117 * config/nds32/pipelines.md: Include n7 settings.
9119 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9120 Chung-Ju Wu <jasonwucj@gmail.com>
9122 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9123 * config/nds32/nds32-e8.md: New file.
9124 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9125 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9127 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9128 * config/nds32/nds32.md (pipeline_model): Add e8.
9129 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9130 * config/nds32/pipelines.md: Include e8 settings.
9132 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9133 Chung-Ju Wu <jasonwucj@gmail.com>
9135 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9136 * config/nds32/nds32-n8.md: New file.
9137 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9138 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9140 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9141 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9142 * config/nds32/nds32.md (pipeline_model): Add n8.
9143 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9144 * config/nds32/pipelines.md: Include n8 settings.
9146 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9147 Chung-Ju Wu <jasonwucj@gmail.com>
9149 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9150 * config/nds32/nds32-n9-2r1w.md: New file.
9151 * config/nds32/nds32-n9-3r2w.md: New file.
9152 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9153 nds32_register_ports): New or modify for cpu n9.
9154 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9156 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9157 * config/nds32/nds32-utils.c: New file.
9158 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9159 TARGET_MUL_SLOW): Define.
9160 * config/nds32/nds32.md (pipeline_model): New attribute.
9161 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9162 New options that support cpu n9.
9163 * config/nds32/pipelines.md: Include n9 settings.
9164 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9166 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9168 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9169 information if necessary.
9170 (output_cond_branch_compare_zero): Likewise.
9171 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9172 (nds32_target_alignment): Refine for alignment.
9173 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9174 (FUNCTION_BOUNDARY): Modify.
9175 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9177 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9179 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9181 * config/nds32/constants.md (unspec_volatile_element): Add values for
9182 TLB operation and data prefetch.
9183 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9184 functions for TLB operation and data prefetch.
9185 * config/nds32/nds32-intrinsic.md: Likewise.
9186 * config/nds32/nds32_intrinsic.h: Likewise.
9187 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9188 (nds32_print_operand): Likewise.
9189 * config/nds32/nds32.h (nds32_builtins): Likewise.
9191 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9192 Andrew Pinski <pinsika@gcc.gnu.org>
9195 * match.pd: Use constant_boolean_node of correct type instead of
9196 boolean_true_node or boolean_false_node for simplifying
9197 pointer comparisons to zero.
9199 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9201 PR tree-optimization/80021
9202 * tree.c (verify_type_variant): Make error call in verify_variant_match
9203 translatable and remove final full stop.
9205 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9207 * config/nds32/constants.md (unspec_volatile_element): Add
9208 UNSPEC_VOLATILE_EH_RETURN.
9209 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9210 nds32_output_stack_pop): Support dwarf exception handling process.
9211 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9212 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9213 exception handling process.
9214 (nds32_compute_stack_frame): Likewise.
9215 (nds32_return_addr_rtx): Likewise.
9216 (nds32_initial_elimination_offset): Likewise.
9217 (nds32_expand_prologue): Likewise.
9218 (nds32_expand_epilogue): Likewise.
9219 (nds32_dynamic_chain_address): New function.
9220 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9222 (DYNAMIC_CHAIN_ADDRESS): Define.
9223 (EH_RETURN_DATA_REGNO): Define.
9224 (EH_RETURN_STACKADJ_RTX): Define.
9225 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9226 patterns for dwarf exception handling.
9228 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9230 * config/nds32/nds32.h: Clean up obsolete macros.
9232 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9234 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9235 Add enum values for particular instructions.
9236 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9237 particular intrinsic functions.
9238 * config/nds32/nds32-intrinsic.md: Likewise.
9239 * config/nds32/nds32_intrinsic.h: Likewise.
9240 * config/nds32/nds32.h (nds32_builtins): Likewise.
9241 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9242 (btst, ave): New patterns for particular instructions.
9244 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9246 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9247 Add enum values for atomic load/store and memory sync.
9248 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9250 * config/nds32/nds32-intrinsic.md: Likewise.
9251 * config/nds32/nds32_intrinsic.h: Likewise.
9252 * config/nds32/nds32.h (nds32_builtins): Likewise.
9254 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9256 PR tree-optimization/85257
9257 * fold-const.c (native_encode_vector): If not all elts could fit
9258 and off is -1, return 0 rather than offset.
9259 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9260 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9261 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9262 adjust buffer in native_interpret_expr call.
9264 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9266 * config/nds32/constants.md (unspec_volatile_element): Add cache
9267 control enum values.
9268 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9269 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9270 * config/nds32/nds32.c (nds32_cctl_names): New.
9271 (nds32_print_operand): Handle cache control register names.
9272 * config/nds32/nds32.h (nds32_builtins): New enum values.
9273 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9275 * config/nds32/nds32.md (type): Add mmu.
9276 * config/nds32/pipelines.md (simple_insn): Add mmu.
9278 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9280 * config/nds32/nds32.md (type): Remove call.
9281 * config/nds32/pipelines.md (simple_insn): Likewise.
9283 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9285 * config/nds32/constants.md (unspec_volatile_element): Add
9286 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9287 UNSPEC_VOLATILE_FMFCFG.
9288 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9289 description for fmfcfg and fmfcsr.
9290 (bdesc_1arg): Add fmtcsr.
9291 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9292 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9293 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9294 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9295 unspec_fmfcfg): New patterns.
9296 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9297 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9298 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9299 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9300 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9301 __nds32__fmfcfg): Define.
9303 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9305 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9306 intrinsic register names.
9307 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9308 intrinsic register enum values and macros.
9310 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9312 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9313 for load/store addressing form.
9314 (nds32_print_operand_address): Likewise.
9316 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9319 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9320 based on LABEL_REF. Remove useless assertion.
9321 (pic_address_needs_scratch): Fix formatting.
9322 (sparc_legitimize_pic_address): Minor tweaks.
9323 (sparc_delegitimize_address): Adjust assertion accordingly.
9324 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9325 into symbolic_operand.
9326 (movsi_high_pic_label_ref): Likewise.
9327 (movsi_lo_sum_pic_label_ref): Likewise.
9328 (movdi_pic_label_ref): Likewise.
9329 (movdi_high_pic_label_ref): Likewise.
9330 (movdi_lo_sum_pic_label_ref): Likewise.
9332 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9334 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9335 custom LIB_SPEC setup.
9337 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9338 Kito Cheng <kito.cheng@gmail.com>
9340 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9341 * config/riscv/freebsd.h: New.
9343 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9345 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9346 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9349 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9350 Kito Cheng <kito.cheng@gmail.com>
9352 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9353 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9354 * config/nds32/nds32-protos.h (nds32_output_call,
9355 nds32_output_return): Declare.
9356 * config/nds32/nds32.md: Refine all the call and return patterns.
9358 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9361 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9362 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9364 PR rtl-optimization/84872
9365 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9366 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9369 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9371 * expr.c (copy_blkmode_to_reg): Revert 254862.
9372 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9374 2018-04-06 Richard Biener <rguenther@suse.de>
9377 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9378 after seeing a component reference with an adjacent field. Treat
9379 refs to arrays at struct end of external decls similar to
9380 refs to unconstrained commons.
9382 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9385 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9386 look through SAVE_EXPRs with non-side-effects argument. Adjust
9388 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9391 2018-04-06 Richard Biener <rguenther@suse.de>
9394 * alias.c (find_base_term): New wrapper around find_base_term
9395 unwinding CSELIB_VAL_PTR changes.
9396 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9399 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9401 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9403 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9404 constant definitions.
9405 ("nop"): lr 0,0 -> nopr r0
9406 ("nop_lr0", "nop_lr1"): New insn definitions.
9408 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9410 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9411 NDS32_V3PUSH_AVAILABLE_P macro.
9413 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9414 Chung-Ju Wu <jasonwucj@gmail.com>
9416 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9417 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9418 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9419 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9420 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9421 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9422 * config/nds32/constraints.md: New constraints and checking for hard
9423 float configuration.
9424 * config/nds32/iterators.md: New mode iterator and attribute for hard
9425 float configuration.
9426 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9428 * config/nds32/nds32-fpu.md: New file.
9429 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9430 deal with hard float code generation.
9431 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9433 (abi_type, float_reg_number): New enum type.
9434 * config/nds32/nds32-predicates.c: New predicates for hard float.
9435 * config/nds32/nds32-protos.h: Declare functions for hard float.
9436 * config/nds32/nds32.c: Implementation for hard float configuration.
9437 * config/nds32/nds32.h: Definitions for hard float configuration.
9438 * config/nds32/nds32.md: Include hard float machine description and
9439 modify patterns for hard float configuration.
9440 * config/nds32/nds32.opt: New options for hard float configuration.
9441 * config/nds32/predicates.md: New predicates for hard float
9444 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9446 * common/config/nds32/nds32-common.c
9447 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9449 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9452 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9453 CONSTRUCTOR_ELT (ctor, ...)->value.
9455 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9458 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9460 2018-04-05 Tom de Vries <tom@codesourcery.com>
9463 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9466 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9467 Kito Cheng <kito.cheng@gmail.com>
9469 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9470 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9471 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9472 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9474 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9475 Kito Cheng <kito.cheng@gmail.com>
9477 * config/nds32/constraints.md (Ufe): New memory constraint.
9478 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9479 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9480 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9482 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9483 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9485 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9487 * config/nds32/nds32.md: Use optimize_size in the condition for
9488 alu-shift instructions.
9490 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9492 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9494 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9496 * config/nds32/nds32.md (negsi2): Refine pattern.
9498 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9499 Chung-Ju Wu <jasonwucj@gmail.com>
9501 * config/nds32/iterators.md (shift_rotate): New code iterator.
9502 (shift): New code attribute.
9503 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9504 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9505 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9506 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9507 bit-wise operations.
9508 (andsi3, *andsi3): Ditto.
9509 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9510 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9511 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9512 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9513 nds32_ior_operand, nds32_xor_operand): New predicates.
9515 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9517 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9518 (addsi3, subsi3): ... this.
9520 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9522 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9524 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9526 * config/nds32/nds32.md: Adjust indention.
9528 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9530 * config/nds32/nds32.md (feature): New attribute.
9532 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9534 * config/nds32/nds32.md (subtype): New attribute.
9536 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9539 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9540 expansion to perform a bitwise AND of the argument followed by a
9541 boolean negation of the result.
9543 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9545 PR rtl-optimization/84878
9546 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9547 the basic block. Assert the use reference is not artificial and that
9548 it has an associated insn.
9550 2018-04-04 Michael Matz <matz@suse.de>
9552 * builtins.c (compute_objsize): Pass correct operand
9553 to array_at_struct_end_p.
9555 2018-04-04 Richard Biener <rguenther@suse.de>
9558 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9559 from contexts for DINFO_LEVEL_TERSE and below.
9561 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9563 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9564 resiter_operand condition.
9565 * config/nds32/nds32.md (*move<mode>): Ditto.
9567 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9568 Monk Chiang <sh.chiang04@gmail.com>
9570 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9572 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9574 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9576 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9577 Kito Cheng <kito.cheng@gmail.com>
9579 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9580 nds32_cond_code_str, output_cond_branch,
9581 output_cond_branch_compare_zero, nds32_expand_cbranch,
9582 nds32_expand_cstore, nds32_expand_movcc,
9583 nds32_output_cbranchsi4_equality_zero,
9584 nds32_output_cbranchsi4_equality_reg,
9585 nds32_output_cbranchsi4_equality_reg_or_const_int,
9586 nds32_output_cbranchsi4_greater_less_zero: New functions.
9587 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9588 nds32_expand_cstore, nds32_expand_movcc,
9589 nds32_output_cbranchsi4_equality_zero,
9590 nds32_output_cbranchsi4_equality_reg,
9591 nds32_output_cbranchsi4_equality_reg_or_const_int,
9592 nds32_output_cbranchsi4_greater_less_zero): Declare.
9593 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9594 nds32_rimm11s_operand): New predicates.
9595 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9596 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9599 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9601 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9602 * config/nds32/nds32.md: Ditto.
9603 * config/nds32/pipelines.md: Ditto.
9605 2018-04-04 Richard Biener <rguenther@suse.de>
9607 PR tree-optimization/85168
9608 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9609 propagating abnormals.
9611 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9613 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9615 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9616 Kito Cheng <kito.cheng@gmail.com>
9618 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9619 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9620 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9621 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9622 * config/nds32/nds32.md (sibcall_internal): New.
9623 (sibcall_register): Remove.
9624 (sibcall_immediate): Remove.
9625 (sibcall_value_internal): New.
9626 (sibcall_value_register): Remove.
9627 (sibcall_value_immediate): Remove.
9628 * config/nds32/predicates.md (nds32_general_register_operand): New.
9629 (nds32_call_address_operand): New.
9631 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9633 PR rtl-optimization/85167
9634 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9635 bb_defs if *split_p, instead preinitialize it to NULL.
9637 PR tree-optimization/85156
9638 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9639 evaluating the argument multiple times.
9641 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9643 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9645 (_mm_cvtpd_ps): Likewise.
9646 (_mm_cvttpd_epi32): Likewise.
9647 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9648 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9649 vector, pixel, and bool following altivec.h include.
9651 2018-04-03 Martin Sebor <msebor@redhat.com>
9653 * doc/extend.texi (Common Function Attributes): Clarify.
9654 (const attribute): Likewise.
9655 (pure attribute): Likewise.
9657 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9660 * config/i386/i386.c (ix86_expand_vector_set): Use
9661 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9663 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9665 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9666 instructions when changing rounding bits to preserve precision bits
9667 in the x87 control word.
9669 2018-04-03 Martin Liska <mliska@suse.cz>
9671 PR tree-optimization/82491
9672 * rtl.h (strip_offset_and_add): Replace += suboffset with
9673 poly_uint64 () + suboffset.
9675 2018-03-29 Martin Liska <mliska@suse.cz>
9676 Martin Jambor <mjambor@suse.cz>
9679 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9680 param_type is not an integral or pointer type.
9682 2018-04-03 Richard Biener <rguenther@suse.de>
9684 * sese.h (recompute_all_dominators): Remove.
9686 2018-04-02 Martin Sebor <msebor@redhat.com>
9688 * doc/invoke.texi (-Wrestrict): Fix typos.
9690 2018-04-02 Jim Wilson <jimw@sifive.com>
9692 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9693 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9694 (<optab>di3, <optab>si3_extend): Likewise.
9695 (<optab>si3_mask, <optab>si3_mask_1): New.
9696 (<optab>di3_mask, <optab>di3_mask_1): New.
9697 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9698 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9699 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9701 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9703 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9706 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9708 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9709 (nds32_canonicalize_comparison): New function.
9711 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9712 Kito Cheng <kito.cheng@gmail.com>
9713 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9715 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9716 * config/nds32/constants.md (unspec_volatile_element): Add
9717 UNSPEC_VOLATILE_RELAX_GROUP.
9718 * config/nds32/nds32-relax-opt.c: New file.
9719 * config/nds32/nds32-predicates.c
9720 (nds32_symbol_load_store_p): New function.
9721 * config/nds32/nds32-protos.h
9722 (nds32_symbol_load_store_p): Declare function.
9723 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9724 * config/nds32/nds32.c
9725 (nds32_register_pass): New function to register pass.
9726 (nds32_register_passes): New function to register passes.
9727 * config/nds32/nds32.md (relax_group): New pattern.
9728 * config/nds32/nds32.opt (mrelax-hint): New option.
9729 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9731 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9733 * config/nds32/t-nds32: Modify files dependency.
9735 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9737 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9738 (PROFILE_HOOK): Define its implementation.
9740 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9742 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9743 type and 32-bit size.
9745 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9748 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9749 (V_128_256): New mode iterator.
9750 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9751 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9752 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9754 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9755 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9757 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9760 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9761 NaN inputs correctly.
9763 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9766 * config/rs6000/vsx.md (??r): New mode attribute.
9767 (*vsx_mov<mode>_64bit): Use it.
9768 (*vsx_mov<mode>_32bit): Likewise.
9770 2018-03-30 Martin Sebor <msebor@redhat.com>
9772 PR tree-optimization/84818
9773 * builtins.c (check_access): Use warning_n.
9775 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9778 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9780 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9783 2018-03-30 Julia Koval <julia.koval@intel.com>
9786 * x86-tune.def (movx, partial_reg_dependency): Enable for
9789 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9792 * lra-constraints.c (process_alt_operands): Move setting
9793 this_alternative_matches below.
9795 2018-03-29 Martin Liska <mliska@suse.cz>
9798 * doc/invoke.texi: Document how LTO works with debug info.
9799 Describe auto-load support of binutils. Mention 'x86-64'
9800 as valid option value of -march option.
9802 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9804 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9807 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9808 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9809 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9812 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9815 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9816 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9817 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9818 for RS6000_BTM_POWERPC64.
9819 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9820 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9821 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9826 2018-03-28 Carl Love <cel@us.ibm.com>
9829 2017-09-27 Carl Love <cel@us.ibm.com>
9831 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9832 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9833 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9836 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9838 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9839 instead of __vector bool.
9840 (_mm_max_pu8): Likewise.
9841 (_mm_min_pi16): Likewise.
9843 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9846 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9847 (DIVWEUO): Likewise.
9849 (DIVDEUO): Likewise.
9850 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9851 DIVWEUO and DIVDEUO.
9852 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9853 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9854 (div_extend): Likewise.
9855 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9857 (__builtin_divweuo): Likewise.
9858 (__builtin_divdeo): Likewise.
9859 (__builtin_divdeuo): Likewise.
9861 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9864 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9865 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9867 PR tree-optimization/82004
9868 * gimple-match-head.c (optimize_pow_to_exp): New function.
9869 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9870 Don't fold to exp if optimize_pow_to_exp is false.
9872 2018-03-28 Martin Liska <mliska@suse.cz>
9875 * calls.c (initialize_argument_information): Fix trailing space.
9876 * common.opt: Fix typo and provide better explanation for
9877 -fsanitize-coverage option.
9878 * config/i386/i386.opt: Fix typo.
9880 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9881 Martin Liska <mliska@suse.cz>
9884 * gimplify.c (asan_poison_variable): Don't do the check for
9885 gimplify_omp_ctxp here.
9886 (gimplify_decl_expr): Do it here.
9887 (gimplify_target_expr): Likewise.
9889 2018-03-28 Martin Liska <mliska@suse.cz>
9892 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9893 chkp_type_bounds_count if MPX is not enabled.
9895 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9897 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9899 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9902 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9903 function to create the function decl for complex long double
9904 multiply and divide for -mabi=ieeelongdouble.
9905 (init_float128_ieee): Call it.
9907 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9910 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9911 -fcf-protection=branch -mibt.
9912 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9914 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9917 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9918 arm_word_relocations.
9920 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9923 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9924 extern array declarations.
9926 2018-03-27 Richard Biener <rguenther@suse.de>
9929 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9930 explicit single_use checks.
9932 2018-03-27 Richard Biener <rguenther@suse.de>
9934 PR tree-optimization/85082
9935 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9938 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9940 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9941 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9942 Turn on fasynchronous-unwind-tables and funwind-tables.
9944 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9947 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9948 (*bmi_blsr_<mode>_ccz): Ditto.
9950 2018-03-26 Tom de Vries <tom@codesourcery.com>
9952 PR tree-optimization/85063
9953 * omp-general.c (offloading_function_p): New function. Factor out
9955 * omp-offload.c (pass_omp_target_link::gate): ... here.
9956 * omp-general.h (offloading_function_p): Declare.
9957 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9958 with attribute omp declare target for offloading functions.
9960 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9962 PR tree-optimization/84005
9963 * tree-data-ref.h (get_base_for_alignment): Declare.
9964 * tree-data-ref.c (get_base_for_alignment_1): New function.
9965 (get_base_for_alignment): Likewise.
9966 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9967 get_base_for_alignment to find a suitable base object, instead
9968 of always using drb->base_address.
9970 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9973 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9974 known size by default.
9976 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9979 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9980 and non BLKmode operands.
9982 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9985 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9986 Clean up attributes.
9988 2018-03-23 Richard Biener <rguenther@suse.de>
9991 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9992 we are going to emit early debug for LTO.
9994 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9997 * function.c (match_asm_constraints_1): Don't optimize if input
9998 doesn't satisfy general_operand predicate for output's mode.
10000 PR inline-asm/85022
10001 * alias.c (write_dependence_p): Don't require for x_canonicalized
10002 non-VOIDmode if x has VOIDmode.
10005 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10006 just don't try to optimize it rather than assert it never happens.
10008 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10010 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10011 macro expansions for definition of ST_INTERNAL_<mode> and
10012 LD_INTERNAL_<mode> builtins.
10013 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10015 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10017 (altivec_expand_st_builtin): Likewise.
10018 (altivec_expand_builtin): Remove calls to deleted functions.
10019 (rs6000_address_for_altivec): Delete this function.
10020 * config/rs6000/vector.md: Remove expands for
10021 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10023 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10026 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10027 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10028 re-computing once computed.
10029 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10030 (arm_init_machine_status): Initialize
10031 machine->static_chain_stack_bytes.
10033 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10036 * doc/extend.texi: Add four new prototypes for vec_ld.
10037 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10038 definitions for more logical presentation.
10039 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10040 entries for V1TI variants of __builtin_altivec_ld builtin.
10041 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10042 handling of V1TI variant of LVX icode pattern.
10043 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10044 (rs6000_gimple_fold_builtin): Likewise.
10045 (altivec_init_builtins): Add code to define
10046 __builtin_altivec_lvx_v1ti function.
10048 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10050 PR inline-asm/84941
10051 * function.c (match_asm_constraints_1): Don't do the optimization
10052 if input isn't a REG, SUBREG, MEM or constant.
10054 2018-03-22 Tom de Vries <tom@codesourcery.com>
10056 PR tree-optimization/84956
10057 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10058 bb_has_abnormal_pred.
10060 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10063 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10064 DECL_INITIAL (decl) to decl at the end.
10065 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10066 adjust the comment.
10068 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10070 * doc/extend.texi (__builtin_tgmath): Document when complex
10071 integer types are treated as _Complex _Float64.
10073 2018-03-21 Tom de Vries <tom@codesourcery.com>
10075 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10077 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10079 PR tree-optimization/84960
10080 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10081 if it is ENTRY block, move them into single succ of ENTRY in that case.
10083 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10085 PR tree-optimization/84811
10086 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10087 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10088 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10089 (known_subrange_p): Update accordingly. Cast each value involved
10090 in the size comparison, rather than casting the result of the
10093 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10095 PR tree-optimization/84982
10096 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10097 by flipping the least significant bit rather than all bits from
10098 bitpos to bitpos + bitsize - 1.
10100 2018-03-21 Nathan Sidwell <nathan@acm.org>
10102 * doc/extend.texi (Deprecated Features): Remove mention of
10103 long-deleted deprecations.
10105 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10108 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10109 * configure: Regenerate.
10111 2018-03-21 Tom de Vries <tom@codesourcery.com>
10113 PR tree-optimization/83126
10114 * tree-parloops.c (num_phis): New function.
10115 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10117 2018-03-21 Nathan Sidwell <nathan@acm.org>
10119 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10120 mention anon-struct/union members and trailing attributes.
10122 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10124 PR tree-optimization/84969
10125 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10126 builtin memset partitions if they set different rhs values.
10128 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10130 PR rtl-optimization/84989
10131 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10132 VEC_DUPLICATE with scalar result mode.
10134 2018-03-21 Martin Liska <mliska@suse.cz>
10137 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10138 not intended return statement.
10140 2018-03-21 Martin Liska <mliska@suse.cz>
10143 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10144 (chkp_find_bound_slots_1): Limit number of iterations.
10146 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10149 * Minor grammar fixes for x86 options.
10151 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10154 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10155 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10157 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10160 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10161 (altivec_lvx_<mode>_1op): Likewise.
10162 (altivec_stvx_<mode>_2op): Likewise.
10163 (altivec_stvx_<mode>_1op): Likewise.
10164 (altivec_lvx_<VM2:mode>): New define_expand.
10165 (altivec_stvx_<VM2:mode>): Likewise.
10166 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10167 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10168 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10169 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10170 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10171 (rs6000_gen_lvx): Likewise.
10172 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10173 (altivec_expand_stv_builtin): Likewise.
10174 (altivec_expand_builtin): Likewise.
10175 * config/rs6000/vector.md: Likewise.
10177 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10180 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10183 2018-03-20 Richard Biener <rguenther@suse.de>
10186 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10187 sign-conversions as zero, fall back to standard scalar_stmt
10190 2018-03-20 Martin Liska <mliska@suse.cz>
10193 * predict.c (rebuild_frequencies): Handle case when we have
10194 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10196 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10199 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10200 flag_section_anchors.
10201 * varasm.c (use_blocks_for_decl_p): Remove hack for
10202 dw2_force_const_mem.
10205 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10207 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10208 be created, use lowpart_subreg of operands[0] rather than operands[0]
10210 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10211 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10212 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10213 and n constraint instead of aarch64_shift_imm_di and Usd.
10214 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10215 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10217 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10220 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10221 to favor GPR over NEON registers.
10222 (<shift>di3_neon): Likewise.
10224 2018-03-20 Tom de Vries <tom@codesourcery.com>
10227 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10228 (nvptx_process_pars): Emit bar.sync asap and alap.
10230 2018-03-20 Tom de Vries <tom@codesourcery.com>
10233 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10234 seen_label if seen_label is already set.
10236 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10239 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10240 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10241 Use 1U instead of 1. Formatting fixes.
10244 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10245 instead of TREE_TYPE (s1) for the return value.
10247 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10249 PR tree-optimization/84946
10250 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10251 bitsize + bitsize in poly_uint64 rather than poly_int64.
10254 * dwarf2asm.c: Include fold-const.c.
10255 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10256 of decl rather than decl itself.
10258 PR rtl-optimization/84643
10259 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10261 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10264 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10265 calling assemble_variable.
10267 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10270 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10271 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10273 2018-03-19 Jim Wilson <jimw@sifive.com>
10276 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10277 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10278 (riscv_first_stack_step): Likewise.
10279 (riscv_option_override): Use STACK_BOUNDARY instead of
10280 MIN_STACK_BOUNDARY.
10281 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10282 MIN_STACK_BOUNDARY.
10283 (BIGGEST_ALIGNMENT): Set to 128.
10284 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10285 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10288 2018-03-19 Richard Biener <rguenther@suse.de>
10290 PR tree-optimization/84933
10291 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10292 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10294 2018-03-19 Richard Biener <rguenther@suse.de>
10296 PR tree-optimization/84859
10297 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10298 (cond_if_else_store_replacement): Perform sinking operation on
10299 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10300 Generalize what a BB with a single eligible store is.
10302 2018-03-19 Richard Biener <rguenther@suse.de>
10304 PR tree-optimization/84929
10305 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10306 chrec_is_positive against non-chrec arg.
10308 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10311 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10313 2018-03-18 Martin Liska <mliska@suse.cz>
10315 PR rtl-optimization/84635
10316 * regrename.c (build_def_use): Use matches_mode only when
10319 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10321 PR tree-optimization/84913
10322 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10323 vectorize chains of COND_EXPRs.
10325 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10327 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10329 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10331 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10333 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10335 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10337 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10338 Kito Cheng <kito.cheng@gmail.com>
10340 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10341 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10342 (nds32_adjust_reg_alloc_order): New function.
10343 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10345 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10347 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10348 nds32_print_operand, nds32_print_operand_address): Use
10349 HOST_WIDE_INT_PRINT_DEC instead.
10351 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10353 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10355 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10358 * config/i386/i386.c (initial_ix86_tune_features,
10359 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10360 unsigned long long.
10361 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10362 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10363 rather than 1u << ix86_tune. Formatting fix.
10364 (ix86_option_override_internal): Change ix86_arch_mask from
10365 unsigned int to unsigned HOST_WIDE_INT, initialize to
10366 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10367 (ix86_function_specific_restore): Likewise.
10369 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10372 * postreload.c (reload_combine_recognize_pattern): Perform
10373 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10374 truncate_int_for_mode the result for the destination's mode.
10377 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10378 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10380 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10383 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10386 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10387 regno_allocno_class_array and sorted_pseudos.
10388 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10389 insns where regno is used.
10391 2018-03-16 Martin Liska <mliska@suse.cz>
10394 * multiple_target.c (create_dispatcher_calls): Redirect
10395 reference in the symbol table.
10397 2018-03-16 Martin Liska <mliska@suse.cz>
10400 * multiple_target.c (create_dispatcher_calls): Redirect also
10403 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10407 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10408 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10409 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10411 2018-03-16 Julia Koval <julia.koval@intel.com>
10413 * doc/invoke.texi (Skylake Server): Add CLWB.
10414 Cannonlake): Remove CLWB.
10416 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10418 PR tree-optimization/84841
10419 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10421 (FLOAT_ONE_CONST_TYPE): Define.
10422 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10423 (sort_by_operand_rank): Put entries with higher constant_type last
10424 rather than first to match comments.
10426 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10428 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10431 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10434 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10435 check and use address of target always.
10437 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10440 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10441 (indirect_thunk_bnd_needed): Likewise.
10442 (indirect_thunks_used): Likewise.
10443 (indirect_thunks_bnd_used): Likewise.
10444 (indirect_return_needed): New.
10445 (indirect_return_bnd_needed): Likewise.
10446 (output_indirect_thunk_function): Add a bool argument for
10448 (output_indirect_thunk_function): Don't generate alias for
10449 function return thunk.
10450 (ix86_code_end): Call output_indirect_thunk_function to generate
10451 function return thunks.
10452 (ix86_output_function_return): Set indirect_return_bnd_needed
10453 and indirect_return_needed instead of indirect_thunk_bnd_needed
10454 and indirect_thunk_needed.
10456 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10458 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10459 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10460 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10462 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10463 Paul Hua <paul.hua.gm@gmail.com>
10466 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10468 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10470 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10471 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10472 resp. SFmode cases.
10474 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10477 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10478 instead of GET_MODE_SIZE when comparing Units.
10480 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10483 * varasm.c (hash_section): Return an unchangeble hash value
10484 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10485 Return !aarch64_can_use_per_function_literal_pools_p ().
10487 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10490 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10491 rather than address of cmode as last argument to prepare_cmp_insn.
10493 2018-03-15 Julia Koval <julia.koval@intel.com>
10495 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10496 F_AVX512VNNI, F_AVX512BITALG): New.
10498 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10501 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10502 insn for floating-point loads and stores.
10504 2018-03-14 Carl Love <cel@us.ibm.com>
10506 * config/rs6000/rs6000-c.c: Add macro definitions for
10507 ALTIVEC_BUILTIN_VEC_PERMXOR.
10508 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10509 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10510 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10511 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10513 * config/doc/extend.texi: Add prototypes for vec_permxor.
10515 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10518 * diagnostic-show-locus.c (class layout_point): Convert m_line
10519 from int to linenum_type.
10520 (line_span::comparator): Use linenum "compare" function when
10521 comparing line numbers.
10522 (test_line_span): New function.
10523 (layout_range::contains_point): Convert param "row" from int to
10525 (layout_range::intersects_line_p): Likewise.
10526 (layout::will_show_line_p): Likewise.
10527 (layout::print_source_line): Likewise.
10528 (layout::should_print_annotation_line_p): Likewise.
10529 (layout::print_annotation_line): Likewise.
10530 (layout::print_leading_fixits): Likewise.
10531 (layout::annotation_line_showed_range_p): Likewise.
10532 (struct line_corrections): Likewise for field m_row.
10533 (line_corrections::line_corrections): Likewise for param "row".
10534 (layout::print_trailing_fixits): Likewise.
10535 (layout::get_state_at_point): Likewise.
10536 (layout::get_x_bound_for_row): Likewise.
10537 (layout::print_line): Likewise.
10538 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10539 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10540 * input.c (selftest::test_linenum_comparisons): New function.
10541 (selftest::input_c_tests): Call it.
10542 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10543 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10544 * selftest.h (ASSERT_GT): New macro.
10545 (ASSERT_GT_AT): New macro.
10546 (ASSERT_LT): New macro.
10547 (ASSERT_LT_AT): New macro.
10549 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10551 PR rtl-optimization/84780
10552 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10554 2018-03-14 Martin Liska <mliska@suse.cz>
10556 * tree.c (record_node_allocation_statistics): Use
10557 get_stats_node_kind.
10558 (get_stats_node_kind): New function extracted from
10559 record_node_allocation_statistics.
10560 (free_node): Use get_stats_node_kind.
10562 2018-03-14 Richard Biener <rguenther@suse.de>
10564 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10565 that the value-set of ANTIC_IN doesn't grow.
10568 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10570 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10571 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10573 2018-03-14 Julia Koval <julia.koval@intel.com>
10575 * config.gcc (icelake-client, icelake-server): New.
10577 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10578 (initial_ix86_arch_features): Ditto.
10579 (PTA_SKYLAKE): Add SGX.
10580 (PTA_ICELAKE): Remove.
10581 (PTA_ICELAKE_CLIENT): New.
10582 (PTA_ICELAKE_SERVER): New.
10583 (ix86_option_override_internal): Split up icelake on icelake client and
10585 (get_builtin_code_for_version): Ditto.
10586 (fold_builtin_cpu): Ditto.
10587 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10588 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10589 * config/i386/i386.h (processor_type): Ditto.
10590 * doc/invoke.texi: Ditto.
10592 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10595 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10596 INTEGER_CST offset, add it together with bitpos / 8 and
10597 sign extend based on POINTER_SIZE.
10601 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10604 * config/i386/constraints.md (Yc): New register constraint.
10605 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10606 Use Yc constraint for alternative 2 of operand 0. Remove
10607 preferred_for_speed attribute.
10609 2018-03-14 Richard Biener <rguenther@suse.de>
10611 PR tree-optimization/84830
10612 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10613 with the old one to avoid oscillations.
10615 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10618 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10620 (assign_by_spills): Return a flag of reload assignment failure.
10621 Do not process the reload assignment failures. Do not spill other
10622 reload pseudos if they has the same reg class. Update n if
10624 (lra_assign): Add a return arg. Set up from the result of
10625 assign_by_spills call.
10626 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10627 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10628 usage_insns if it is not NULL.
10629 (spill_hard_reg_in_range): New function.
10630 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10631 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10632 function prototypes.
10633 (lra_assign): Change prototype.
10634 * lra.c (lra): Add code to deal with fails by splitting hard reg
10637 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10639 * config/riscv/riscv.opt (mrelax): New option.
10640 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10641 "norelax" when riscv_mrelax is disabled.
10642 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10644 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10647 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10648 reassociation for int modes.
