1 2018-04-02 Jim Wilson <jimw@sifive.com>
3 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5 (<optab>di3, <optab>si3_extend): Likewise.
6 (<optab>si3_mask, <optab>si3_mask_1): New.
7 (<optab>di3_mask, <optab>di3_mask_1): New.
8 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
12 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
14 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
17 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
19 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20 (nds32_canonicalize_comparison): New function.
22 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
23 Kito Cheng <kito.cheng@gmail.com>
24 Kuan-Lin Chen <kuanlinchentw@gmail.com>
26 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
27 * config/nds32/constants.md (unspec_volatile_element): Add
28 UNSPEC_VOLATILE_RELAX_GROUP.
29 * config/nds32/nds32-relax-opt.c: New file.
30 * config/nds32/nds32-predicates.c
31 (nds32_symbol_load_store_p): New function.
32 * config/nds32/nds32-protos.h
33 (nds32_symbol_load_store_p): Declare function.
34 (make_pass_nds32_relax_opt): Declare new rtl pass function.
35 * config/nds32/nds32.c
36 (nds32_register_pass): New function to register pass.
37 (nds32_register_passes): New function to register passes.
38 * config/nds32/nds32.md (relax_group): New pattern.
39 * config/nds32/nds32.opt (mrelax-hint): New option.
40 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
42 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
44 * config/nds32/t-nds32: Modify files dependency.
46 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
48 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
49 (PROFILE_HOOK): Define its implementation.
51 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
53 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
56 2018-04-01 Jakub Jelinek <jakub@redhat.com>
59 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
60 (V_128_256): New mode iterator.
61 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
62 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
63 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
65 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
66 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
68 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
71 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
74 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
77 * config/rs6000/vsx.md (??r): New mode attribute.
78 (*vsx_mov<mode>_64bit): Use it.
79 (*vsx_mov<mode>_32bit): Likewise.
81 2018-03-30 Martin Sebor <msebor@redhat.com>
83 PR tree-optimization/84818
84 * builtins.c (check_access): Use warning_n.
86 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
89 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
91 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
94 2018-03-30 Julia Koval <julia.koval@intel.com>
97 * x86-tune.def (movx, partial_reg_dependency): Enable for
100 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
103 * lra-constraints.c (process_alt_operands): Move setting
104 this_alternative_matches below.
106 2018-03-29 Martin Liska <mliska@suse.cz>
109 * doc/invoke.texi: Document how LTO works with debug info.
110 Describe auto-load support of binutils. Mention 'x86-64'
111 as valid option value of -march option.
113 2018-03-29 Jakub Jelinek <jakub@redhat.com>
115 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
118 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
119 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
120 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
123 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
126 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
127 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
128 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
129 for RS6000_BTM_POWERPC64.
130 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
131 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
132 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
137 2018-03-28 Carl Love <cel@us.ibm.com>
140 2017-09-27 Carl Love <cel@us.ibm.com>
142 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
143 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
144 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
147 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
149 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
150 instead of __vector bool.
151 (_mm_max_pu8): Likewise.
152 (_mm_min_pi16): Likewise.
154 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
157 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
161 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
163 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
164 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
165 (div_extend): Likewise.
166 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
168 (__builtin_divweuo): Likewise.
169 (__builtin_divdeo): Likewise.
170 (__builtin_divdeuo): Likewise.
172 2018-03-28 Jakub Jelinek <jakub@redhat.com>
175 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
176 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
178 PR tree-optimization/82004
179 * gimple-match-head.c (optimize_pow_to_exp): New function.
180 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
181 Don't fold to exp if optimize_pow_to_exp is false.
183 2018-03-28 Martin Liska <mliska@suse.cz>
186 * calls.c (initialize_argument_information): Fix trailing space.
187 * common.opt: Fix typo and provide better explanation for
188 -fsanitize-coverage option.
189 * config/i386/i386.opt: Fix typo.
191 2018-03-28 Jakub Jelinek <jakub@redhat.com>
192 Martin Liska <mliska@suse.cz>
195 * gimplify.c (asan_poison_variable): Don't do the check for
196 gimplify_omp_ctxp here.
197 (gimplify_decl_expr): Do it here.
198 (gimplify_target_expr): Likewise.
200 2018-03-28 Martin Liska <mliska@suse.cz>
203 * config/i386/i386.c (ix86_function_arg_advance): Do not call
204 chkp_type_bounds_count if MPX is not enabled.
206 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
208 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
210 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
213 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
214 function to create the function decl for complex long double
215 multiply and divide for -mabi=ieeelongdouble.
216 (init_float128_ieee): Call it.
218 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
221 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
222 -fcf-protection=branch -mibt.
223 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
225 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
228 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
230 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
233 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
234 extern array declarations.
236 2018-03-27 Richard Biener <rguenther@suse.de>
239 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
240 explicit single_use checks.
242 2018-03-27 Richard Biener <rguenther@suse.de>
244 PR tree-optimization/85082
245 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
248 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
250 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
251 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
252 Turn on fasynchronous-unwind-tables and funwind-tables.
254 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
257 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
258 (*bmi_blsr_<mode>_ccz): Ditto.
260 2018-03-26 Tom de Vries <tom@codesourcery.com>
262 PR tree-optimization/85063
263 * omp-general.c (offloading_function_p): New function. Factor out
265 * omp-offload.c (pass_omp_target_link::gate): ... here.
266 * omp-general.h (offloading_function_p): Declare.
267 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
268 with attribute omp declare target for offloading functions.
270 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
272 PR tree-optimization/84005
273 * tree-data-ref.h (get_base_for_alignment): Declare.
274 * tree-data-ref.c (get_base_for_alignment_1): New function.
275 (get_base_for_alignment): Likewise.
276 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
277 get_base_for_alignment to find a suitable base object, instead
278 of always using drb->base_address.
280 2018-03-23 Jakub Jelinek <jakub@redhat.com>
283 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
284 known size by default.
286 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
289 * lra-constraints.c (process_alt_operands): Don't match BLKmode
290 and non BLKmode operands.
292 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
295 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
298 2018-03-23 Richard Biener <rguenther@suse.de>
301 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
302 we are going to emit early debug for LTO.
304 2018-03-23 Jakub Jelinek <jakub@redhat.com>
307 * function.c (match_asm_constraints_1): Don't optimize if input
308 doesn't satisfy general_operand predicate for output's mode.
311 * alias.c (write_dependence_p): Don't require for x_canonicalized
312 non-VOIDmode if x has VOIDmode.
315 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
316 just don't try to optimize it rather than assert it never happens.
318 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
320 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
321 macro expansions for definition of ST_INTERNAL_<mode> and
322 LD_INTERNAL_<mode> builtins.
323 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
325 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
327 (altivec_expand_st_builtin): Likewise.
328 (altivec_expand_builtin): Remove calls to deleted functions.
329 (rs6000_address_for_altivec): Delete this function.
330 * config/rs6000/vector.md: Remove expands for
331 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
333 2018-03-22 Sudakshina Das <sudi.das@arm.com>
336 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
337 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
338 re-computing once computed.
339 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
340 (arm_init_machine_status): Initialize
341 machine->static_chain_stack_bytes.
343 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
346 * doc/extend.texi: Add four new prototypes for vec_ld.
347 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
348 definitions for more logical presentation.
349 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
350 entries for V1TI variants of __builtin_altivec_ld builtin.
351 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
352 handling of V1TI variant of LVX icode pattern.
353 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
354 (rs6000_gimple_fold_builtin): Likewise.
355 (altivec_init_builtins): Add code to define
356 __builtin_altivec_lvx_v1ti function.
358 2018-03-22 Jakub Jelinek <jakub@redhat.com>
361 * function.c (match_asm_constraints_1): Don't do the optimization
362 if input isn't a REG, SUBREG, MEM or constant.
364 2018-03-22 Tom de Vries <tom@codesourcery.com>
366 PR tree-optimization/84956
367 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
368 bb_has_abnormal_pred.
370 2018-03-22 Jakub Jelinek <jakub@redhat.com>
373 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
374 DECL_INITIAL (decl) to decl at the end.
375 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
378 2018-03-21 Joseph Myers <joseph@codesourcery.com>
380 * doc/extend.texi (__builtin_tgmath): Document when complex
381 integer types are treated as _Complex _Float64.