10650 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10652 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10653 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10655 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10656 * config/aarch64/aarch64-sve.md
10657 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10658 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10659 (*extend<mode><Vwide>2): Rename to...
10660 (aarch64_sve_extend<mode><Vwide>2): ...this.
10661 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10662 renaming the old pattern to...
10663 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10665 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10666 define_expand, renaming the old pattern to...
10667 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10668 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10669 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10670 account when deciding which SVE instruction the optab should use.
10671 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10673 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10675 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10676 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10677 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10678 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10679 (tlsdesc_small_<mode>): Turn a define_expand and use
10680 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10681 (tlsdesc_small_advsimd_<mode>): ...this.
10682 (tlsdesc_small_sve_<mode>): New pattern.
10684 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10686 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10687 (UNSPEC_UMUL_HIGHPART): New constants.
10688 (MUL_HIGHPART): New int iteraor.
10689 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10690 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10692 (*<su>mul<mode>3_highpart): New define_insn.
10694 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10697 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10700 2018-03-13 Martin Liska <mliska@suse.cz>
10703 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10705 (sem_item_optimizer::~sem_item_optimizer): Release it.
10706 (sem_item_optimizer::merge_classes): Register variable aliases.
10707 (sem_item_optimizer::fixup_pt_set): New function.
10708 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10709 * ipa-icf.h: Declare new variables and functions.
10711 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10713 PR middle-end/84834
10714 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10715 integer_pow2p@2 and test integer_pow2p in condition.
10716 (A < 0 ? C : 0): Similarly for @1.
10718 PR middle-end/84831
10719 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10720 characters starting at p contain '\0' character, don't look beyond
10724 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10725 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10728 * reg-stack.c (change_stack): Change update_end var from int to
10729 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10730 also call set_block_for_insn on the newly added insns and rescan.
10733 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10734 on the last operand.
10737 * tree.c (stabilize_reference_1): Return save_expr (e) for
10738 STATEMENT_LIST even if it doesn't have side-effects.
10740 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10742 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10744 2018-03-12 Renlin Li <renlin.li@arm.com>
10746 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10747 aarch64_output_scalar_simd_mov_immediate.
10749 2018-03-12 Martin Sebor <msebor@redhat.com>
10751 PR tree-optimization/83456
10752 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10753 for perfectly overlapping calls to memcpy.
10754 (gimple_fold_builtin_memory_chk): Same.
10755 (gimple_fold_builtin_strcpy): Handle no-warning.
10756 (gimple_fold_builtin_stxcpy_chk): Same.
10757 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10759 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10761 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10762 parameter. Use it for SFmode.
10763 (rs6000_function_arg_advance_1): Adjust.
10764 (rs6000_function_arg): Adjust.
10765 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10767 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10769 PR rtl-optimization/84169
10770 PR rtl-optimization/84780
10771 * combine.c (can_combine_p): Check for a 2-insn combination whether
10772 the destination register is used between the two insns, too.
10774 2018-03-12 Richard Biener <rguenther@suse.de>
10776 PR tree-optimization/84803
10777 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10778 for refs DR analysis didn't process.
10780 2018-03-12 Richard Biener <rguenther@suse.de>
10782 PR tree-optimization/84777
10783 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10784 force-vectorize loops ignore whether we are optimizing for size.
10786 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10788 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10789 (TARGET_MD_ASM_ADJUST): Define.
10791 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10792 Kito Cheng <kito.cheng@gmail.com>
10793 Chung-Ju Wu <jasonwucj@gmail.com>
10795 * config/nds32/nds32.c (nds32_compute_stack_frame,
10796 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10797 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10798 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10799 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10800 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10801 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10802 * config/nds32/nds32.md (prologue, epilogue): Use macro
10803 NDS32_V3PUSH_AVAILABLE_P to do checking.
10805 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10808 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10809 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10810 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10811 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10812 addition of most attributes on !orig_type_die or the attribute not
10813 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10815 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10816 Chung-Ju Wu <jasonwucj@gmail.com>
10818 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10819 __NDS32_VH__ macro.
10820 * config/nds32/nds32.opt (mvh): New option.
10822 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10823 Chung-Ju Wu <jasonwucj@gmail.com>
10825 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10827 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10828 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10831 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10832 Chung-Ju Wu <jasonwucj@gmail.com>
10834 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10836 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10837 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10839 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10840 Kito Cheng <kito.cheng@gmail.com>
10841 Chung-Ju Wu <jasonwucj@gmail.com>
10843 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10844 UNSPEC_FFMISM and UNSPEC_FLMISM.
10845 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10846 for ffb, ffmism and flmism.
10847 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10848 (unspec_ffmism): Ditto.
10849 (unspec_flmism): Ditto.
10850 (nds32_expand_builtin_impl): Check if string extension is available.
10851 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10852 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10854 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
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 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10880 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10882 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10885 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10886 (dw_val_node): Add val_symbolic_view.
10887 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10888 (symview_upper_bound): New.
10889 (new_line_info_table): Initialize symviews_since_reset.
10890 (dwarf2out_source_line): Count symviews_since_reset and set
10891 symview_upper_bound.
10892 (dw_val_equal_p): Handle symview.
10893 (add_AT_symview): New.
10894 (print_dw_val): Handle symview.
10895 (attr_checksum, attr_checksum_ordered): Likewise.
10896 (same_dw_val_p, size_of_die): Likewise.
10897 (value_format, output_die): Likewise.
10898 (add_high_low_attributes): Use add_AT_symview for entry_view.
10899 (dwarf2out_finish): Reset symview_upper_bound, clear
10902 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10905 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10906 Add strict argument and use it.
10907 (rs6000_split_multireg_move): Update for new strict argument.
10908 (mem_operand_gpr): Disallow all non-offsettable addresses.
10909 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10911 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10914 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10915 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10916 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10919 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10920 decl, use remap_type if we want to use the type.
10922 2018-03-09 Martin Sebor <msebor@redhat.com>
10924 PR tree-optimization/84526
10925 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10927 (builtin_access::generic_overlap): Be prepared to handle non-array
10930 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10932 PR rtl-optimization/84682
10933 * lra-constraints.c (process_address_1): Check is_address flag
10934 for address constraints.
10935 (process_alt_operands): Likewise.
10936 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10937 preprocess_constraints.
10938 * recog.h (preprocess_constraints): Add oploc parameter.
10940 * recog.c (preprocess_constraints): Test address_operand for
10941 CT_ADDRESS constraints.
10943 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10946 * lra-assigns.c (assign_by_spills): Return a flag of reload
10947 assignment failure. Do not process the reload assignment
10948 failures. Do not spill other reload pseudos if they has the same
10950 (lra_assign): Add a return arg. Set up from the result of
10951 assign_by_spills call.
10952 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10953 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10954 usage_insns if it is not NULL.
10955 (spill_hard_reg_in_range): New function.
10956 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10957 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10958 function prototypes.
10959 (lra_assign): Change prototype.
10960 * lra.c (lra): Add code to deal with fails by splitting hard reg
10963 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10966 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10967 Accept complain bool parameter. Only emit errors if it is true.
10968 (arm_parse_cpu_option_name): Likewise.
10969 (arm_target_thumb_only): Adjust callers of the above.
10970 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10971 prototype to take a default true bool parameter.
10972 (arm_parse_arch_option_name): Likewise.
10974 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10975 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10979 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10980 * configure: Regenerate.
10981 * configure.ac ("linker --version-script option"): New.
10982 ("linker soname option"): New.
10984 2018-03-09 Richard Biener <rguenther@suse.de>
10986 PR tree-optimization/84775
10987 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10988 immediate uses of predicate stmts and mark them modified.
10991 PR tree-optimization/84178
10992 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10994 (version_loop_for_if_conversion): Delay update_ssa call.
10995 (tree_if_conversion): Delay update_ssa until after predicate
10998 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11001 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11002 when the function accesses prior frames.
11004 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11007 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11008 gen_llsym, otherwise call maybe_gen_llsym.
11010 PR inline-asm/84742
11011 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11012 has ',' character inside of it.
11014 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11017 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11018 as clobbering CC_REGNUM.
11020 2018-03-08 Richard Biener <rguenther@suse.de>
11022 PR middle-end/84552
11023 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11024 (follow_copies_to_constant): Do not follow SSA names registered
11027 2018-03-08 Richard Biener <rguenther@suse.de>
11029 PR tree-optimization/84178
11030 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11032 (version_loop_for_if_conversion): Delay update_ssa call.
11033 (tree_if_conversion): Delay update_ssa until after predicate
11036 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11038 PR tree-optimization/84178
11039 * tree-if-conv.c (release_bb_predicate): Remove the
11040 the assertion that the stmts have NULL use_ops.
11041 Discard the statements, asserting that they haven't
11042 yet been added to a BB.
11044 2018-03-08 Richard Biener <rguenther@suse.de>
11046 PR tree-optimization/84746
11047 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11048 (phi_translate): Pass in destination ANTIC_OUT set.
11049 (phi_translate_1): Likewise. For a simplified result lookup
11050 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11051 (phi_translate_set): Adjust.
11052 (do_pre_regular_insertion): Likewise.
11053 (do_pre_partial_partial_insertion): Likewise.
11055 2018-03-08 Martin Liska <mliska@suse.cz>
11057 PR gcov-profile/84735
11058 * doc/gcov.texi: Document usage of profile files.
11059 * gcov-io.h: Document changes in the format.
11061 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11065 * dwarf2out.c (struct dw_line_info_table): Update comments for
11067 (FORCE_RESET_NEXT_VIEW): New.
11068 (FORCE_RESETTING_VIEW_P): New.
11069 (RESETTING_VIEW_P): Check for -1 too.
11070 (ZERO_VIEW_P): Likewise.
11071 (new_line_info_table): Force-reset next view.
11072 (dwarf2out_begin_function): Likewise.
11073 (dwarf2out_source_line): Simplify zero_view_p initialization.
11074 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11075 view directly. Omit view when omitting .loc at line 0.
11077 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11079 PR tree-optimization/84740
11080 * tree-switch-conversion.c (process_switch): Call build_constructors
11081 only if info.phi_count is non-zero.
11083 PR tree-optimization/84739
11084 * tree-tailcall.c (find_tail_calls): Check call arguments against
11085 DECL_ARGUMENTS (current_function_decl) rather than
11086 DECL_ARGUMENTS (func) when checking for tail recursion.
11088 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11090 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11091 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11092 Volker Reichelt's entry and add entries for people that perform
11093 GCC fuzzy testing and report numerous bugs.
11095 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11098 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11099 readonly data in sdata, if that is disabled.
11100 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11101 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11102 -mreadonly-in-sdata option.
11104 2018-03-07 Martin Sebor <msebor@redhat.com>
11106 PR tree-optimization/84468
11107 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11108 basic block when looking for nul assignment.
11110 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11113 * except.h (output_function_exception_table): Adjust prototype.
11114 * except.c (output_function_exception_table): Remove FNNAME parameter
11115 and add SECTION parameter. Ouput one part of the table at a time.
11116 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11117 the first part of the exception table and emit unwind directives.
11118 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11119 (i386_pe_seh_cold_init): Likewise.
11120 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11121 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11122 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11123 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11124 * config/i386/winnt.c: Include except.h.
11125 (struct seh_frame_state): Add reg_offset, after_prologue and
11126 in_cold_section fields.
11127 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11128 (i386_pe_seh_cold_init): New function.
11129 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11130 to seh->in_cold_section.
11131 (seh_emit_push): Record the offset of the push.
11132 (seh_emit_save): Record the offet of the save.
11133 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11134 Test seh->after_prologue to disregard the epilogue.
11135 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11136 (i386_pe_end_cold_function): New function.
11138 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11141 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11142 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11145 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11147 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11148 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11150 PR middle-end/84723
11151 * multiple_target.c: Include tree-inline.h and intl.h.
11152 (expand_target_clones): Diagnose and fail if node->definition and
11153 !tree_versionable_function_p (node->decl).
11155 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11157 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11159 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11160 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11161 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11163 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11166 * combine.c (try_combine): Use reg_or_subregno instead of handling
11167 just paradoxical SUBREGs and REGs.
11169 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11171 * config/arc/arc.c (arc_finalize_pic): Remove function.
11172 (arc_must_save_register): We use single base PIC register, remove
11173 checks to save/restore the PIC register.
11174 (arc_expand_prologue): Likewise.
11175 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11177 (arc_verify_short): Likewise.
11178 (arc_attr_type): Likewise.
11179 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11180 (walk_stores): Likewise.
11181 (arc_address_cost): Make it static.
11182 (arc_verify_short): Likewise.
11183 (branch_dest): Likewise.
11184 (arc_attr_type): Likewise.
11185 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11186 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11187 (arc_final_prescan_insn): Remove inserting the nops due to
11188 hardware hazards. It is done in reorg step.
11189 (insn_length_variant_t): Remove.
11190 (insn_length_parameters_t): Likewise.
11191 (arc_insn_length_parameters): Likewise.
11192 (arc_get_insn_variants): Likewise.
11193 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11195 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11197 PR inline-asm/84683
11198 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11201 PR tree-optimization/84687
11202 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11204 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11206 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11208 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11209 Rename to ppc_speculation_barrier.
11210 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11211 __builtin_ppc_speculation_barrier.
11213 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11216 * combine.c (combine_simplify_rtx): Don't try to simplify if
11217 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11220 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11222 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11223 to 32 bytes when compiling for POWER9.
11225 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11228 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11229 regparm >= 3 with no arg reg available also for calls with
11230 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11233 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11235 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11237 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11240 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11242 2018-03-05 Richard Biener <rguenther@suse.de>
11244 PR tree-optimization/84486
11245 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11246 When inserting a __builtin_assume_aligned call set the LHS
11247 SSA name alignment info accordingly.
11249 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11251 PR tree-optimization/84114
11252 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11253 Avoid reassociation of FLOAT_MODE addition.
11255 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11257 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11258 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11259 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11260 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11261 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11262 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11263 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11265 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11266 __builtin_ia32_wbinvd): New builtins.
11267 (SPECIAL_ARGS2): New.
11268 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11269 (SPECIAL_ARGS2): New.
11270 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11271 (ix86_valid_target_attribute_inner_p): Ditto.
11272 (ix86_init_mmx_sse_builtins): Add special_args2.
11273 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11274 TARGET_WBNOINVD_P): New.
11275 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11276 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11277 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11278 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11279 * config/i386/pconfigintrin.h: New file.
11280 * config/i386/wbnoinvdintrin.h: Ditto.
11281 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11283 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11285 2018-03-05 Richard Biener <rguenther@suse.de>
11287 PR tree-optimization/84670
11288 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11290 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11291 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11292 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11293 doesn't grow if all successors (recursively) were visited at least
11296 2018-03-05 Richard Biener <rguenther@suse.de>
11298 PR tree-optimization/84650
11299 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11300 if executed in the loop pipeline.
11302 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11304 * doc/configfiles.texi (Configuration Files): Move info about
11305 conditionalizing $target-protos.h to...
11306 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11307 differs from $target-protos.h.
11309 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11310 Chung-Ju Wu <jasonwucj@gmail.com>
11312 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11313 * config/nds32/nds32-multiple.md (setmemsi): Define.
11314 * config/nds32/nds32-memory-manipulation.c
11315 (nds32_gen_dup_4_byte_to_word_value): New.
11316 (emit_setmem_word_loop): New.
11317 (emit_setmem_byte_loop): New.
11318 (nds32_expand_setmem_loop): New.
11319 (nds32_expand_setmem_loop_v3m): New.
11320 (nds32_expand_setmem_unroll): New.
11321 (nds32_expand_setmem): New.
11323 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11324 Chung-Ju Wu <jasonwucj@gmail.com>
11326 * config/nds32/nds32-memory-manipulation.c
11327 (nds32_emit_load_store): New.
11328 (nds32_emit_post_inc_load_store): New.
11329 (nds32_emit_mem_move): New.
11330 (nds32_emit_mem_move_block): New.
11331 (nds32_expand_movmemsi_loop_unknown_size): New.
11332 (nds32_expand_movmemsi_loop_known_size): New.
11333 (nds32_expand_movmemsi_loop): New.
11334 (nds32_expand_movmemsi_unroll): New.
11335 (nds32_expand_movmemqi): Rename ...
11336 (nds32_expand_movmemsi): ... to this.
11337 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11338 (movmemsi): ... to this.
11339 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11340 (nds32_expand_movmemsi): ... to this.
11342 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11343 Monk Chiang <sh.chiang04@gmail.com>
11344 Chung-Ju Wu <jasonwucj@gmail.com>
11346 * config/nds32/nds32-protos.h
11347 (nds32_expand_load_multiple): New arguments.
11348 (nds32_expand_store_multiple): Ditto.
11349 (nds32_valid_multiple_load_store): Rename ...
11350 (nds32_valid_multiple_load_store_p): ... to this.
11351 * config/nds32/nds32-memory-manipulation.c
11352 (nds32_expand_load_multiple): Refine implementation.
11353 (nds32_expand_store_multiple): Ditto.
11354 * config/nds32/nds32-multiple.md
11355 (load_multiple): Update nds32_expand_load_multiple interface.
11356 (store_multiple): Update nds32_expand_store_multiple interface.
11357 * config/nds32/nds32-predicates.c
11358 (nds32_valid_multiple_load_store): Rename ...
11359 (nds32_valid_multiple_load_store_p): ... to this and refine
11361 * config/nds32/predicates.md
11362 (nds32_load_multiple_and_update_address_operation): New predicate.
11363 (nds32_store_multiple_and_update_address_operation): New predicate.
11365 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11366 Chung-Ju Wu <jasonwucj@gmail.com>
11368 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11369 (combo): New attribute.
11370 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11372 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11374 * config/nds32/nds32.opt: Change -mcmodel= default value.
11376 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11377 Monk Chiang <sh.chiang04@gmail.com>
11378 Chung-Ju Wu <jasonwucj@gmail.com>
11380 * config/nds32/constants.md (unspec_element): New enum.
11381 * config/nds32/constraints.md (Umw): New constraint.
11382 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11383 * config/nds32/nds32-intrinsic.md: Likewise.
11384 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11385 (nds32_valid_smw_lwm_base_p): New.
11386 (nds32_output_smw_single_word): New.
11387 (nds32_output_lmw_single_word): New.
11388 (nds32_expand_unaligned_load): New.
11389 (nds32_expand_unaligned_store): New.
11390 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11391 (nds32_output_smw_single_word): Declare.
11392 (nds32_output_lmw_single_word): Declare.
11393 (nds32_expand_unaligned_load): Declare.
11394 (nds32_expand_unaligned_store): Declare.
11395 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11396 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11397 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11398 NDS32_BUILTIN_UASTORE_DW.
11399 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11402 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11403 Kito Cheng <kito.cheng@gmail.com>
11404 Chung-Ju Wu <jasonwucj@gmail.com>
11406 * config/nds32/nds32-intrinsic.c
11407 (nds32_expand_builtin_null_ftype_reg): Delete.
11408 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11409 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11410 (nds32_read_argument): New.
11411 (nds32_legitimize_target): Ditto.
11412 (nds32_legitimize_argument): Ditto.
11413 (nds32_check_constant_argument): Ditto.
11414 (nds32_expand_unop_builtin): Ditto.
11415 (nds32_expand_unopimm_builtin): Ditto.
11416 (nds32_expand_binop_builtin): Ditto.
11417 (nds32_builtin_decl_impl): Ditto.
11418 (builtin_description): Ditto.
11419 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11420 (nds32_init_builtins_impl): Ditto.
11421 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11422 (nds32_builtin_decl): New.
11423 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11424 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11426 2018-03-02 Jeff Law <law@redhat.com>
11428 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11429 (redundant_insn, fill_simple_delay_slots): Likewise.
11430 (fill_slots_from_thread): Likewise.
11431 * resource.c (mark_referenced_resources): Likewise.
11432 (mark_set_resources, find_dead_or_set_registers): Likewise.
11434 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11436 * substring-locations.h (format_warning_va): Formatting fix for
11437 ATTRIBUTE_GCC_DIAG.
11438 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11440 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11441 * substring-locations.c: Include intl.h.
11442 (format_warning_va): Turned into small wrapper around
11443 format_warning_n_va, renamed to ...
11444 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11445 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11447 (format_warning_at_substring_n): New function.
11448 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11449 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11450 format_warning_at_substring with just a shorter name instead of
11451 const function pointer.
11452 (fmtwarn_n): New function.
11453 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11454 appropriate, get rid of all the fmtstr temporaries, move conditionals
11455 with G_() wrapped string literals directly into fmtwarn arguments,
11456 cast dir.len to (int), formatting fixes.
11458 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11460 * doc/invoke.texi: Remove "Cilk Plus" references.
11462 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11463 Richard Biener <rguenther@suse.de>
11466 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11467 for error or warning attributes if CALL_FROM_THUNK_P is set.
11470 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11473 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11474 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11477 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11478 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11480 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11481 instead of -1U in last predictors element's probability member.
11483 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11486 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11487 arguments if they are comparable.
11489 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11491 PR tree-optimization/84634
11492 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11493 masks and masked_loop_p with a single loop_masks, making sure it's
11494 null for bb vectorization.
11496 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11498 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11499 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11500 loop->force_vectorize to check whether there is no alias.
11502 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11505 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11507 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11509 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11510 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11511 instead of a loop around prev_real_insn.
11512 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11515 PR inline-asm/84625
11516 * config/i386/i386.c (ix86_print_operand): Use conditional
11517 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11520 2018-03-02 Richard Biener <rguenther@suse.de>
11522 PR tree-optimization/84427
11523 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11524 (bitmap_set_subtract_values): Rewrite to handle multiple
11527 (prune_clobbered_mems): Likewise.
11528 (phi_translate): Take edge instead of pred/phiblock.
11529 (phi_translate_1): Likewise.
11530 (phi_translate_set): Likewise. Insert all translated
11531 exprs for a value into the set, keeping possibly multiple
11532 expressions per value.
11533 (compute_antic_aux): Adjust for phi_translate changes.
11534 When intersecting union the expressions and prune those
11535 not in the final value set, keeping possibly multiple
11536 expressions per value. Do not use value-insertion
11537 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11538 all expressions. Add verification that the value-sets
11539 only shrink during iteration.
11540 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11541 (do_pre_regular_insertion): Likewise.
11542 (do_pre_partial_partial_insertion): Likewise.
11544 2018-03-02 Richard Biener <rguenther@suse.de>
11547 * config/darwin.c (saved_debug_info_level): New static global.
11548 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11549 (darwin_asm_lto_end): Restore debug info generation settings.
11551 2018-03-01 Martin Liska <mliska@suse.cz>
11554 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11555 volatile arguments.
11557 2018-03-01 Richard Biener <rguenther@suse.de>
11560 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11561 type annotation with LTO when debug was disabled at compile-time.
11563 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11565 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11567 (mips_final_postscan_insn): Likewise.
11569 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11571 PR rtl-optimization/84528
11572 * alias.c (init_alias_target): Add commentary.
11573 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11574 a unique base value if the frame pointer is not eliminated
11575 to the stack pointer.
11577 2018-03-01 Tom de Vries <tom@codesourcery.com>
11579 PR rtl-optimization/83327
11580 * lra-int.h (hard_regs_spilled_into): Declare.
11581 * lra.c (hard_regs_spilled_into): Define.
11582 (init_reg_info): Init hard_regs_spilled_into.
11583 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11584 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11585 (process_bb_lives): Handle hard_regs_spilled_into.
11586 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11587 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11589 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11591 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11592 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11593 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11594 * config/rs6000/aix72.h: New file.
11596 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11598 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11599 instead of warning_at with conditional singular and plural messages
11603 * stor-layout.c (update_alignment_for_field): For
11604 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11605 && !DECL_PACKED (field), do the alignment update, just use
11606 only desired_align instead of MAX (type_align, desired_align)
11608 (place_field): Don't do known_align < desired_align handling
11609 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11610 is non-NULL, instead do it after rli->prev_field handling and
11611 only if not within a bitfield word. For DECL_PACKED (field)
11612 use type_align of BITS_PER_UNIT.
11614 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11616 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11617 superfluous parentheses and trailing spaces.
11619 2018-02-28 Richard Biener <rguenther@suse.de>
11621 PR tree-optimization/84584
11622 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11623 SCoPs with fake exit edge.
11625 2018-02-28 Martin Liska <mliska@suse.cz>
11628 * timevar.c (timer::print): Fix format to properly print 100%
11631 2018-02-28 Richard Biener <rguenther@suse.de>
11633 PR middle-end/84607
11634 * genmatch.c (capture_info::walk_match): Do not mark
11635 captured expressions without operands as expr_p given
11636 they act more like predicates and should be subject to
11637 "lost tail" side-effect preserving.
11639 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11641 PR rtl-optimization/81611
11642 * auto-inc-dec.c (attempt_change): Move dead note from
11643 mem_insn if it's the next use of regno
11644 (find_address): Take address use of reg holding
11645 non-incremented value. Add parm to limit search to the named
11647 (merge_in_block): Attempt to use a mem insn that is the next
11648 use of the original regno.
11650 2018-02-27 Martin Sebor <msebor@redhat.com>
11653 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11654 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11656 2018-02-27 Martin Sebor <msebor@redhat.com>
11658 PR translation/84207
11659 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11660 n argument to unsigned HOST_WIDE_INT.
11661 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11662 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11663 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11664 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11666 2018-02-27 Richard Biener <rguenther@suse.de>
11668 PR tree-optimization/84512
11669 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11670 Do not use the estimate returned from record_stmt_cost for
11671 the scalar iteration cost but sum properly using add_stmt_cost.
11673 2018-02-27 Richard Biener <rguenther@suse.de>
11675 PR tree-optimization/84466
11676 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11677 Adjust last change to less strictly validate use operands.
11679 2018-02-27 Martin Liska <mliska@suse.cz>
11681 PR gcov-profile/84548
11682 * gcov.c (process_file): Allow partial overlap and consider it
11683 also as group functions.
11684 (output_lines): Properly calculate range of lines for a group.
11686 2018-02-27 Martin Liska <mliska@suse.cz>
11688 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11689 'ggc' suffixes. Change first column width.
11690 (timer::print): Fix formatting of the column.
11692 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11694 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11695 preserve inline entry blocks for the sake of debug inline
11696 entry point markers alone.
11697 (remove_unused_locals): Suggest in comments a better place to
11698 force the preservation of inline entry blocks that are
11699 otherwise unused, but do not preserve them.
11701 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11703 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11705 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11708 * config/i386/constraints.md (Bs): Replace
11709 ix86_indirect_branch_register with
11710 TARGET_INDIRECT_BRANCH_REGISTER.
11712 * config/i386/i386.md (indirect_jump): Likewise.
11713 (tablejump): Likewise.
11714 (*sibcall_memory): Likewise.
11715 (*sibcall_value_memory): Likewise.
11716 Peepholes of indirect call and jump via memory: Likewise.
11717 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11718 (*sibcall_value_GOT_32): Likewise.
11719 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11720 (GOT_memory_operand): Likewise.
11721 (call_insn_operand): Likewise.
11722 (sibcall_insn_operand): Likewise.
11723 (GOT32_symbol_operand): Likewise.
11724 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11726 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11728 PR rtl-optimization/83496
11729 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11730 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11731 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11732 redundant insn, if any.
11733 (relax_delay_slots): Likewise.
11734 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11736 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11738 PR tree-optimization/83965
11739 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11740 that grouped statements are part of a reduction chain. Return
11741 true if the statement is not marked as a reduction itself but
11742 is part of a group.
11743 (vect_recog_dot_prod_pattern): Don't check whether the statement
11744 is part of a group here.
11745 (vect_recog_sad_pattern): Likewise.
11746 (vect_recog_widen_sum_pattern): Likewise.
11748 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11751 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11753 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11756 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11758 (ix86_output_indirect_function_return): New prototype.
11759 (ix86_split_simple_return_pop_internal): Likewise.
11760 * config/i386/i386.c (indirect_return_via_cx): New.
11761 (indirect_return_via_cx_bnd): Likewise.
11762 (indirect_thunk_name): Handle return va CX_REG.
11763 (output_indirect_thunk_function): Create alias for
11764 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11765 (ix86_output_indirect_jmp): Remove the bool argument.
11766 (ix86_output_indirect_function_return): New function.
11767 (ix86_split_simple_return_pop_internal): Likewise.
11768 * config/i386/i386.md (*indirect_jump): Don't pass false
11769 to ix86_output_indirect_jmp.
11770 (*tablejump_1): Likewise.
11771 (simple_return_pop_internal): Change it to define_insn_and_split.
11772 Call ix86_split_simple_return_pop_internal to split it for
11773 -mfunction-return=.
11774 (simple_return_indirect_internal): Call
11775 ix86_output_indirect_function_return instead of
11776 ix86_output_indirect_jmp.
11778 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11781 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11782 memset and value initialization afterwards.
11784 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11786 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11788 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11791 * common/config/aarch64/aarch64-common.c
11792 (aarch_option_optimization_table[]): Switch
11793 off fomit-frame-pointer
11795 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11796 Chung-Ju Wu <jasonwucj@gmail.com>
11798 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11800 (store_multiple): Ditto.
11802 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11804 * config.gcc: Add --with-cpu support for nds32 target.
11805 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11806 * config/nds32/nds32.opt: Add -mcpu= option.
11808 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11810 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11811 isel=yes): Warn for these deprecated options.
11813 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11815 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11816 ISA_2_5_MASKS_EMBEDDED.
11818 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11820 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11821 p->max as pointers rather than using iterative_hash_expr.
11823 2018-02-23 Carl Love <cel@us.ibm.com>
11825 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11826 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11828 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11829 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11830 P8V_BUILTIN_VEC_VUNSIGNED2.
11832 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11835 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11836 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11838 * lra-constraints.c (curr_insn_transform): Set up
11839 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11841 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11842 * lra-eliminations.c (spill_pseudos): Ditto.
11843 (process_insn_for_elimination): Ditto.
11844 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11845 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11846 LRA_NON_CLOBBERED_ALT.
11848 2018-02-22 Martin Sebor <msebor@redhat.com>
11850 PR tree-optimization/84480
11851 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11852 to maybe_diag_stxncpy_trunc. Call it.
11853 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11854 from gimple_fold_builtin_strcpy. Print inlining stack.
11855 (handle_builtin_stxncpy): Print inlining stack.
11856 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11858 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11861 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11862 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11863 and -fcheck-pointer-bounds are used together.
11864 (indirect_thunk_prefix): New enum.
11865 (indirect_thunk_need_prefix): New function.
11866 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11867 "_nt" instead of "_bnd" for NOTRACK prefix.
11868 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11869 (output_indirect_thunk_function): Likewise.
11871 (ix86_code_end): Update output_indirect_thunk_function calls.
11872 (ix86_output_indirect_branch_via_reg): Replace
11873 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11874 (ix86_output_indirect_branch_via_push): Likewise.
11875 (ix86_output_function_return): Likewise.
11876 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11877 incompatible with -fcf-protection=branch and
11878 -fcheck-pointer-bounds.
11880 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11883 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11884 Change gcc_assert call to output_operand_lossage.
11886 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11888 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11890 2018-02-22 DJ Delorie <dj@redhat.com>
11891 Sebastian Perta <sebastian.perta@renesas.com>
11892 Oleg Endo <olegendo@gcc.gnu.org>
11894 * config/rx/rx.c (rx_rtx_costs): New function.
11895 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11897 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11899 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11901 2018-02-22 Martin Liska <mliska@suse.cz>
11904 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11905 Add "native" as a possible value.
11907 2018-02-22 Martin Liska <mliska@suse.cz>
11910 * config/i386/i386.c (ix86_option_override_internal):
11911 Add "native" as a possible value for -march and -mtune.
11913 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11916 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11917 to all type variants.
11919 PR tree-optimization/84503
11920 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11921 width as info->bitpos + info->bitsize - start.
11922 (merged_store_group::merge_overlapping): Simplify width computation.
11923 (check_no_overlap): New function.
11924 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11925 start + width and last_order of the group, fail if check_no_overlap
11927 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11928 to group if check_no_overlap fails.
11930 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11932 * config/rs6000/altivec.md: Delete contraint arguments to
11933 define_expand, define_split, and define_peephole2, and in
11934 define_insn_and_split if always unused.
11935 * config/rs6000/darwin.md: Ditto.
11936 * config/rs6000/dfp.md: Ditto.
11937 * config/rs6000/rs6000.md: Ditto.
11938 * config/rs6000/sync.md: Ditto.
11939 * config/rs6000/vector.md: Ditto.
11940 * config/rs6000/vsx.md: Ditto.
11942 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11944 * config/rs6000/altivec.md: Write output control strings as braced
11945 blocks instead of double-quoted strings.
11946 * config/rs6000/darwin.md: Ditto.
11947 * config/rs6000/rs6000.md: Ditto.
11948 * config/rs6000/vector.md: Ditto.
11949 * config/rs6000/vsx.md: Ditto.
11951 2018-02-21 Jason Merrill <jason@redhat.com>
11953 PR c++/84314 - ICE with templates and fastcall attribute.
11954 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11956 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11958 * ipa-cp.c (determine_versionability): Fix comment typos.
11960 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11963 * ipa-cp.c (determine_versionability): Do not version functions caling
11966 2018-02-21 Martin Liska <mliska@suse.cz>
11969 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11970 Add "native" as a possible value.
11971 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11972 the macro when native cpu detection is available.
11974 2018-02-21 Martin Liska <mliska@suse.cz>
11977 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11978 Add "native" as a possible value.
11979 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11980 when native cpu detection is available.
11982 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11983 Martin Sebor <msebor@redhat.com>
11985 PR tree-optimization/84478
11986 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11988 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11989 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11990 support which is conservatively correct, for 2 only stay conservative
11991 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11992 argument to the 2 argument get_range_strlen, adjust 6 arg
11993 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11995 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11996 (gimple_fold_builtin_strlen): Pass true as last argument to
11999 2018-02-20 Martin Sebor <msebor@redhat.com>
12001 PR middle-end/84095
12002 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12003 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12004 (builtin_memref::builtin_memref): Factor out parts into
12005 set_base_and_offset and call it.
12007 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12009 PR middle-end/84406
12010 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12011 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12012 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12013 search at the associated MODE_INT.
12015 2018-02-20 Jeff Law <law@redhat.com>
12017 PR middle-end/82123
12018 PR tree-optimization/81592
12019 PR middle-end/79257
12020 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12021 for range data rather than using global data.
12022 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12023 range data rather than using global data.
12024 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12025 pass it to children as needed.