383 2018-03-21 Tom de Vries <tom@codesourcery.com>
385 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
387 2018-03-21 Jakub Jelinek <jakub@redhat.com>
389 PR tree-optimization/84960
390 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
391 if it is ENTRY block, move them into single succ of ENTRY in that case.
393 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
395 PR tree-optimization/84811
396 * poly-int.h (poly_span_traits): Remove the T3 parameter and
397 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
398 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
399 (known_subrange_p): Update accordingly. Cast each value involved
400 in the size comparison, rather than casting the result of the
403 2018-03-21 Jakub Jelinek <jakub@redhat.com>
405 PR tree-optimization/84982
406 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
407 by flipping the least significant bit rather than all bits from
408 bitpos to bitpos + bitsize - 1.
410 2018-03-21 Nathan Sidwell <nathan@acm.org>
412 * doc/extend.texi (Deprecated Features): Remove mention of
413 long-deleted deprecations.
415 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
418 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
419 * configure: Regenerate.
421 2018-03-21 Tom de Vries <tom@codesourcery.com>
423 PR tree-optimization/83126
424 * tree-parloops.c (num_phis): New function.
425 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
427 2018-03-21 Nathan Sidwell <nathan@acm.org>
429 * doc/extend.texi (Deprecated Features): Update deprecated flags,
430 mention anon-struct/union members and trailing attributes.
432 2018-03-21 Bin Cheng <bin.cheng@arm.com>
434 PR tree-optimization/84969
435 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
436 builtin memset partitions if they set different rhs values.
438 2018-03-21 Jakub Jelinek <jakub@redhat.com>
440 PR rtl-optimization/84989
441 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
442 VEC_DUPLICATE with scalar result mode.
444 2018-03-21 Martin Liska <mliska@suse.cz>
447 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
448 not intended return statement.
450 2018-03-21 Martin Liska <mliska@suse.cz>
453 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
454 (chkp_find_bound_slots_1): Limit number of iterations.
456 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
459 * Minor grammar fixes for x86 options.
461 2018-03-20 Jakub Jelinek <jakub@redhat.com>
464 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
465 holding REG_CFA_RESTORE notes, instead turn them into a USE.
467 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
470 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
471 (altivec_lvx_<mode>_1op): Likewise.
472 (altivec_stvx_<mode>_2op): Likewise.
473 (altivec_stvx_<mode>_1op): Likewise.
474 (altivec_lvx_<VM2:mode>): New define_expand.
475 (altivec_stvx_<VM2:mode>): Likewise.
476 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
477 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
478 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
479 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
480 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
481 (rs6000_gen_lvx): Likewise.
482 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
483 (altivec_expand_stv_builtin): Likewise.
484 (altivec_expand_builtin): Likewise.
485 * config/rs6000/vector.md: Likewise.
487 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
490 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
493 2018-03-20 Richard Biener <rguenther@suse.de>
496 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
497 sign-conversions as zero, fall back to standard scalar_stmt
500 2018-03-20 Martin Liska <mliska@suse.cz>
503 * predict.c (rebuild_frequencies): Handle case when we have
504 PROFILE_ABSENT, but flag_guess_branch_prob is false.
506 2018-03-20 Jakub Jelinek <jakub@redhat.com>
509 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
510 flag_section_anchors.
511 * varasm.c (use_blocks_for_decl_p): Remove hack for
515 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
517 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
518 be created, use lowpart_subreg of operands[0] rather than operands[0]
520 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
521 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
522 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
523 and n constraint instead of aarch64_shift_imm_di and Usd.
524 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
525 (*aarch64_<optab>_reg_minus<mode>3): ... this.
527 2018-03-20 Sudakshina Das <sudi.das@arm.com>
530 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
531 to favor GPR over NEON registers.
532 (<shift>di3_neon): Likewise.
534 2018-03-20 Tom de Vries <tom@codesourcery.com>
537 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
538 (nvptx_process_pars): Emit bar.sync asap and alap.
540 2018-03-20 Tom de Vries <tom@codesourcery.com>
543 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
544 seen_label if seen_label is already set.
546 2018-03-20 Jakub Jelinek <jakub@redhat.com>
549 * config/i386/i386.c (fold_builtin_cpu): For features above 31
550 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
551 Use 1U instead of 1. Formatting fixes.
554 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
555 instead of TREE_TYPE (s1) for the return value.
557 2018-03-19 Jakub Jelinek <jakub@redhat.com>
559 PR tree-optimization/84946
560 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
561 bitsize + bitsize in poly_uint64 rather than poly_int64.
564 * dwarf2asm.c: Include fold-const.c.
565 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
566 of decl rather than decl itself.
568 PR rtl-optimization/84643
569 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
571 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
574 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
575 calling assemble_variable.
577 2018-03-19 Sudakshina Das <sudi.das@arm.com>
580 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
581 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
583 2018-03-19 Jim Wilson <jimw@sifive.com>
586 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
587 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
588 (riscv_first_stack_step): Likewise.
589 (riscv_option_override): Use STACK_BOUNDARY instead of
591 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
593 (BIGGEST_ALIGNMENT): Set to 128.
594 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
595 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
598 2018-03-19 Richard Biener <rguenther@suse.de>
600 PR tree-optimization/84933
601 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
602 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
604 2018-03-19 Richard Biener <rguenther@suse.de>
606 PR tree-optimization/84859
607 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
608 (cond_if_else_store_replacement): Perform sinking operation on
609 single-store BBs regardless of MAX_STORES_TO_SINK setting.
610 Generalize what a BB with a single eligible store is.
612 2018-03-19 Richard Biener <rguenther@suse.de>
614 PR tree-optimization/84929
615 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
616 chrec_is_positive against non-chrec arg.
618 2018-03-19 Tamar Christina <tamar.christina@arm.com>
621 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
623 2018-03-18 Martin Liska <mliska@suse.cz>
625 PR rtl-optimization/84635
626 * regrename.c (build_def_use): Use matches_mode only when
629 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
631 PR tree-optimization/84913
632 * tree-vect-loop.c (vectorizable_reduction): Don't try to
633 vectorize chains of COND_EXPRs.
635 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
637 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
639 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
641 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
643 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
645 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
647 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
648 Kito Cheng <kito.cheng@gmail.com>
650 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
651 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
652 (nds32_adjust_reg_alloc_order): New function.
653 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
655 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
657 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
658 nds32_print_operand, nds32_print_operand_address): Use
659 HOST_WIDE_INT_PRINT_DEC instead.
661 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
663 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
665 2018-03-17 Jakub Jelinek <jakub@redhat.com>
668 * config/i386/i386.c (initial_ix86_tune_features,
669 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
671 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
672 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
673 rather than 1u << ix86_tune. Formatting fix.
674 (ix86_option_override_internal): Change ix86_arch_mask from
675 unsigned int to unsigned HOST_WIDE_INT, initialize to
676 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
677 (ix86_function_specific_restore): Likewise.
679 2018-03-16 Jakub Jelinek <jakub@redhat.com>
682 * postreload.c (reload_combine_recognize_pattern): Perform
683 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
684 truncate_int_for_mode the result for the destination's mode.
687 * hsa-gen.c (mem_type_for_type): Fix comment typo.
688 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
690 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
693 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
696 * lra-assigns.c (lra_split_hard_reg_for): Don't use
697 regno_allocno_class_array and sorted_pseudos.
698 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
699 insns where regno is used.
701 2018-03-16 Martin Liska <mliska@suse.cz>
704 * multiple_target.c (create_dispatcher_calls): Redirect
705 reference in the symbol table.
707 2018-03-16 Martin Liska <mliska@suse.cz>
710 * multiple_target.c (create_dispatcher_calls): Redirect also
713 2018-03-16 Jakub Jelinek <jakub@redhat.com>
717 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
718 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
719 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
721 2018-03-16 Julia Koval <julia.koval@intel.com>
723 * doc/invoke.texi (Skylake Server): Add CLWB.
724 Cannonlake): Remove CLWB.
726 2018-03-16 Jakub Jelinek <jakub@redhat.com>
728 PR tree-optimization/84841
729 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
731 (FLOAT_ONE_CONST_TYPE): Define.