12026 (struct directive::fmtresult): Similarly.
12027 (struct directive::set_width): Similarly.
12028 (struct directive::set_precision): Similarly.
12029 (format_integer, format_directive, parse_directive): Similarly.
12030 (format_none): Accept unnamed vr_values parameter.
12031 (format_percent, format_floating, format_character): Similarly.
12032 (format_string, format_plain): Similarly.
12033 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12034 the EVRP range analyzer for range data rather than using global data.
12035 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12036 gimple-ssa-evrp-analyze.h
12037 (class sprintf_dom_walker): Add after_dom_children member function.
12038 Add evrp_range_analyzer member.
12039 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12040 range analyzer as needed.
12041 (sprintf_dom_walker::after_dom_children): New member function.
12042 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12044 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12045 (evrp_range_analyzer::pop_to_marker): Likewise.
12047 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12049 PR tree-optimization/84419
12050 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12051 with the required type if its current type is compatible but
12054 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12056 PR middle-end/82004
12057 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12058 after vectorization.
12060 2018-02-20 Martin Liska <mliska@suse.cz>
12063 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12064 possible values if we don't have a hint.
12066 2018-02-20 Martin Liska <mliska@suse.cz>
12070 * final.c (shorten_branches): Build align_tab array with one
12072 * opts.c (finish_options): Add alignment option limit check.
12073 (MAX_CODE_ALIGN): Likewise.
12074 (MAX_CODE_ALIGN_VALUE): Likewise.
12075 * doc/invoke.texi: Document maximum allowed option value for
12076 all -falign-* options.
12078 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12081 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12082 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12083 * var-tracking.c (emit_note_insn_var_location): Remove all references
12084 to NOTE_INSN_CALL_ARG_LOCATION.
12085 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12086 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12087 Use copy_rtx_if_shared.
12088 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12089 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12090 (dwarf2out_var_location): Remove handling of
12091 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12093 * final.c (final_scan_insn): Remove all references to
12094 NOTE_INSN_CALL_ARG_LOCATION.
12095 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12096 before dumping final insns.
12097 * except.c (emit_note_eh_region_end): Remove all references to
12098 NOTE_INSN_CALL_ARG_LOCATION.
12099 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12100 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12101 * config/arc/arc.c (hwloop_optimize): Likewise.
12102 * config/arm/arm.c (create_fix_barrier): Likewise.
12103 * config/s390/s390.c (s390_chunkify_start): Likewise.
12104 * config/sh/sh.c (find_barrier): Likewise.
12105 * config/i386/i386.c (rest_of_insert_endbranch,
12106 ix86_seh_fixup_eh_fallthru): Likewise.
12107 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12108 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12109 * config/frv/frv.c (frv_function_prologue): Likewise.
12110 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12112 (note_outside_basic_block_p): Remove all references to
12113 NOTE_INSN_CALL_ARG_LOCATION.
12114 * gengtype.c (adjust_field_rtx_def): Likewise.
12115 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12117 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12118 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12121 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12124 PR tree-optimization/84452
12125 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12126 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12129 2018-02-19 Martin Liska <mliska@suse.cz>
12132 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12134 2018-02-19 Martin Liska <mliska@suse.cz>
12135 Richard Sandiford <richard.sandiford@linaro.org>
12137 PR tree-optimization/82491
12138 * gimple-fold.c (get_base_constructor): Make earlier bail out
12141 2018-02-19 Carl Love <cel@us.ibm.com>
12143 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12144 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12146 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12147 P8V_BUILTIN_VEC_NEG.
12149 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12151 * config/rl78/rl78.md (movdf): New define expand.
12153 2018-02-19 Martin Liska <mliska@suse.cz>
12156 * doc/invoke.texi: Fix typo.
12157 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12159 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12161 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12162 handle rs6000_single_float and rs6000_double_float specially for
12165 2018-02-16 Jeff Law <law@redhat.com>
12167 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12168 (pop_regs): Use it.
12170 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12173 * ipa-inline.c (inline_small_functions): Fix a typo.
12175 2018-02-16 Nathan Sidwell <nathan@acm.org>
12177 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12179 2018-02-16 Carl Love <cel@us.ibm.com>
12181 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12182 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12183 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12184 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12185 expansion to P8V_BUILTIN_VEC_FLOAT2.
12187 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12189 PR rtl-optimization/70023
12190 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12191 src_regno into account.
12193 2018-02-16 Carl Love <cel@us.ibm.com>
12195 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12196 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12197 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12198 * config/rs6000/rs6000.c: Remove case statements for
12199 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12200 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12201 and P9V_BUILTIN_VEC_VINSERT4B.
12202 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12203 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12204 * config/rs6000/vsx.md:
12205 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12208 2018-02-16 Carl Love <cel@us.ibm.com>
12210 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12212 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12214 * config/rs6000/rs6000-c.c: Add the definitions for
12215 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12216 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12217 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12218 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12219 definition for insert4b and define insn *insert3b_internal.
12220 * doc/extend.texi: Add documentation for vec_extract4b.
12222 2018-02-16 Nathan Sidwell <nathan@acm.org>
12224 * doc/extend.texi (Backwards Compatibility): Mention friend
12225 injection. Note for-scope is deprecated.
12226 * doc/invoke.texi (-ffriend-injection): Deprecate.
12228 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12230 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12231 that moved to I2, also allow destinations that are a paradoxical
12232 subreg (instead of a normal reg).
12234 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12237 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12240 2018-02-16 Richard Biener <rguenther@suse.de>
12242 PR tree-optimization/84037
12243 PR tree-optimization/84016
12245 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12246 Adjust vec_construct for the fact we need additional higher latency
12247 128bit inserts for AVX256 and AVX512 vector builds.
12248 (ix86_add_stmt_cost): Scale vector construction cost for
12251 2018-02-16 Richard Biener <rguenther@suse.de>
12253 PR tree-optimization/84417
12254 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12255 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12256 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12258 2018-02-16 Martin Liska <mliska@suse.cz>
12261 * internal-fn.def (ASAN_CHECK): Set proper flags.
12262 (ASAN_MARK): Likewise.
12264 2018-02-16 Julia Koval <julia.koval@intel.com>
12266 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12267 from PTA_CANNONLAKE.
12269 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12272 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12273 Use ++iter rather than iter++ for std::list iterators.
12274 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12275 defer deleting them until all nodes in the forest are processed. Do
12276 free even leaf nodes. Change to_process into auto_vec.
12279 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12280 * vec.h (vec_default_construct): Use memset instead of placement new
12281 if BROKEN_VALUE_INITIALIZATION is defined.
12282 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12283 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12286 PR rtl-optimization/83723
12287 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12288 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12289 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12291 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12293 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12295 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12297 PR rtl-optimization/81443
12298 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12299 from inner REGs to paradoxical SUBREGs.
12301 2018-02-16 Richard Biener <rguenther@suse.de>
12303 PR tree-optimization/84399
12304 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12305 For operands we can analyze at their definition make sure we can
12306 analyze them at each use as well.
12308 2018-02-16 Richard Biener <rguenther@suse.de>
12310 PR tree-optimization/84190
12311 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12312 volatile accesses if the decl isn't volatile.
12314 2018-02-15 Jason Merrill <jason@redhat.com>
12316 PR c++/84314 - ICE with templates and fastcall attribute.
12317 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12318 TYPE_CANONICAL on an existing type.
12320 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12322 PR tree-optimization/84383
12323 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12324 dstoff nor call operand_equal_p if dstbase is NULL.
12326 PR tree-optimization/84334
12327 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12328 also a CONSTANT_CLASS_P, punt.
12330 2018-02-14 Jim Wilson <jimw@sifive.com>
12332 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12333 first SMALL_OPERAND check. New local min_second_step. Move assert
12334 to where locals are set. Add TARGET_RVC support.
12335 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12337 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12339 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12341 2018-02-14 Martin Sebor <msebor@redhat.com>
12343 PR tree-optimization/83698
12344 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12345 arrays constrain the offset range to their bounds.
12346 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12347 (builtin_access::overlap): Avoid setting the size of overlap if it's
12349 (maybe_diag_overlap): Also consider arrays when deciding what values
12350 of offsets to include in diagnostics.
12352 2018-02-14 Martin Sebor <msebor@redhat.com>
12355 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12356 that correspond to the kind of a declaration.
12358 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12361 * config/pa/pa.md: Load address of PIC label using the linkage table
12362 if the label is nonlocal.
12364 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12366 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12367 warning message if user requests -maltivec=be.
12368 * doc/invoke.texi: Document deprecation of -maltivec=be.
12370 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12373 * config/rs6000/rs6000-c.c: Update definitions for
12374 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12375 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12377 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12380 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12381 add _get_ssp intrinsics. Remove argument from
12382 __builtin_ia32_rdssp[d|q].
12383 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12384 * config/i386/i386-builtin.def: Remove argument from
12385 __builtin_ia32_rdssp[d|q].
12386 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12387 ix86_expand_special_args_builtin for _rdssp[d|q].
12388 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12389 Clear register before usage.
12390 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12391 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12393 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12395 PR tree-optimization/84357
12396 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12397 operand 1 of an ARRAY_REF too.
12399 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12402 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12403 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12405 (set_of_reg): New struct.
12406 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12407 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12408 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12410 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12411 Split into bitclr, bitset, bitinvert patterns if appropriate.
12412 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12413 use rx_fuse_in_memory_bitop.
12414 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12415 to named insn, correct maximum insn length.
12417 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12420 * machmode.def: Define a complex mode for PARTIAL_INT.
12421 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12423 * doc/rtl.texi: Document CSPImode.
12424 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12426 (msp430_hard_regno_nregs_with_padding): Likewise.
12428 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12431 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12433 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12435 PR rtl-optimization/84169
12436 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12437 we generated a parallel as new i3 and we split that to new i2 and i3
12438 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12439 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12440 those to i2, not i1. Partially rewrite this scan code.
12442 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12445 * stor-layout.c (place_field): For variable length fields, adjust
12446 offset_align afterwards not just based on the field's alignment,
12447 but also on the size.
12449 PR middle-end/84309
12450 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12451 of exps and logs in the use_exp2 case.
12453 2018-02-13 Jeff Law <law@redhat.com>
12455 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12456 entry for "vector".
12458 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12461 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12465 * common.opt (gas-loc-support, gas-locview-support): New.
12466 (ginline-points, ginternal-reset-location-views): New.
12467 * doc/invoke.texi: Document them. Use @itemx where intended.
12468 (gvariable-location-views): Adjust.
12469 * target.def (reset_location_view): New.
12470 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12471 (TARGET_RESET_LOCATION_VIEW): New.
12472 * doc/tm.texi: Rebuilt.
12473 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12474 (dwarf2out_default_as_locview_support): New.
12475 (output_asm_line_debug_info): Use option variables.
12476 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12477 (output_loc_list): Likewise.
12478 (add_high_low_attributes): Check option variables.
12479 Don't output entry view attribute in strict mode.
12480 (gen_inlined_subroutine_die): Check option variables.
12481 (dwarf2out_inline_entry): Likewise.
12482 (init_sections_and_labels): Likewise.
12483 (dwarf2out_early_finish): Likewise.
12484 (maybe_reset_location_view): New, from...
12485 (dwarf2out_var_location): ... here. Call it.
12486 * debug.h (dwarf2out_default_as_loc_support): Declare.
12487 (dwarf2out_default_as_locview_support): Declare.
12488 * hooks.c (hook_int_rtx_insn_0): New.
12489 * hooks.h (hook_int_rtx_insn_0): Declare.
12490 * toplev.c (process_options): Take -gas-loc-support and
12491 -gas-locview-support from dwarf2out. Enable
12492 -gvariable-location-views by default only with locview
12493 assembler support. Enable -ginternal-reset-location-views by
12494 default only if the target defines the corresponding hook.
12495 Enable -ginline-points by default if location views are
12496 enabled; force it disabled if statement frontiers are
12498 * tree-inline.c (expand_call_inline): Check option variables.
12499 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12501 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12503 PR tree-optimization/84321
12504 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12505 handling. Also check whether the anti-range contains any values
12506 that satisfy the mask; switch to a VR_RANGE if not.
12508 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12511 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12513 2018-02-13 Martin Jambor <mjambor@suse.cz>
12516 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12517 of call statements, also set location of a load to a temporary.
12519 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12521 * config/rl78/rl78.c (add_vector_labels): New function.
12522 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12523 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12524 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12525 which checks that no arguments are passed.
12526 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12527 * doc/extend.texi: Documentation for the new attribute.
12529 2018-02-13 Andreas Schwab <schwab@suse.de>
12531 * config/riscv/linux.h (CPP_SPEC): Define.
12533 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12536 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12537 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12538 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12539 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12540 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12541 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12542 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12543 temporarily for AES and PCLMUL builtins.
12545 PR tree-optimization/84339
12546 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12547 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12550 PR middle-end/84309
12551 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12552 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12553 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12555 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12557 * omp-simd-clone.h: New file.
12558 * omp-simd-clone.c: Include omp-simd-clone.h.
12559 (expand_simd_clones): No longer static.
12560 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12561 cgraph.h and omp-simd-clone.h.
12562 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12563 (vect_recog_widen_shift_pattern): Formatting fix.
12564 (vect_pattern_recog_1): Don't check optab for calls.
12567 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12568 operands[2] into a REG before using gen_lowpart on it.
12570 2018-02-12 Jeff Law <law@redhat.com>
12573 * config/sh/sh.c (find_barrier): Consider a sibling call
12576 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12577 successfully back substituting a reg.
12579 2018-02-12 Richard Biener <rguenther@suse.de>
12581 PR tree-optimization/84037
12582 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12583 parameter, move visited init to caller.
12584 (vect_slp_analyze_operations): Separate cost from validity
12585 check, initialize visited once for all instances.
12586 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12588 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12590 (vect_model_promotion_demotion_cost): Likewise.
12591 (vectorizable_bswap): Guard cost modeling with !slp_node
12592 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12594 (vectorizable_call): Likewise.
12595 (vectorizable_conversion): Likewise.
12596 (vectorizable_assignment): Likewise.
12597 (vectorizable_shift): Likewise.
12598 (vectorizable_operation): Likewise.
12599 (vectorizable_store): Likewise.
12600 (vectorizable_load): Likewise.
12601 (vectorizable_condition): Likewise.
12602 (vectorizable_comparison): Likewise.
12604 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12607 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12608 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12609 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12611 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12613 PR middle-end/83665
12614 * params.def (inline-min-speedup): Increase from 8 to 15.
12615 (max-inline-insns-auto): Decrease from 40 to 30.
12616 * ipa-split.c (consider_split): Add some buffer for function to
12617 be considered inlining candidate.
12618 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12621 2018-02-12 Richard Biener <rguenther@suse.de>
12623 PR tree-optimization/84037
12624 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12625 matched stmts if we cannot swap the non-matched ones.
12627 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12629 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12630 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12631 _mm_maskz_scalef_round_ss): New intrinsics.
12632 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12633 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12634 __builtin_ia32_scalefss_round): Remove.
12635 (__builtin_ia32_scalefsd_mask_round,
12636 __builtin_ia32_scalefss_mask_round): New intrinsics.
12637 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12638 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12639 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12640 "<round_constraint>")): Changed to ...
12641 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12642 "<round_scalar_constraint>")): ... this.
12643 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12644 %0, %1, %2<round_op3>}"): Changed to ...
12645 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12646 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12647 %2<round_scalar_mask_op3>}"): ... this.
12648 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12650 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12652 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12653 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12654 (_mm_maskz_sqrt_round_ss): New intrinsics.
12655 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12656 (__builtin_ia32_sqrtsd_mask_round)
12657 (__builtin_ia32_sqrtss_mask_round): New builtins.
12658 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12659 (__builtin_ia32_sqrtss_round): Remove.
12660 (__builtin_ia32_sqrtsd_mask_round)
12661 (__builtin_ia32_sqrtss_mask_round): New builtins.
12662 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12663 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12664 ((match_operand:VF_128 1 "vector_operand"
12665 "xBm,<round_constraint>")): Changed to ...
12666 ((match_operand:VF_128 1 "vector_operand"
12667 "xBm,<round_scalar_constraint>")): ... this.
12668 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12669 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12670 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12671 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12672 %<iptr>1<round_scalar_mask_op3>}): ... this.
12673 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12674 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12676 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12679 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12680 Cast vec_cmpeq result to correct type.
12681 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12682 Cast vec_cmpgt result to correct type.
12684 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12686 * final.c (final_scan_insn_1): Renamed from...
12687 (final_scan_insn): ... this. New wrapper, to recover
12688 seen from the outermost call in recursive ones.
12689 * config/sparc/sparc.c (output_return): Drop seen from call.
12690 (output_sibcall): Likewise.
12691 * config/visium/visium.c (output_branch): Likewise.
12693 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12695 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12698 2018-02-10 Alan Modra <amodra@gmail.com>
12701 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12702 Specify LR as an input.
12704 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12707 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12708 remove_member_access_dummy_vars): New functions.
12709 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12710 lower_omp_1, execute_lower_omp): Use them.
12712 PR rtl-optimization/84308
12713 * shrink-wrap.c (spread_components): Release todo vector.
12715 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12717 PR rtl-optimization/57193
12718 * ira-color.c (struct allocno_color_data): Add member
12719 conflict_allocno_hard_prefs.
12720 (update_conflict_allocno_hard_prefs): New.
12721 (bucket_allocno_compare_func): Add a preference based on
12722 conflict_allocno_hard_prefs.
12723 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12724 (color_allocnos): Remove a dead code. Initiate
12725 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12727 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12730 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12731 constraint from =wa to wa. Avoid a subreg on the output operand,
12732 instead use a pseudo and subreg it in a move.
12733 (p9_xxbrd_<mode>): Changed to ...
12734 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12735 (p9_xxbrd_v2df): New expander.
12736 (p9_xxbrw_<mode>): Changed to ...
12737 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12738 (p9_xxbrw_v4sf): New expander.
12740 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12742 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12744 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12747 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12748 multiply in 32-bit mode.
12749 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12750 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12753 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12755 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12756 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12757 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12758 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12760 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12763 * dwarf2out.c (is_trivial_indirect_ref): New function.
12764 (dwarf2out_late_global_decl): Do not generate a location
12765 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12766 and that are not defined in the current unit.
12768 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12770 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12771 instead of a libcall for UNORDERED.
12773 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12776 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12777 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12779 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12782 * config/s390/s390.c (s390_set_current_function): Invoke
12783 s390_indirect_branch_settings also if fndecl didn't change.
12785 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12787 * config/rs6000/rs6000.md (blockage): Set length to zero.
12789 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12791 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12793 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12796 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12797 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12801 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12802 PARALLEL incoming that failed vt_get_decl_and_offset check.
12804 PR middle-end/84237
12805 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12806 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12808 (get_variable_section): For decls in named .bss* sections pass true as
12809 second argument to bss_initializer_p.
12811 2018-02-09 Marek Polacek <polacek@redhat.com>
12812 Jakub Jelinek <jakub@redhat.com>
12815 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12816 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12817 Sync some changes from cxx_fold_indirect_ref.
12819 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12821 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12823 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12824 (BLOCK_INLINE_ENTRY_LABEL): New.
12825 (dwarf2out_var_location): Disregard inline entry markers.
12826 (inline_entry_data): New struct.
12827 (inline_entry_data_hasher): New hashtable type.
12828 (inline_entry_data_hasher::hash): New.
12829 (inline_entry_data_hasher::equal): New.
12830 (inline_entry_data_table): New variable.
12831 (add_high_low_attributes): Add DW_AT_entry_pc and
12832 DW_AT_GNU_entry_view attributes if a pending entry is found
12833 in inline_entry_data_table. Add old entry_pc attribute only
12834 if debug nonbinding markers are disabled.
12835 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12836 markers are enabled.
12837 (block_within_block_p, dwarf2out_inline_entry): New.
12838 (dwarf2out_finish): Check that no entries remained in
12839 inline_entry_data_table.
12840 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12841 (final_scan_insn, notice_source_line): Likewise.
12842 (rest_of_clean_state): Skip inline entry markers.
12843 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12845 * gimple.c (gimple_build_debug_inline_entry): New.
12846 * gimple.h (enum gimple_debug_subcode): Add
12847 GIMPLE_DEBUG_INLINE_ENTRY.
12848 (gimple_build_debug_inline_entry): Declare.
12849 (gimple_debug_inline_entry_p): New.
12850 (gimple_debug_nonbind_marker_p): Adjust.
12851 * insn-notes.def (INLINE_ENTRY): New.
12852 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12853 inline entry marker notes.
12854 (print_insn): Likewise.
12855 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12856 (INSN_DEBUG_MARKER_KIND): Likewise.
12857 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12858 * tree-inline.c (expand_call_inline): Build and insert
12859 debug_inline_entry stmt.
12860 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12861 inline entry blocks early, if nonbind markers are enabled.
12862 (dump_scope_block): Dump fragment info.
12863 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12864 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12865 (gimple_build_debug_inline_entry): New.
12866 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12867 Enable/disable inline entry points too.
12868 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12869 (DEBUG_INSN): Describe inline entry markers.
12871 * common.opt (gvariable-location-views): New.
12872 (gvariable-location-views=incompat5): New.
12873 * config.in: Rebuilt.
12874 * configure: Rebuilt.
12875 * configure.ac: Test assembler for view support.
12876 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12877 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12878 * dwarf2out.c (var_loc_view): New typedef.
12879 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12880 (dwarf2out_locviews_in_attribute): New.
12881 (dwarf2out_locviews_in_loclist): New.
12882 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12883 (enum dw_line_info_opcode): Add LI_adv_address.
12884 (struct dw_line_info_table): Add view.
12885 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12886 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12887 (zero_view_p): New variable.
12888 (ZERO_VIEW_P): New macro.
12889 (output_asm_line_debug_info): New.
12890 (struct var_loc_node): Add view.
12891 (add_AT_view_list, AT_loc_list): New.
12892 (add_var_loc_to_decl): Add view param. Test it against last.
12893 (new_loc_list): Add view params. Record them.
12894 (AT_loc_list_ptr): Handle loc and view lists.
12895 (view_list_to_loc_list_val_node): New.
12896 (print_dw_val): Handle dw_val_class_view_list.
12897 (size_of_die): Likewise.
12898 (value_format): Likewise.
12899 (loc_list_has_views): New.
12900 (gen_llsym): Set vl_symbol too.
12901 (maybe_gen_llsym, skip_loc_list_entry): New.
12902 (dwarf2out_maybe_output_loclist_view_pair): New.
12903 (output_loc_list): Output view list or entries too.
12904 (output_view_list_offset): New.
12905 (output_die): Handle dw_val_class_view_list.
12906 (output_dwarf_version): New.
12907 (output_compilation_unit_header): Use it.
12908 (output_skeleton_debug_sections): Likewise.
12909 (output_rnglists, output_line_info): Likewise.
12910 (output_pubnames, output_aranges): Update version comments.
12911 (output_one_line_info_table): Output view numbers in asm comments.
12912 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12913 Call maybe_gen_llsym.
12914 (loc_list_from_tree_1): Adjust.
12915 (add_AT_location_description): Create view list attribute if
12916 needed, check it's absent otherwise.
12917 (convert_cfa_to_fb_loc_list): Adjust.
12918 (maybe_emit_file): Call output_asm_line_debug_info for test.
12919 (dwarf2out_var_location): Reset views as needed. Precompute
12920 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12921 attribute. Set view.
12922 (new_line_info_table): Reset next view.
12923 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12924 (dwarf2out_source_line): Likewise. Output view resets and labels to
12925 the assembler, or select appropriate line info opcodes.
12926 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12927 (optimize_string_length): Catch it. Adjust.
12928 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12929 dw_val_class_view_list, and remove it if no longer needed.
12930 (hash_loc_list): Hash view numbers.
12931 (loc_list_hasher::equal): Compare them.
12932 (optimize_location_lists): Check whether a view list symbol is
12933 needed, and whether the locview attribute is present, and
12934 whether they match. Remove the locview attribute if no longer
12936 (index_location_lists): Call skip_loc_list_entry for test.
12937 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12938 Use output_dwarf_version.
12939 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12940 (struct dw_val_node): Add val_view_list.
12941 * final.c (SEEN_NEXT_VIEW): New.
12942 (set_next_view_needed): New.
12943 (clear_next_view_needed): New.
12944 (maybe_output_next_view): New.
12945 (final_start_function): Rename to...
12946 (final_start_function_1): ... this. Take pointer to FIRST,
12947 add SEEN parameter. Emit param bindings in the initial view.
12948 (final_start_function): Reintroduce SEEN-less interface.
12949 (final): Rename to...
12950 (final_1): ... this. Take SEEN parameter. Output final pending
12951 next view at the end.
12952 (final): Reintroduce seen-less interface.
12953 (final_scan_insn): Output pending next view before switching
12954 sections or ending a block. Mark the next view as needed when
12955 outputting variable locations. Notify debug backend of section
12956 changes, and of location view changes.
12957 (rest_of_handle_final): Adjust.
12958 * toplev.c (process_options): Autodetect value for debug variable
12959 location views option. Warn on incompat5 without -gdwarf-5.
12960 * doc/invoke.texi (gvariable-location-views): New.
12961 (gvariable-location-views=incompat5): New.
12962 (gno-variable-location-views): New.
12964 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12966 PR tree-optimization/84136
12967 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12968 that the result of find_edge is non-NULL.
12970 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12973 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12974 storing integer register in SImode. Fix cost of 256 and 512
12975 byte aligned SSE register store.
12977 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12979 * config/i386/i386.c (ix86_multiplication_cost): Fix
12980 multiplication cost for TARGET_AVX512DQ.
12982 2018-02-08 Marek Polacek <polacek@redhat.com>
12984 PR tree-optimization/84238
12985 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12988 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12990 PR tree-optimization/84265
12991 * tree-vect-stmts.c (vectorizable_store): Don't treat
12992 VMAT_CONTIGUOUS accesses as grouped.
12993 (vectorizable_load): Likewise.
12995 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12997 PR tree-optimization/81635
12998 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12999 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13000 (test_round_for_mask): New functions.
13001 (wide_int_cc_tests): Call test_round_for_mask.
13002 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13003 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13004 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13005 range returned by get_range_info.
13007 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13010 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13011 * symtab.c: Include builtins.h
13012 (symtab_node::output_to_lto_symbol_table_p): Move here
13013 from lto-streamer-out.c:output_symbol_p.
13014 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13015 (output_symbol_p): Move all logic to symtab.c
13016 (produce_symtab): Update.
13018 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13020 * config/s390/s390-opts.h (enum indirect_branch): Define.
13021 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13022 (s390_indirect_branch_via_thunk)
13023 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13024 (enum s390_indirect_branch_type): Define.
13025 * config/s390/s390.c (struct s390_frame_layout, struct
13026 machine_function): Remove.
13027 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13028 (indirect_branch_table_label_no, indirect_branch_table_name):
13030 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13031 (enum s390_indirect_branch_option): Define.
13032 (s390_return_addr_from_memory): New function.
13033 (s390_handle_string_attribute): New function.
13034 (s390_attribute_table): Add new attribute handler.
13035 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13036 (s390_indirect_branch_via_thunk): New function.
13037 (s390_indirect_branch_via_inline_thunk): New function.
13038 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13039 sibling call optimization for non z10 compiles.
13040 (s390_emit_call): Force indirect branch target to be a single
13041 register. Add r1 clobber for non-z10 compiles.
13042 (s390_emit_epilogue): Emit return jump via return_use expander.
13043 (s390_reorg): Handle JUMP_INSNs as execute targets.
13044 (s390_option_override_internal): Perform validity checks for the
13045 new command line options.
13046 (s390_indirect_branch_attrvalue): New function.
13047 (s390_indirect_branch_settings): New function.
13048 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13049 (s390_output_indirect_thunk_function): New function.
13050 (s390_code_end): Implement target hook.
13051 (s390_case_values_threshold): Implement target hook.
13052 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13054 * config/s390/s390.h (struct s390_frame_layout)
13055 (struct machine_function): Move here from s390.c.
13056 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13057 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13058 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13059 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13060 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13061 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13062 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13063 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13064 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13065 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13066 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13067 (mnemonic attribute): Add values which aren't recognized
13069 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13070 pattern for branch conversion. Fix mnemonic attribute.
13071 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13072 indirect branch via thunk if requested.
13073 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13074 ("*indirect_jump"): Disable for branch conversion using out of
13076 ("indirect_jump_via_thunk<mode>_z10")
13077 ("indirect_jump_via_thunk<mode>")
13078 ("indirect_jump_via_inlinethunk<mode>_z10")
13079 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13080 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13081 ("casesi_jump_via_inlinethunk<mode>_z10")
13082 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13083 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13084 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13085 ("*indirect2_jump"): Disable for branch conversion.
13086 ("casesi_jump"): Turn into expander and expand patterns for branch
13088 ("return_use"): New expander.
13089 ("*return"): Emit return via thunk and rename it to ...
13090 ("*return<mode>"): ... this one.
13091 * config/s390/s390.opt: Add new options and and enum for the
13094 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13096 * lra-constraints.c (match_reload): Unconditionally use
13097 gen_lowpart_SUBREG, rather than selecting between that
13098 and equivalent gen_rtx_SUBREG code.
13100 2018-02-08 Richard Biener <rguenther@suse.de>
13102 PR tree-optimization/84233
13103 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13104 changed flag instead of boguously re-using phi_inserted.
13106 2018-02-08 Martin Jambor <mjambor@suse.cz>
13108 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13109 static local variables.
13111 2018-02-08 Richard Biener <rguenther@suse.de>
13113 PR tree-optimization/84278
13114 * tree-vect-stmts.c (vectorizable_store): When looking for
13115 smaller vector types to perform grouped strided loads/stores
13116 make sure the mode is supported by the target.
13117 (vectorizable_load): Likewise.
13119 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13121 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13122 Increase LDP/STP opportunities by adding adjacent callee-saves.
13124 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13126 PR rtl-optimization/84068
13127 PR rtl-optimization/83459
13128 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13130 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13132 PR tree-optimization/84224
13133 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13134 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13135 non-zero arguments.
13137 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13140 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13141 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13142 count, remove one USE.
13144 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13146 * doc/install.texi (Configuration): Document the
13147 --with-long-double-format={ibm,ieee} PowerPC configuration
13151 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13152 Convert from define_expand to be define_insn_and_split. Rework
13153 float/double/_Float128 conversions to QI/HI/SImode to work with
13154 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13155 conversions to QI/HImode types did a store and then a load to
13156 truncate the value. For conversions to VSX registers, don't split
13157 the insn, instead emit the code directly. Use the code iterator
13158 any_fix to combine signed and unsigned conversions.
13159 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13160 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13161 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13162 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13163 (fix_<mode>di2_hw): Likewise.
13164 (fixuns_<mode>di2_hw): Likewise.
13165 (fix_<mode>si2_hw): Likewise.
13166 (fixuns_<mode>si2_hw): Likewise.
13167 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13168 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13169 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13170 fix<uns>_trunc<SFDF:mode>si2_p8.
13171 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13172 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13173 (fix<uns>_<mode>_mem): Likewise.
13174 (fctiw<u>z_<mode>_mem): Likewise.
13175 (fix<uns>_<mode>_mem): Likewise.
13176 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13177 the register allocator from doing a direct move to the GPRs to do
13178 a store, and instead use the ISA 3.0 store byte/half-word from
13179 vector register instruction. For IEEE 128-bit floating point,
13180 also optimize stores of 32-bit ints.
13181 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13183 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13185 * genextract.c (push_pathstr_operand): New function to support
13187 (walk_rtx): Call push_pathstr_operand.
13188 (print_path): Support [a-zA-Z].
13190 2018-02-07 Richard Biener <rguenther@suse.de>
13192 PR tree-optimization/84037
13193 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13194 (cse_and_gimplify_to_preheader): Declare.
13195 (vect_get_place_in_interleaving_chain): Likewise.
13196 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13198 (_loop_vec_info::~_loop_vec_info): Delete it.
13199 (cse_and_gimplify_to_preheader): New function.
13200 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13201 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13202 (vectorizable_load): Likewise. For grouped stores always base
13203 the IV on the first element.
13204 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13205 condition before gimplifying.
13207 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13209 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13210 *DIV_EXPR and *MOD_EXPR.
13212 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13215 * config/i386/i386.c (ix86_option_override_internal): Mask out
13216 the CF_SET bit when checking -fcf-protection.
13218 2018-02-07 Tom de Vries <tom@codesourcery.com>
13221 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13224 2018-02-07 Richard Biener <rguenther@suse.de>
13226 PR tree-optimization/84204
13227 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13230 PR tree-optimization/84205
13231 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13232 special-case isl_ast_op_zdiv_r.
13234 PR tree-optimization/84223
13235 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13236 Only add conditions from within the region.
13237 (gather_bbs::after_dom_children): Adjust.
13239 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13242 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13243 * config/avr/avr.md: Only post-reload split REG-REG moves if
13244 either register is GENERAL_REG_P.
13246 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13248 PR tree-optimization/84235
13249 * tree-ssa-scopedtables.c
13250 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13251 if the subtraction is performed in floating point type where NaNs are
13252 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13253 build 1. Formatting fix.
13255 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13258 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13259 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13260 and skip it regardless of bb boundaries. Use CALL_P macro,
13261 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13262 unnecessarily, formatting fix.
13264 2018-02-06 Michael Collison <michael.collison@arm.com>
13266 * config/arm/thumb2.md:
13267 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13268 (*thumb_mov_notscc): Ditto.
13270 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13273 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13274 unsigned_fix, not "s".
13276 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13278 * configure.ac (gcc_fn_eh_frame_ro): New function.
13279 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13280 correct .eh_frame permissions.
13281 * configure: Regenerate.
13283 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13285 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13286 irrelevant options.
13288 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13290 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13291 Display warning message for -mno-speculate-indirect-jumps.
13293 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13295 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13297 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13299 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13301 PR tree-optimization/84225
13302 * tree-eh.c (find_trapping_overflow): Only call
13303 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13305 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13308 * config/i386/i386.c: Reimplement the check of possible options
13309 -mibt/-mshstk conbination. Change error messages.
13310 * doc/invoke.texi: Fix a typo: remove extra '='.
13312 2018-02-06 Marek Polacek <polacek@redhat.com>
13314 PR tree-optimization/84228
13315 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13317 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13320 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13321 emitted arch directives.
13322 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13323 __ARM_FEATURE_COPROC before changing architectures.
13325 2018-02-06 Richard Biener <rguenther@suse.de>
13327 * config/i386/i386.c (print_reg): Fix typo.
13328 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13330 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13332 * configure: Regenerate.
13334 2018-02-05 Martin Sebor <msebor@redhat.com>
13336 PR tree-optimization/83369
13337 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13340 2018-02-05 Martin Liska <mliska@suse.cz>
13342 * doc/invoke.texi: Cherry-pick upstream r323995.
13344 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13346 * ira.c (ira_init_register_move_cost): Adjust comment.
13348 2018-02-05 Martin Liska <mliska@suse.cz>
13350 PR gcov-profile/84137
13351 * doc/gcov.texi: Fix typo in documentation.
13353 2018-02-05 Martin Liska <mliska@suse.cz>
13355 PR gcov-profile/83879
13356 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13357 using dlopen functionality.
13359 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13361 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13362 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13363 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13364 _mm_maskz_range_round_ss): New intrinsics.
13365 (__builtin_ia32_rangesd128_round)
13366 (__builtin_ia32_rangess128_round): Remove.
13367 (__builtin_ia32_rangesd128_mask_round,
13368 __builtin_ia32_rangess128_mask_round): New builtins.
13369 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13370 __builtin_ia32_rangess128_round): Remove.
13371 (__builtin_ia32_rangesd128_mask_round,
13372 __builtin_ia32_rangess128_mask_round): New builtins.
13373 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13374 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13375 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13376 "<round_saeonly_constraint>")): Changed to ...
13377 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13378 "<round_saeonly_scalar_constraint>")): ... this.
13379 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13380 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13381 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13382 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13383 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13385 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13387 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13389 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13390 Remove all values except native, 8540 and 8548.
13392 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13394 * config/i386/i386.c (ix86_output_function_return): Pass
13395 INVALID_REGNUM, instead of -1, as invalid register number to
13396 indirect_thunk_name and output_indirect_thunk.
13398 2018-02-02 Julia Koval <julia.koval@intel.com>
13400 * config.gcc: Add -march=icelake.
13401 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13402 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13403 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13404 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13405 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13406 (processor_target_table): Add icelake.
13407 (ix86_option_override_internal): Handle new PTAs.
13408 (get_builtin_code_for_version): Handle icelake.
13409 (M_INTEL_COREI7_ICELAKE): New.
13410 (fold_builtin_cpu): Handle icelake.
13411 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13412 * doc/invoke.texi: Add -march=icelake.
13414 2018-02-02 Julia Koval <julia.koval@intel.com>
13416 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13417 to wide_int_bitmask.
13418 * wide-int-bitmask.h: New.
13420 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13423 * config/i386/i386.md: Replace Pmode with word_mode in
13424 builtin_setjmp_setup and builtin_longjmp to support x32.
13426 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13430 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13432 (rs6000_supported_cpu_names): New static variable.
13433 (linux_cpu_translation_table): Likewise.
13434 (elf_platform) <cpu>: Define new static variable and use it.
13435 Translate kernel AT_PLATFORM name to canonical name if needed.
13436 Error if platform name is unknown.
13438 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13441 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13443 2018-02-01 Jeff Law <law@redhat.com>
13446 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13447 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13448 the scratch if RELEASE_VIA_POP is false.
13449 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13450 If we have to save a temporary register, decrement SIZE appropriately.
13451 Pass new arguments to release_scratch_register_on_entry.
13452 (ix86_adjust_stack_and_probe): Likewise.
13453 (ix86_emit_probe_stack_range): Pass new arguments to
13454 release_scratch_register_on_entry.
13456 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13458 PR rtl-optimization/84157
13459 * combine.c (change_zero_ext): Use REG_P predicate in
13460 front of HARD_REGISTER_P predicate.
13462 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13464 * config/avr/avr.c (avr_option_override): Move disabling of
13465 -fdelete-null-pointer-checks to...
13466 * common/config/avr/avr-common.c (avr_option_optimization_table):
13469 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13471 PR tree-optimization/81635
13472 * tree-data-ref.c (split_constant_offset_1): For types that
13473 wrap on overflow, try to use range info to prove that wrapping
13476 2018-02-01 Renlin Li <renlin.li@arm.com>
13479 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13480 TAILCALL_ADDR_REGS.
13481 (aarch64_register_move_cost): Likewise.
13482 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13483 TAILCALL_ADDR_REGS.
13484 (REG_CLASS_NAMES): Likewise.
13485 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13486 TAILCALL_ADDR_REGS. Remove IP registers.
13487 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13489 2018-02-01 Richard Biener <rguenther@suse.de>
13491 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13492 for specifying RPO order and allow NULL for that.
13493 * domwalk.c (dom_walker::dom_walker): Likewise.
13494 (dom_walker::walk): Handle NULL RPO order.
13495 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13497 (rewrite_update_dom_walker): Likewise.
13498 (mark_def_dom_walker): Likewise.
13500 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13502 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13503 (aarch64_maybe_expand_sve_subreg_move): Declare.
13504 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13505 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13507 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13508 that are semantically a reverse operation.
13509 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13510 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13511 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13513 (aarch64_can_change_mode_class): For big-endian, forbid changes
13514 between two SVE modes if they have different element sizes.
13516 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13518 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13519 the TImode handling for big-endian targets.
13521 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13523 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13524 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13526 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13527 Remove BSWAP handing for big-endian targets and use the form of
13528 LD1RQ appropariate for the mode.
13530 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13532 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13533 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13534 duplicated element.
13536 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13539 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13540 check for operands that need to go through aarch64_sve_reload_be.
13542 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13544 PR tree-optimization/81661
13545 PR tree-optimization/84117
13546 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13547 * tree-eh.c: Include gimplify.h.
13548 (find_trapping_overflow, replace_trapping_overflow,
13549 rewrite_to_non_trapping_overflow): New functions.
13550 * tree-vect-loop.c: Include tree-eh.h.
13551 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13552 * tree-data-ref.c: Include tree-eh.h.
13553 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13555 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13557 PR rtl-optimization/84123
13558 * combine.c (change_zero_ext): Check if hard register satisfies
13559 can_change_dest_mode before calling gen_lowpart_SUBREG.
13561 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13564 * ira.c (ira_init_register_move_cost): Remove assert.
13566 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13568 PR rtl-optimization/84071
13569 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13570 * doc/tm.texi: Regenerate.
13572 2018-01-31 Richard Biener <rguenther@suse.de>
13574 PR tree-optimization/84132
13575 * tree-data-ref.c (analyze_miv_subscript): Properly
13576 check whether evolution_function_is_affine_multivariate_p
13577 before calling gcd_of_steps_may_divide_p.
13579 2018-01-31 Julia Koval <julia.koval@intel.com>
13582 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13583 * config/i386/i386.md (rdpid_rex64) New.
13584 (rdpid): Make 32bit only.
13586 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13589 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13590 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13592 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13595 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13597 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13599 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13601 PR rtl-optimization/84071
13602 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13603 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13605 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13607 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13608 (arc_attribute_table): Add 'aux' attribute.
13609 (arc_in_small_data_p): Consider aux like variables.
13610 (arc_is_aux_reg_p): New function.
13611 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13612 (arc_get_aux_arg): New function.
13613 (prepare_move_operands): Handle aux-register access.
13614 (arc_handle_aux_attribute): New function.
13615 * doc/extend.texi (ARC Variable attributes): Add subsection.
13617 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13619 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13620 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13621 (arc_attribute_table): Add 'uncached' attribute.
13622 (arc_print_operand): Print '.di' flag for uncached memory
13624 (arc_in_small_data_p): Do not consider for small data the uncached
13626 (arc_is_uncached_mem_p): New function.
13627 * config/arc/predicates.md (compact_store_memory_operand): Check
13628 for uncached memory accesses.
13629 (nonvol_nonimm_operand): Likewise.
13630 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13632 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13635 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13636 falign-loops=): Add Optimization flag.
13638 2018-01-30 Jeff Law <law@redhat.com>
13641 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13642 INT_REGISTERS_SAVED. Check it prior to calling
13643 get_scratch_register_on_entry.
13644 (ix86_adjust_stack_and_probe): Similarly.
13645 (ix86_emit_probe_stack_range): Similarly.
13646 (ix86_expand_prologue): Corresponding changes.
13648 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13651 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13652 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13654 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13657 * lra-constraints.c (curr_insn_transform): Process AND in the
13660 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13662 PR rtl-optimization/83986
13663 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13664 dependence against last_pending_memory_flush in addition to
13665 pending_jump_insns.
13667 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13669 PR tree-optimization/81611
13670 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13673 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13676 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13679 2018-01-30 Richard Biener <rguenther@suse.de>
13680 Jakub Jelinek <jakub@redhat.com>
13682 PR tree-optimization/84111
13683 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13684 inner loops added during recursion, as they don't have up-to-date
13687 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13690 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13691 (can_inline_edge_by_limits_p): ... here.
13692 (can_early_inline_edge_p, check_callers,
13693 update_caller_keys, update_callee_keys, recursive_inlining,
13694 add_new_edges_to_heap, speculation_useful_p,
13695 inline_small_functions,
13696 inline_small_functions, flatten_function,
13697 inline_to_all_callers_1): Update.
13699 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13701 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13704 2018-01-30 Richard Biener <rguenther@suse.de>
13706 PR tree-optimization/83008
13707 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13708 invariant and constant vector uses in stmts when they need
13709 more than one stmt.
13711 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13714 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13715 * configure: Regenerate.
13717 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13719 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13721 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13722 Use gen_rtx_REG rather than gen_lowpart.
13724 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13726 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13727 rather than 0 when creating partial subregs.
13729 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13731 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13734 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13737 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13738 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13739 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13740 flags. This restores the settings used before the 2017-07-24.
13741 Turning off pre increment/decrement/modify allows IVOPTS to
13742 optimize DF/SF loops where the index is an int.
13744 2018-01-29 Richard Biener <rguenther@suse.de>
13745 Kelvin Nilsen <kelvin@gcc.gnu.org>
13748 * tree-vect-stmts.c (vectorizable_call): Don't call
13749 targetm.vectorize_builtin_md_vectorized_function if callee is
13752 2018-01-22 Carl Love <cel@us.ibm.com>
13754 * doc/extend.tex: Fix typo in second arg in
13755 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13757 2018-01-29 Richard Biener <rguenther@suse.de>
13759 PR tree-optimization/84086
13760 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13761 (flush_ssaname_freelist): When SSA names were released reset
13762 the SCEV hash table.
13764 2018-01-29 Richard Biener <rguenther@suse.de>
13766 PR tree-optimization/84057
13767 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13768 removed paths when removing edges.
13770 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13772 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13773 -mfunction-return=@var{choice}.
13775 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13777 PR diagnostic/84034
13778 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13779 Handle CR like TAB.
13780 (layout::print_source_line): Likewise.
13781 (test_get_line_width_without_trailing_whitespace): Add test cases.
13783 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13785 PR middle-end/84040
13786 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13789 2018-01-26 Jim Wilson <jimw@sifive.com>
13791 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13793 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13796 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13798 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13799 and CMP + SUB-immediate -> SUBS.
13801 2018-01-26 Martin Sebor <msebor@redhat.com>
13803 PR tree-optimization/83896
13804 * tree-ssa-strlen.c (get_string_len): Rename...
13805 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13806 Avoid assuming length is constant.
13807 (handle_char_store): Use HOST_WIDE_INT for string length.
13809 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13812 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13813 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13815 2018-01-26 Richard Biener <rguenther@suse.de>
13817 PR rtl-optimization/84003
13818 * dse.c (record_store): Only record redundant stores when
13819 the earlier store aliases at least all accesses the later one does.
13821 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13823 PR rtl-optimization/83985
13824 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13825 REG_CFA_RESTORE insns.
13826 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13827 REG_CFA_RESTORE insns here.
13830 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13831 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13833 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13835 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13836 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13837 (arc_init): Likewise.
13838 (arc_override_options): Likewise.
13839 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13841 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13843 * config/arc/arc.h (TARGET_DBNZ): Define.
13844 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13845 properly set the tune attribute.
13846 (dbnz): Use TARGET_DBNZ guard.
13847 * config/arc/arc.opt (mtune): Add core3 option.
13849 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13851 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13852 recognize new pic like addresses.
13853 (arc_delegitimize_address): Clean up.
13855 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13857 * config/arc/arc-arches.def: Option mrf16 valid for all
13859 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13860 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13861 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13862 * config/arc/arc-tables.opt: Regenerate.
13863 * config/arc/arc.c (arc_conditional_register_usage): Handle
13864 reduced register file case.
13865 (arc_file_start): Set must have build attributes.
13866 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13867 mrf16 option value.
13868 * config/arc/arc.opt (mrf16): Add new option.
13869 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13870 * config/arc/genmultilib.awk: Handle new mrf16 option.
13871 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13872 * config/arc/t-multilib: Regenerate.
13873 * doc/invoke.texi (ARC Options): Document mrf16 option.
13875 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13877 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13878 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13879 (arc_attribute_table): Add 'secure_call' attribute.
13880 (arc_print_operand): Print secure call operand.
13881 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13883 (arc_is_secure_call_p): New function. * config/arc/arc.md
13884 (call_i): Add support for sjli instruction.
13885 (call_value_i): Likewise.
13886 * config/arc/constraints.md (Csc): New constraint.
13888 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13889 John Eric Martin <John.Martin@emmicro-us.com>
13891 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13892 * config/arc/arc.c (_arc_jli_section): New struct.
13893 (arc_jli_section): New type.
13894 (rc_jli_sections): New static variable.
13895 (arc_handle_jli_attribute): New function.
13896 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13897 (arc_file_end): New function.
13898 (TARGET_ASM_FILE_END): Define.
13899 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13900 (arc_add_jli_section): New function.
13901 (jli_call_scan): Likewise.
13902 (arc_reorg): Call jli_call_scan.
13903 (arc_output_addsi): Remove 'S' from printing asm operand.
13904 (arc_is_jli_call_p): New function.
13905 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13907 (movhi_insn): Likewise.
13908 (movsi_insn): Likewise.
13909 (movsi_set_cc_insn): Likewise.
13910 (loadqi_update): Likewise.
13911 (load_zeroextendqisi_update): Likewise.
13912 (load_signextendqisi_update): Likewise.
13913 (loadhi_update): Likewise.
13914 (load_zeroextendhisi_update): Likewise.
13915 (load_signextendhisi_update): Likewise.
13916 (loadsi_update): Likewise.
13917 (loadsf_update): Likewise.
13918 (movsicc_insn): Likewise.
13919 (bset_insn): Likewise.
13920 (bxor_insn): Likewise.
13921 (bclr_insn): Likewise.
13922 (bmsk_insn): Likewise.
13923 (bicsi3_insn): Likewise.
13924 (cmpsi_cc_c_insn): Likewise.
13925 (movsi_ne): Likewise.
13926 (movsi_cond_exec): Likewise.
13927 (clrsbsi2): Likewise.
13928 (norm_f): Likewise.
13938 (call_i): Remove 'S' asm letter, add jli instruction.
13939 (call_value_i): Likewise.
13940 * config/arc/arc.op (mjli-always): New option.
13941 * config/arc/constraints.md (Cji): New constraint.
13942 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13944 (subsf3_fpx): Likewise.
13945 (mulsf3_fpx): Likewise.
13946 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13948 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13949 function attrbutes.
13950 * doc/invoke.texi (ARC): Document mjli-always option.
13952 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13954 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13955 avoid addition with 0 and use incw and decw where possible.
13957 2018-01-26 Richard Biener <rguenther@suse.de>
13959 PR tree-optimization/81082
13960 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13961 association if it requires casting to unsigned.
13962 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13963 from fold_plusminus_mult_expr to catch important cases late when
13964 range info is available.
13966 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13968 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13969 * configure.ac (hidden_linkonce): New test.
13970 * configure: Regenerate.
13971 * config.in: Regenerate.
13973 2018-01-26 Julia Koval <julia.koval@intel.com>
13975 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13976 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13977 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13978 _mm_mask_bitshuffle_epi64_mask): Fix type.
13979 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13980 USI_FTYPE_V4DI_V4DI_USI): Remove.
13981 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13982 __builtin_ia32_vpshufbitqmb256_mask,
13983 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13984 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13985 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13987 2018-01-26 Alan Modra <amodra@gmail.com>
13990 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13991 UNSPEC_VBPERMQ. Sort other unspecs.
13993 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13995 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13997 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13999 PR middle-end/83055
14000 * predict.c (drop_profile): Do not push/pop cfun; update also
14002 (handle_missing_profiles): Fix logic looking for zero profiles.
14004 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14006 PR middle-end/83977
14007 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14008 on functions with #pragma omp declare simd or functions with simd
14010 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14011 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14012 Remove trailing \n from warning_at calls.
14014 2018-01-25 Tom de Vries <tom@codesourcery.com>
14017 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14018 for neutered workers.
14020 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14023 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14024 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14026 2018-01-24 Jeff Law <law@redhat.com>
14029 * i386.c (get_probe_interval): Move to earlier point.
14030 (ix86_compute_frame_layout): If -fstack-clash-protection and
14031 the frame is larger than the probe interval, then use pushes
14032 to save registers rather than reg->mem moves.
14033 (ix86_expand_prologue): Remove conditional for int_registers_saved
14036 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14039 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14040 min/max for never referenced object.
14042 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14044 PR middle-end/83977
14045 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14047 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14048 attributes from DECL_ATTRIBUTES (decl) without affecting
14049 DECL_ATTRIBUTES (current_function_decl).
14050 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14051 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14053 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14055 PR tree-optimization/83979
14056 * fold-const.c (fold_comparison): Use constant_boolean_node
14057 instead of boolean_{true,false}_node.
14059 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14061 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14064 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14066 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14067 Simplify the clause that sets the length attribute.
14068 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14069 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14070 clause that sets the length attribute.
14071 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14073 2018-01-24 Tom de Vries <tom@codesourcery.com>
14076 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14077 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14078 Add strict parameter.
14079 (prevent_branch_around_nothing): Insert dummy insn between branch to
14080 label and label with no ptx insn inbetween.
14081 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14083 2018-01-24 Tom de Vries <tom@codesourcery.com>
14086 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14087 for neutered threads in warp.
14088 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14090 2018-01-24 Richard Biener <rguenther@suse.de>
14092 PR tree-optimization/83176
14093 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14096 2018-01-24 Richard Biener <rguenther@suse.de>
14098 PR tree-optimization/82819
14099 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14100 code generating pluses that are no-ops in the target precision.
14102 2018-01-24 Richard Biener <rguenther@suse.de>
14104 PR middle-end/84000
14105 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14107 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14109 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14110 to merge probabilities.
14111 * predict.c (probably_never_executed): Also mark as cold functions
14112 with global 0 profile and guessed local profile.
14113 * profile-count.c (profile_probability::combine_with_count): New
14115 * profile-count.h (profile_probability::operator*,
14116 profile_probability::operator*=, profile_probability::operator/,
14117 profile_probability::operator/=): Reduce precision to adjusted
14118 and set value to guessed on contradictory divisions.
14119 (profile_probability::combine_with_freq): Remove.
14120 (profile_probability::combine_wiht_count): Declare.
14121 (profile_count::force_nonzero):: Set to adjusted.
14122 (profile_count::probability_in):: Set quality to adjusted.
14123 * tree-ssa-tail-merge.c (replace_block_by): Use
14124 combine_with_count.
14126 2018-01-23 Andrew Waterman <andrew@sifive.com>
14127 Jim Wilson <jimw@sifive.com>
14129 * config/riscv/riscv.c (riscv_stack_boundary): New.
14130 (riscv_option_override): Set riscv_stack_boundary. Handle
14131 riscv_preferred_stack_boundary_arg.
14132 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14133 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14134 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14135 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14136 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14137 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14139 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14142 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14143 of struct ix86_frame.
14144 (ix86_expand_epilogue): Likewise. Add a local variable for
14145 the reg_save_offset field in struct ix86_frame.
14147 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14149 PR tree-optimization/82604
14150 * tree-loop-distribution.c (enum partition_kind): New enum item
14151 PKIND_PARTIAL_MEMSET.
14152 (partition_builtin_p): Support above new enum item.
14153 (generate_code_for_partition): Ditto.
14154 (compute_access_range): Differentiate cases that equality can be
14155 proven at all loops, the innermost loops or no loops.
14156 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14157 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14158 (finalize_partitions, distribute_loop): Don't fuse partition of
14159 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14160 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14161 parloop is enabled.
14163 2018-01-23 Martin Liska <mliska@suse.cz>
14165 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14166 order to ignore the predictor.
14167 (PRED_POLYMORPHIC_CALL): Likewise.
14168 (PRED_RECURSIVE_CALL): Likewise.
14170 2018-01-23 Martin Liska <mliska@suse.cz>
14172 * tree-profile.c (tree_profiling): Print function header to
14173 aware reader which function we are working on.
14174 * value-prof.c (gimple_find_values_to_profile): Do not print
14175 not interesting value histograms.
14177 2018-01-23 Martin Liska <mliska@suse.cz>
14179 * profile-count.h (enum profile_quality): Add
14180 profile_uninitialized as the first value. Do not number values
14181 as they are zero based.
14182 (profile_count::verify): Update sanity check.
14183 (profile_probability::verify): Likewise.
14185 2018-01-23 Nathan Sidwell <nathan@acm.org>
14187 * doc/invoke.texi (ffor-scope): Deprecate.
14189 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14191 PR tree-optimization/83510
14192 * domwalk.c (set_all_edges_as_executable): New function.
14193 (dom_walker::dom_walker): Convert bool param
14194 "skip_unreachable_blocks" to enum reachability. Move setup of
14195 edge flags to set_all_edges_as_executable and only do it when
14196 reachability is REACHABLE_BLOCKS.
14197 * domwalk.h (enum dom_walker::reachability): New enum.
14198 (dom_walker::dom_walker): Convert bool param
14199 "skip_unreachable_blocks" to enum reachability.
14200 (set_all_edges_as_executable): New decl.
14201 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14202 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14204 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14205 but converting true to REACHABLE_BLOCKS.
14206 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14208 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14209 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14210 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14212 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14213 if check_all_array_refs will be called.
14215 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14217 * tree.c (selftest::test_location_wrappers): Add more test
14220 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14222 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14223 (selftest::test_bit_in_range): Likewise.
14225 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14228 * doc/sourcebuild.texi (vect_float): Say that the selector
14229 only describes the situation when -funsafe-math-optimizations is on.
14230 (vect_float_strict): Document.
14232 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14234 PR tree-optimization/83965
14235 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14236 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14237 instead of checking only for a reduction.
14238 (vect_recog_widen_sum_pattern): Likewise.
14240 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14242 * predict.c (probably_never_executed): Only use precise profile info.
14243 (compute_function_frequency): Skip after inlining hack since we now
14244 have quality checking.
14246 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14248 * profile-count.h (profile_probability::very_unlikely,
14249 profile_probability::unlikely, profile_probability::even): Set
14250 precision to guessed.
14252 2018-01-23 Richard Biener <rguenther@suse.de>
14254 PR tree-optimization/83963
14255 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14256 Properly terminate dominator walk when crossing the exit edge not
14257 when visiting its source block.
14259 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14262 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14263 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14265 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14267 PR tree-optimization/83957
14268 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14269 semicolon after for body surrounded by braces.
14271 PR tree-optimization/83081
14272 * profile-count.h (profile_probability::split): New method.
14273 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14274 Use profile_probability::split.
14275 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14276 when splitting a single conditional jump into 2.
14278 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14280 PR tree-optimization/69452
14281 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14284 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14286 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14287 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14288 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14290 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14292 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14294 * config/rl78/rl78.md (movdi): New define_expand.
14295 * config/rl78/rl78.c (rl78_split_movdi): New function.
14297 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14300 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14302 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14303 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14304 128-bit to produce an UNSPEC move to get the double word with the
14305 signbit and then a shift directly to do signbit.
14306 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14307 implementation with a new version that just does either a direct
14308 move or a regular move. Move memory interface to separate insns.
14309 Move insns so they are next to the expander.
14310 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14311 with signbit move. Split big and little endian case.
14312 (signbit<mode>2_dm_mem_le): Likewise.
14313 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14314 (signbit<mode>2_dm2): Likewise.
14316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14318 * config/rl78/rl78.md (anddi3): New define_expand.
14320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14322 * config/rl78/rl78.md (umindi3): New define_expand.
14324 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14326 * config/rl78/rl78.md (smindi3): New define_expand.
14328 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14330 * config/rl78/rl78.md (smaxdi3): New define_expand.
14332 2018-01-22 Carl Love <cel@us.ibm.com>
14334 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14335 LVX_V1TI): Add macro expansion.
14336 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14337 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14338 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14339 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14340 Change check to determine if the instruction is a byte reversing
14341 entry. Fix typo in comment.
14342 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14343 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14344 Add def_builtin calls for new builtins.
14345 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14346 Add define_insn expansion.
14348 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14350 * config/rl78/rl78.md (umaxdi3): New define_expand.
14352 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14354 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14355 for non-QImode registers.
14357 2018-01-22 Richard Biener <rguenther@suse.de>
14359 PR tree-optimization/83963
14360 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14361 including the loop exit block.
14362 (scop_detection::merge_sese): Likewise.
14363 (scop_detection::add_scop): Do it here instead.
14365 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14367 * doc/sourcebuild.texi (arm_softfloat): Document.
14369 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14372 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14373 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14374 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14376 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14377 David Edelsohn <dje.gcc@gmail.com>
14380 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14381 Change "crset eq" to "crset 2".
14382 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14383 (*call_indirect_aix<mode>_nospec): Likewise.
14384 (*call_value_indirect_aix<mode>_nospec): Likewise.
14385 (*call_indirect_elfv2<mode>_nospec): Likewise.
14386 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14387 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14388 change assembly output from . to $.
14389 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14390 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14391 (*tablejump<mode>_internal1_nospec): Likewise.
14393 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14396 * config/sh/sh_optimize_sett_clrt.cc:
14397 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14399 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14401 PR tree-optimization/83940
14402 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14403 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14404 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14405 use it to pass back the definition type.
14406 (vect_check_store_rhs): Likewise rhs_dt_out.
14407 (vect_build_gather_load_calls): Add a mask_dt argument and use
14408 it instead of a call to vect_is_simple_use.
14409 (vectorizable_store): Update calls to vect_check_load_store_mask
14410 and vect_check_store_rhs. Use the dt returned by the latter instead
14411 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14412 instead of calls to vect_is_simple_use. Pass the scalar rather
14413 than the vector operand to vect_is_simple_use when handling
14414 second and subsequent copies of an rhs value.
14415 (vectorizable_load): Update calls to vect_check_load_store_mask
14416 and vect_build_gather_load_calls. Use the cached mask_dt and
14417 gs_info.offset_dt instead of calls to vect_is_simple_use.
14419 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14421 PR middle-end/83945
14422 * tree-emutls.c: Include gimplify.h.
14423 (lower_emutls_2): New function.
14424 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14425 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14426 it before further processing.
14429 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14430 UINTVAL (trueop1) instead of INTVAL (op1).
14432 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14436 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14437 INCOMING_FRAME_SP_OFFSET if not defined.
14438 (scan_trace): Add ENTRY argument. If true and
14439 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14440 emit a note to adjust the CFA offset.
14441 (create_cfi_notes): Adjust scan_trace callers.
14442 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14443 INCOMING_FRAME_SP_OFFSET in the CIE.
14444 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14445 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14447 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14448 * doc/tm.texi: Regenerated.
14450 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14452 PR rtl-optimization/83147
14453 * lra-constraints.c (remove_inheritance_pseudos): Use
14454 lra_substitute_pseudo_within_insn.
14456 2018-01-19 Tom de Vries <tom@codesourcery.com>
14457 Cesar Philippidis <cesar@codesourcery.com>
14460 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14462 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14465 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14466 spaces for function labels.
14468 2018-01-19 Martin Liska <mliska@suse.cz>
14470 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14471 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14472 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14473 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14474 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14475 (PRED_CONST_RETURN): Change from 69 to 65.
14476 (PRED_NULL_RETURN): Change from 91 to 71.
14477 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14478 (PRED_LOOP_GUARD): Change from 66 to 73.
14480 2018-01-19 Martin Liska <mliska@suse.cz>
14482 * predict.c (predict_insn_def): Add new assert.
14483 (struct branch_predictor): Change type to signed integer.
14484 (test_prediction_value_range): Amend test to cover
14485 PROB_UNINITIALIZED.
14486 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14487 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14488 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14489 (PRED_LOOP_IV_COMPARE): Likewise.
14490 * predict.h (PROB_UNINITIALIZED): Define new constant.
14492 2018-01-19 Martin Liska <mliska@suse.cz>
14494 * predict.c (dump_prediction): Add new format for
14495 analyze_brprob.py script which is enabled with -details
14497 * profile-count.h (precise_p): New function.
14499 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14501 PR tree-optimization/83922
14502 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14503 there are no statements that need masking.
14504 (vect_active_double_reduction_p): New function.
14505 (vect_analyze_loop_operations): Use it when handling phis that
14506 are not in the loop header.
14508 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14510 PR tree-optimization/83914
14511 * tree-vect-loop.c (vectorizable_induction): Don't convert
14512 init_expr or apply the peeling adjustment for inductions
14513 that are nested within the vectorized loop.
14515 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14517 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14520 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14524 * function.h (gimplify_parameters): Add gimple_seq * argument.
14525 * function.c: Include gimple.h and options.h.
14526 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14527 for the added local temporaries if needed.
14528 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14529 if there are any parameter cleanups, wrap whole body into a
14530 try/finally with the cleanups.
14532 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14535 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14536 Use GET_MODE_CLASS for scalar floating point.
14538 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14542 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14543 Fix call of call_cgraph_insertion_hooks.
14545 2018-01-18 Martin Sebor <msebor@redhat.com>
14547 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14549 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14552 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14555 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14558 * common.opt: (-ffile-prefix-map): New option.
14559 * opts.c (common_handle_option): Defer it.
14560 * opts-global.c (handle_common_deferred_options): Handle it.
14561 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14562 * file-prefix-map.h: New file.
14563 (remap_debug_filename, add_debug_prefix_map): ...here.
14564 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14565 * final.c (debug_prefix_map, add_debug_prefix_map
14566 remap_debug_filename): Move to...
14567 * file-prefix-map.c: New file.
14568 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14569 generalize, get rid of alloca(), use strrchr() instead of strchr().
14570 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14571 Implement in terms of add_prefix_map().
14572 (remap_macro_filename, remap_debug_filename): Implement in term of
14574 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14575 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14576 * dbxout.c: Include file-prefix-map.h.
14577 * varasm.c: Likewise.
14578 * vmsdbgout.c: Likewise.
14579 * xcoffout.c: Likewise.
14580 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14581 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14582 * doc/invoke.texi (-ffile-prefix-map): Document.
14583 (-fdebug-prefix-map): Update description.
14585 2018-01-18 Martin Liska <mliska@suse.cz>
14587 * config/i386/i386.c (indirect_thunk_name): Document that also
14589 (output_indirect_thunk): Document why both instructions
14590 (pause and lfence) are generated.
14592 2018-01-18 Richard Biener <rguenther@suse.de>
14594 PR tree-optimization/83887
14595 * graphite-scop-detection.c
14596 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14597 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14598 (scop_detection::merge_sese): Re-implement with a flood-fill
14599 algorithm that properly finds a SESE region if it exists.
14601 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14604 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14605 pointer_diff optimizations use view_convert instead of convert.
14607 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14609 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14610 Generate different code for -mno-speculate-indirect-jumps.
14611 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14612 (*call_indirect_aix<mode>): Disable for
14613 -mno-speculate-indirect-jumps.
14614 (*call_indirect_aix<mode>_nospec): New define_insn.
14615 (*call_value_indirect_aix<mode>): Disable for
14616 -mno-speculate-indirect-jumps.
14617 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14618 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14619 -mno-speculate-indirect-jumps.
14620 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14622 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14624 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14625 long double type, set the flags for noting the default long double
14626 type, even if we don't pass or return a long double type.
14628 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14631 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14634 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14636 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14637 support for merge[hl].
14638 (fold_mergehl_helper): New helper function.
14639 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14640 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14641 (altivec_vmrglw_direct): Add xxmrglw insn.
14643 2018-01-17 Andrew Waterman <andrew@sifive.com>
14645 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14646 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14648 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14651 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14652 call the lto_location_cache before reading the
14653 DECL_SOURCE_LOCATION of the types.
14655 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14656 Richard Sandiford <richard.sandiford@linaro.org>
14658 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14659 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14660 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14661 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14662 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14664 * config/aarch64/constraints.md (aarch64_movti_operand):
14666 * config/aarch64/predicates.md (Uti): Add new constraint.
14668 2018-01-17 Carl Love <cel@us.ibm.com>
14670 * config/rs6000/vsx.md (define_expand xl_len_r,
14671 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14672 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14674 (define_expand, define_insn): Move the shift left from the
14675 define_insn to the define_expand for lxvl and stxvl instructions.
14676 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14677 and XL_LEN_R definitions to PURE.
14679 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14681 * config/i386/i386.c (indirect_thunk_name): Declare regno
14682 as unsigned int. Compare regno with INVALID_REGNUM.
14683 (output_indirect_thunk): Ditto.
14684 (output_indirect_thunk_function): Ditto.
14685 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14686 in the call to output_indirect_thunk_function.
14688 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14690 PR middle-end/83884
14691 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14692 rather than the size of inner_type to determine the stack slot size
14693 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14695 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14698 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14701 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14703 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14704 endian Linux systems to optionally enable multilibs for selecting
14705 the long double type if the user configured an explicit type.
14706 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14707 have no long double multilibs if not defined.
14708 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14709 warn if the user used -mabi={ieee,ibm}longdouble and we built
14710 multilibs for long double.
14711 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14712 appropriate multilib option.
14713 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14715 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14716 for building long double multilibs.
14717 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14719 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14721 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14724 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14726 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14729 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14732 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14735 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14737 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14738 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14740 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14742 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14743 different rtl trees depending on TARGET_64BIT.
14744 (rs6000_gen_lvx): Likewise.
14746 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14748 * config/visium/visium.md (nop): Tweak comment.
14749 (hazard_nop): Likewise.
14751 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14753 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14754 -mspeculate-indirect-jumps.
14755 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14756 for -mno-speculate-indirect-jumps.
14757 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14758 (*call_value_indirect_elfv2<mode>): Disable for
14759 -mno-speculate-indirect-jumps.
14760 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14761 (indirect_jump): Emit different RTL for
14762 -mno-speculate-indirect-jumps.
14763 (*indirect_jump<mode>): Disable for
14764 -mno-speculate-indirect-jumps.