732 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
733 (sort_by_operand_rank): Put entries with higher constant_type last
734 rather than first to match comments.
736 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
738 * config/nios2/nios2.md (movsi_internal): Fix thinko in
741 2018-03-15 Jakub Jelinek <jakub@redhat.com>
744 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
745 check and use address of target always.
747 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
750 * config/i386/i386.c (indirect_thunk_needed): Update comments.
751 (indirect_thunk_bnd_needed): Likewise.
752 (indirect_thunks_used): Likewise.
753 (indirect_thunks_bnd_used): Likewise.
754 (indirect_return_needed): New.
755 (indirect_return_bnd_needed): Likewise.
756 (output_indirect_thunk_function): Add a bool argument for
758 (output_indirect_thunk_function): Don't generate alias for
759 function return thunk.
760 (ix86_code_end): Call output_indirect_thunk_function to generate
761 function return thunks.
762 (ix86_output_function_return): Set indirect_return_bnd_needed
763 and indirect_return_needed instead of indirect_thunk_bnd_needed
764 and indirect_thunk_needed.
766 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
768 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
769 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
770 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
772 2018-03-15 David Malcolm <dmalcolm@redhat.com>
773 Paul Hua <paul.hua.gm@gmail.com>
776 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
778 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
780 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
781 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
784 2018-03-15 Tamar Christina <tamar.christina@arm.com>
787 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
788 instead of GET_MODE_SIZE when comparing Units.
790 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
793 * varasm.c (hash_section): Return an unchangeble hash value
794 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
795 Return !aarch64_can_use_per_function_literal_pools_p ().
797 2018-03-15 Jakub Jelinek <jakub@redhat.com>
800 * optabs.c (emit_conditional_move): Pass address of cmode's copy
801 rather than address of cmode as last argument to prepare_cmp_insn.
803 2018-03-15 Julia Koval <julia.koval@intel.com>
805 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
806 F_AVX512VNNI, F_AVX512BITALG): New.
808 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
811 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
812 insn for floating-point loads and stores.
814 2018-03-14 Carl Love <cel@us.ibm.com>
816 * config/rs6000/rs6000-c.c: Add macro definitions for
817 ALTIVEC_BUILTIN_VEC_PERMXOR.
818 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
819 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
820 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
821 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
823 * config/doc/extend.texi: Add prototypes for vec_permxor.
825 2018-03-14 David Malcolm <dmalcolm@redhat.com>
828 * diagnostic-show-locus.c (class layout_point): Convert m_line
829 from int to linenum_type.
830 (line_span::comparator): Use linenum "compare" function when
831 comparing line numbers.
832 (test_line_span): New function.
833 (layout_range::contains_point): Convert param "row" from int to
835 (layout_range::intersects_line_p): Likewise.
836 (layout::will_show_line_p): Likewise.
837 (layout::print_source_line): Likewise.
838 (layout::should_print_annotation_line_p): Likewise.
839 (layout::print_annotation_line): Likewise.
840 (layout::print_leading_fixits): Likewise.
841 (layout::annotation_line_showed_range_p): Likewise.
842 (struct line_corrections): Likewise for field m_row.
843 (line_corrections::line_corrections): Likewise for param "row".
844 (layout::print_trailing_fixits): Likewise.
845 (layout::get_state_at_point): Likewise.
846 (layout::get_x_bound_for_row): Likewise.
847 (layout::print_line): Likewise.
848 (diagnostic_show_locus): Likewise for locals "last_line" and
850 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
851 * input.c (selftest::test_linenum_comparisons): New function.
852 (selftest::input_c_tests): Call it.
853 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
854 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
855 * selftest.h (ASSERT_GT): New macro.
856 (ASSERT_GT_AT): New macro.
857 (ASSERT_LT): New macro.
858 (ASSERT_LT_AT): New macro.
860 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
862 PR rtl-optimization/84780
863 * combine.c (distribute_links): Don't make a link based on pc_rtx.
865 2018-03-14 Martin Liska <mliska@suse.cz>
867 * tree.c (record_node_allocation_statistics): Use
869 (get_stats_node_kind): New function extracted from
870 record_node_allocation_statistics.
871 (free_node): Use get_stats_node_kind.
873 2018-03-14 Richard Biener <rguenther@suse.de>
875 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
876 that the value-set of ANTIC_IN doesn't grow.
879 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
881 (BB_VISITED_WITH_VISITED_SUCCS): New define.
882 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
884 2018-03-14 Julia Koval <julia.koval@intel.com>
886 * config.gcc (icelake-client, icelake-server): New.
888 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
889 (initial_ix86_arch_features): Ditto.
890 (PTA_SKYLAKE): Add SGX.
891 (PTA_ICELAKE): Remove.
892 (PTA_ICELAKE_CLIENT): New.
893 (PTA_ICELAKE_SERVER): New.
894 (ix86_option_override_internal): Split up icelake on icelake client and
896 (get_builtin_code_for_version): Ditto.
897 (fold_builtin_cpu): Ditto.
898 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
899 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
900 * config/i386/i386.h (processor_type): Ditto.
901 * doc/invoke.texi: Ditto.
903 2018-03-14 Jakub Jelinek <jakub@redhat.com>
906 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
907 INTEGER_CST offset, add it together with bitpos / 8 and
908 sign extend based on POINTER_SIZE.
912 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
915 * config/i386/constraints.md (Yc): New register constraint.
916 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
917 Use Yc constraint for alternative 2 of operand 0. Remove
918 preferred_for_speed attribute.
920 2018-03-14 Richard Biener <rguenther@suse.de>
922 PR tree-optimization/84830
923 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
924 with the old one to avoid oscillations.
926 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
929 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
931 (assign_by_spills): Return a flag of reload assignment failure.
932 Do not process the reload assignment failures. Do not spill other
933 reload pseudos if they has the same reg class. Update n if
935 (lra_assign): Add a return arg. Set up from the result of
936 assign_by_spills call.
937 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
938 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
939 usage_insns if it is not NULL.
940 (spill_hard_reg_in_range): New function.
941 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
942 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
944 (lra_assign): Change prototype.
945 * lra.c (lra): Add code to deal with fails by splitting hard reg
948 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
950 * config/riscv/riscv.opt (mrelax): New option.
951 * config/riscv/riscv.c (riscv_file_start): Emit ".option
952 "norelax" when riscv_mrelax is disabled.
953 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
955 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
958 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
959 reassociation for int modes.
961 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
963 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
964 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
966 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
967 * config/aarch64/aarch64-sve.md
968 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
969 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
970 (*extend<mode><Vwide>2): Rename to...
971 (aarch64_sve_extend<mode><Vwide>2): ...this.
972 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
973 renaming the old pattern to...
974 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
976 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
977 define_expand, renaming the old pattern to...
978 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
979 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
980 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
981 account when deciding which SVE instruction the optab should use.
982 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
984 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
986 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
987 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
988 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
989 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
990 (tlsdesc_small_<mode>): Turn a define_expand and use
991 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
992 (tlsdesc_small_advsimd_<mode>): ...this.
993 (tlsdesc_small_sve_<mode>): New pattern.
995 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
997 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
998 (UNSPEC_UMUL_HIGHPART): New constants.
999 (MUL_HIGHPART): New int iteraor.
1000 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1001 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1003 (*<su>mul<mode>3_highpart): New define_insn.
1005 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1008 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1011 2018-03-13 Martin Liska <mliska@suse.cz>
1014 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1016 (sem_item_optimizer::~sem_item_optimizer): Release it.
1017 (sem_item_optimizer::merge_classes): Register variable aliases.
1018 (sem_item_optimizer::fixup_pt_set): New function.
1019 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1020 * ipa-icf.h: Declare new variables and functions.
1022 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1025 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1026 integer_pow2p@2 and test integer_pow2p in condition.
1027 (A < 0 ? C : 0): Similarly for @1.
1030 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1031 characters starting at p contain '\0' character, don't look beyond
1035 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1036 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1039 * reg-stack.c (change_stack): Change update_end var from int to
1040 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1041 also call set_block_for_insn on the newly added insns and rescan.
1044 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1045 on the last operand.
1048 * tree.c (stabilize_reference_1): Return save_expr (e) for
1049 STATEMENT_LIST even if it doesn't have side-effects.
1051 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1053 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1055 2018-03-12 Renlin Li <renlin.li@arm.com>
1057 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1058 aarch64_output_scalar_simd_mov_immediate.