14765 (*indirect_jump<mode>_nospec): New define_insn.
14766 (tablejump): Emit different RTL for
14767 -mno-speculate-indirect-jumps.
14768 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14769 (tablejumpsi_nospec): New define_expand.
14770 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14771 (tablejumpdi_nospec): New define_expand.
14772 (*tablejump<mode>_internal1): Disable for
14773 -mno-speculate-indirect-jumps.
14774 (*tablejump<mode>_internal1_nospec): New define_insn.
14775 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14778 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14780 * caller-save.c (insert_save): Drop unnecessary parameter. All
14783 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14784 Richard Biener <rguenth@suse.de>
14787 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14788 return early, inline manually is_gimple_sizepos. Make sure if we
14789 call gimplify_expr we don't end up with a gimple constant.
14790 * tree.c (variably_modified_type_p): Don't return true for
14791 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14792 * gimplify.h (is_gimple_sizepos): Remove.
14794 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14796 PR tree-optimization/83857
14797 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14798 vectorizable_live_operation for pure SLP statements.
14799 (vectorizable_live_operation): Handle PHIs.
14801 2018-01-16 Richard Biener <rguenther@suse.de>
14803 PR tree-optimization/83867
14804 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14805 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14807 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14810 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14811 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14812 If off is not INTEGER_CST, issue a may not be aligned warning
14813 rather than isn't aligned. Use isn%'t rather than isn't.
14814 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14816 <case MULT_EXPR>: Improve the case when bottom and one of the
14817 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14818 operand, in that case check if the other operand is multiple of
14819 bottom divided by the INTEGER_CST operand.
14821 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14824 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14825 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14826 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14827 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14828 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14829 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14830 (pa_function_arg_size): New function.
14832 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14834 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14835 in a separate statement.
14837 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14839 PR tree-optimization/83847
14840 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14841 group gathers and scatters.
14843 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14845 PR rtl-optimization/86620
14846 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14848 PR rtl-optimization/83213
14849 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14850 to last if both are JUMP_INSNs.
14852 PR tree-optimization/83843
14853 * gimple-ssa-store-merging.c
14854 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14855 store_immediate_info for bswap/nop orig_stores.
14857 2018-01-15 Andrew Waterman <andrew@sifive.com>
14859 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14861 <UDIV>: Increase cost if !TARGET_DIV.
14863 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14865 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14866 (define_attr "cr_logical_3op"): New.
14867 (cceq_ior_compare): Adjust.
14868 (cceq_ior_compare_complement): Adjust.
14869 (*cceq_rev_compare): Adjust.
14870 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14871 (is_cracked_insn): Adjust.
14872 (insn_must_be_first_in_group): Adjust.
14873 * config/rs6000/40x.md: Adjust.
14874 * config/rs6000/440.md: Adjust.
14875 * config/rs6000/476.md: Adjust.
14876 * config/rs6000/601.md: Adjust.
14877 * config/rs6000/603.md: Adjust.
14878 * config/rs6000/6xx.md: Adjust.
14879 * config/rs6000/7450.md: Adjust.
14880 * config/rs6000/7xx.md: Adjust.
14881 * config/rs6000/8540.md: Adjust.
14882 * config/rs6000/cell.md: Adjust.
14883 * config/rs6000/e300c2c3.md: Adjust.
14884 * config/rs6000/e500mc.md: Adjust.
14885 * config/rs6000/e500mc64.md: Adjust.
14886 * config/rs6000/e5500.md: Adjust.
14887 * config/rs6000/e6500.md: Adjust.
14888 * config/rs6000/mpc.md: Adjust.
14889 * config/rs6000/power4.md: Adjust.
14890 * config/rs6000/power5.md: Adjust.
14891 * config/rs6000/power6.md: Adjust.
14892 * config/rs6000/power7.md: Adjust.
14893 * config/rs6000/power8.md: Adjust.
14894 * config/rs6000/power9.md: Adjust.
14895 * config/rs6000/rs64.md: Adjust.
14896 * config/rs6000/titan.md: Adjust.
14898 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14900 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14901 ix86_indirect_branch_register logic.
14903 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14905 * config/i386/constraints.md (Bs): Update
14906 ix86_indirect_branch_register check. Don't check
14907 ix86_indirect_branch_register with GOT_memory_operand.
14909 * config/i386/predicates.md (GOT_memory_operand): Don't check
14910 ix86_indirect_branch_register here.
14911 (GOT32_symbol_operand): Likewise.
14913 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14915 * config/i386/predicates.md (constant_call_address_operand):
14916 Rewrite ix86_indirect_branch_register logic.
14917 (sibcall_insn_operand): Likewise.
14919 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14921 * config/i386/constraints.md (Bs): Replace
14922 ix86_indirect_branch_thunk_register with
14923 ix86_indirect_branch_register.
14925 * config/i386/i386.md (indirect_jump): Likewise.
14926 (tablejump): Likewise.
14927 (*sibcall_memory): Likewise.
14928 (*sibcall_value_memory): Likewise.
14929 Peepholes of indirect call and jump via memory: Likewise.
14930 * config/i386/i386.opt: Likewise.
14931 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14932 (GOT_memory_operand): Likewise.
14933 (call_insn_operand): Likewise.
14934 (sibcall_insn_operand): Likewise.
14935 (GOT32_symbol_operand): Likewise.
14937 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14939 PR middle-end/83837
14940 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14941 type rather than type addr's type points to.
14942 (expand_omp_atomic_mutex): Likewise.
14943 (expand_omp_atomic): Likewise.
14945 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14948 * config/i386/i386.c (output_indirect_thunk_function): Use
14949 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14950 for __x86_return_thunk.
14952 2018-01-15 Richard Biener <rguenther@suse.de>
14954 PR middle-end/83850
14955 * expmed.c (extract_bit_field_1): Fix typo.
14957 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14960 * config/arm/iterators.md (VF): New mode iterator.
14961 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14962 Remove integer-related logic from pattern.
14963 (neon_vabd<mode>_3): Likewise.
14965 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14967 PR middle-end/82694
14968 * common.opt (fstrict-overflow): No longer an alias.
14969 (fwrapv-pointer): New option.
14970 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14971 also for pointer types based on flag_wrapv_pointer.
14972 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14973 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14974 opts->x_flag_wrapv got set.
14975 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14976 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14977 POINTER_TYPE_OVERFLOW_UNDEFINED.
14978 * match.pd: Likewise in address comparison pattern.
14979 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14981 2018-01-15 Richard Biener <rguenther@suse.de>
14984 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14985 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14986 Reset type names to their identifier if their TYPE_DECL doesn't
14987 have linkage (and thus is used for ODR and devirt).
14988 (save_debug_info_for_decl): Remove.
14989 (save_debug_info_for_type): Likewise.
14990 (add_tree_to_fld_list): Adjust.
14991 * tree-pretty-print.c (dump_generic_node): Make dumping of
14992 type names more robust.
14994 2018-01-15 Richard Biener <rguenther@suse.de>
14996 * BASE-VER: Bump to 8.0.1.
14998 2018-01-14 Martin Sebor <msebor@redhat.com>
15001 * builtins.c (check_access): Avoid warning when the no-warning bit
15004 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15006 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15007 * ira-color (allocno_hard_regs_compare): Likewise.
15009 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15012 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15013 Use .pushsection/.popsection.
15015 2018-01-14 Martin Sebor <msebor@redhat.com>
15018 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15020 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15022 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15023 entry from extra_headers.
15024 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15025 extra_headers, make the list bitwise identical to the i?86-*-* one.
15027 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15029 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15030 -mcmodel=large with -mindirect-branch=thunk,
15031 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15032 -mfunction-return=thunk-extern.
15033 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15034 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15035 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15037 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15039 * config/i386/i386.c (print_reg): Print the name of the full
15040 integer register without '%'.
15041 (ix86_print_operand): Handle 'V'.
15042 * doc/extend.texi: Document 'V' modifier.
15044 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15046 * config/i386/constraints.md (Bs): Disallow memory operand for
15047 -mindirect-branch-register.
15049 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15050 (GOT_memory_operand): Likewise.
15051 (call_insn_operand): Likewise.
15052 (sibcall_insn_operand): Likewise.
15053 (GOT32_symbol_operand): Likewise.
15054 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15055 for -mindirect-branch-register.
15056 (tablejump): Likewise.
15057 (*sibcall_memory): Likewise.
15058 (*sibcall_value_memory): Likewise.
15059 Disallow peepholes of indirect call and jump via memory for
15060 -mindirect-branch-register.
15061 (*call_pop): Replace m with Bw.
15062 (*call_value_pop): Likewise.
15063 (*sibcall_pop_memory): Replace m with Bs.
15064 * config/i386/i386.opt (mindirect-branch-register): New option.
15065 * doc/invoke.texi: Document -mindirect-branch-register option.
15067 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15069 * config/i386/i386-protos.h (ix86_output_function_return): New.
15070 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15071 set function_return_type.
15072 (indirect_thunk_name): Add ret_p to indicate thunk for function
15074 (output_indirect_thunk_function): Pass false to
15075 indirect_thunk_name.
15076 (ix86_output_indirect_branch_via_reg): Likewise.
15077 (ix86_output_indirect_branch_via_push): Likewise.
15078 (output_indirect_thunk_function): Create alias for function
15079 return thunk if regno < 0.
15080 (ix86_output_function_return): New function.
15081 (ix86_handle_fndecl_attribute): Handle function_return.
15082 (ix86_attribute_table): Add function_return.
15083 * config/i386/i386.h (machine_function): Add
15084 function_return_type.
15085 * config/i386/i386.md (simple_return_internal): Use
15086 ix86_output_function_return.
15087 (simple_return_internal_long): Likewise.
15088 * config/i386/i386.opt (mfunction-return=): New option.
15089 (indirect_branch): Mention -mfunction-return=.
15090 * doc/extend.texi: Document function_return function attribute.
15091 * doc/invoke.texi: Document -mfunction-return= option.
15093 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15095 * config/i386/i386-opts.h (indirect_branch): New.
15096 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15097 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15098 with local indirect jump when converting indirect call and jump.
15099 (ix86_set_indirect_branch_type): New.
15100 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15101 (indirectlabelno): New.
15102 (indirect_thunk_needed): Likewise.
15103 (indirect_thunk_bnd_needed): Likewise.
15104 (indirect_thunks_used): Likewise.
15105 (indirect_thunks_bnd_used): Likewise.
15106 (INDIRECT_LABEL): Likewise.
15107 (indirect_thunk_name): Likewise.
15108 (output_indirect_thunk): Likewise.
15109 (output_indirect_thunk_function): Likewise.
15110 (ix86_output_indirect_branch_via_reg): Likewise.
15111 (ix86_output_indirect_branch_via_push): Likewise.
15112 (ix86_output_indirect_branch): Likewise.
15113 (ix86_output_indirect_jmp): Likewise.
15114 (ix86_code_end): Call output_indirect_thunk_function if needed.
15115 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15117 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15118 (ix86_attribute_table): Add indirect_branch.
15119 * config/i386/i386.h (machine_function): Add indirect_branch_type
15120 and has_local_indirect_jump.
15121 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15123 (tablejump): Likewise.
15124 (*indirect_jump): Use ix86_output_indirect_jmp.
15125 (*tablejump_1): Likewise.
15126 (simple_return_indirect_internal): Likewise.
15127 * config/i386/i386.opt (mindirect-branch=): New option.
15128 (indirect_branch): New.
15131 (thunk-inline): Likewise.
15132 (thunk-extern): Likewise.
15133 * doc/extend.texi: Document indirect_branch function attribute.
15134 * doc/invoke.texi: Document -mindirect-branch= option.
15136 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15139 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15141 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15143 * ipa-inline.c (want_inline_small_function_p): Return false if
15144 inlining has already failed with CIF_FINAL_ERROR.
15145 (update_caller_keys): Call want_inline_small_function_p before
15147 (update_callee_keys): Likewise.
15149 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15151 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15153 (rs6000_quadword_masked_address_p): Likewise.
15154 (quad_aligned_load_p): Likewise.
15155 (quad_aligned_store_p): Likewise.
15156 (const_load_sequence_p): Add comment to describe the outer-most loop.
15157 (mimic_memory_attributes_and_flags): New function.
15158 (rs6000_gen_stvx): Likewise.
15159 (replace_swapped_aligned_store): Likewise.
15160 (rs6000_gen_lvx): Likewise.
15161 (replace_swapped_aligned_load): Likewise.
15162 (replace_swapped_load_constant): Capitalize argument name in
15163 comment describing this function.
15164 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15165 and stores that access quad-word aligned addresses and replace
15166 with stvx or lvx instructions when appropriate.
15167 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15168 New function prototype.
15169 (rs6000_quadword_masked_address_p): Likewise.
15170 (rs6000_gen_lvx): Likewise.
15171 (rs6000_gen_stvx): Likewise.
15172 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15173 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15174 when memory address is aligned.
15175 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15176 this split to select lvx instruction when memory address is aligned.
15177 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15178 instruction when memory address is aligned.
15179 (*vsx_le_perm_load_v16qi): Likewise.
15180 (four unnamed splitters): Modify to select the stvx instruction
15181 when memory is aligned.
15183 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15185 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15186 which appears in the queue multiple times.
15188 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15189 Alan Hayward <alan.hayward@arm.com>
15190 David Sherwood <david.sherwood@arm.com>
15192 * tree-vectorizer.h (vec_lower_bound): New structure.
15193 (_loop_vec_info): Add check_nonzero and lower_bounds.
15194 (LOOP_VINFO_CHECK_NONZERO): New macro.
15195 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15196 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15197 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15198 fields. Make seg_len the distance travelled, not including the
15200 (dr_direction_indicator): Declare.
15201 (dr_zero_step_indicator): Likewise.
15202 (dr_known_forward_stride_p): Likewise.
15203 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15205 (runtime_alias_check_p): Allow runtime alias checks with
15207 (operator ==): Compare access_size and align.
15208 (prune_runtime_alias_test_list): Rework for new distinction between
15209 the access_size and seg_len.
15210 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15212 (get_segment_min_max): New function.
15213 (create_intersect_range_checks): Use it.
15214 (dr_step_indicator): New function.
15215 (dr_direction_indicator): Likewise.
15216 (dr_zero_step_indicator): Likewise.
15217 (dr_known_forward_stride_p): Likewise.
15218 * tree-loop-distribution.c (data_ref_segment_size): Return
15219 DR_STEP * (niters - 1).
15220 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15222 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15223 (vect_preserves_scalar_order_p): New function, split out from...
15224 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15225 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15226 (vect_vfa_access_size): New function.
15227 (vect_vfa_align): Likewise.
15228 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15229 (dump_lower_bound): New function.
15230 (vect_check_lower_bound): Likewise.
15231 (vect_small_gap_p): Likewise.
15232 (vectorizable_with_step_bound_p): Likewise.
15233 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15234 depencies if the vectorization factor is 1. Convert the checks
15235 for nonzero steps into checks on the bounds of DR_STEP. Try using
15236 a bunds check for variable steps if the minimum required step is
15237 relatively small. Update calls to the dr_with_seg_len
15238 constructor and to vect_compile_time_alias.
15239 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15241 (vect_loop_versioning): Call it.
15242 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15244 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15247 Alan Hayward <alan.hayward@arm.com>
15248 David Sherwood <david.sherwood@arm.com>
15250 * doc/sourcebuild.texi (vect_scatter_store): Document.
15251 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15253 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15255 * genopinit.c (main): Add supports_vec_scatter_store and
15256 supports_vec_scatter_store_cached to target_optabs.
15257 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15258 IFN_MASK_SCATTER_STORE.
15259 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15261 * internal-fn.h (internal_store_fn_p): Declare.
15262 (internal_fn_stored_value_index): Likewise.
15263 * internal-fn.c (scatter_store_direct): New macro.
15264 (expand_scatter_store_optab_fn): New function.
15265 (direct_scatter_store_optab_supported_p): New macro.
15266 (internal_store_fn_p): New function.
15267 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15268 IFN_MASK_SCATTER_STORE.
15269 (internal_fn_mask_index): Likewise.
15270 (internal_fn_stored_value_index): New function.
15271 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15272 for scatter stores.
15273 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15274 * optabs-query.c (supports_vec_scatter_store_p): New function.
15275 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15276 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15277 true for scatter stores.
15278 (vect_gather_scatter_fn_p): Handle scatter stores too.
15279 (vect_check_gather_scatter): Consider using scatter stores if
15280 supports_vec_scatter_store_p.
15281 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15282 scatter stores too.
15283 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15284 internal_fn_stored_value_index.
15285 (check_load_store_masking): Handle scatter stores too.
15286 (vect_get_store_rhs): Make public.
15287 (vectorizable_call): Use internal_store_fn_p.
15288 (vectorizable_store): Handle scatter store internal functions.
15289 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15290 when deciding whether the end of the group has been reached.
15291 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15292 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15293 (mask_scatter_store<mode>): New insns.
15295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15296 Alan Hayward <alan.hayward@arm.com>
15297 David Sherwood <david.sherwood@arm.com>
15299 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15300 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15301 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15303 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15304 Use vect_truncate_gather_scatter_offset if we can't treat the
15305 operation as a normal gather load or scatter store.
15306 (get_group_load_store_type): Take the gather_scatter_info
15307 as argument. Try using a gather load or scatter store for
15308 single-element groups.
15309 (get_load_store_type): Update calls to get_group_load_store_type
15310 and vect_use_strided_gather_scatters_p.
15312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15313 Alan Hayward <alan.hayward@arm.com>
15314 David Sherwood <david.sherwood@arm.com>
15316 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15317 optional tree argument.
15318 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15320 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15321 but continue to use the current value as a fallback.
15322 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15323 to compare the updates.
15324 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15325 (get_load_store_type): Use it when handling a strided access.
15326 (vect_get_strided_load_store_ops): New function.
15327 (vect_get_data_ptr_increment): Likewise.
15328 (vectorizable_load): Handle strided gather loads. Always pass
15329 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15332 Alan Hayward <alan.hayward@arm.com>
15333 David Sherwood <david.sherwood@arm.com>
15335 * doc/md.texi (gather_load@var{m}): Document.
15336 (mask_gather_load@var{m}): Likewise.
15337 * genopinit.c (main): Add supports_vec_gather_load and
15338 supports_vec_gather_load_cached to target_optabs.
15339 * optabs-tree.c (init_tree_optimization_optabs): Use
15340 ggc_cleared_alloc to allocate target_optabs.
15341 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15342 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15344 * internal-fn.h (internal_load_fn_p): Declare.
15345 (internal_gather_scatter_fn_p): Likewise.
15346 (internal_fn_mask_index): Likewise.
15347 (internal_gather_scatter_fn_supported_p): Likewise.
15348 * internal-fn.c (gather_load_direct): New macro.
15349 (expand_gather_load_optab_fn): New function.
15350 (direct_gather_load_optab_supported_p): New macro.
15351 (direct_internal_fn_optab): New function.
15352 (internal_load_fn_p): Likewise.
15353 (internal_gather_scatter_fn_p): Likewise.
15354 (internal_fn_mask_index): Likewise.
15355 (internal_gather_scatter_fn_supported_p): Likewise.
15356 * optabs-query.c (supports_at_least_one_mode_p): New function.
15357 (supports_vec_gather_load_p): Likewise.
15358 * optabs-query.h (supports_vec_gather_load_p): Declare.
15359 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15360 and memory_type field.
15361 (NUM_PATTERNS): Bump to 15.
15362 * tree-vect-data-refs.c: Include internal-fn.h.
15363 (vect_gather_scatter_fn_p): New function.
15364 (vect_describe_gather_scatter_call): Likewise.
15365 (vect_check_gather_scatter): Try using internal functions for
15366 gather loads. Recognize existing calls to a gather load function.
15367 (vect_analyze_data_refs): Consider using gather loads if
15368 supports_vec_gather_load_p.
15369 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15370 (vect_get_gather_scatter_offset_type): Likewise.
15371 (vect_convert_mask_for_vectype): Likewise.
15372 (vect_add_conversion_to_patterm): Likewise.
15373 (vect_try_gather_scatter_pattern): Likewise.
15374 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15375 (vect_vect_recog_func_ptrs): Add it.
15376 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15377 internal_fn_mask_index and internal_gather_scatter_fn_p.
15378 (check_load_store_masking): Take the gather_scatter_info as an
15379 argument and handle gather loads.
15380 (vect_get_gather_scatter_ops): New function.
15381 (vectorizable_call): Check internal_load_fn_p.
15382 (vectorizable_load): Likewise. Handle gather load internal
15384 (vectorizable_store): Update call to check_load_store_masking.
15385 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15386 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15387 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15388 (aarch64_gather_scale_operand_d): New predicates.
15389 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15390 (mask_gather_load<mode>): New insns.
15392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15393 Alan Hayward <alan.hayward@arm.com>
15394 David Sherwood <david.sherwood@arm.com>
15396 * optabs.def (fold_left_plus_optab): New optab.
15397 * doc/md.texi (fold_left_plus_@var{m}): Document.
15398 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15399 * internal-fn.c (fold_left_direct): Define.
15400 (expand_fold_left_optab_fn): Likewise.
15401 (direct_fold_left_optab_supported_p): Likewise.
15402 * fold-const-call.c (fold_const_fold_left): New function.
15403 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15404 * tree-parloops.c (valid_reduction_p): New function.
15405 (gather_scalar_reductions): Use it.
15406 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15407 (vect_finish_replace_stmt): Declare.
15408 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15409 (needs_fold_left_reduction_p): New function, split out from...
15410 (vect_is_simple_reduction): ...here. Accept reductions that
15411 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15412 (vect_force_simple_reduction): Also store the reduction type in
15413 the assignment's STMT_VINFO_REDUC_TYPE.
15414 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15415 (merge_with_identity): New function.
15416 (vect_expand_fold_left): Likewise.
15417 (vectorize_fold_left_reduction): Likewise.
15418 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15419 scalar phi in place for it. Check for target support and reject
15420 cases that would reassociate the operation. Defer the transform
15421 phase to vectorize_fold_left_reduction.
15422 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15423 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15424 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15428 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15429 call to ifc_temp_var.
15431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15432 Alan Hayward <alan.hayward@arm.com>
15433 David Sherwood <david.sherwood@arm.com>
15435 * target.def (legitimize_address_displacement): Take the original
15436 offset as a poly_int.
15437 * targhooks.h (default_legitimize_address_displacement): Update
15439 * targhooks.c (default_legitimize_address_displacement): Likewise.
15440 * doc/tm.texi: Regenerate.
15441 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15442 as an argument, moving assert of ad->disp == ad->disp_term to...
15443 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15444 Try calling targetm.legitimize_address_displacement before expanding
15445 the address rather than afterwards, and adjust for the new interface.
15446 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15447 Match the new hook interface. Handle SVE addresses.
15448 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15449 new hook interface.
15451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15453 * Makefile.in (OBJS): Add early-remat.o.
15454 * target.def (select_early_remat_modes): New hook.
15455 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15456 * doc/tm.texi: Regenerate.
15457 * targhooks.h (default_select_early_remat_modes): Declare.
15458 * targhooks.c (default_select_early_remat_modes): New function.
15459 * timevar.def (TV_EARLY_REMAT): New timevar.
15460 * passes.def (pass_early_remat): New pass.
15461 * tree-pass.h (make_pass_early_remat): Declare.
15462 * early-remat.c: New file.
15463 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15465 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15472 vfm1 with a bound_epilog parameter.
15473 (vect_do_peeling): Update calls accordingly, and move the prologue
15474 call earlier in the function. Treat the base bound_epilog as 0 for
15475 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15476 this base when peeling for gaps.
15477 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15478 with fully-masked loops.
15479 (vect_estimate_min_profitable_iters): Handle the single peeled
15480 iteration in that case.
15482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15483 Alan Hayward <alan.hayward@arm.com>
15484 David Sherwood <david.sherwood@arm.com>
15486 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15487 single-element interleaving even if the size is not a power of 2.
15488 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15489 accesses for single-element interleaving if the group size is
15492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15493 Alan Hayward <alan.hayward@arm.com>
15494 David Sherwood <david.sherwood@arm.com>
15496 * doc/md.texi (fold_extract_last_@var{m}): Document.
15497 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15498 * optabs.def (fold_extract_last_optab): New optab.
15499 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15500 * internal-fn.c (fold_extract_direct): New macro.
15501 (expand_fold_extract_optab_fn): Likewise.
15502 (direct_fold_extract_optab_supported_p): Likewise.
15503 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15504 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15505 EXTRACT_LAST_REDUCTION.
15506 (get_initial_def_for_reduction): Do not create an initial vector
15507 for EXTRACT_LAST_REDUCTION reductions.
15508 (vectorizable_reduction): Leave the scalar phi in place for
15509 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15510 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15511 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15512 transform phase to vectorizable_condition.
15513 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15515 (vect_finish_stmt_generation): ...here.
15516 (vect_finish_replace_stmt): New function.
15517 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15518 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15520 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15523 Alan Hayward <alan.hayward@arm.com>
15524 David Sherwood <david.sherwood@arm.com>
15526 * doc/md.texi (extract_last_@var{m}): Document.
15527 * optabs.def (extract_last_optab): New optab.
15528 * internal-fn.def (EXTRACT_LAST): New internal function.
15529 * internal-fn.c (cond_unary_direct): New macro.
15530 (expand_cond_unary_optab_fn): Likewise.
15531 (direct_cond_unary_optab_supported_p): Likewise.
15532 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15533 loops using EXTRACT_LAST.
15534 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15535 (extract_last_<mode>): ...this optab.
15536 (vec_extract<mode><Vel>): Update accordingly.
15538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15539 Alan Hayward <alan.hayward@arm.com>
15540 David Sherwood <david.sherwood@arm.com>
15542 * target.def (empty_mask_is_expensive): New hook.
15543 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15544 * doc/tm.texi: Regenerate.
15545 * targhooks.h (default_empty_mask_is_expensive): Declare.
15546 * targhooks.c (default_empty_mask_is_expensive): New function.
15547 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15548 if the target says that empty masks are expensive.
15549 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15551 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15554 Alan Hayward <alan.hayward@arm.com>
15555 David Sherwood <david.sherwood@arm.com>
15557 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15558 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15559 (vect_use_loop_mask_for_alignment_p): New function.
15560 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15561 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15562 niters_skip argument. Make sure that the first niters_skip elements
15563 of the first iteration are inactive.
15564 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15565 Update call to vect_set_loop_masks_directly.
15566 (get_misalign_in_elems): New function, split out from...
15567 (vect_gen_prolog_loop_niters): ...here.
15568 (vect_update_init_of_dr): Take a code argument that specifies whether
15569 the adjustment should be added or subtracted.
15570 (vect_update_init_of_drs): Likewise.
15571 (vect_prepare_for_masked_peels): New function.
15572 (vect_do_peeling): Skip prologue peeling if we're using a mask
15573 instead. Update call to vect_update_inits_of_drs.
15574 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15576 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15577 alignment. Do not include the number of peeled iterations in
15578 the minimum threshold in that case.
15579 (vectorizable_induction): Adjust the start value down by
15580 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15581 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15582 Take the number of skipped iterations into account when calculating
15584 * tree-vect-stmts.c (vect_gen_while_not): New function.
15586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15587 Alan Hayward <alan.hayward@arm.com>
15588 David Sherwood <david.sherwood@arm.com>
15590 * doc/sourcebuild.texi (vect_fully_masked): Document.
15591 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15592 default value to 0.
15593 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15595 (vect_analyze_loop_2): ...here. Don't check the vectorization
15596 factor against the number of loop iterations if the loop is
15599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15600 Alan Hayward <alan.hayward@arm.com>
15601 David Sherwood <david.sherwood@arm.com>
15603 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15604 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15605 (dump_groups): Update accordingly.
15606 (iv_use::mem_type): New member variable.
15607 (address_p): New function.
15608 (record_use): Add a mem_type argument and initialize the new
15610 (record_group_use): Add a mem_type argument. Use address_p.
15611 Remove obsolete null checks of base_object. Update call to record_use.
15612 (find_interesting_uses_op): Update call to record_group_use.
15613 (find_interesting_uses_cond): Likewise.
15614 (find_interesting_uses_address): Likewise.
15615 (get_mem_type_for_internal_fn): New function.
15616 (find_address_like_use): Likewise.
15617 (find_interesting_uses_stmt): Try find_address_like_use before
15618 calling find_interesting_uses_op.
15619 (addr_offset_valid_p): Use the iv mem_type field as the type
15620 of the addressed memory.
15621 (add_autoinc_candidates): Likewise.
15622 (get_address_cost): Likewise.
15623 (split_small_address_groups_p): Use address_p.
15624 (split_address_groups): Likewise.
15625 (add_iv_candidate_for_use): Likewise.
15626 (autoinc_possible_for_pair): Likewise.
15627 (rewrite_groups): Likewise.
15628 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15629 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15630 (get_alias_ptr_type_for_ptr_address): New function.
15631 (rewrite_use_address): Rewrite address uses in calls that were
15632 identified by find_address_like_use.
15634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15635 Alan Hayward <alan.hayward@arm.com>
15636 David Sherwood <david.sherwood@arm.com>
15638 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15640 * gimple-expr.h (is_gimple_addressable: Likewise.
15641 * gimple-expr.c (is_gimple_address): Likewise.
15642 * internal-fn.c (expand_call_mem_ref): New function.
15643 (expand_mask_load_optab_fn): Use it.
15644 (expand_mask_store_optab_fn): Likewise.
15646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15647 Alan Hayward <alan.hayward@arm.com>
15648 David Sherwood <david.sherwood@arm.com>
15650 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15651 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15652 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15653 (cond_umax@var{mode}): Document.
15654 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15655 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15656 (cond_umin_optab, cond_umax_optab): New optabs.
15657 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15658 (COND_IOR, COND_XOR): New internal functions.
15659 * internal-fn.h (get_conditional_internal_fn): Declare.
15660 * internal-fn.c (cond_binary_direct): New macro.
15661 (expand_cond_binary_optab_fn): Likewise.
15662 (direct_cond_binary_optab_supported_p): Likewise.
15663 (get_conditional_internal_fn): New function.
15664 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15665 Cope with reduction statements that are vectorized as calls rather
15667 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15668 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15669 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15670 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15671 (UNSPEC_COND_EOR): New unspecs.
15672 (optab): Add mappings for them.
15673 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15674 (sve_int_op, sve_fp_op): New int attributes.
15676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15677 Alan Hayward <alan.hayward@arm.com>
15678 David Sherwood <david.sherwood@arm.com>
15680 * optabs.def (while_ult_optab): New optab.
15681 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15682 * internal-fn.def (WHILE_ULT): New internal function.
15683 * internal-fn.h (direct_internal_fn_supported_p): New override
15684 that takes two types as argument.
15685 * internal-fn.c (while_direct): New macro.
15686 (expand_while_optab_fn): New function.
15687 (convert_optab_supported_p): Likewise.
15688 (direct_while_optab_supported_p): New macro.
15689 * wide-int.h (wi::udiv_ceil): New function.
15690 * tree-vectorizer.h (rgroup_masks): New structure.
15691 (vec_loop_masks): New typedef.
15692 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15693 and fully_masked_p.
15694 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15695 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15696 (vect_max_vf): New function.
15697 (slpeel_make_loop_iterate_ntimes): Delete.
15698 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15699 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15700 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15701 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15702 internal-fn.h, stor-layout.h and optabs-query.h.
15703 (vect_set_loop_mask): New function.
15704 (add_preheader_seq): Likewise.
15705 (add_header_seq): Likewise.
15706 (interleave_supported_p): Likewise.
15707 (vect_maybe_permute_loop_masks): Likewise.
15708 (vect_set_loop_masks_directly): Likewise.
15709 (vect_set_loop_condition_masked): Likewise.
15710 (vect_set_loop_condition_unmasked): New function, split out from
15711 slpeel_make_loop_iterate_ntimes.
15712 (slpeel_make_loop_iterate_ntimes): Rename to..
15713 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15714 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15715 (vect_do_peeling): Update call accordingly.
15716 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15718 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15719 mask_compare_type, can_fully_mask_p and fully_masked_p.
15720 (release_vec_loop_masks): New function.
15721 (_loop_vec_info): Use it to free the loop masks.
15722 (can_produce_all_loop_masks_p): New function.
15723 (vect_get_max_nscalars_per_iter): Likewise.
15724 (vect_verify_full_masking): Likewise.
15725 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15726 retries, and free the mask rgroups before retrying. Check loop-wide
15727 reasons for disallowing fully-masked loops. Make the final decision
15728 about whether use a fully-masked loop or not.
15729 (vect_estimate_min_profitable_iters): Do not assume that peeling
15730 for the number of iterations will be needed for fully-masked loops.
15731 (vectorizable_reduction): Disable fully-masked loops.
15732 (vectorizable_live_operation): Likewise.
15733 (vect_halve_mask_nunits): New function.
15734 (vect_double_mask_nunits): Likewise.
15735 (vect_record_loop_mask): Likewise.
15736 (vect_get_loop_mask): Likewise.
15737 (vect_transform_loop): Handle the case in which the final loop
15738 iteration might handle a partial vector. Call vect_set_loop_condition
15739 instead of slpeel_make_loop_iterate_ntimes.
15740 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15741 (check_load_store_masking): New function.
15742 (prepare_load_store_mask): Likewise.
15743 (vectorizable_store): Handle fully-masked loops.
15744 (vectorizable_load): Likewise.
15745 (supportable_widening_operation): Use vect_halve_mask_nunits for
15747 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15748 (vect_gen_while): New function.
15749 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15750 (aarch64_uqdec<mode>): New insn.
15752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15753 Alan Hayward <alan.hayward@arm.com>
15754 David Sherwood <david.sherwood@arm.com>
15756 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15757 (reduc_xor_scal_optab): New optabs.
15758 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15759 (reduc_xor_scal_@var{m}): Document.
15760 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15761 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15762 internal functions.
15763 * fold-const-call.c (fold_const_call): Handle them.
15764 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15765 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15766 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15767 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15768 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15769 (UNSPEC_XORV): New unspecs.
15770 (optab): Add entries for them.
15771 (BITWISEV): New int iterator.
15772 (bit_reduc_op): New int attributes.
15774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15775 Alan Hayward <alan.hayward@arm.com>
15776 David Sherwood <david.sherwood@arm.com>
15778 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15779 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15780 * optabs.def (vec_shl_insert_optab): New optab.
15781 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15782 (duplicate_and_interleave): Likewise.
15783 * tree-vect-loop.c: Include internal-fn.h.