1060 2018-03-12 Martin Sebor <msebor@redhat.com>
1062 PR tree-optimization/83456
1063 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1064 for perfectly overlapping calls to memcpy.
1065 (gimple_fold_builtin_memory_chk): Same.
1066 (gimple_fold_builtin_strcpy): Handle no-warning.
1067 (gimple_fold_builtin_stxcpy_chk): Same.
1068 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1070 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1072 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1073 parameter. Use it for SFmode.
1074 (rs6000_function_arg_advance_1): Adjust.
1075 (rs6000_function_arg): Adjust.
1076 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1078 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1080 PR rtl-optimization/84169
1081 PR rtl-optimization/84780
1082 * combine.c (can_combine_p): Check for a 2-insn combination whether
1083 the destination register is used between the two insns, too.
1085 2018-03-12 Richard Biener <rguenther@suse.de>
1087 PR tree-optimization/84803
1088 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1089 for refs DR analysis didn't process.
1091 2018-03-12 Richard Biener <rguenther@suse.de>
1093 PR tree-optimization/84777
1094 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1095 force-vectorize loops ignore whether we are optimizing for size.
1097 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1100 (TARGET_MD_ASM_ADJUST): Define.
1102 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1103 Kito Cheng <kito.cheng@gmail.com>
1104 Chung-Ju Wu <jasonwucj@gmail.com>
1106 * config/nds32/nds32.c (nds32_compute_stack_frame,
1107 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1108 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1109 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1110 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1111 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1112 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1113 * config/nds32/nds32.md (prologue, epilogue): Use macro
1114 NDS32_V3PUSH_AVAILABLE_P to do checking.
1116 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1119 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1120 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1121 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1122 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1123 addition of most attributes on !orig_type_die or the attribute not
1124 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1126 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1127 Chung-Ju Wu <jasonwucj@gmail.com>
1129 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1131 * config/nds32/nds32.opt (mvh): New option.
1133 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1134 Chung-Ju Wu <jasonwucj@gmail.com>
1136 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1138 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1139 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1142 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1143 Chung-Ju Wu <jasonwucj@gmail.com>
1145 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1147 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1148 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1150 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1151 Kito Cheng <kito.cheng@gmail.com>
1152 Chung-Ju Wu <jasonwucj@gmail.com>
1154 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1155 UNSPEC_FFMISM and UNSPEC_FLMISM.
1156 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1157 for ffb, ffmism and flmism.
1158 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1159 (unspec_ffmism): Ditto.
1160 (unspec_flmism): Ditto.
1161 (nds32_expand_builtin_impl): Check if string extension is available.
1162 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1163 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1165 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1168 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1171 * lra-assigns.c (assign_by_spills): Return a flag of reload
1172 assignment failure. Do not process the reload assignment
1173 failures. Do not spill other reload pseudos if they has the same
1175 (lra_assign): Add a return arg. Set up from the result of
1176 assign_by_spills call.
1177 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1178 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1179 usage_insns if it is not NULL.
1180 (spill_hard_reg_in_range): New function.
1181 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1182 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1183 function prototypes.
1184 (lra_assign): Change prototype.
1185 * lra.c (lra): Add code to deal with fails by splitting hard reg
1188 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1191 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1193 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1196 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1197 (dw_val_node): Add val_symbolic_view.
1198 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1199 (symview_upper_bound): New.
1200 (new_line_info_table): Initialize symviews_since_reset.
1201 (dwarf2out_source_line): Count symviews_since_reset and set
1202 symview_upper_bound.
1203 (dw_val_equal_p): Handle symview.
1204 (add_AT_symview): New.
1205 (print_dw_val): Handle symview.
1206 (attr_checksum, attr_checksum_ordered): Likewise.
1207 (same_dw_val_p, size_of_die): Likewise.
1208 (value_format, output_die): Likewise.
1209 (add_high_low_attributes): Use add_AT_symview for entry_view.
1210 (dwarf2out_finish): Reset symview_upper_bound, clear
1213 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1216 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1217 Add strict argument and use it.
1218 (rs6000_split_multireg_move): Update for new strict argument.
1219 (mem_operand_gpr): Disallow all non-offsettable addresses.
1220 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1222 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1225 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1226 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1227 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1230 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1231 decl, use remap_type if we want to use the type.
1233 2018-03-09 Martin Sebor <msebor@redhat.com>
1235 PR tree-optimization/84526
1236 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1238 (builtin_access::generic_overlap): Be prepared to handle non-array
1241 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1243 PR rtl-optimization/84682
1244 * lra-constraints.c (process_address_1): Check is_address flag
1245 for address constraints.
1246 (process_alt_operands): Likewise.
1247 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1248 preprocess_constraints.
1249 * recog.h (preprocess_constraints): Add oploc parameter.
1251 * recog.c (preprocess_constraints): Test address_operand for
1252 CT_ADDRESS constraints.
1254 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1257 * lra-assigns.c (assign_by_spills): Return a flag of reload
1258 assignment failure. Do not process the reload assignment
1259 failures. Do not spill other reload pseudos if they has the same
1261 (lra_assign): Add a return arg. Set up from the result of
1262 assign_by_spills call.
1263 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1264 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1265 usage_insns if it is not NULL.
1266 (spill_hard_reg_in_range): New function.
1267 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1268 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1269 function prototypes.
1270 (lra_assign): Change prototype.
1271 * lra.c (lra): Add code to deal with fails by splitting hard reg
1274 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1277 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1278 Accept complain bool parameter. Only emit errors if it is true.
1279 (arm_parse_cpu_option_name): Likewise.
1280 (arm_target_thumb_only): Adjust callers of the above.
1281 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1282 prototype to take a default true bool parameter.
1283 (arm_parse_arch_option_name): Likewise.
1285 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1286 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1290 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1291 * configure: Regenerate.
1292 * configure.ac ("linker --version-script option"): New.
1293 ("linker soname option"): New.
1295 2018-03-09 Richard Biener <rguenther@suse.de>
1297 PR tree-optimization/84775
1298 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1299 immediate uses of predicate stmts and mark them modified.
1302 PR tree-optimization/84178
1303 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1305 (version_loop_for_if_conversion): Delay update_ssa call.
1306 (tree_if_conversion): Delay update_ssa until after predicate
1309 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1312 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1313 when the function accesses prior frames.
1315 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1318 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1319 gen_llsym, otherwise call maybe_gen_llsym.
1322 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1323 has ',' character inside of it.
1325 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1328 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1329 as clobbering CC_REGNUM.
1331 2018-03-08 Richard Biener <rguenther@suse.de>
1334 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1335 (follow_copies_to_constant): Do not follow SSA names registered
1338 2018-03-08 Richard Biener <rguenther@suse.de>
1340 PR tree-optimization/84178
1341 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1343 (version_loop_for_if_conversion): Delay update_ssa call.
1344 (tree_if_conversion): Delay update_ssa until after predicate
1347 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1349 PR tree-optimization/84178
1350 * tree-if-conv.c (release_bb_predicate): Remove the
1351 the assertion that the stmts have NULL use_ops.
1352 Discard the statements, asserting that they haven't
1353 yet been added to a BB.
1355 2018-03-08 Richard Biener <rguenther@suse.de>
1357 PR tree-optimization/84746
1358 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1359 (phi_translate): Pass in destination ANTIC_OUT set.
1360 (phi_translate_1): Likewise. For a simplified result lookup
1361 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1362 (phi_translate_set): Adjust.
1363 (do_pre_regular_insertion): Likewise.
1364 (do_pre_partial_partial_insertion): Likewise.
1366 2018-03-08 Martin Liska <mliska@suse.cz>
1368 PR gcov-profile/84735
1369 * doc/gcov.texi: Document usage of profile files.
1370 * gcov-io.h: Document changes in the format.
1372 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1376 * dwarf2out.c (struct dw_line_info_table): Update comments for
1378 (FORCE_RESET_NEXT_VIEW): New.
1379 (FORCE_RESETTING_VIEW_P): New.
1380 (RESETTING_VIEW_P): Check for -1 too.
1381 (ZERO_VIEW_P): Likewise.
1382 (new_line_info_table): Force-reset next view.
1383 (dwarf2out_begin_function): Likewise.
1384 (dwarf2out_source_line): Simplify zero_view_p initialization.
1385 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1386 view directly. Omit view when omitting .loc at line 0.
1388 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1390 PR tree-optimization/84740
1391 * tree-switch-conversion.c (process_switch): Call build_constructors
1392 only if info.phi_count is non-zero.
1394 PR tree-optimization/84739
1395 * tree-tailcall.c (find_tail_calls): Check call arguments against
1396 DECL_ARGUMENTS (current_function_decl) rather than
1397 DECL_ARGUMENTS (func) when checking for tail recursion.
1399 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1401 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1402 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1403 Volker Reichelt's entry and add entries for people that perform
1404 GCC fuzzy testing and report numerous bugs.
1406 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1409 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1410 readonly data in sdata, if that is disabled.
1411 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1412 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1413 -mreadonly-in-sdata option.
1415 2018-03-07 Martin Sebor <msebor@redhat.com>
1417 PR tree-optimization/84468
1418 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1419 basic block when looking for nul assignment.
1421 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1424 * except.h (output_function_exception_table): Adjust prototype.
1425 * except.c (output_function_exception_table): Remove FNNAME parameter
1426 and add SECTION parameter. Ouput one part of the table at a time.
1427 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1428 the first part of the exception table and emit unwind directives.
1429 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1430 (i386_pe_seh_cold_init): Likewise.
1431 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1432 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1433 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1434 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1435 * config/i386/winnt.c: Include except.h.
1436 (struct seh_frame_state): Add reg_offset, after_prologue and
1437 in_cold_section fields.
1438 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1439 (i386_pe_seh_cold_init): New function.
1440 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1441 to seh->in_cold_section.
1442 (seh_emit_push): Record the offset of the push.
1443 (seh_emit_save): Record the offet of the save.
1444 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1445 Test seh->after_prologue to disregard the epilogue.
1446 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1447 (i386_pe_end_cold_function): New function.
1449 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1452 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1453 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1456 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1458 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1459 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1462 * multiple_target.c: Include tree-inline.h and intl.h.
1463 (expand_target_clones): Diagnose and fail if node->definition and
1464 !tree_versionable_function_p (node->decl).
1466 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1468 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1470 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1471 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1472 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1474 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1477 * combine.c (try_combine): Use reg_or_subregno instead of handling
1478 just paradoxical SUBREGs and REGs.
1480 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1482 * config/arc/arc.c (arc_finalize_pic): Remove function.
1483 (arc_must_save_register): We use single base PIC register, remove
1484 checks to save/restore the PIC register.
1485 (arc_expand_prologue): Likewise.
1486 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1488 (arc_verify_short): Likewise.
1489 (arc_attr_type): Likewise.
1490 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1491 (walk_stores): Likewise.
1492 (arc_address_cost): Make it static.
1493 (arc_verify_short): Likewise.
1494 (branch_dest): Likewise.
1495 (arc_attr_type): Likewise.
1496 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1497 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1498 (arc_final_prescan_insn): Remove inserting the nops due to
1499 hardware hazards. It is done in reorg step.
1500 (insn_length_variant_t): Remove.
1501 (insn_length_parameters_t): Likewise.
1502 (arc_insn_length_parameters): Likewise.
1503 (arc_get_insn_variants): Likewise.
1504 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1506 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1509 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1512 PR tree-optimization/84687
1513 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1515 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1517 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1519 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1520 Rename to ppc_speculation_barrier.
1521 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1522 __builtin_ppc_speculation_barrier.
1524 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1527 * combine.c (combine_simplify_rtx): Don't try to simplify if
1528 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1531 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1533 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1534 to 32 bytes when compiling for POWER9.
1536 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1539 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1540 regparm >= 3 with no arg reg available also for calls with
1541 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1544 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1546 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1548 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1551 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1553 2018-03-05 Richard Biener <rguenther@suse.de>
1555 PR tree-optimization/84486
1556 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1557 When inserting a __builtin_assume_aligned call set the LHS
1558 SSA name alignment info accordingly.
1560 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1562 PR tree-optimization/84114
1563 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1564 Avoid reassociation of FLOAT_MODE addition.
1566 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1568 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1569 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1570 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1571 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1572 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1573 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1574 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1576 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1577 __builtin_ia32_wbinvd): New builtins.
1578 (SPECIAL_ARGS2): New.
1579 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1580 (SPECIAL_ARGS2): New.
1581 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1582 (ix86_valid_target_attribute_inner_p): Ditto.
1583 (ix86_init_mmx_sse_builtins): Add special_args2.
1584 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1585 TARGET_WBNOINVD_P): New.
1586 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1587 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1588 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1589 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1590 * config/i386/pconfigintrin.h: New file.
1591 * config/i386/wbnoinvdintrin.h: Ditto.
1592 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1593 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1595 2018-03-05 Richard Biener <rguenther@suse.de>
1597 PR tree-optimization/84670
1598 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1600 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1601 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1602 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1603 doesn't grow if all successors (recursively) were visited at least
1606 2018-03-05 Richard Biener <rguenther@suse.de>
1608 PR tree-optimization/84650
1609 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1610 if executed in the loop pipeline.
1612 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1614 * doc/configfiles.texi (Configuration Files): Move info about
1615 conditionalizing $target-protos.h to...
1616 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1617 differs from $target-protos.h.
1619 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1620 Chung-Ju Wu <jasonwucj@gmail.com>
1622 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1623 * config/nds32/nds32-multiple.md (setmemsi): Define.
1624 * config/nds32/nds32-memory-manipulation.c
1625 (nds32_gen_dup_4_byte_to_word_value): New.
1626 (emit_setmem_word_loop): New.
1627 (emit_setmem_byte_loop): New.
1628 (nds32_expand_setmem_loop): New.
1629 (nds32_expand_setmem_loop_v3m): New.
1630 (nds32_expand_setmem_unroll): New.
1631 (nds32_expand_setmem): New.
1633 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1634 Chung-Ju Wu <jasonwucj@gmail.com>
1636 * config/nds32/nds32-memory-manipulation.c
1637 (nds32_emit_load_store): New.
1638 (nds32_emit_post_inc_load_store): New.
1639 (nds32_emit_mem_move): New.
1640 (nds32_emit_mem_move_block): New.
1641 (nds32_expand_movmemsi_loop_unknown_size): New.
1642 (nds32_expand_movmemsi_loop_known_size): New.
1643 (nds32_expand_movmemsi_loop): New.
1644 (nds32_expand_movmemsi_unroll): New.
1645 (nds32_expand_movmemqi): Rename ...
1646 (nds32_expand_movmemsi): ... to this.
1647 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1648 (movmemsi): ... to this.
1649 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1650 (nds32_expand_movmemsi): ... to this.
1652 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1653 Monk Chiang <sh.chiang04@gmail.com>
1654 Chung-Ju Wu <jasonwucj@gmail.com>
1656 * config/nds32/nds32-protos.h
1657 (nds32_expand_load_multiple): New arguments.
1658 (nds32_expand_store_multiple): Ditto.
1659 (nds32_valid_multiple_load_store): Rename ...
1660 (nds32_valid_multiple_load_store_p): ... to this.
1661 * config/nds32/nds32-memory-manipulation.c
1662 (nds32_expand_load_multiple): Refine implementation.
1663 (nds32_expand_store_multiple): Ditto.
1664 * config/nds32/nds32-multiple.md
1665 (load_multiple): Update nds32_expand_load_multiple interface.
1666 (store_multiple): Update nds32_expand_store_multiple interface.
1667 * config/nds32/nds32-predicates.c
1668 (nds32_valid_multiple_load_store): Rename ...
1669 (nds32_valid_multiple_load_store_p): ... to this and refine
1671 * config/nds32/predicates.md
1672 (nds32_load_multiple_and_update_address_operation): New predicate.
1673 (nds32_store_multiple_and_update_address_operation): New predicate.
1675 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1676 Chung-Ju Wu <jasonwucj@gmail.com>
1678 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1679 (combo): New attribute.
1680 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1682 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1684 * config/nds32/nds32.opt: Change -mcmodel= default value.