15784 (neutral_op_for_slp_reduction): New function, split out from
15785 get_initial_defs_for_reduction.
15786 (get_initial_def_for_reduction): Handle option 2 for variable-length
15787 vectors by loading the neutral value into a vector and then shifting
15788 the initial value into element 0.
15789 (get_initial_defs_for_reduction): Replace the code argument with
15790 the neutral value calculated by neutral_op_for_slp_reduction.
15791 Use gimple_build_vector for constant-length vectors.
15792 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15793 but the first group_size elements have a neutral value.
15794 Use duplicate_and_interleave otherwise.
15795 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15796 Update call to get_initial_defs_for_reduction. Handle SLP
15797 reductions for variable-length vectors by creating one vector
15798 result for each scalar result, with the elements associated
15799 with other scalar results stubbed out with the neutral value.
15800 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15801 Require IFN_VEC_SHL_INSERT for double reductions on
15802 variable-length vectors, or SLP reductions that have
15803 a neutral value. Require can_duplicate_and_interleave_p
15804 support for variable-length unchained SLP reductions if there
15805 is no neutral value, such as for MIN/MAX reductions. Also require
15806 the number of vector elements to be a multiple of the number of
15807 SLP statements when doing variable-length unchained SLP reductions.
15808 Update call to vect_create_epilog_for_reduction.
15809 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15810 and remove initial values.
15811 (duplicate_and_interleave): Make public.
15812 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15813 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15816 Alan Hayward <alan.hayward@arm.com>
15817 David Sherwood <david.sherwood@arm.com>
15819 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15820 (can_duplicate_and_interleave_p): New function.
15821 (vect_get_and_check_slp_defs): Take the vector of statements
15822 rather than just the current one. Remove excess parentheses.
15823 Restriction rejectinon of vect_constant_def and vect_external_def
15824 for variable-length vectors to boolean types, or types for which
15825 can_duplicate_and_interleave_p is false.
15826 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15827 (duplicate_and_interleave): New function.
15828 (vect_get_constant_vectors): Use gimple_build_vector for
15829 constant-length vectors and suitable variable-length constant
15830 vectors. Use duplicate_and_interleave for other variable-length
15831 vectors. Don't defer the update when inserting new statements.
15833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15834 Alan Hayward <alan.hayward@arm.com>
15835 David Sherwood <david.sherwood@arm.com>
15837 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15838 min_profitable_iters doesn't go negative.
15840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15841 Alan Hayward <alan.hayward@arm.com>
15842 David Sherwood <david.sherwood@arm.com>
15844 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15845 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15846 * optabs.def (vec_mask_load_lanes_optab): New optab.
15847 (vec_mask_store_lanes_optab): Likewise.
15848 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15849 (MASK_STORE_LANES): Likewise.
15850 * internal-fn.c (mask_load_lanes_direct): New macro.
15851 (mask_store_lanes_direct): Likewise.
15852 (expand_mask_load_optab_fn): Handle masked operations.
15853 (expand_mask_load_lanes_optab_fn): New macro.
15854 (expand_mask_store_optab_fn): Handle masked operations.
15855 (expand_mask_store_lanes_optab_fn): New macro.
15856 (direct_mask_load_lanes_optab_supported_p): Likewise.
15857 (direct_mask_store_lanes_optab_supported_p): Likewise.
15858 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15860 (vect_load_lanes_supported): Likewise.
15861 * tree-vect-data-refs.c (strip_conversion): New function.
15862 (can_group_stmts_p): Likewise.
15863 (vect_analyze_data_ref_accesses): Use it instead of checking
15864 for a pair of assignments.
15865 (vect_store_lanes_supported): Take a masked_p parameter.
15866 (vect_load_lanes_supported): Likewise.
15867 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15868 vect_store_lanes_supported and vect_load_lanes_supported.
15869 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15870 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15871 parameter. Don't allow gaps for masked accesses.
15872 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15873 and vect_load_lanes_supported.
15874 (get_load_store_type): Take a masked_p parameter and update
15875 call to get_group_load_store_type.
15876 (vectorizable_store): Update call to get_load_store_type.
15877 Handle IFN_MASK_STORE_LANES.
15878 (vectorizable_load): Update call to get_load_store_type.
15879 Handle IFN_MASK_LOAD_LANES.
15881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15882 Alan Hayward <alan.hayward@arm.com>
15883 David Sherwood <david.sherwood@arm.com>
15885 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15887 * config/aarch64/aarch64-protos.h
15888 (aarch64_sve_struct_memory_operand_p): Declare.
15889 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15890 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15891 (VPRED, vpred): Handle SVE structure modes.
15892 * config/aarch64/constraints.md (Utx): New constraint.
15893 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15894 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15895 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15896 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15897 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15898 structure modes. Split into pieces after RA.
15899 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15900 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15902 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15903 SVE structure modes.
15904 (aarch64_classify_address): Likewise.
15905 (sizetochar): Move earlier in file.
15906 (aarch64_print_operand): Handle SVE register lists.
15907 (aarch64_array_mode): New function.
15908 (aarch64_sve_struct_memory_operand_p): Likewise.
15909 (TARGET_ARRAY_MODE): Redefine.
15911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15912 Alan Hayward <alan.hayward@arm.com>
15913 David Sherwood <david.sherwood@arm.com>
15915 * target.def (array_mode): New target hook.
15916 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15917 * doc/tm.texi: Regenerate.
15918 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15919 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15920 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15921 targetm.array_mode.
15922 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15925 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15926 Alan Hayward <alan.hayward@arm.com>
15927 David Sherwood <david.sherwood@arm.com>
15929 * fold-const.c (fold_binary_loc): Check the argument types
15930 rather than the result type when testing for a vector operation.
15932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15934 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15935 * doc/tm.texi: Regenerate.
15937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15938 Alan Hayward <alan.hayward@arm.com>
15939 David Sherwood <david.sherwood@arm.com>
15941 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15942 (sve): Document new AArch64 extension.
15943 * doc/md.texi (w): Extend the description of the AArch64
15944 constraint to include SVE vectors.
15945 (Upl, Upa): Document new AArch64 predicate constraints.
15946 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15948 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15949 (msve-vector-bits=): New option.
15950 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15951 SVE when these are disabled.
15952 (sve): New extension.
15953 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15954 modes. Adjust their number of units based on aarch64_sve_vg.
15955 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15956 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15957 aarch64_addr_query_type.
15958 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15959 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15960 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15961 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15962 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15963 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15964 (aarch64_simd_imm_zero_p): Delete.
15965 (aarch64_check_zero_based_sve_index_immediate): Declare.
15966 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15967 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15968 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15969 (aarch64_sve_float_mul_immediate_p): Likewise.
15970 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15971 rather than an rtx.
15972 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15973 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15974 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15975 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15976 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15977 (aarch64_regmode_natural_size): Likewise.
15978 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15979 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15981 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15982 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15983 for VG and the SVE predicate registers.
15984 (V_ALIASES): Add a "z"-prefixed alias.
15985 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15986 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15987 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15988 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15989 (REG_CLASS_NAMES): Add entries for them.
15990 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15991 and the predicate registers.
15992 (aarch64_sve_vg): Declare.
15993 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15994 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15995 (REGMODE_NATURAL_SIZE): Define.
15996 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15998 * config/aarch64/aarch64.c: Include cfgrtl.h.
15999 (simd_immediate_info): Add a constructor for series vectors,
16000 and an associated step field.
16001 (aarch64_sve_vg): New variable.
16002 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16003 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16004 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16005 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16006 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16007 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16008 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16009 (aarch64_get_mask_mode): New functions.
16010 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16011 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16012 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16013 predicate modes and predicate registers. Explicitly restrict
16014 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16015 to store a vector mode if it is recognized by
16016 aarch64_classify_vector_mode.
16017 (aarch64_regmode_natural_size): New function.
16018 (aarch64_hard_regno_caller_save_mode): Return the original mode
16020 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16021 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16022 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16023 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16025 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16026 does not overlap dest if the function is frame-related. Handle
16028 (aarch64_split_add_offset): New function.
16029 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16030 them aarch64_add_offset.
16031 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16032 and update call to aarch64_sub_sp.
16033 (aarch64_add_cfa_expression): New function.
16034 (aarch64_expand_prologue): Pass extra temporary registers to the
16035 functions above. Handle the case in which we need to emit new
16036 DW_CFA_expressions for registers that were originally saved
16037 relative to the stack pointer, but now have to be expressed
16038 relative to the frame pointer.
16039 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16041 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16042 IP0 and IP1 values for SVE frames.
16043 (aarch64_expand_vec_series): New function.
16044 (aarch64_expand_sve_widened_duplicate): Likewise.
16045 (aarch64_expand_sve_const_vector): Likewise.
16046 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16047 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16048 into the register, rather than emitting a SET directly.
16049 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16050 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16051 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16052 (offset_9bit_signed_scaled_p): New functions.
16053 (aarch64_replicate_bitmask_imm): New function.
16054 (aarch64_bitmask_imm): Use it.
16055 (aarch64_cannot_force_const_mem): Reject expressions involving
16056 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16057 (aarch64_classify_index): Handle SVE indices, by requiring
16058 a plain register index with a scale that matches the element size.
16059 (aarch64_classify_address): Handle SVE addresses. Assert that
16060 the mode of the address is VOIDmode or an integer mode.
16061 Update call to aarch64_classify_symbol.
16062 (aarch64_classify_symbolic_expression): Update call to
16063 aarch64_classify_symbol.
16064 (aarch64_const_vec_all_in_range_p): New function.
16065 (aarch64_print_vector_float_operand): Likewise.
16066 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16067 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16068 and the FP immediates 1.0 and 0.5.
16069 (aarch64_print_address_internal): Handle SVE addresses.
16070 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16071 (aarch64_regno_regclass): Handle predicate registers.
16072 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16074 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16075 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16076 (aarch64_convert_sve_vector_bits): New function.
16077 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16078 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16079 rather than an rtx.
16080 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16081 Handle SVE vector and predicate modes. Accept VL-based constants
16082 that need only one temporary register, and VL offsets that require
16083 no temporary registers.
16084 (aarch64_conditional_register_usage): Mark the predicate registers
16085 as fixed if SVE isn't available.
16086 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16087 Return true for SVE vector and predicate modes.
16088 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16089 rather than an unsigned int. Handle SVE modes.
16090 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16092 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16094 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16095 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16096 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16097 (aarch64_sve_float_mul_immediate_p): New functions.
16098 (aarch64_sve_valid_immediate): New function.
16099 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16100 Explicitly reject structure modes. Check for INDEX constants.
16101 Handle PTRUE and PFALSE constants.
16102 (aarch64_check_zero_based_sve_index_immediate): New function.
16103 (aarch64_simd_imm_zero_p): Delete.
16104 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16105 vector modes. Accept constants in the range of CNT[BHWD].
16106 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16107 ask for an Advanced SIMD mode.
16108 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16109 (aarch64_simd_vector_alignment): Handle SVE predicates.
16110 (aarch64_vectorize_preferred_vector_alignment): New function.
16111 (aarch64_simd_vector_alignment_reachable): Use it instead of
16113 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16114 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16116 (MAX_VECT_LEN): Delete.
16117 (expand_vec_perm_d): Add a vec_flags field.
16118 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16119 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16120 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16122 (aarch64_evpc_rev): Rename to...
16123 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16124 (aarch64_evpc_rev_global): New function.
16125 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16126 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16128 (aarch64_evpc_sve_tbl): New function.
16129 (aarch64_expand_vec_perm_const_1): Update after rename of
16130 aarch64_evpc_rev. Handle SVE permutes too, trying
16131 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16132 than aarch64_evpc_tbl.
16133 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16134 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16135 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16136 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16137 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16138 (aarch64_expand_sve_vcond): New functions.
16139 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16140 of aarch64_vector_mode_p.
16141 (aarch64_dwarf_poly_indeterminate_value): New function.
16142 (aarch64_compute_pressure_classes): Likewise.
16143 (aarch64_can_change_mode_class): Likewise.
16144 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16145 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16146 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16147 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16148 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16149 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16150 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16151 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16153 (Dn, Dl, Dr): Accept const as well as const_vector.
16154 (Dz): Likewise. Compare against CONST0_RTX.
16155 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16156 of "vector" where appropriate.
16157 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16158 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16159 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16160 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16161 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16162 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16163 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16164 (v_int_equiv): Extend to SVE modes.
16165 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16167 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16168 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16169 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16170 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16171 (SVE_COND_FP_CMP): New int iterators.
16172 (perm_hilo): Handle the new unpack unspecs.
16173 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16175 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16176 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16177 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16178 (aarch64_equality_operator, aarch64_constant_vector_operand)
16179 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16180 (aarch64_sve_nonimmediate_operand): Likewise.
16181 (aarch64_sve_general_operand): Likewise.
16182 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16183 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16184 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16185 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16186 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16187 (aarch64_sve_float_arith_immediate): Likewise.
16188 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16189 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16190 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16191 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16192 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16193 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16194 (aarch64_sve_float_arith_operand): Likewise.
16195 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16196 (aarch64_sve_float_mul_operand): Likewise.
16197 (aarch64_sve_vec_perm_operand): Likewise.
16198 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16199 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16200 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16201 as well as const_vector.
16202 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16203 in file. Use CONST0_RTX and CONSTM1_RTX.
16204 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16205 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16206 Use aarch64_simd_imm_zero.
16207 * config/aarch64/aarch64-sve.md: New file.
16208 * config/aarch64/aarch64.md: Include it.
16209 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16210 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16211 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16212 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16213 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16214 (sve): New attribute.
16215 (enabled): Disable instructions with the sve attribute unless
16217 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16218 aarch64_expand_mov_immediate.
16219 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16220 CNT[BHSD] immediates.
16221 (movti): Split CONST_POLY_INT moves into two halves.
16222 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16223 Split additions that need a temporary here if the destination
16224 is the stack pointer.
16225 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16226 (*add<mode>3_poly_1): New instruction.
16227 (set_clobber_cc): New expander.
16229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16231 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16232 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16233 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16234 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16235 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16236 Change innermode from fixed_mode_size to machine_mode.
16237 (simplify_subreg): Update call accordingly. Handle a constant-sized
16238 subreg of a variable-length CONST_VECTOR.
16240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16241 Alan Hayward <alan.hayward@arm.com>
16242 David Sherwood <david.sherwood@arm.com>
16244 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16245 (add_offset_to_base): New function, split out from...
16246 (create_mem_ref): ...here. When handling a scale other than 1,
16247 check first whether the address is valid without the offset.
16248 Add it into the base if so, leaving the index and scale as-is.
16250 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16253 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16254 fold_for_warn before checking if arg2 is INTEGER_CST.
16256 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16258 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16259 (store_multiple_operation): Delete.
16260 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16261 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16262 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16263 guarded by TARGET_STRING.
16264 (rs6000_output_load_multiple): Delete.
16265 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16266 OPTION_MASK_STRING / TARGET_STRING handling.
16267 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16268 (const rs6000_opt_masks) <"string">: Change mask to 0.
16269 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16270 (MASK_STRING): Delete.
16271 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16273 (load_multiple): Delete.
16280 (store_multiple): Delete.
16287 (movmemsi_8reg): Delete.
16288 (corresponding unnamed define_insn): Delete.
16289 (movmemsi_6reg): Delete.
16290 (corresponding unnamed define_insn): Delete.
16291 (movmemsi_4reg): Delete.
16292 (corresponding unnamed define_insn): Delete.
16293 (movmemsi_2reg): Delete.
16294 (corresponding unnamed define_insn): Delete.
16295 (movmemsi_1reg): Delete.
16296 (corresponding unnamed define_insn): Delete.
16297 * config/rs6000/rs6000.opt (mno-string): New.
16298 (mstring): Replace by deprecation warning stub.
16299 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16301 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16303 * regrename.c (regrename_do_replace): If replacing the same
16304 reg multiple times, try to reuse last created gen_raw_REG.
16307 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16308 main to workaround a bug in GDB.
16310 2018-01-12 Tom de Vries <tom@codesourcery.com>
16313 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16315 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16317 PR rtl-optimization/80481
16318 * ira-color.c (get_cap_member): New function.
16319 (allocnos_conflict_by_live_ranges_p): Use it.
16320 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16321 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16323 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16326 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16327 (*saddl_se_1): Ditto.
16328 (*ssubsi_1): Ditto.
16329 (*ssubl_se_1): Ditto.
16331 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16333 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16334 rather than wi::to_widest for DR_INITs.
16335 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16336 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16337 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16339 (vect_analyze_group_access_1): Note that here.
16341 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16343 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16344 polynomial type sizes.
16346 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16348 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16349 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16350 (gimple_add_tmp_var): Likewise.
16352 2018-01-12 Martin Liska <mliska@suse.cz>
16354 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16355 (gimple_alloc_sizes): Likewise.
16356 (dump_gimple_statistics): Use PRIu64 in printf format.
16357 * gimple.h: Change uint64_t to int.
16359 2018-01-12 Martin Liska <mliska@suse.cz>
16361 * tree-core.h: Use uint64_t instead of int.
16362 * tree.c (tree_node_counts): Likewise.
16363 (tree_node_sizes): Likewise.
16364 (dump_tree_statistics): Use PRIu64 in printf format.
16366 2018-01-12 Martin Liska <mliska@suse.cz>
16368 * Makefile.in: As qsort_chk is implemented in vec.c, add
16369 vec.o to linkage of gencfn-macros.
16370 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16371 passing the info to record_node_allocation_statistics.
16372 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16374 * ggc-common.c (struct ggc_usage): Add operator== and use
16375 it in operator< and compare function.
16376 * mem-stats.h (struct mem_usage): Likewise.
16377 * vec.c (struct vec_usage): Remove operator< and compare
16378 function. Can be simply inherited.
16380 2018-01-12 Martin Jambor <mjambor@suse.cz>
16383 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16384 * tree-ssa-math-opts.c: Include domwalk.h.
16385 (convert_mult_to_fma_1): New function.
16386 (fma_transformation_info): New type.
16387 (fma_deferring_state): Likewise.
16388 (cancel_fma_deferring): New function.
16389 (result_of_phi): Likewise.
16390 (last_fma_candidate_feeds_initial_phi): Likewise.
16391 (convert_mult_to_fma): Added deferring logic, split actual
16392 transformation to convert_mult_to_fma_1.
16393 (math_opts_dom_walker): New type.
16394 (math_opts_dom_walker::after_dom_children): New method, body moved
16395 here from pass_optimize_widening_mul::execute, added deferring logic
16397 (pass_optimize_widening_mul::execute): Moved most of code to
16398 math_opts_dom_walker::after_dom_children.
16399 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16400 * config/i386/i386.c (ix86_option_override_internal): Added
16401 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16403 2018-01-12 Richard Biener <rguenther@suse.de>
16406 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16407 inline instance vars.
16409 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16412 * config/rx/rx.c (rx_is_restricted_memory_address):
16413 Handle SUBREG case.
16415 2018-01-12 Richard Biener <rguenther@suse.de>
16417 PR tree-optimization/80846
16418 * target.def (split_reduction): New target hook.
16419 * targhooks.c (default_split_reduction): New function.
16420 * targhooks.h (default_split_reduction): Declare.
16421 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16422 target requests first reduce vectors by combining low and high
16424 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16425 (get_vectype_for_scalar_type_and_size): Export.
16426 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16427 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16428 * doc/tm.texi: Regenerate.
16429 * config/i386/i386.c (ix86_split_reduction): Implement
16430 TARGET_VECTORIZE_SPLIT_REDUCTION.
16432 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16435 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16436 in PIC mode except for TARGET_VXWORKS_RTP.
16437 * config/sparc/sparc.c: Include cfgrtl.h.
16438 (TARGET_INIT_PIC_REG): Define.
16439 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16440 (sparc_pic_register_p): New predicate.
16441 (sparc_legitimate_address_p): Use it.
16442 (sparc_legitimize_pic_address): Likewise.
16443 (sparc_delegitimize_address): Likewise.
16444 (sparc_mode_dependent_address_p): Likewise.
16445 (gen_load_pcrel_sym): Remove 4th parameter.
16446 (load_got_register): Adjust call to above. Remove obsolete stuff.
16447 (sparc_expand_prologue): Do not call load_got_register here.
16448 (sparc_flat_expand_prologue): Likewise.
16449 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16450 (sparc_use_pseudo_pic_reg): New function.
16451 (sparc_init_pic_reg): Likewise.
16452 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16453 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16455 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16457 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16458 Add item for branch_cost.
16460 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16462 PR rtl-optimization/83565
16463 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16464 not extend the result to a larger mode for rotate operations.
16465 (num_sign_bit_copies1): Likewise.
16467 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16470 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16472 Use values-Xc.o for -pedantic.
16473 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16475 2018-01-12 Martin Liska <mliska@suse.cz>
16478 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16480 (possible_polymorphic_call_targets): Use it.
16481 (ipa_devirt): Likewise.
16483 2018-01-12 Martin Liska <mliska@suse.cz>
16485 * profile-count.h (enum profile_quality): Use 0 as invalid
16486 enum value of profile_quality.
16488 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16490 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16491 -mext-string options.
16493 2018-01-12 Richard Biener <rguenther@suse.de>
16495 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16496 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16497 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16499 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16501 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16503 * configure.ac (--with-long-double-format): Add support for the
16504 configuration option to change the default long double format on
16506 * config.gcc (powerpc*-linux*-*): Likewise.
16507 * configure: Regenerate.
16508 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16509 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16510 used without modification.
16512 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16514 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16515 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16516 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16517 MISC_BUILTIN_SPEC_BARRIER.
16518 (rs6000_init_builtins): Likewise.
16519 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16521 (speculation_barrier): New define_insn.
16522 * doc/extend.texi: Document __builtin_speculation_barrier.
16524 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16527 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16528 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16529 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16531 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16532 integral modes instead of "ss" and "sd".
16533 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16534 vectors with 32-bit and 64-bit elements.
16535 (vecdupssescalarmodesuffix): New mode attribute.
16536 (vec_dup<mode>): Use it.
16538 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16541 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16542 frame if argument is passed on stack.
16544 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16547 * ree.c (combine_reaching_defs): Optimize also
16548 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16549 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16551 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16553 PR middle-end/83189
16554 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16556 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16558 PR middle-end/83718
16559 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16560 after they are computed.
16562 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16564 PR tree-optimization/83695
16565 * gimple-loop-linterchange.cc
16566 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16567 reset cached scev information after interchange.
16568 (pass_linterchange::execute): Remove call to scev_reset_htab.
16570 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16572 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16573 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16574 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16575 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16576 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16577 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16578 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16579 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16580 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16581 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16582 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16583 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16584 (V_lane_reg): Likewise.
16585 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16587 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16588 (vfmal_lane_low<mode>_intrinsic,
16589 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16590 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16591 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16592 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16593 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16594 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16596 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16598 * config/arm/arm-cpus.in (fp16fml): New feature.
16599 (ALL_SIMD): Add fp16fml.
16600 (armv8.2-a): Add fp16fml as an option.
16601 (armv8.3-a): Likewise.
16602 (armv8.4-a): Add fp16fml as part of fp16.
16603 * config/arm/arm.h (TARGET_FP16FML): Define.
16604 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16606 * config/arm/arm-modes.def (V2HF): Define.
16607 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16608 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16609 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16610 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16611 vfmsl_low, vfmsl_high): New set of builtins.
16612 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16613 (vfml_op): New code attribute.
16614 (VFMLHALVES): New int iterator.
16615 (VFML, VFMLSEL): New mode attributes.
16616 (V_reg): Define mapping for V2HF.
16617 (V_hi, V_lo): New mode attributes.
16618 (VF_constraint): Likewise.
16619 (vfml_half, vfml_half_selector): New int attributes.
16620 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16622 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16623 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16625 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16626 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16627 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16628 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16630 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16631 Document new effective target and option set.
16633 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16635 * config/arm/arm-cpus.in (armv8_4): New feature.
16636 (ARMv8_4a): New fgroup.
16637 (armv8.4-a): New arch.
16638 * config/arm/arm-tables.opt: Regenerate.
16639 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16640 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16641 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16642 Add matching rules for -march=armv8.4-a and extensions.
16643 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16645 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16648 * config/rx/rx.md (BW): New mode attribute.
16649 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16651 2018-01-11 Richard Biener <rguenther@suse.de>
16653 PR tree-optimization/83435
16654 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16655 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16656 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16658 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16659 Alan Hayward <alan.hayward@arm.com>
16660 David Sherwood <david.sherwood@arm.com>
16662 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16664 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16665 (aarch64_print_address_internal): Use it to check for a zero offset.
16667 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16668 Alan Hayward <alan.hayward@arm.com>
16669 David Sherwood <david.sherwood@arm.com>
16671 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16672 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16673 Return a poly_int64 rather than a HOST_WIDE_INT.
16674 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16675 rather than a HOST_WIDE_INT.
16676 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16677 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16678 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16679 final_offset from HOST_WIDE_INT to poly_int64.
16680 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16681 to_constant when getting the number of units in an Advanced SIMD
16683 (aarch64_builtin_vectorized_function): Check for a constant number
16685 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16687 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16688 attribute instead of GET_MODE_NUNITS.
16689 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16690 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16691 GET_MODE_SIZE for fixed-size registers.
16692 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16693 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16694 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16695 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16696 (aarch64_print_operand, aarch64_print_address_internal)
16697 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16698 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16699 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16700 Handle polynomial GET_MODE_SIZE.
16701 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16702 wider than SImode without modification.
16703 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16704 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16705 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16706 passing and returning SVE modes.
16707 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16708 rather than GEN_INT.
16709 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16710 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16711 (aarch64_allocate_and_probe_stack_space): Likewise.
16712 (aarch64_layout_frame): Cope with polynomial offsets.
16713 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16714 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16715 polynomial offsets.
16716 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16717 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16718 poly_int64 rather than a HOST_WIDE_INT.
16719 (aarch64_get_separate_components, aarch64_process_components)
16720 (aarch64_expand_prologue, aarch64_expand_epilogue)
16721 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16722 (aarch64_anchor_offset): New function, split out from...
16723 (aarch64_legitimize_address): ...here.
16724 (aarch64_builtin_vectorization_cost): Handle polynomial
16725 TYPE_VECTOR_SUBPARTS.
16726 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16728 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16729 number of elements from the PARALLEL rather than the mode.
16730 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16731 rather than GET_MODE_BITSIZE.
16732 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16733 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16734 (aarch64_expand_vec_perm_const_1): Handle polynomial
16735 d->perm.length () and d->perm elements.
16736 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16737 Apply to_constant to d->perm elements.
16738 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16739 polynomial CONST_VECTOR_NUNITS.
16740 (aarch64_move_pointer): Take amount as a poly_int64 rather
16742 (aarch64_progress_pointer): Avoid temporary variable.
16743 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16744 the mode attribute instead of GET_MODE.
16746 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16747 Alan Hayward <alan.hayward@arm.com>
16748 David Sherwood <david.sherwood@arm.com>
16750 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16751 x exists before using it.
16752 (aarch64_add_constant_internal): Rename to...
16753 (aarch64_add_offset_1): ...this. Replace regnum with separate
16754 src and dest rtxes. Handle the case in which they're different,
16755 including when the offset is zero. Replace scratchreg with an rtx.
16756 Use 2 additions if there is no spare register into which we can
16757 move a 16-bit constant.
16758 (aarch64_add_constant): Delete.
16759 (aarch64_add_offset): Replace reg with separate src and dest
16760 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16761 Use aarch64_add_offset_1.
16762 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16763 an rtx rather than an int. Take the delta as a poly_int64
16764 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16765 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16766 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16767 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16768 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16769 and aarch64_add_sp.
16770 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16771 aarch64_add_constant.
16773 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16775 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16776 Use scalar_float_mode.
16778 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16780 * config/aarch64/aarch64-simd.md
16781 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16782 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16783 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16784 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16785 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16786 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16787 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16788 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16789 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16790 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16792 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16795 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16796 targ_options->x_arm_arch_string is non NULL.
16798 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16800 * config/aarch64/aarch64.h
16801 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16803 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16806 * expmed.c (emit_store_flag_force): Swap if const op0
16807 and change VOIDmode to mode of op0.
16809 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16811 PR rtl-optimization/83761
16812 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16813 than bytes to mode_for_size.
16815 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16817 PR middle-end/83189
16818 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16819 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16822 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16824 PR middle-end/83575
16825 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16826 when in layout mode.
16827 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16828 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16831 2018-01-10 Michael Collison <michael.collison@arm.com>
16833 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16834 * config/aarch64/aarch64-option-extension.def: Add
16835 AARCH64_OPT_EXTENSION of 'fp16fml'.
16836 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16837 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16838 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16839 * config/aarch64/constraints.md (Ui7): New constraint.
16840 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16841 (VFMLA_SEL_W): Ditto.
16844 (VFMLA16_LOW): New int iterator.
16845 (VFMLA16_HIGH): Ditto.
16846 (UNSPEC_FMLAL): New unspec.
16847 (UNSPEC_FMLSL): Ditto.
16848 (UNSPEC_FMLAL2): Ditto.
16849 (UNSPEC_FMLSL2): Ditto.
16850 (f16mac): New code attribute.
16851 * config/aarch64/aarch64-simd-builtins.def
16852 (aarch64_fmlal_lowv2sf): Ditto.
16853 (aarch64_fmlsl_lowv2sf): Ditto.
16854 (aarch64_fmlalq_lowv4sf): Ditto.
16855 (aarch64_fmlslq_lowv4sf): Ditto.
16856 (aarch64_fmlal_highv2sf): Ditto.
16857 (aarch64_fmlsl_highv2sf): Ditto.
16858 (aarch64_fmlalq_highv4sf): Ditto.
16859 (aarch64_fmlslq_highv4sf): Ditto.
16860 (aarch64_fmlal_lane_lowv2sf): Ditto.
16861 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16862 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16863 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16864 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16865 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16866 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16867 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16868 (aarch64_fmlal_lane_highv2sf): Ditto.
16869 (aarch64_fmlsl_lane_highv2sf): Ditto.
16870 (aarch64_fmlal_laneq_highv2sf): Ditto.
16871 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16872 (aarch64_fmlalq_lane_highv4sf): Ditto.
16873 (aarch64_fmlsl_lane_highv4sf): Ditto.
16874 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16875 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16876 * config/aarch64/aarch64-simd.md:
16877 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16878 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16879 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16880 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16881 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16882 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16883 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16884 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16885 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16886 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16887 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16888 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16889 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16890 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16891 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16892 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16893 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16894 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16895 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16896 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16897 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16898 (vfmlsl_low_u32): Ditto.
16899 (vfmlalq_low_u32): Ditto.
16900 (vfmlslq_low_u32): Ditto.
16901 (vfmlal_high_u32): Ditto.
16902 (vfmlsl_high_u32): Ditto.
16903 (vfmlalq_high_u32): Ditto.
16904 (vfmlslq_high_u32): Ditto.
16905 (vfmlal_lane_low_u32): Ditto.
16906 (vfmlsl_lane_low_u32): Ditto.
16907 (vfmlal_laneq_low_u32): Ditto.
16908 (vfmlsl_laneq_low_u32): Ditto.
16909 (vfmlalq_lane_low_u32): Ditto.
16910 (vfmlslq_lane_low_u32): Ditto.
16911 (vfmlalq_laneq_low_u32): Ditto.
16912 (vfmlslq_laneq_low_u32): Ditto.
16913 (vfmlal_lane_high_u32): Ditto.
16914 (vfmlsl_lane_high_u32): Ditto.
16915 (vfmlal_laneq_high_u32): Ditto.
16916 (vfmlsl_laneq_high_u32): Ditto.
16917 (vfmlalq_lane_high_u32): Ditto.
16918 (vfmlslq_lane_high_u32): Ditto.
16919 (vfmlalq_laneq_high_u32): Ditto.
16920 (vfmlslq_laneq_high_u32): Ditto.
16921 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16922 (AARCH64_FL_FOR_ARCH8_4): New.
16923 (AARCH64_ISA_F16FML): New ISA flag.
16924 (TARGET_F16FML): New feature flag for fp16fml.
16925 (doc/invoke.texi): Document new fp16fml option.
16927 2018-01-10 Michael Collison <michael.collison@arm.com>
16929 * config/aarch64/aarch64-builtins.c:
16930 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16931 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16932 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16933 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16934 (AARCH64_ISA_SHA3): New ISA flag.
16935 (TARGET_SHA3): New feature flag for sha3.
16936 * config/aarch64/iterators.md (sha512_op): New int attribute.
16937 (CRYPTO_SHA512): New int iterator.
16938 (UNSPEC_SHA512H): New unspec.
16939 (UNSPEC_SHA512H2): Ditto.
16940 (UNSPEC_SHA512SU0): Ditto.
16941 (UNSPEC_SHA512SU1): Ditto.
16942 * config/aarch64/aarch64-simd-builtins.def
16943 (aarch64_crypto_sha512hqv2di): New builtin.
16944 (aarch64_crypto_sha512h2qv2di): Ditto.
16945 (aarch64_crypto_sha512su0qv2di): Ditto.
16946 (aarch64_crypto_sha512su1qv2di): Ditto.
16947 (aarch64_eor3qv8hi): Ditto.
16948 (aarch64_rax1qv2di): Ditto.
16949 (aarch64_xarqv2di): Ditto.
16950 (aarch64_bcaxqv8hi): Ditto.
16951 * config/aarch64/aarch64-simd.md:
16952 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16953 (aarch64_crypto_sha512su0qv2di): Ditto.
16954 (aarch64_crypto_sha512su1qv2di): Ditto.
16955 (aarch64_eor3qv8hi): Ditto.
16956 (aarch64_rax1qv2di): Ditto.
16957 (aarch64_xarqv2di): Ditto.
16958 (aarch64_bcaxqv8hi): Ditto.
16959 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16960 (vsha512h2q_u64): Ditto.
16961 (vsha512su0q_u64): Ditto.
16962 (vsha512su1q_u64): Ditto.
16963 (veor3q_u16): Ditto.
16964 (vrax1q_u64): Ditto.
16965 (vxarq_u64): Ditto.
16966 (vbcaxq_u16): Ditto.
16967 * config/arm/types.md (crypto_sha512): New type attribute.
16968 (crypto_sha3): Ditto.
16969 (doc/invoke.texi): Document new sha3 option.
16971 2018-01-10 Michael Collison <michael.collison@arm.com>
16973 * config/aarch64/aarch64-builtins.c:
16974 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16975 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16976 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16977 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16978 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16979 (AARCH64_ISA_SM4): New ISA flag.