1686 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1687 Monk Chiang <sh.chiang04@gmail.com>
1688 Chung-Ju Wu <jasonwucj@gmail.com>
1690 * config/nds32/constants.md (unspec_element): New enum.
1691 * config/nds32/constraints.md (Umw): New constraint.
1692 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1693 * config/nds32/nds32-intrinsic.md: Likewise.
1694 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1695 (nds32_valid_smw_lwm_base_p): New.
1696 (nds32_output_smw_single_word): New.
1697 (nds32_output_lmw_single_word): New.
1698 (nds32_expand_unaligned_load): New.
1699 (nds32_expand_unaligned_store): New.
1700 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1701 (nds32_output_smw_single_word): Declare.
1702 (nds32_output_lmw_single_word): Declare.
1703 (nds32_expand_unaligned_load): Declare.
1704 (nds32_expand_unaligned_store): Declare.
1705 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1706 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1707 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1708 NDS32_BUILTIN_UASTORE_DW.
1709 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1712 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1713 Kito Cheng <kito.cheng@gmail.com>
1714 Chung-Ju Wu <jasonwucj@gmail.com>
1716 * config/nds32/nds32-intrinsic.c
1717 (nds32_expand_builtin_null_ftype_reg): Delete.
1718 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1719 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1720 (nds32_read_argument): New.
1721 (nds32_legitimize_target): Ditto.
1722 (nds32_legitimize_argument): Ditto.
1723 (nds32_check_constant_argument): Ditto.
1724 (nds32_expand_unop_builtin): Ditto.
1725 (nds32_expand_unopimm_builtin): Ditto.
1726 (nds32_expand_binop_builtin): Ditto.
1727 (nds32_builtin_decl_impl): Ditto.
1728 (builtin_description): Ditto.
1729 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1730 (nds32_init_builtins_impl): Ditto.
1731 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1732 (nds32_builtin_decl): New.
1733 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1734 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1736 2018-03-02 Jeff Law <law@redhat.com>
1738 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1739 (redundant_insn, fill_simple_delay_slots): Likewise.
1740 (fill_slots_from_thread): Likewise.
1741 * resource.c (mark_referenced_resources): Likewise.
1742 (mark_set_resources, find_dead_or_set_registers): Likewise.
1744 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1746 * substring-locations.h (format_warning_va): Formatting fix for
1748 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1750 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1751 * substring-locations.c: Include intl.h.
1752 (format_warning_va): Turned into small wrapper around
1753 format_warning_n_va, renamed to ...
1754 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1755 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1757 (format_warning_at_substring_n): New function.
1758 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1759 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1760 format_warning_at_substring with just a shorter name instead of
1761 const function pointer.
1762 (fmtwarn_n): New function.
1763 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1764 appropriate, get rid of all the fmtstr temporaries, move conditionals
1765 with G_() wrapped string literals directly into fmtwarn arguments,
1766 cast dir.len to (int), formatting fixes.
1768 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1770 * doc/invoke.texi: Remove "Cilk Plus" references.
1772 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1773 Richard Biener <rguenther@suse.de>
1776 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1777 for error or warning attributes if CALL_FROM_THUNK_P is set.
1780 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1783 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1784 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1787 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1788 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1790 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1791 instead of -1U in last predictors element's probability member.
1793 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1796 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1797 arguments if they are comparable.
1799 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1801 PR tree-optimization/84634
1802 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1803 masks and masked_loop_p with a single loop_masks, making sure it's
1804 null for bb vectorization.
1806 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1808 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1809 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1810 loop->force_vectorize to check whether there is no alias.
1812 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1815 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1817 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1819 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1820 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1821 instead of a loop around prev_real_insn.
1822 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1826 * config/i386/i386.c (ix86_print_operand): Use conditional
1827 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1830 2018-03-02 Richard Biener <rguenther@suse.de>
1832 PR tree-optimization/84427
1833 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1834 (bitmap_set_subtract_values): Rewrite to handle multiple
1837 (prune_clobbered_mems): Likewise.
1838 (phi_translate): Take edge instead of pred/phiblock.
1839 (phi_translate_1): Likewise.
1840 (phi_translate_set): Likewise. Insert all translated
1841 exprs for a value into the set, keeping possibly multiple
1842 expressions per value.
1843 (compute_antic_aux): Adjust for phi_translate changes.
1844 When intersecting union the expressions and prune those
1845 not in the final value set, keeping possibly multiple
1846 expressions per value. Do not use value-insertion
1847 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1848 all expressions. Add verification that the value-sets
1849 only shrink during iteration.
1850 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1851 (do_pre_regular_insertion): Likewise.
1852 (do_pre_partial_partial_insertion): Likewise.
1854 2018-03-02 Richard Biener <rguenther@suse.de>
1857 * config/darwin.c (saved_debug_info_level): New static global.
1858 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1859 (darwin_asm_lto_end): Restore debug info generation settings.
1861 2018-03-01 Martin Liska <mliska@suse.cz>
1864 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1867 2018-03-01 Richard Biener <rguenther@suse.de>
1870 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1871 type annotation with LTO when debug was disabled at compile-time.
1873 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1875 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1877 (mips_final_postscan_insn): Likewise.
1879 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1881 PR rtl-optimization/84528
1882 * alias.c (init_alias_target): Add commentary.
1883 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1884 a unique base value if the frame pointer is not eliminated
1885 to the stack pointer.
1887 2018-03-01 Tom de Vries <tom@codesourcery.com>
1889 PR rtl-optimization/83327
1890 * lra-int.h (hard_regs_spilled_into): Declare.
1891 * lra.c (hard_regs_spilled_into): Define.
1892 (init_reg_info): Init hard_regs_spilled_into.
1893 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1894 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1895 (process_bb_lives): Handle hard_regs_spilled_into.
1896 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1897 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1899 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1901 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1902 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1903 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1904 * config/rs6000/aix72.h: New file.
1906 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1908 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1909 instead of warning_at with conditional singular and plural messages
1913 * stor-layout.c (update_alignment_for_field): For
1914 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1915 && !DECL_PACKED (field), do the alignment update, just use
1916 only desired_align instead of MAX (type_align, desired_align)
1918 (place_field): Don't do known_align < desired_align handling
1919 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1920 is non-NULL, instead do it after rli->prev_field handling and
1921 only if not within a bitfield word. For DECL_PACKED (field)
1922 use type_align of BITS_PER_UNIT.
1924 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1926 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1927 superfluous parentheses and trailing spaces.
1929 2018-02-28 Richard Biener <rguenther@suse.de>
1931 PR tree-optimization/84584
1932 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1933 SCoPs with fake exit edge.
1935 2018-02-28 Martin Liska <mliska@suse.cz>
1938 * timevar.c (timer::print): Fix format to properly print 100%
1941 2018-02-28 Richard Biener <rguenther@suse.de>
1944 * genmatch.c (capture_info::walk_match): Do not mark
1945 captured expressions without operands as expr_p given
1946 they act more like predicates and should be subject to
1947 "lost tail" side-effect preserving.
1949 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1951 PR rtl-optimization/81611
1952 * auto-inc-dec.c (attempt_change): Move dead note from
1953 mem_insn if it's the next use of regno
1954 (find_address): Take address use of reg holding
1955 non-incremented value. Add parm to limit search to the named
1957 (merge_in_block): Attempt to use a mem insn that is the next
1958 use of the original regno.
1960 2018-02-27 Martin Sebor <msebor@redhat.com>
1963 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1964 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1966 2018-02-27 Martin Sebor <msebor@redhat.com>
1968 PR translation/84207
1969 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1970 n argument to unsigned HOST_WIDE_INT.
1971 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1972 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1973 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1974 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1976 2018-02-27 Richard Biener <rguenther@suse.de>
1978 PR tree-optimization/84512
1979 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1980 Do not use the estimate returned from record_stmt_cost for
1981 the scalar iteration cost but sum properly using add_stmt_cost.
1983 2018-02-27 Richard Biener <rguenther@suse.de>
1985 PR tree-optimization/84466
1986 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1987 Adjust last change to less strictly validate use operands.
1989 2018-02-27 Martin Liska <mliska@suse.cz>
1991 PR gcov-profile/84548
1992 * gcov.c (process_file): Allow partial overlap and consider it
1993 also as group functions.
1994 (output_lines): Properly calculate range of lines for a group.