16980 (TARGET_SM4): New feature flag for sm4.
16981 * config/aarch64/aarch64-simd-builtins.def
16982 (aarch64_sm3ss1qv4si): Ditto.
16983 (aarch64_sm3tt1aq4si): Ditto.
16984 (aarch64_sm3tt1bq4si): Ditto.
16985 (aarch64_sm3tt2aq4si): Ditto.
16986 (aarch64_sm3tt2bq4si): Ditto.
16987 (aarch64_sm3partw1qv4si): Ditto.
16988 (aarch64_sm3partw2qv4si): Ditto.
16989 (aarch64_sm4eqv4si): Ditto.
16990 (aarch64_sm4ekeyqv4si): Ditto.
16991 * config/aarch64/aarch64-simd.md:
16992 (aarch64_sm3ss1qv4si): Ditto.
16993 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16994 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16995 (aarch64_sm4eqv4si): Ditto.
16996 (aarch64_sm4ekeyqv4si): Ditto.
16997 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16998 (sm3part_op): Ditto.
16999 (CRYPTO_SM3TT): Ditto.
17000 (CRYPTO_SM3PART): Ditto.
17001 (UNSPEC_SM3SS1): New unspec.
17002 (UNSPEC_SM3TT1A): Ditto.
17003 (UNSPEC_SM3TT1B): Ditto.
17004 (UNSPEC_SM3TT2A): Ditto.
17005 (UNSPEC_SM3TT2B): Ditto.
17006 (UNSPEC_SM3PARTW1): Ditto.
17007 (UNSPEC_SM3PARTW2): Ditto.
17008 (UNSPEC_SM4E): Ditto.
17009 (UNSPEC_SM4EKEY): Ditto.
17010 * config/aarch64/constraints.md (Ui2): New constraint.
17011 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17012 * config/arm/types.md (crypto_sm3): New type attribute.
17013 (crypto_sm4): Ditto.
17014 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17015 (vsm3tt1aq_u32): Ditto.
17016 (vsm3tt1bq_u32): Ditto.
17017 (vsm3tt2aq_u32): Ditto.
17018 (vsm3tt2bq_u32): Ditto.
17019 (vsm3partw1q_u32): Ditto.
17020 (vsm3partw2q_u32): Ditto.
17021 (vsm4eq_u32): Ditto.
17022 (vsm4ekeyq_u32): Ditto.
17023 (doc/invoke.texi): Document new sm4 option.
17025 2018-01-10 Michael Collison <michael.collison@arm.com>
17027 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17028 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17029 (AARCH64_FL_FOR_ARCH8_4): New.
17030 (AARCH64_FL_V8_4): New flag.
17031 (doc/invoke.texi): Document new armv8.4-a option.
17033 2018-01-10 Michael Collison <michael.collison@arm.com>
17035 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17036 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17037 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17038 * config/aarch64/aarch64-option-extension.def: Add
17039 AARCH64_OPT_EXTENSION of 'sha2'.
17040 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17041 (crypto): Disable sha2 and aes if crypto disabled.
17042 (crypto): Enable aes and sha2 if enabled.
17043 (simd): Disable sha2 and aes if simd disabled.
17044 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17046 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17047 (TARGET_SHA2): New feature flag for sha2.
17048 (TARGET_AES): New feature flag for aes.
17049 * config/aarch64/aarch64-simd.md:
17050 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17051 conditional on TARGET_AES.
17052 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17053 (aarch64_crypto_sha1hsi): Make pattern conditional
17055 (aarch64_crypto_sha1hv4si): Ditto.
17056 (aarch64_be_crypto_sha1hv4si): Ditto.
17057 (aarch64_crypto_sha1su1v4si): Ditto.
17058 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17059 (aarch64_crypto_sha1su0v4si): Ditto.
17060 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17061 (aarch64_crypto_sha256su0v4si): Ditto.
17062 (aarch64_crypto_sha256su1v4si): Ditto.
17063 (doc/invoke.texi): Document new aes and sha2 options.
17065 2018-01-10 Martin Sebor <msebor@redhat.com>
17067 PR tree-optimization/83781
17068 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17071 2018-01-11 Martin Sebor <msebor@gmail.com>
17072 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17074 PR tree-optimization/83501
17075 PR tree-optimization/81703
17077 * tree-ssa-strlen.c (get_string_cst): Rename...
17078 (get_string_len): ...to this. Handle global constants.
17079 (handle_char_store): Adjust.
17081 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17082 Jim Wilson <jimw@sifive.com>
17084 * config/riscv/riscv-protos.h (riscv_output_return): New.
17085 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17086 (riscv_attribute_table, riscv_output_return),
17087 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17088 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17089 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17091 (riscv_expand_prologue): Add early return for naked function.
17092 (riscv_expand_epilogue): Likewise.
17093 (riscv_function_ok_for_sibcall): Return false for naked function.
17094 (riscv_set_current_function): New.
17095 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17096 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17097 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17098 * doc/extend.texi (RISC-V Function Attributes): New.
17100 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17102 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17103 check for 128-bit long double before checking TCmode.
17104 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17105 128-bit long doubles before checking TFmode or TCmode.
17106 (FLOAT128_IBM_P): Likewise.
17108 2018-01-10 Martin Sebor <msebor@redhat.com>
17110 PR tree-optimization/83671
17111 * builtins.c (c_strlen): Unconditionally return zero for the empty
17113 Use -Warray-bounds for warnings.
17114 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17115 for non-constant array indices with COMPONENT_REF, arrays of
17116 arrays, and pointers to arrays.
17117 (gimple_fold_builtin_strlen): Determine and set length range for
17118 non-constant character arrays.
17120 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17122 PR middle-end/81897
17123 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17126 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17128 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17130 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17133 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17134 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17135 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17136 indexed_or_indirect_operand predicate.
17137 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17138 (*vsx_le_perm_load_v8hi): Likewise.
17139 (*vsx_le_perm_load_v16qi): Likewise.
17140 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17141 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17142 (*vsx_le_perm_store_v8hi): Likewise.
17143 (*vsx_le_perm_store_v16qi): Likewise.
17144 (eight unnamed splitters): Likewise.
17146 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17148 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17149 * config/rs6000/emmintrin.h: Likewise.
17150 * config/rs6000/mmintrin.h: Likewise.
17151 * config/rs6000/xmmintrin.h: Likewise.
17153 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17156 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17158 * tree.c (tree_nop_conversion): Return true for location wrapper
17160 (maybe_wrap_with_location): New function.
17161 (selftest::check_strip_nops): New function.
17162 (selftest::test_location_wrappers): New function.
17163 (selftest::tree_c_tests): Call it.
17164 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17165 (maybe_wrap_with_location): New decl.
17166 (EXPR_LOCATION_WRAPPER_P): New macro.
17167 (location_wrapper_p): New inline function.
17168 (tree_strip_any_location_wrapper): New inline function.
17170 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17173 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17174 stack_realign_offset for the largest alignment of stack slot
17176 (ix86_find_max_used_stack_alignment): New function.
17177 (ix86_finalize_stack_frame_flags): Use it. Set
17178 max_used_stack_alignment if we don't realign stack.
17179 * config/i386/i386.h (machine_function): Add
17180 max_used_stack_alignment.
17182 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17184 * config/arm/arm.opt (-mbranch-cost): New option.
17185 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17188 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17191 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17192 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17194 2018-01-10 Richard Biener <rguenther@suse.de>
17197 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17198 early out so it also covers the case where we have a non-NULL
17201 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17203 PR tree-optimization/83753
17204 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17205 for non-strided grouped accesses if the number of elements is 1.
17207 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17210 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17211 * i386.h (TARGET_USE_GATHER): Define.
17212 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17214 2018-01-10 Martin Liska <mliska@suse.cz>
17217 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17218 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17220 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17221 CLEANUP_NO_PARTITIONING is not set.
17223 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17225 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17226 for vectors, as a partial revert of r254296.
17227 * rtl.h (const_vec_p): Delete.
17228 (const_vec_duplicate_p): Don't test for vector CONSTs.
17229 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17230 * expmed.c (make_tree): Likewise.
17233 * common.md (E, F): Use CONSTANT_P instead of checking for
17235 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17236 checking for CONST_VECTOR.
17238 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17240 PR middle-end/83575
17241 * predict.c (force_edge_cold): Handle in more sane way edges
17242 with no prediction.
17244 2018-01-09 Carl Love <cel@us.ibm.com>
17246 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17248 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17249 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17250 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17251 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17252 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17253 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17254 * config/rs6000/rs6000-protos.h: Add extern defition for
17255 rs6000_generate_float2_double_code.
17256 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17258 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17259 (float2_v2df): Add define_expand.
17261 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17264 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17265 op_mode in the force_to_mode call.
17267 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17269 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17270 instead of checking each element individually.
17271 (aarch64_evpc_uzp): Likewise.
17272 (aarch64_evpc_zip): Likewise.
17273 (aarch64_evpc_ext): Likewise.
17274 (aarch64_evpc_rev): Likewise.
17275 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17276 instead of checking each element individually. Return true without
17278 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17279 whether all selected elements come from the same input, instead of
17280 checking each element individually. Remove calls to gen_rtx_REG,
17281 start_sequence and end_sequence and instead assert that no rtl is
17284 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17286 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17287 order of HIGH and CONST checks.
17289 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17291 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17292 if the destination isn't an SSA_NAME.
17294 2018-01-09 Richard Biener <rguenther@suse.de>
17296 PR tree-optimization/83668
17297 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17299 (canonicalize_loop_form): ... here, renamed from ...
17300 (canonicalize_loop_closed_ssa_form): ... this and amended to
17301 swap successor edges for loop exit blocks to make us use
17302 the RPO order we need for initial schedule generation.
17304 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17306 PR tree-optimization/64811
17307 * match.pd: When optimizing comparisons with Inf, avoid
17308 introducing or losing exceptions from comparisons with NaN.
17310 2018-01-09 Martin Liska <mliska@suse.cz>
17313 * asan.c (shadow_mem_size): Add gcc_assert.
17315 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17317 Don't save registers in main().
17320 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17321 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17322 * config/avr/avr.c (avr_set_current_function): Don't error if
17323 naked, OS_task or OS_main are specified at the same time.
17324 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17326 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17328 * common/config/avr/avr-common.c (avr_option_optimization_table):
17329 Switch on -mmain-is-OS_task for optimizing compilations.
17331 2018-01-09 Richard Biener <rguenther@suse.de>
17333 PR tree-optimization/83572
17334 * graphite.c: Include cfganal.h.
17335 (graphite_transform_loops): Connect infinite loops to exit
17336 and remove fake edges at the end.
17338 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17340 * ipa-inline.c (edge_badness): Revert accidental checkin.
17342 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17345 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17346 symbols; not inline clones.
17348 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17351 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17352 hard registers. Formatting fixes.
17354 PR preprocessor/83722
17355 * gcc.c (try_generate_repro): Pass
17356 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17357 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17360 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17361 Kito Cheng <kito.cheng@gmail.com>
17363 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17364 (riscv_leaf_function_p): Delete.
17365 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17367 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17369 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17371 (do_ifelse): New function.
17372 (do_isel): New function.
17373 (do_sub3): New function.
17374 (do_add3): New function.
17375 (do_load_mask_compare): New function.
17376 (do_overlap_load_compare): New function.
17377 (expand_compare_loop): New function.
17378 (expand_block_compare): Call expand_compare_loop() when appropriate.
17379 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17380 option description.
17381 (-mblock-compare-inline-loop-limit): New option.
17383 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17386 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17387 Reverse order of second and third operands in first alternative.
17388 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17389 of first and second elements in UNSPEC_VPERMR vector.
17390 (altivec_expand_vec_perm_le): Likewise.
17392 2018-01-08 Jeff Law <law@redhat.com>
17394 PR rtl-optimizatin/81308
17395 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17396 (process_switch): If group_case_labels makes a change, then set
17398 (pass_convert_switch::execute): If a switch is converted, then
17399 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17401 PR rtl-optimization/81308
17402 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17405 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17407 PR target/83663 - Revert r255946
17408 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17409 generation for cases where splatting a value is not useful.
17410 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17411 across a vec_duplicate and a paradoxical subreg forming a vector
17412 mode to a vec_concat.
17414 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17416 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17417 -march=armv8.3-a variants.
17418 * config/arm/t-multilib: Likewise.
17419 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17421 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17423 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17425 (cceq_ior_compare_complement): Give it a name so I can use it, and
17426 change boolean_or_operator predicate to boolean_operator so it can
17427 be used to generate a crand.
17428 (eqne): New code iterator.
17429 (bd/bd_neg): New code_attrs.
17430 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17431 a single define_insn.
17432 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17433 decrement (bdnzt/bdnzf/bdzt/bdzf).
17434 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17435 with the new names of the branch decrement patterns, and added the
17436 names of the branch decrement conditional patterns.
17438 2018-01-08 Richard Biener <rguenther@suse.de>
17440 PR tree-optimization/83563
17441 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17444 2018-01-08 Richard Biener <rguenther@suse.de>
17446 PR middle-end/83713
17447 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17449 2018-01-08 Richard Biener <rguenther@suse.de>
17451 PR tree-optimization/83685
17452 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17453 references to abnormals.
17455 2018-01-08 Richard Biener <rguenther@suse.de>
17458 * dwarf2out.c (output_indirect_strings): Handle empty
17459 skeleton_debug_str_hash.
17460 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17462 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17464 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17465 (emit_store_direct): Likewise.
17466 (arc_trampoline_adjust_address): Likewise.
17467 (arc_asm_trampoline_template): New function.
17468 (arc_initialize_trampoline): Use asm_trampoline_template.
17469 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17470 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17471 * config/arc/arc.md (flush_icache): Delete pattern.
17473 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17475 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17476 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17479 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17482 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17483 by not USED_FOR_TARGET.
17484 (make_pass_resolve_sw_modes): Likewise.
17486 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17488 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17491 2018-01-08 Richard Biener <rguenther@suse.de>
17493 PR middle-end/83580
17494 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17496 2018-01-08 Richard Biener <rguenther@suse.de>
17498 PR middle-end/83517
17499 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17501 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17503 PR middle-end/81897
17504 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17505 basic blocks with a small number of successors.
17506 (convert_control_dep_chain_into_preds): Improve handling of
17508 (dump_predicates): Split apart into...
17509 (dump_pred_chain): ...here...
17510 (dump_pred_info): ...and here.
17511 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17512 (can_chain_union_be_invalidated_p): Improve check for invalidation
17514 (uninit_uses_cannot_happen): Avoid unnecessary if
17515 convert_control_dep_chain_into_preds yielded nothing.
17517 2018-01-06 Martin Sebor <msebor@redhat.com>
17519 PR tree-optimization/83640
17520 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17521 subtracting negative offset from size.
17522 (builtin_access::overlap): Adjust offset bounds of the access to fall
17523 within the size of the object if possible.
17525 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17527 PR rtl-optimization/83699
17528 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17529 extract_bit_field_as_subreg to cases in which the extracted
17530 value is also a vector.
17532 * lra-constraints.c (process_alt_operands): Test for the equivalence
17533 substitutions when detecting a possible reload cycle.
17535 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17538 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17539 by default if flag_selective_schedling{,2}. Formatting fixes.
17541 PR rtl-optimization/83682
17542 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17543 if it has non-VECTOR_MODE element mode.
17544 (vec_duplicate_p): Likewise.
17546 PR middle-end/83694
17547 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17548 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17550 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17553 * config/i386/i386-builtin.def
17554 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17555 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17556 Require also OPTION_MASK_ISA_AVX512F in addition to
17557 OPTION_MASK_ISA_GFNI.
17558 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17559 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17560 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17561 to OPTION_MASK_ISA_GFNI.
17562 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17563 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17564 OPTION_MASK_ISA_AVX512BW.
17565 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17566 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17567 addition to OPTION_MASK_ISA_GFNI.
17568 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17569 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17570 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17571 to OPTION_MASK_ISA_GFNI.
17572 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17573 a requirement for all ISAs rather than any of them with a few
17575 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17577 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17578 bitmasks to be enabled with 3 exceptions, instead of requiring any
17579 enabled ISA with lots of exceptions.
17580 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17581 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17582 Change avx512bw in isa attribute to avx512f.
17583 * config/i386/sgxintrin.h: Add license boilerplate.
17584 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17585 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17586 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17587 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17589 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17590 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17591 temporarily sse2 rather than sse if not enabled already.
17594 * config/i386/sse.md (VI248_VLBW): Rename to ...
17595 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17596 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17597 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17598 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17599 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17600 mode iterator instead of VI248_VLBW.
17602 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17604 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17605 (record_modified): Skip clobbers; add debug output.
17606 (param_change_prob): Use sreal frequencies.
17608 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17610 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17611 punt for user-aligned variables.
17613 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17615 * tree-chrec.c (chrec_contains_symbols): Return true for
17618 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17621 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17622 of (x|y) == x for BICS pattern.
17624 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17626 PR tree-optimization/83605
17627 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17628 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17631 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17633 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17634 * config/epiphany/rtems.h: New file.
17636 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17637 Uros Bizjak <ubizjak@gmail.com>
17640 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17641 QIreg_operand instead of register_operand predicate.
17642 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17643 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17644 comments instead of -fmitigate[-_]rop.
17646 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17649 * cgraphunit.c (symbol_table::compile): Switch to text_section
17650 before calling assembly_start debug hook.
17651 * run-rtl-passes.c (run_rtl_passes): Likewise.
17654 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17656 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17657 range_int_cst_p rather than !symbolic_range_p before calling
17658 extract_range_from_multiplicative_op_1.
17660 2018-01-04 Jeff Law <law@redhat.com>
17662 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17663 redundant test in assertion.
17665 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17667 * doc/rtl.texi: Document machine_mode wrapper classes.
17669 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17671 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17672 using tree_to_uhwi.
17674 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17676 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17677 the VEC_PERM_EXPR fold to fail.
17679 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17682 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17683 to switched_sections.
17685 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17688 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17689 test for d.testing.
17691 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17694 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17695 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17697 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17700 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17701 is BLKmode and bitpos not zero or mode change is needed.
17703 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17706 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17709 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17712 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17713 instead of MULT rtx. Update all corresponding splitters.
17714 (*saddl_se): Ditto.
17715 (*ssub<modesuffix>): Ditto.
17716 (*ssubl_se): Ditto.
17717 (*cmp_sadd_di): Update split patterns.
17718 (*cmp_sadd_si): Ditto.
17719 (*cmp_sadd_sidi): Ditto.
17720 (*cmp_ssub_di): Ditto.
17721 (*cmp_ssub_si): Ditto.
17722 (*cmp_ssub_sidi): Ditto.
17723 * config/alpha/predicates.md (const23_operand): New predicate.
17724 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17725 Look for ASHIFT, not MULT inner operand.
17726 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17728 2018-01-04 Martin Liska <mliska@suse.cz>
17730 PR gcov-profile/83669
17731 * gcov.c (output_intermediate_file): Add version to intermediate
17733 * doc/gcov.texi: Document new field 'version' in intermediate
17734 file format. Fix location of '-k' option of gcov command.
17736 2018-01-04 Martin Liska <mliska@suse.cz>
17739 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17741 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17743 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17745 2018-01-03 Martin Sebor <msebor@redhat.com>
17747 PR tree-optimization/83655
17748 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17749 checking calls with invalid arguments.
17751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17753 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17754 (vectorizable_mask_load_store): Delete.
17755 (vectorizable_call): Return false for masked loads and stores.
17756 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17757 instead of gimple_assign_rhs1.
17758 (vectorizable_load): Handle IFN_MASK_LOAD.
17759 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17763 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17765 (vectorizable_mask_load_store): ...here.
17766 (vectorizable_load): ...and here.
17768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17770 * tree-vect-stmts.c (vect_build_all_ones_mask)
17771 (vect_build_zero_merge_argument): New functions, split out from...
17772 (vectorizable_load): ...here.
17774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17776 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17778 (vectorizable_mask_load_store): ...here.
17779 (vectorizable_store): ...and here.
17781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17783 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17785 (vectorizable_mask_load_store): ...here.
17787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17789 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17790 (vect_model_store_cost): Take a vec_load_store_type instead of a
17792 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17793 (vect_model_store_cost): Take a vec_load_store_type instead of a
17795 (vectorizable_mask_load_store): Update accordingly.
17796 (vectorizable_store): Likewise.
17797 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17801 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17802 IFN_MASK_LOAD calls here rather than...
17803 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17806 Alan Hayward <alan.hayward@arm.com>
17807 David Sherwood <david.sherwood@arm.com>
17809 * expmed.c (extract_bit_field_1): For vector extracts,
17810 fall back to extract_bit_field_as_subreg if vec_extract
17813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17814 Alan Hayward <alan.hayward@arm.com>
17815 David Sherwood <david.sherwood@arm.com>
17817 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17818 they are variable or constant sized.
17819 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17820 slots for constant-sized data.
17822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17823 Alan Hayward <alan.hayward@arm.com>
17824 David Sherwood <david.sherwood@arm.com>
17826 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17827 handling COND_EXPRs with boolean comparisons, try to find a better
17828 basis for the mask type than the boolean itself.
17830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17832 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17833 is calculated and how it can be overridden.
17834 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17835 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17837 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17841 Alan Hayward <alan.hayward@arm.com>
17842 David Sherwood <david.sherwood@arm.com>
17844 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17845 Remove the mode argument.
17846 (aarch64_simd_valid_immediate): Remove the mode and inverse
17848 * config/aarch64/iterators.md (bitsize): New iterator.
17849 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17850 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17851 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17852 aarch64_simd_valid_immediate.
17853 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17854 (aarch64_reg_or_bic_imm): Likewise.
17855 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17856 with an insn_type enum and msl with a modifier_type enum.
17857 Replace element_width with a scalar_mode. Change the shift
17858 to unsigned int. Add constructors for scalar_float_mode and
17859 scalar_int_mode elements.
17860 (aarch64_vect_float_const_representable_p): Delete.
17861 (aarch64_can_const_movi_rtx_p)
17862 (aarch64_simd_scalar_immediate_valid_for_move)
17863 (aarch64_simd_make_constant): Update call to
17864 aarch64_simd_valid_immediate.
17865 (aarch64_advsimd_valid_immediate_hs): New function.
17866 (aarch64_advsimd_valid_immediate): Likewise.
17867 (aarch64_simd_valid_immediate): Remove mode and inverse
17868 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17869 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17870 and aarch64_float_const_representable_p on the result.
17871 (aarch64_output_simd_mov_immediate): Remove mode argument.
17872 Update call to aarch64_simd_valid_immediate and use of
17873 simd_immediate_info.
17874 (aarch64_output_scalar_simd_mov_immediate): Update call
17877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17878 Alan Hayward <alan.hayward@arm.com>
17879 David Sherwood <david.sherwood@arm.com>
17881 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17882 (mode_nunits): Likewise CONST_MODE_NUNITS.
17883 * machmode.def (ADJUST_NUNITS): Document.
17884 * genmodes.c (mode_data::need_nunits_adj): New field.
17885 (blank_mode): Update accordingly.
17886 (adj_nunits): New variable.
17887 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17889 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17890 listed in adj_nunits.
17891 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17892 listed in adj_nunits. Don't emit case statements for such modes.
17893 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17894 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17895 nothing if adj_nunits is nonnull.
17896 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17897 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17898 (emit_mode_fbit): Update use of print_maybe_const_decl.
17899 (emit_move_size): Likewise. Treat the array as non-const
17901 (emit_mode_adjustments): Handle adj_nunits.
17903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17905 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17906 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17907 (VECTOR_MODES): Use it.
17908 (make_vector_modes): Take the prefix as an argument.
17910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17911 Alan Hayward <alan.hayward@arm.com>
17912 David Sherwood <david.sherwood@arm.com>
17914 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17915 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17916 for MODE_VECTOR_BOOL.
17917 * machmode.def (VECTOR_BOOL_MODE): Document.
17918 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17919 (make_vector_bool_mode): New function.
17920 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17922 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17923 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17925 * stor-layout.c (int_mode_for_mode): Likewise.
17926 * tree.c (build_vector_type_for_mode): Likewise.
17927 * varasm.c (output_constant_pool_2): Likewise.
17928 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17929 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17930 for MODE_VECTOR_BOOL.
17931 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17932 of mode class checks.
17933 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17934 instead of a list of mode class checks.
17935 (expand_vector_scalar_condition): Likewise.
17936 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17939 Alan Hayward <alan.hayward@arm.com>
17940 David Sherwood <david.sherwood@arm.com>
17942 * machmode.h (mode_size): Change from unsigned short to
17944 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17945 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17946 or if measurement_type is not polynomial.
17947 (fixed_size_mode::includes_p): Check for constant-sized modes.
17948 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17949 return a poly_uint16 rather than an unsigned short.
17950 (emit_mode_size): Change the type of mode_size from unsigned short
17951 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17952 (emit_mode_adjustments): Cope with polynomial vector sizes.
17953 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17955 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17957 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17958 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17959 * caller-save.c (setup_save_areas): Likewise.
17960 (replace_reg_with_saved_mem): Likewise.
17961 * calls.c (emit_library_call_value_1): Likewise.
17962 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17963 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17964 (gen_lowpart_for_combine): Likewise.
17965 * convert.c (convert_to_integer_1): Likewise.
17966 * cse.c (equiv_constant, cse_insn): Likewise.
17967 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17968 (cselib_subst_to_values): Likewise.
17969 * dce.c (word_dce_process_block): Likewise.
17970 * df-problems.c (df_word_lr_mark_ref): Likewise.
17971 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17972 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17973 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17974 (rtl_for_decl_location): Likewise.
17975 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17976 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17977 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17978 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17979 (expand_expr_real_1): Likewise.
17980 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17981 (pad_below): Likewise.
17982 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17983 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17984 * ira.c (get_subreg_tracking_sizes): Likewise.
17985 * ira-build.c (ira_create_allocno_objects): Likewise.
17986 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17987 (ira_sort_regnos_for_alter_reg): Likewise.
17988 * ira-costs.c (record_operand_costs): Likewise.
17989 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17990 (resolve_simple_move): Likewise.
17991 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17992 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17993 (lra_constraints): Likewise.
17994 (CONST_POOL_OK_P): Reject variable-sized modes.
17995 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17996 (add_pseudo_to_slot, lra_spill): Likewise.
17997 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17998 * optabs-query.c (get_best_extraction_insn): Likewise.
17999 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18000 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18001 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18002 * recog.c (offsettable_address_addr_space_p): Likewise.
18003 * regcprop.c (maybe_mode_change): Likewise.
18004 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18005 * regrename.c (build_def_use): Likewise.
18006 * regstat.c (dump_reg_info): Likewise.
18007 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18008 (find_reloads, find_reloads_subreg_address): Likewise.
18009 * reload1.c (eliminate_regs_1): Likewise.
18010 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18011 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18012 (simplify_binary_operation_1, simplify_subreg): Likewise.
18013 * targhooks.c (default_function_arg_padding): Likewise.
18014 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18015 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18016 (verify_gimple_assign_ternary): Likewise.
18017 * tree-inline.c (estimate_move_cost): Likewise.
18018 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18019 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18020 (get_address_cost_ainc): Likewise.
18021 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18022 (vect_supportable_dr_alignment): Likewise.
18023 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18024 (vectorizable_reduction): Likewise.
18025 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18026 (vectorizable_operation, vectorizable_load): Likewise.
18027 * tree.c (build_same_sized_truth_vector_type): Likewise.
18028 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18029 * var-tracking.c (emit_note_insn_var_location): Likewise.
18030 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18031 (ADDR_VEC_ALIGN): Likewise.
18033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18034 Alan Hayward <alan.hayward@arm.com>
18035 David Sherwood <david.sherwood@arm.com>
18037 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18039 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18040 or if measurement_type is polynomial.
18041 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18042 * combine.c (make_extraction): Likewise.
18043 * dse.c (find_shift_sequence): Likewise.
18044 * dwarf2out.c (mem_loc_descriptor): Likewise.
18045 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18046 (extract_bit_field, extract_low_bits): Likewise.
18047 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18048 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18049 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18050 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18051 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18052 * reload.c (find_reloads): Likewise.
18053 * reload1.c (alter_reg): Likewise.
18054 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18055 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18056 * tree-if-conv.c (predicate_mem_writes): Likewise.
18057 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18058 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18059 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18060 * valtrack.c (dead_debug_insert_temp): Likewise.
18061 * varasm.c (mergeable_constant_section): Likewise.
18062 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18065 Alan Hayward <alan.hayward@arm.com>
18066 David Sherwood <david.sherwood@arm.com>
18068 * expr.c (expand_assignment): Cope with polynomial mode sizes
18069 when assigning to a CONCAT.
18071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18072 Alan Hayward <alan.hayward@arm.com>
18073 David Sherwood <david.sherwood@arm.com>
18075 * machmode.h (mode_precision): Change from unsigned short to
18077 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18079 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18080 or if measurement_type is not polynomial.
18081 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18082 in which the mode is already known to be a scalar_int_mode.
18083 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18084 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18086 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18087 for GET_MODE_PRECISION.
18088 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18089 for GET_MODE_PRECISION.
18090 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18092 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18093 (expand_field_assignment, make_extraction): Likewise.
18094 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18095 (get_last_value): Likewise.
18096 * convert.c (convert_to_integer_1): Likewise.
18097 * cse.c (cse_insn): Likewise.
18098 * expr.c (expand_expr_real_1): Likewise.
18099 * lra-constraints.c (simplify_operand_subreg): Likewise.
18100 * optabs-query.c (can_atomic_load_p): Likewise.
18101 * optabs.c (expand_atomic_load): Likewise.
18102 (expand_atomic_store): Likewise.
18103 * ree.c (combine_reaching_defs): Likewise.
18104 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18105 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18106 * tree.h (type_has_mode_precision_p): Likewise.
18107 * ubsan.c (instrument_si_overflow): Likewise.
18109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18110 Alan Hayward <alan.hayward@arm.com>
18111 David Sherwood <david.sherwood@arm.com>
18113 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18114 polynomial numbers of units.
18115 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18116 (valid_vector_subparts_p): New function.
18117 (build_vector_type): Remove temporary shim and take the number
18118 of units as a poly_uint64 rather than an int.
18119 (build_opaque_vector_type): Take the number of units as a
18120 poly_uint64 rather than an int.
18121 * tree.c (build_vector_from_ctor): Handle polynomial
18122 TYPE_VECTOR_SUBPARTS.
18123 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18124 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18125 (build_vector_from_val): If the number of units is variable,
18126 use build_vec_duplicate_cst for constant operands and
18127 VEC_DUPLICATE_EXPR otherwise.
18128 (make_vector_type): Remove temporary is_constant ().
18129 (build_vector_type, build_opaque_vector_type): Take the number of
18130 units as a poly_uint64 rather than an int.
18131 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18133 * cfgexpand.c (expand_debug_expr): Likewise.
18134 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18135 (store_constructor, expand_expr_real_1): Likewise.
18136 (const_scalar_mask_from_tree): Likewise.
18137 * fold-const-call.c (fold_const_reduction): Likewise.
18138 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18139 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18140 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18141 (fold_relational_const): Likewise.
18142 (native_interpret_vector): Likewise. Change the size from an
18143 int to an unsigned int.
18144 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18145 TYPE_VECTOR_SUBPARTS.
18146 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18147 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18148 duplicating a non-constant operand into a variable-length vector.
18149 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18150 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18151 * ipa-icf.c (sem_variable::equals): Likewise.
18152 * match.pd: Likewise.
18153 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18154 * print-tree.c (print_node): Likewise.
18155 * stor-layout.c (layout_type): Likewise.
18156 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18157 * tree-cfg.c (verify_gimple_comparison): Likewise.
18158 (verify_gimple_assign_binary): Likewise.
18159 (verify_gimple_assign_ternary): Likewise.
18160 (verify_gimple_assign_single): Likewise.
18161 * tree-pretty-print.c (dump_generic_node): Likewise.
18162 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18163 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18164 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18165 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18166 (vect_shift_permute_load_chain): Likewise.
18167 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18168 (expand_vector_condition, optimize_vector_constructor): Likewise.
18169 (lower_vec_perm, get_compute_type): Likewise.
18170 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18171 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18172 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18173 (vect_recog_mask_conversion_pattern): Likewise.
18174 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18175 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18176 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18177 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18178 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18179 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18180 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18181 (supportable_widening_operation): Likewise.
18182 (supportable_narrowing_operation): Likewise.
18183 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18185 * varasm.c (output_constant): Likewise.
18187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18188 Alan Hayward <alan.hayward@arm.com>
18189 David Sherwood <david.sherwood@arm.com>
18191 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18192 so that both the length == 3 and length != 3 cases set up their
18193 own permute vectors. Add comments explaining why we know the
18194 number of elements is constant.
18195 (vect_permute_load_chain): Likewise.
18197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18198 Alan Hayward <alan.hayward@arm.com>
18199 David Sherwood <david.sherwood@arm.com>
18201 * machmode.h (mode_nunits): Change from unsigned char to
18203 (ONLY_FIXED_SIZE_MODES): New macro.
18204 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18205 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18206 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18208 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18209 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18210 or if measurement_type is not polynomial.
18211 * genmodes.c (ZERO_COEFFS): New macro.
18212 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18214 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18215 Use ZERO_COEFFS when emitting initializers.
18216 * data-streamer.h (bp_pack_poly_value): New function.
18217 (bp_unpack_poly_value): Likewise.
18218 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18219 for GET_MODE_NUNITS.
18220 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18221 for GET_MODE_NUNITS.
18222 * tree.c (make_vector_type): Remove temporary shim and make
18223 the real function take the number of units as a poly_uint64
18224 rather than an int.
18225 (build_vector_type_for_mode): Handle polynomial nunits.
18226 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18227 * emit-rtl.c (const_vec_series_p_1): Likewise.
18228 (gen_rtx_CONST_VECTOR): Likewise.
18229 * fold-const.c (test_vec_duplicate_folding): Likewise.
18230 * genrecog.c (validate_pattern): Likewise.
18231 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18232 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18233 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18234 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18235 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18236 * rtlanal.c (subreg_get_info): Likewise.
18237 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18238 (vect_grouped_load_supported): Likewise.
18239 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18240 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18241 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18242 (simplify_const_unary_operation, simplify_binary_operation_1)
18243 (simplify_const_binary_operation, simplify_ternary_operation)
18244 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18245 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18246 instead of CONST_VECTOR_NUNITS.
18247 * varasm.c (output_constant_pool_2): Likewise.