1996 2018-02-27 Martin Liska <mliska@suse.cz>
1998 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1999 'ggc' suffixes. Change first column width.
2000 (timer::print): Fix formatting of the column.
2002 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2004 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2005 preserve inline entry blocks for the sake of debug inline
2006 entry point markers alone.
2007 (remove_unused_locals): Suggest in comments a better place to
2008 force the preservation of inline entry blocks that are
2009 otherwise unused, but do not preserve them.
2011 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2013 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2015 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2018 * config/i386/constraints.md (Bs): Replace
2019 ix86_indirect_branch_register with
2020 TARGET_INDIRECT_BRANCH_REGISTER.
2022 * config/i386/i386.md (indirect_jump): Likewise.
2023 (tablejump): Likewise.
2024 (*sibcall_memory): Likewise.
2025 (*sibcall_value_memory): Likewise.
2026 Peepholes of indirect call and jump via memory: Likewise.
2027 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2028 (*sibcall_value_GOT_32): Likewise.
2029 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2030 (GOT_memory_operand): Likewise.
2031 (call_insn_operand): Likewise.
2032 (sibcall_insn_operand): Likewise.
2033 (GOT32_symbol_operand): Likewise.
2034 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2036 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2038 PR rtl-optimization/83496
2039 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2040 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2041 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2042 redundant insn, if any.
2043 (relax_delay_slots): Likewise.
2044 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2046 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2048 PR tree-optimization/83965
2049 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2050 that grouped statements are part of a reduction chain. Return
2051 true if the statement is not marked as a reduction itself but
2053 (vect_recog_dot_prod_pattern): Don't check whether the statement
2054 is part of a group here.
2055 (vect_recog_sad_pattern): Likewise.
2056 (vect_recog_widen_sum_pattern): Likewise.
2058 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2061 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2063 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2066 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2068 (ix86_output_indirect_function_return): New prototype.
2069 (ix86_split_simple_return_pop_internal): Likewise.
2070 * config/i386/i386.c (indirect_return_via_cx): New.
2071 (indirect_return_via_cx_bnd): Likewise.
2072 (indirect_thunk_name): Handle return va CX_REG.
2073 (output_indirect_thunk_function): Create alias for
2074 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2075 (ix86_output_indirect_jmp): Remove the bool argument.
2076 (ix86_output_indirect_function_return): New function.
2077 (ix86_split_simple_return_pop_internal): Likewise.
2078 * config/i386/i386.md (*indirect_jump): Don't pass false
2079 to ix86_output_indirect_jmp.
2080 (*tablejump_1): Likewise.
2081 (simple_return_pop_internal): Change it to define_insn_and_split.
2082 Call ix86_split_simple_return_pop_internal to split it for
2084 (simple_return_indirect_internal): Call
2085 ix86_output_indirect_function_return instead of
2086 ix86_output_indirect_jmp.
2088 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2091 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2092 memset and value initialization afterwards.
2094 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2096 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2098 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2101 * common/config/aarch64/aarch64-common.c
2102 (aarch_option_optimization_table[]): Switch
2103 off fomit-frame-pointer
2105 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2106 Chung-Ju Wu <jasonwucj@gmail.com>
2108 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2110 (store_multiple): Ditto.
2112 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2114 * config.gcc: Add --with-cpu support for nds32 target.
2115 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2116 * config/nds32/nds32.opt: Add -mcpu= option.
2118 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2120 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2121 isel=yes): Warn for these deprecated options.
2123 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2125 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2126 ISA_2_5_MASKS_EMBEDDED.
2128 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2130 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2131 p->max as pointers rather than using iterative_hash_expr.
2133 2018-02-23 Carl Love <cel@us.ibm.com>
2135 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2136 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2138 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2139 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2140 P8V_BUILTIN_VEC_VUNSIGNED2.
2142 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2145 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2146 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2148 * lra-constraints.c (curr_insn_transform): Set up
2149 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2151 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2152 * lra-eliminations.c (spill_pseudos): Ditto.
2153 (process_insn_for_elimination): Ditto.
2154 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2155 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2156 LRA_NON_CLOBBERED_ALT.
2158 2018-02-22 Martin Sebor <msebor@redhat.com>
2160 PR tree-optimization/84480
2161 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2162 to maybe_diag_stxncpy_trunc. Call it.
2163 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2164 from gimple_fold_builtin_strcpy. Print inlining stack.
2165 (handle_builtin_stxncpy): Print inlining stack.
2166 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2168 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2171 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2172 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2173 and -fcheck-pointer-bounds are used together.
2174 (indirect_thunk_prefix): New enum.
2175 (indirect_thunk_need_prefix): New function.
2176 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2177 "_nt" instead of "_bnd" for NOTRACK prefix.
2178 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2179 (output_indirect_thunk_function): Likewise.
2181 (ix86_code_end): Update output_indirect_thunk_function calls.
2182 (ix86_output_indirect_branch_via_reg): Replace
2183 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2184 (ix86_output_indirect_branch_via_push): Likewise.
2185 (ix86_output_function_return): Likewise.
2186 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2187 incompatible with -fcf-protection=branch and
2188 -fcheck-pointer-bounds.
2190 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2193 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2194 Change gcc_assert call to output_operand_lossage.
2196 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2198 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2200 2018-02-22 DJ Delorie <dj@redhat.com>
2201 Sebastian Perta <sebastian.perta@renesas.com>
2202 Oleg Endo <olegendo@gcc.gnu.org>
2204 * config/rx/rx.c (rx_rtx_costs): New function.
2205 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2207 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2209 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2211 2018-02-22 Martin Liska <mliska@suse.cz>
2214 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2215 Add "native" as a possible value.
2217 2018-02-22 Martin Liska <mliska@suse.cz>
2220 * config/i386/i386.c (ix86_option_override_internal):
2221 Add "native" as a possible value for -march and -mtune.
2223 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2226 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2227 to all type variants.
2229 PR tree-optimization/84503
2230 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2231 width as info->bitpos + info->bitsize - start.
2232 (merged_store_group::merge_overlapping): Simplify width computation.
2233 (check_no_overlap): New function.
2234 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2235 start + width and last_order of the group, fail if check_no_overlap
2237 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2238 to group if check_no_overlap fails.
2240 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2242 * config/rs6000/altivec.md: Delete contraint arguments to
2243 define_expand, define_split, and define_peephole2, and in
2244 define_insn_and_split if always unused.
2245 * config/rs6000/darwin.md: Ditto.
2246 * config/rs6000/dfp.md: Ditto.
2247 * config/rs6000/rs6000.md: Ditto.
2248 * config/rs6000/sync.md: Ditto.
2249 * config/rs6000/vector.md: Ditto.
2250 * config/rs6000/vsx.md: Ditto.
2252 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2254 * config/rs6000/altivec.md: Write output control strings as braced
2255 blocks instead of double-quoted strings.
2256 * config/rs6000/darwin.md: Ditto.
2257 * config/rs6000/rs6000.md: Ditto.
2258 * config/rs6000/vector.md: Ditto.
2259 * config/rs6000/vsx.md: Ditto.
2261 2018-02-21 Jason Merrill <jason@redhat.com>
2263 PR c++/84314 - ICE with templates and fastcall attribute.
2264 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2266 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2268 * ipa-cp.c (determine_versionability): Fix comment typos.
2270 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2273 * ipa-cp.c (determine_versionability): Do not version functions caling
2276 2018-02-21 Martin Liska <mliska@suse.cz>
2279 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2280 Add "native" as a possible value.
2281 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2282 the macro when native cpu detection is available.
2284 2018-02-21 Martin Liska <mliska@suse.cz>
2287 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2288 Add "native" as a possible value.
2289 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2290 when native cpu detection is available.
2292 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2293 Martin Sebor <msebor@redhat.com>
2295 PR tree-optimization/84478
2296 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2298 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2299 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2300 support which is conservatively correct, for 2 only stay conservative
2301 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2302 argument to the 2 argument get_range_strlen, adjust 6 arg
2303 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2305 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2306 (gimple_fold_builtin_strlen): Pass true as last argument to
2309 2018-02-20 Martin Sebor <msebor@redhat.com>
2312 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2313 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2314 (builtin_memref::builtin_memref): Factor out parts into
2315 set_base_and_offset and call it.