18248 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18249 explicit-encoded elements in the XVEC for variable-length vectors.
18251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18253 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18256 Alan Hayward <alan.hayward@arm.com>
18257 David Sherwood <david.sherwood@arm.com>
18259 * coretypes.h (fixed_size_mode): Declare.
18260 (fixed_size_mode_pod): New typedef.
18261 * builtins.h (target_builtins::x_apply_args_mode)
18262 (target_builtins::x_apply_result_mode): Change type to
18263 fixed_size_mode_pod.
18264 * builtins.c (apply_args_size, apply_result_size, result_vector)
18265 (expand_builtin_apply_args_1, expand_builtin_apply)
18266 (expand_builtin_return): Update accordingly.
18268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18270 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18271 * cselib.c (cselib_hash_rtx): Likewise.
18272 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18273 CONST_VECTOR encoding.
18275 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18276 Jeff Law <law@redhat.com>
18279 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18280 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18281 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18282 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18285 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18286 explicitly probe *sp in a noreturn function if there were any callee
18287 register saves or frame pointer is needed.
18289 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18292 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18293 BLKmode for ternary, binary or unary expressions.
18296 * var-tracking.c (delete_vta_debug_insn): New inline function.
18297 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18298 insns from get_insns () to NULL instead of each bb separately.
18299 Use delete_vta_debug_insn. No longer static.
18300 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18301 delete_vta_debug_insns callers.
18302 * rtl.h (delete_vta_debug_insns): Declare.
18303 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18304 instead of variable_tracking_main.
18306 2018-01-03 Martin Sebor <msebor@redhat.com>
18308 PR tree-optimization/83603
18309 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18310 arguments past the endof the argument list in functions declared
18311 without a prototype.
18312 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18313 Avoid checking when arguments are null.
18315 2018-01-03 Martin Sebor <msebor@redhat.com>
18318 * doc/extend.texi (attribute const): Fix a typo.
18319 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18320 issuing -Wsuggest-attribute for void functions.
18322 2018-01-03 Martin Sebor <msebor@redhat.com>
18324 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18325 offset_int::from instead of wide_int::to_shwi.
18326 (maybe_diag_overlap): Remove assertion.
18327 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18328 * gimple-ssa-sprintf.c (format_directive): Same.
18329 (parse_directive): Same.
18330 (sprintf_dom_walker::compute_format_length): Same.
18331 (try_substitute_return_value): Same.
18333 2018-01-03 Jeff Law <law@redhat.com>
18335 PR middle-end/83654
18336 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18337 non-constant residual for zero at runtime and avoid probing in
18338 that case. Reorganize code for trailing problem to mirror handling
18341 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18343 PR tree-optimization/83501
18344 * tree-ssa-strlen.c (get_string_cst): New.
18345 (handle_char_store): Call get_string_cst.
18347 2018-01-03 Martin Liska <mliska@suse.cz>
18349 PR tree-optimization/83593
18350 * tree-ssa-strlen.c: Include tree-cfg.h.
18351 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18352 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18353 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18355 (strlen_dom_walker::before_dom_children): Call
18356 gimple_purge_dead_eh_edges. Dump tranformation with details
18358 (strlen_dom_walker::before_dom_children): Update call by adding
18359 new argument cleanup_eh.
18360 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18362 2018-01-03 Martin Liska <mliska@suse.cz>
18365 * cif-code.def (VARIADIC_THUNK): New enum value.
18366 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18369 2018-01-03 Jan Beulich <jbeulich@suse.com>
18371 * sse.md (mov<mode>_internal): Tighten condition for when to use
18372 vmovdqu<ssescalarsize> for TI and OI modes.
18374 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18376 Update copyright years.
18378 2018-01-03 Martin Liska <mliska@suse.cz>
18381 * ipa-visibility.c (function_and_variable_visibility): Skip
18382 functions with noipa attribure.
18384 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18386 * gcc.c (process_command): Update copyright notice dates.
18387 * gcov-dump.c (print_version): Ditto.
18388 * gcov.c (print_version): Ditto.
18389 * gcov-tool.c (print_version): Ditto.
18390 * gengtype.c (create_file): Ditto.
18391 * doc/cpp.texi: Bump @copying's copyright year.
18392 * doc/cppinternals.texi: Ditto.
18393 * doc/gcc.texi: Ditto.
18394 * doc/gccint.texi: Ditto.
18395 * doc/gcov.texi: Ditto.
18396 * doc/install.texi: Ditto.
18397 * doc/invoke.texi: Ditto.
18399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18401 * vector-builder.h (vector_builder::m_full_nelts): Change from
18402 unsigned int to poly_uint64.
18403 (vector_builder::full_nelts): Update prototype accordingly.
18404 (vector_builder::new_vector): Likewise.
18405 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18406 (vector_builder::operator ==): Likewise.
18407 (vector_builder::finalize): Likewise.
18408 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18409 Take the number of elements as a poly_uint64 rather than an
18411 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18412 from unsigned int to poly_uint64.
18413 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18414 (vec_perm_indices::new_vector): Likewise.
18415 (vec_perm_indices::length): Likewise.
18416 (vec_perm_indices::nelts_per_input): Likewise.
18417 (vec_perm_indices::input_nelts): Likewise.
18418 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18419 number of elements per input as a poly_uint64 rather than an
18420 unsigned int. Use the original encoding for variable-length
18421 vectors, rather than clamping each individual element.
18422 For the second and subsequent elements in each pattern,
18423 clamp the step and base before clamping their sum.
18424 (vec_perm_indices::series_p): Handle polynomial element counts.
18425 (vec_perm_indices::all_in_range_p): Likewise.
18426 (vec_perm_indices_to_tree): Likewise.
18427 (vec_perm_indices_to_rtx): Likewise.
18428 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18429 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18430 (tree_vector_builder::new_binary_operation): Handle polynomial
18431 element counts. Return false if we need to know the number
18432 of elements at compile time.
18433 * fold-const.c (fold_vec_perm): Punt if the number of elements
18434 isn't known at compile time.
18436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18438 * vec-perm-indices.h (vec_perm_builder): Change element type
18439 from HOST_WIDE_INT to poly_int64.
18440 (vec_perm_indices::element_type): Update accordingly.
18441 (vec_perm_indices::clamp): Handle polynomial element_types.
18442 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18443 (vec_perm_indices::all_in_range_p): Likewise.
18444 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18446 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18447 polynomial vec_perm_indices element types.
18448 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18449 * fold-const.c (fold_vec_perm): Likewise.
18450 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18451 * tree-vect-generic.c (lower_vec_perm): Likewise.
18452 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18453 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18454 element type to HOST_WIDE_INT.
18456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18457 Alan Hayward <alan.hayward@arm.com>
18458 David Sherwood <david.sherwood@arm.com>
18460 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18461 rather than an int. Use plus_constant.
18462 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18463 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18466 Alan Hayward <alan.hayward@arm.com>
18467 David Sherwood <david.sherwood@arm.com>
18469 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18470 a HOST_WIDE_INT to a poly_int64.
18472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18473 Alan Hayward <alan.hayward@arm.com>
18474 David Sherwood <david.sherwood@arm.com>
18476 * calls.c (load_register_parameters): Cope with polynomial
18477 mode sizes. Require a constant size for BLKmode parameters
18478 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18479 forces a parameter to be padded at the lsb end in order to
18480 fill a complete number of words, require the parameter size
18481 to be ordered wrt UNITS_PER_WORD.
18483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18484 Alan Hayward <alan.hayward@arm.com>
18485 David Sherwood <david.sherwood@arm.com>
18487 * reload1.c (spill_stack_slot_width): Change element type
18488 from unsigned int to poly_uint64_pod.
18489 (alter_reg): Treat mode sizes as polynomial.
18491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18492 Alan Hayward <alan.hayward@arm.com>
18493 David Sherwood <david.sherwood@arm.com>
18495 * reload.c (complex_word_subreg_p): New function.
18496 (reload_inner_reg_of_subreg, push_reload): Use it.
18498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18499 Alan Hayward <alan.hayward@arm.com>
18500 David Sherwood <david.sherwood@arm.com>
18502 * lra-constraints.c (process_alt_operands): Reject matched
18503 operands whose sizes aren't ordered.
18504 (match_reload): Refer to this check here.
18506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18507 Alan Hayward <alan.hayward@arm.com>
18508 David Sherwood <david.sherwood@arm.com>
18510 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18511 that the mode size is in the set {1, 2, 4, 8, 16}.
18513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18514 Alan Hayward <alan.hayward@arm.com>
18515 David Sherwood <david.sherwood@arm.com>
18517 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18518 Use plus_constant instead of gen_rtx_PLUS.
18520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18521 Alan Hayward <alan.hayward@arm.com>
18522 David Sherwood <david.sherwood@arm.com>
18524 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18525 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18526 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18527 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18528 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18529 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18530 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18531 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18532 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18533 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18535 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18536 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18537 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18538 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18539 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18540 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18541 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18542 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18543 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18544 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18546 * expr.c (emit_move_resolve_push): Treat the input and result
18547 of PUSH_ROUNDING as a poly_int64.
18548 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18549 (emit_push_insn): Likewise.
18550 * lra-eliminations.c (mark_not_eliminable): Likewise.
18551 * recog.c (push_operand): Likewise.
18552 * reload1.c (elimination_effects): Likewise.
18553 * rtlanal.c (nonzero_bits1): Likewise.
18554 * calls.c (store_one_arg): Likewise. Require the padding to be
18555 known at compile time.
18557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18558 Alan Hayward <alan.hayward@arm.com>
18559 David Sherwood <david.sherwood@arm.com>
18561 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18562 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18569 rather than an int.
18571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18572 Alan Hayward <alan.hayward@arm.com>
18573 David Sherwood <david.sherwood@arm.com>
18575 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18576 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18577 via stack temporaries. Treat the mode size as polynomial too.
18579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18580 Alan Hayward <alan.hayward@arm.com>
18581 David Sherwood <david.sherwood@arm.com>
18583 * expr.c (expand_expr_real_2): When handling conversions involving
18584 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18585 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18586 as a poly_uint64 too.
18588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18589 Alan Hayward <alan.hayward@arm.com>
18590 David Sherwood <david.sherwood@arm.com>
18592 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18595 Alan Hayward <alan.hayward@arm.com>
18596 David Sherwood <david.sherwood@arm.com>
18598 * combine.c (can_change_dest_mode): Handle polynomial
18599 REGMODE_NATURAL_SIZE.
18600 * expmed.c (store_bit_field_1): Likewise.
18601 * expr.c (store_constructor): Likewise.
18602 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18603 and polynomial REGMODE_NATURAL_SIZE.
18604 (gen_lowpart_common): Likewise.
18605 * reginfo.c (record_subregs_of_mode): Likewise.
18606 * rtlanal.c (read_modify_subreg_p): Likewise.
18608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18609 Alan Hayward <alan.hayward@arm.com>
18610 David Sherwood <david.sherwood@arm.com>
18612 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18613 numbers of elements.
18615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18616 Alan Hayward <alan.hayward@arm.com>
18617 David Sherwood <david.sherwood@arm.com>
18619 * match.pd: Cope with polynomial numbers of vector elements.
18621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18622 Alan Hayward <alan.hayward@arm.com>
18623 David Sherwood <david.sherwood@arm.com>
18625 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18626 in a POINTER_PLUS_EXPR.
18628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18629 Alan Hayward <alan.hayward@arm.com>
18630 David Sherwood <david.sherwood@arm.com>
18632 * omp-simd-clone.c (simd_clone_subparts): New function.
18633 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18634 (ipa_simd_modify_function_body): Likewise.
18636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18637 Alan Hayward <alan.hayward@arm.com>
18638 David Sherwood <david.sherwood@arm.com>
18640 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18641 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18642 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18643 (expand_vector_condition, vector_element): Likewise.
18644 (subparts_gt): New function.
18645 (get_compute_type): Use subparts_gt.
18646 (count_type_subparts): Delete.
18647 (expand_vector_operations_1): Use subparts_gt instead of
18648 count_type_subparts.
18650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18651 Alan Hayward <alan.hayward@arm.com>
18652 David Sherwood <david.sherwood@arm.com>
18654 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18655 (vect_compile_time_alias): ...this new function. Do the calculation
18656 on poly_ints rather than trees.
18657 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
18665 (vect_schedule_slp_instance): Likewise.
18667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18668 Alan Hayward <alan.hayward@arm.com>
18669 David Sherwood <david.sherwood@arm.com>
18671 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18672 constant and extern definitions for variable-length vectors.
18673 (vect_get_constant_vectors): Note that the number of units
18674 is known to be constant.
18676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18677 Alan Hayward <alan.hayward@arm.com>
18678 David Sherwood <david.sherwood@arm.com>
18680 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18681 of units as polynomial. Choose between WIDE and NARROW based
18684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18685 Alan Hayward <alan.hayward@arm.com>
18686 David Sherwood <david.sherwood@arm.com>
18688 * tree-vect-stmts.c (simd_clone_subparts): New function.
18689 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18692 Alan Hayward <alan.hayward@arm.com>
18693 David Sherwood <david.sherwood@arm.com>
18695 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18696 vectors as polynomial. Use build_index_vector for
18697 IFN_GOMP_SIMD_LANE.
18699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18700 Alan Hayward <alan.hayward@arm.com>
18701 David Sherwood <david.sherwood@arm.com>
18703 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18704 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18705 for variable-length vectors.
18706 (vectorizable_mask_load_store): Treat the number of units as
18707 polynomial, asserting that it is constant if the condition has
18708 already been enforced.
18709 (vectorizable_store, vectorizable_load): Likewise.
18711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18712 Alan Hayward <alan.hayward@arm.com>
18713 David Sherwood <david.sherwood@arm.com>
18715 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18716 of units as polynomial. Punt if we can't tell at compile time
18717 which vector contains the final result.
18719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18720 Alan Hayward <alan.hayward@arm.com>
18721 David Sherwood <david.sherwood@arm.com>
18723 * tree-vect-loop.c (vectorizable_induction): Treat the number
18724 of units as polynomial. Punt on SLP inductions. Use an integer
18725 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18726 cast of such a series for variable-length floating-point
18729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18730 Alan Hayward <alan.hayward@arm.com>
18731 David Sherwood <david.sherwood@arm.com>
18733 * tree.h (build_index_vector): Declare.
18734 * tree.c (build_index_vector): New function.
18735 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18736 of units as polynomial, forcibly converting it to a constant if
18737 vectorizable_reduction has already enforced the condition.
18738 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18739 to create a {1,2,3,...} vector.
18740 (vectorizable_reduction): Treat the number of units as polynomial.
18741 Choose vectype_in based on the largest scalar element size rather
18742 than the smallest number of units. Enforce the restrictions
18745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18746 Alan Hayward <alan.hayward@arm.com>
18747 David Sherwood <david.sherwood@arm.com>
18749 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18750 number of units as polynomial.
18752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18753 Alan Hayward <alan.hayward@arm.com>
18754 David Sherwood <david.sherwood@arm.com>
18756 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18757 * target.def (autovectorize_vector_sizes): Return the vector sizes
18758 by pointer, using vector_sizes rather than a bitmask.
18759 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18760 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18761 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18763 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18764 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18765 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18766 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18767 * omp-general.c (omp_max_vf): Likewise.
18768 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18769 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18770 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18771 * tree-vect-slp.c (vect_slp_bb): Likewise.
18772 * doc/tm.texi: Regenerate.
18773 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18775 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18776 the vector size as a poly_uint64 rather than an unsigned int.
18777 (current_vector_size): Change from an unsigned int to a poly_uint64.
18778 (get_vectype_for_scalar_type): Update accordingly.
18779 * tree.h (build_truth_vector_type): Take the size and number of
18780 units as a poly_uint64 rather than an unsigned int.
18781 (build_vector_type): Add a temporary overload that takes
18782 the number of units as a poly_uint64 rather than an unsigned int.
18783 * tree.c (make_vector_type): Likewise.
18784 (build_truth_vector_type): Take the number of units as a poly_uint64
18785 rather than an unsigned int.
18787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18788 Alan Hayward <alan.hayward@arm.com>
18789 David Sherwood <david.sherwood@arm.com>
18791 * target.def (get_mask_mode): Take the number of units and length
18792 as poly_uint64s rather than unsigned ints.
18793 * targhooks.h (default_get_mask_mode): Update accordingly.
18794 * targhooks.c (default_get_mask_mode): Likewise.
18795 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18796 * doc/tm.texi: Regenerate.
18798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18799 Alan Hayward <alan.hayward@arm.com>
18800 David Sherwood <david.sherwood@arm.com>
18802 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18803 * omp-general.c (omp_max_vf): Likewise.
18804 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18805 (expand_omp_simd): Handle polynomial safelen.
18806 * omp-low.c (omplow_simd_context): Add a default constructor.
18807 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18808 (lower_rec_simd_input_clauses): Update accordingly.
18809 (lower_rec_input_clauses): Likewise.
18811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18812 Alan Hayward <alan.hayward@arm.com>
18813 David Sherwood <david.sherwood@arm.com>
18815 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18816 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18817 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18818 (vect_analyze_slp_cost): Likewise.
18819 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18820 (vect_model_load_cost): Likewise.
18822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18823 Alan Hayward <alan.hayward@arm.com>
18824 David Sherwood <david.sherwood@arm.com>
18826 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18827 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18828 from an unsigned int * to a poly_uint64_pod *.
18829 (calculate_unrolling_factor): New function.
18830 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18833 Alan Hayward <alan.hayward@arm.com>
18834 David Sherwood <david.sherwood@arm.com>
18836 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18837 from an unsigned int to a poly_uint64.
18838 (_loop_vec_info::slp_unrolling_factor): Likewise.
18839 (_loop_vec_info::vectorization_factor): Change from an int
18841 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18842 (vect_get_num_vectors): New function.
18843 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18844 (vect_get_num_copies): Use vect_get_num_vectors.
18845 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18846 to an unsigned int *.
18847 (vect_analyze_data_refs): Change min_vf from an int * to a
18849 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18850 than an unsigned HOST_WIDE_INT.
18851 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18852 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18853 to an unsigned int *.
18854 (vect_analyze_data_ref_dependences): Likewise.
18855 (vect_compute_data_ref_alignment): Handle polynomial vf.
18856 (vect_enhance_data_refs_alignment): Likewise.
18857 (vect_prune_runtime_alias_test_list): Likewise.
18858 (vect_shift_permute_load_chain): Likewise.
18859 (vect_supportable_dr_alignment): Likewise.
18860 (dependence_distance_ge_vf): Take the vectorization factor as a
18861 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18862 (vect_analyze_data_refs): Change min_vf from an int * to a
18864 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18865 vfm1 as a poly_uint64 rather than an int. Make the same change
18866 for the returned bound_scalar.
18867 (vect_gen_vector_loop_niters): Handle polynomial vf.
18868 (vect_do_peeling): Likewise. Update call to
18869 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18870 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18872 * tree-vect-loop.c (vect_determine_vectorization_factor)
18873 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18874 (vect_get_known_peeling_cost): Likewise.
18875 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18876 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18877 (vect_transform_loop): Likewise. Use the lowest possible VF when
18878 updating the upper bounds of the loop.
18879 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18880 rather than an int.
18881 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18882 polynomial unroll factors.
18883 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18884 (vect_make_slp_decision): Likewise.
18885 (vect_supported_load_permutation_p): Likewise, and polynomial
18887 (vect_analyze_slp_cost): Handle polynomial vf.
18888 (vect_slp_analyze_node_operations): Likewise.
18889 (vect_slp_analyze_bb_1): Likewise.
18890 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18891 than an unsigned HOST_WIDE_INT.
18892 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18893 (vectorizable_load): Handle polynomial vf.
18894 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18896 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18899 Alan Hayward <alan.hayward@arm.com>
18900 David Sherwood <david.sherwood@arm.com>
18902 * match.pd: Handle bit operations involving three constants
18903 and try to fold one pair.
18905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18907 * tree-vect-loop-manip.c: Include gimple-fold.h.
18908 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18909 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18910 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18911 Add a path that uses a step of VF instead of 1, but disable it
18913 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18914 and niters_no_overflow parameters. Update calls to
18915 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18916 Create a new SSA name if the latter choses to use a ste other
18917 than zero, and return it via niters_vector_mult_vf_var.
18918 * tree-vect-loop.c (vect_transform_loop): Update calls to
18919 vect_do_peeling, vect_gen_vector_loop_niters and
18920 slpeel_make_loop_iterate_ntimes.
18921 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18922 (vect_gen_vector_loop_niters): Update declarations after above changes.
18924 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18926 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18927 128-bit round to integer instructions.
18928 (ceil<mode>2): Likewise.
18929 (btrunc<mode>2): Likewise.
18930 (round<mode>2): Likewise.
18932 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18934 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18935 unaligned VSX load/store on P8/P9.
18936 (expand_block_clear): Allow the use of unaligned VSX
18937 load/store on P8/P9.
18939 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18941 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18943 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18944 swap associated with both a load and a store.
18946 2018-01-02 Andrew Waterman <andrew@sifive.com>
18948 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18949 * config/riscv/riscv.md (clear_cache): Use it.
18951 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18953 * web.c: Remove out-of-date comment.
18955 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18957 * expr.c (fixup_args_size_notes): Check that any existing
18958 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18959 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18960 (emit_single_push_insn): ...here.
18962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18964 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18965 (const_vector_encoded_nelts): New function.
18966 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18967 (const_vector_int_elt, const_vector_elt): Declare.
18968 * emit-rtl.c (const_vector_int_elt_1): New function.
18969 (const_vector_elt): Likewise.
18970 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18971 of CONST_VECTOR_ELT.
18973 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18975 * expr.c: Include rtx-vector-builder.h.
18976 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18977 directly on the tree encoding.
18978 (const_vector_from_tree): Likewise.
18979 * optabs.c: Include rtx-vector-builder.h.
18980 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18981 sequence of "u" values.
18982 * vec-perm-indices.c: Include rtx-vector-builder.h.
18983 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18984 directly on the vec_perm_indices encoding.
18986 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18988 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18989 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18990 * rtx-vector-builder.h: New file.
18991 * rtx-vector-builder.c: Likewise.
18992 * rtl.h (rtx_def::u2): Add a const_vector field.
18993 (CONST_VECTOR_NPATTERNS): New macro.
18994 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18995 (CONST_VECTOR_DUPLICATE_P): Likewise.
18996 (CONST_VECTOR_STEPPED_P): Likewise.
18997 (CONST_VECTOR_ENCODED_ELT): Likewise.
18998 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18999 (unwrap_const_vec_duplicate): Likewise.
19000 (const_vec_series_p): Check for a non-duplicated vector encoding.
19001 Say that the function only returns true for integer vectors.
19002 * emit-rtl.c: Include rtx-vector-builder.h.
19003 (gen_const_vec_duplicate_1): Delete.
19004 (gen_const_vector): Call gen_const_vec_duplicate instead of
19005 gen_const_vec_duplicate_1.
19006 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19007 (gen_const_vec_duplicate): Use rtx_vector_builder.
19008 (gen_const_vec_series): Likewise.
19009 (gen_rtx_CONST_VECTOR): Likewise.
19010 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19011 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19012 Build a new vector rather than modifying a CONST_VECTOR in-place.
19013 (handle_special_swappables): Update call accordingly.
19014 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19015 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19016 Build a new vector rather than modifying a CONST_VECTOR in-place.
19017 (handle_special_swappables): Update call accordingly.
19019 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19021 * simplify-rtx.c (simplify_const_binary_operation): Use
19022 CONST_VECTOR_ELT instead of XVECEXP.
19024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19026 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19027 the selector elements to be different from the data elements
19028 if the selector is a VECTOR_CST.
19029 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19030 ssizetype for the selector.
19032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19034 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19035 before testing each element individually.
19036 * tree-vect-generic.c (lower_vec_perm): Likewise.
19038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19040 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19041 * selftest-run-tests.c (selftest::run_tests): Call it.
19042 * vector-builder.h (vector_builder::operator ==): New function.
19043 (vector_builder::operator !=): Likewise.
19044 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19045 (vec_perm_indices::all_from_input_p): New function.
19046 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19047 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19048 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19049 instead of reading the VECTOR_CST directly. Detect whether both
19050 vector inputs are the same before constructing the vec_perm_indices,
19051 and update the number of inputs argument accordingly. Use the
19052 utility functions added above. Only construct sel2 if we need to.
19054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19056 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19057 the broadcast of the low byte.
19058 (expand_mult_highpart): Use an explicit encoding for the permutes.
19059 * optabs-query.c (can_mult_highpart_p): Likewise.
19060 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19061 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19062 (vectorizable_bswap): Likewise.
19063 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19064 explicit encoding for the power-of-2 permutes.
19065 (vect_permute_store_chain): Likewise.
19066 (vect_grouped_load_supported): Likewise.
19067 (vect_permute_load_chain): Likewise.
19069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19071 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19072 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19073 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19074 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19075 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19076 (vect_gen_perm_mask_any): Likewise.
19078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19080 * int-vector-builder.h: New file.
19081 * vec-perm-indices.h: Include int-vector-builder.h.
19082 (vec_perm_indices): Redefine as an int_vector_builder.
19083 (auto_vec_perm_indices): Delete.
19084 (vec_perm_builder): Redefine as a stand-alone class.
19085 (vec_perm_indices::vec_perm_indices): New function.
19086 (vec_perm_indices::clamp): Likewise.
19087 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19088 (vec_perm_indices::new_vector): New function.
19089 (vec_perm_indices::new_expanded_vector): Update for new
19090 vec_perm_indices class.
19091 (vec_perm_indices::rotate_inputs): New function.
19092 (vec_perm_indices::all_in_range_p): Operate directly on the
19093 encoded form, without computing elided elements.
19094 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19095 encoding. Update for new vec_perm_indices class.
19096 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19097 the given vec_perm_builder.
19098 (expand_vec_perm_var): Update vec_perm_builder constructor.
19099 (expand_mult_highpart): Use vec_perm_builder instead of
19100 auto_vec_perm_indices.
19101 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19102 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19103 or double series encoding as appropriate.
19104 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19105 vec_perm_indices instead of auto_vec_perm_indices.
19106 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19107 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19108 (vect_permute_store_chain): Likewise.
19109 (vect_grouped_load_supported): Likewise.
19110 (vect_permute_load_chain): Likewise.
19111 (vect_shift_permute_load_chain): Likewise.
19112 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19113 (vect_transform_slp_perm_load): Likewise.
19114 (vect_schedule_slp_instance): Likewise.
19115 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19116 (vectorizable_mask_load_store): Likewise.
19117 (vectorizable_bswap): Likewise.
19118 (vectorizable_store): Likewise.
19119 (vectorizable_load): Likewise.
19120 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19121 vec_perm_indices instead of auto_vec_perm_indices. Use
19122 tree_to_vec_perm_builder to read the vector from a tree.
19123 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19124 vec_perm_builder instead of a vec_perm_indices.
19125 (have_whole_vector_shift): Use vec_perm_builder and
19126 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19127 truncation to calc_vec_perm_mask_for_shift.
19128 (vect_create_epilog_for_reduction): Likewise.
19129 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19130 from auto_vec_perm_indices to vec_perm_indices.
19131 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19132 instead of changing individual elements.
19133 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19134 the vector in d.perm.
19135 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19136 from auto_vec_perm_indices to vec_perm_indices.
19137 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19138 instead of changing individual elements.
19139 (arm_vectorize_vec_perm_const): Use new_vector to install
19140 the vector in d.perm.
19141 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19142 Update vec_perm_builder constructor.
19143 (rs6000_expand_interleave): Likewise.
19144 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19145 (rs6000_expand_interleave): Likewise.
19147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19149 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19150 to qimode could truncate the indices.
19151 * optabs.c (expand_vec_perm_var): Likewise.
19153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19155 * Makefile.in (OBJS): Add vec-perm-indices.o.
19156 * vec-perm-indices.h: New file.
19157 * vec-perm-indices.c: Likewise.
19158 * target.h (vec_perm_indices): Replace with a forward class
19160 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19161 * optabs.h: Include vec-perm-indices.h.
19162 (expand_vec_perm): Delete.
19163 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19164 (expand_vec_perm_const): Declare.
19165 * target.def (vec_perm_const_ok): Replace with...
19166 (vec_perm_const): ...this new hook.
19167 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19168 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19169 * doc/tm.texi: Regenerate.
19170 * optabs.def (vec_perm_const): Delete.
19171 * doc/md.texi (vec_perm_const): Likewise.
19172 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19173 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19174 expand_vec_perm for constant permutation vectors. Assert that
19175 the mode of variable permutation vectors is the integer equivalent
19176 of the mode that is being permuted.
19177 * optabs-query.h (selector_fits_mode_p): Declare.
19178 * optabs-query.c: Include vec-perm-indices.h.
19179 (selector_fits_mode_p): New function.
19180 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19181 is defined, instead of checking whether the vec_perm_const_optab
19182 exists. Use targetm.vectorize.vec_perm_const instead of
19183 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19184 fit in the vector mode before using a variable permute.
19185 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19186 vec_perm_indices instead of an rtx.
19187 (expand_vec_perm): Replace with...
19188 (expand_vec_perm_const): ...this new function. Take the selector
19189 as a vec_perm_indices rather than an rtx. Also take the mode of
19190 the selector. Update call to shift_amt_for_vec_perm_mask.
19191 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19192 Use vec_perm_indices::new_expanded_vector to expand the original
19193 selector into bytes. Check whether the indices fit in the vector
19194 mode before using a variable permute.
19195 (expand_vec_perm_var): Make global.
19196 (expand_mult_highpart): Use expand_vec_perm_const.
19197 * fold-const.c: Includes vec-perm-indices.h.
19198 * tree-ssa-forwprop.c: Likewise.
19199 * tree-vect-data-refs.c: Likewise.
19200 * tree-vect-generic.c: Likewise.
19201 * tree-vect-loop.c: Likewise.
19202 * tree-vect-slp.c: Likewise.
19203 * tree-vect-stmts.c: Likewise.
19204 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19206 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19207 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19208 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19209 (aarch64_vectorize_vec_perm_const): ...this new function.
19210 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19211 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19212 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19213 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19214 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19215 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19216 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19218 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19219 check for NEON modes.
19220 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19221 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19222 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19223 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19225 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19226 the old VEC_PERM_CONST conditions.
19227 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19228 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19229 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19230 (ia64_vectorize_vec_perm_const_ok): Merge into...
19231 (ia64_vectorize_vec_perm_const): ...this new function.
19232 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19233 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19234 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19235 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19236 * config/mips/mips.c (mips_expand_vec_perm_const)
19237 (mips_vectorize_vec_perm_const_ok): Merge into...
19238 (mips_vectorize_vec_perm_const): ...this new function.
19239 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19240 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19241 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19242 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19243 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19244 (rs6000_expand_vec_perm_const): Delete.
19245 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19247 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19248 (altivec_expand_vec_perm_const_le): Take each operand individually.
19249 Operate on constant selectors rather than rtxes.
19250 (altivec_expand_vec_perm_const): Likewise. Update call to
19251 altivec_expand_vec_perm_const_le.
19252 (rs6000_expand_vec_perm_const): Delete.
19253 (rs6000_vectorize_vec_perm_const_ok): Delete.
19254 (rs6000_vectorize_vec_perm_const): New function.
19255 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19256 an element count and rtx array.
19257 (rs6000_expand_extract_even): Update call accordingly.
19258 (rs6000_expand_interleave): Likewise.
19259 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19260 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19261 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19262 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19263 (rs6000_expand_vec_perm_const): Delete.
19264 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19265 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19266 (altivec_expand_vec_perm_const_le): Take each operand individually.
19267 Operate on constant selectors rather than rtxes.
19268 (altivec_expand_vec_perm_const): Likewise. Update call to
19269 altivec_expand_vec_perm_const_le.
19270 (rs6000_expand_vec_perm_const): Delete.
19271 (rs6000_vectorize_vec_perm_const_ok): Delete.
19272 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19273 reference to the SPE evmerge intructions.
19274 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19275 an element count and rtx array.
19276 (rs6000_expand_extract_even): Update call accordingly.
19277 (rs6000_expand_interleave): Likewise.
19278 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19279 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19281 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19285 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19286 vector mode and that that mode matches the mode of the data
19288 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19289 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19290 directly using expand_vec_perm_1 when forcing selectors into
19292 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19294 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19296 * optabs-query.h (can_vec_perm_p): Delete.
19297 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19298 * optabs-query.c (can_vec_perm_p): Split into...
19299 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19300 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19301 particular selector is valid.
19302 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19303 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19304 (vect_grouped_load_supported): Likewise.
19305 (vect_shift_permute_load_chain): Likewise.
19306 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19307 (vect_transform_slp_perm_load): Likewise.
19308 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19309 (vectorizable_bswap): Likewise.
19310 (vect_gen_perm_mask_checked): Likewise.
19311 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19312 implementations of variable permutation vectors into account
19313 when deciding which selector to use.
19314 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19315 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19316 with a false third argument.
19317 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19318 to test whether the constant selector is valid and can_vec_perm_var_p
19319 to test whether a variable selector is valid.
19321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19323 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19324 * optabs-query.c (can_vec_perm_p): Likewise.
19325 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19326 instead of vec_perm_indices.
19327 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19328 (vect_gen_perm_mask_checked): Likewise,
19329 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19330 (vect_gen_perm_mask_checked): Likewise,
19332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19334 * optabs-query.h (qimode_for_vec_perm): Declare.
19335 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19336 (qimode_for_vec_perm): ...this new function.
19337 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19339 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19341 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19342 does not have a conditional at the top.
19344 2018-01-02 Richard Biener <rguenther@suse.de>
19346 * ipa-inline.c (big_speedup_p): Fix expression.
19348 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19351 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19354 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19358 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19359 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19360 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19361 cond_taken_branch_cost 3->4.
19363 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19365 PR tree-optimization/83581
19366 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19367 TODO_cleanup_cfg if any changes have been made.
19369 PR middle-end/83608
19370 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19371 convert_modes if target mode has the right side, but different mode
19374 PR middle-end/83609
19375 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19376 last argument when extracting from CONCAT. If either from_real or
19377 from_imag is NULL, use expansion through memory. If result is not
19378 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19379 the parts directly to inner mode, if even that fails, use expansion
19382 PR middle-end/83623
19383 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19384 check for bswap in mode rather than HImode and use that in expand_unop
19387 Copyright (C) 2018 Free Software Foundation, Inc.
19389 Copying and distribution of this file, with or without modification,
19390 are permitted in any medium without royalty provided the copyright
19391 notice and this notice are preserved.