2317 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2320 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2321 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2322 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2323 search at the associated MODE_INT.
2325 2018-02-20 Jeff Law <law@redhat.com>
2328 PR tree-optimization/81592
2330 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2331 for range data rather than using global data.
2332 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2333 range data rather than using global data.
2334 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2335 pass it to children as needed.
2336 (struct directive::fmtresult): Similarly.
2337 (struct directive::set_width): Similarly.
2338 (struct directive::set_precision): Similarly.
2339 (format_integer, format_directive, parse_directive): Similarly.
2340 (format_none): Accept unnamed vr_values parameter.
2341 (format_percent, format_floating, format_character): Similarly.
2342 (format_string, format_plain): Similarly.
2343 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2344 the EVRP range analyzer for range data rather than using global data.
2345 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2346 gimple-ssa-evrp-analyze.h
2347 (class sprintf_dom_walker): Add after_dom_children member function.
2348 Add evrp_range_analyzer member.
2349 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2350 range analyzer as needed.
2351 (sprintf_dom_walker::after_dom_children): New member function.
2352 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2354 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2355 (evrp_range_analyzer::pop_to_marker): Likewise.
2357 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2359 PR tree-optimization/84419
2360 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2361 with the required type if its current type is compatible but
2364 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2367 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2368 after vectorization.
2370 2018-02-20 Martin Liska <mliska@suse.cz>
2373 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2374 possible values if we don't have a hint.
2376 2018-02-20 Martin Liska <mliska@suse.cz>
2380 * final.c (shorten_branches): Build align_tab array with one
2382 * opts.c (finish_options): Add alignment option limit check.
2383 (MAX_CODE_ALIGN): Likewise.
2384 (MAX_CODE_ALIGN_VALUE): Likewise.
2385 * doc/invoke.texi: Document maximum allowed option value for
2386 all -falign-* options.
2388 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2391 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2392 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2393 * var-tracking.c (emit_note_insn_var_location): Remove all references
2394 to NOTE_INSN_CALL_ARG_LOCATION.
2395 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2396 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2397 Use copy_rtx_if_shared.
2398 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2399 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2400 (dwarf2out_var_location): Remove handling of
2401 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2403 * final.c (final_scan_insn): Remove all references to
2404 NOTE_INSN_CALL_ARG_LOCATION.
2405 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2406 before dumping final insns.
2407 * except.c (emit_note_eh_region_end): Remove all references to
2408 NOTE_INSN_CALL_ARG_LOCATION.
2409 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2410 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2411 * config/arc/arc.c (hwloop_optimize): Likewise.
2412 * config/arm/arm.c (create_fix_barrier): Likewise.
2413 * config/s390/s390.c (s390_chunkify_start): Likewise.
2414 * config/sh/sh.c (find_barrier): Likewise.
2415 * config/i386/i386.c (rest_of_insert_endbranch,
2416 ix86_seh_fixup_eh_fallthru): Likewise.
2417 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2418 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2419 * config/frv/frv.c (frv_function_prologue): Likewise.
2420 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2422 (note_outside_basic_block_p): Remove all references to
2423 NOTE_INSN_CALL_ARG_LOCATION.
2424 * gengtype.c (adjust_field_rtx_def): Likewise.
2425 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2427 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2428 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2431 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2434 PR tree-optimization/84452
2435 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2436 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2439 2018-02-19 Martin Liska <mliska@suse.cz>
2442 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2444 2018-02-19 Martin Liska <mliska@suse.cz>
2445 Richard Sandiford <richard.sandiford@linaro.org>
2447 PR tree-optimization/82491
2448 * gimple-fold.c (get_base_constructor): Make earlier bail out
2451 2018-02-19 Carl Love <cel@us.ibm.com>
2453 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2454 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2456 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2457 P8V_BUILTIN_VEC_NEG.
2459 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2461 * config/rl78/rl78.md (movdf): New define expand.
2463 2018-02-19 Martin Liska <mliska@suse.cz>
2466 * doc/invoke.texi: Fix typo.
2467 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2469 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2471 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2472 handle rs6000_single_float and rs6000_double_float specially for
2475 2018-02-16 Jeff Law <law@redhat.com>
2477 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2480 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2483 * ipa-inline.c (inline_small_functions): Fix a typo.
2485 2018-02-16 Nathan Sidwell <nathan@acm.org>
2487 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2489 2018-02-16 Carl Love <cel@us.ibm.com>
2491 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2492 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2493 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2494 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2495 expansion to P8V_BUILTIN_VEC_FLOAT2.
2497 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2499 PR rtl-optimization/70023
2500 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2501 src_regno into account.
2503 2018-02-16 Carl Love <cel@us.ibm.com>
2505 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2506 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2507 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2508 * config/rs6000/rs6000.c: Remove case statements for
2509 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2510 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2511 and P9V_BUILTIN_VEC_VINSERT4B.
2512 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2513 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2514 * config/rs6000/vsx.md:
2515 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2518 2018-02-16 Carl Love <cel@us.ibm.com>
2520 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2522 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2524 * config/rs6000/rs6000-c.c: Add the definitions for
2525 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2526 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2527 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2528 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2529 definition for insert4b and define insn *insert3b_internal.
2530 * doc/extend.texi: Add documentation for vec_extract4b.
2532 2018-02-16 Nathan Sidwell <nathan@acm.org>
2534 * doc/extend.texi (Backwards Compatibility): Mention friend
2535 injection. Note for-scope is deprecated.
2536 * doc/invoke.texi (-ffriend-injection): Deprecate.
2538 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2540 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2541 that moved to I2, also allow destinations that are a paradoxical
2542 subreg (instead of a normal reg).
2544 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2547 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2550 2018-02-16 Richard Biener <rguenther@suse.de>
2552 PR tree-optimization/84037
2553 PR tree-optimization/84016
2555 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2556 Adjust vec_construct for the fact we need additional higher latency
2557 128bit inserts for AVX256 and AVX512 vector builds.
2558 (ix86_add_stmt_cost): Scale vector construction cost for
2561 2018-02-16 Richard Biener <rguenther@suse.de>
2563 PR tree-optimization/84417
2564 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2565 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2566 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2568 2018-02-16 Martin Liska <mliska@suse.cz>
2571 * internal-fn.def (ASAN_CHECK): Set proper flags.
2572 (ASAN_MARK): Likewise.
2574 2018-02-16 Julia Koval <julia.koval@intel.com>
2576 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2577 from PTA_CANNONLAKE.
2579 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2582 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2583 Use ++iter rather than iter++ for std::list iterators.
2584 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2585 defer deleting them until all nodes in the forest are processed. Do
2586 free even leaf nodes. Change to_process into auto_vec.
2589 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2590 * vec.h (vec_default_construct): Use memset instead of placement new
2591 if BROKEN_VALUE_INITIALIZATION is defined.
2592 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2593 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2596 PR rtl-optimization/83723
2597 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2598 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2599 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2601 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2603 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2605 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2607 PR rtl-optimization/81443
2608 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2609 from inner REGs to paradoxical SUBREGs.
2611 2018-02-16 Richard Biener <rguenther@suse.de>
2613 PR tree-optimization/84399
2614 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2615 For operands we can analyze at their definition make sure we can
2616 analyze them at each use as well.
2618 2018-02-16 Richard Biener <rguenther@suse.de>
2620 PR tree-optimization/84190
2621 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2622 volatile accesses if the decl isn't volatile.
2624 2018-02-15 Jason Merrill <jason@redhat.com>
2626 PR c++/84314 - ICE with templates and fastcall attribute.
2627 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2628 TYPE_CANONICAL on an existing type.
2630 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2632 PR tree-optimization/84383
2633 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2634 dstoff nor call operand_equal_p if dstbase is NULL.
2636 PR tree-optimization/84334
2637 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2638 also a CONSTANT_CLASS_P, punt.
2640 2018-02-14 Jim Wilson <jimw@sifive.com>
2642 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2643 first SMALL_OPERAND check. New local min_second_step. Move assert
2644 to where locals are set. Add TARGET_RVC support.
2645 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2647 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2649 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2651 2018-02-14 Martin Sebor <msebor@redhat.com>
2653 PR tree-optimization/83698
2654 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2655 arrays constrain the offset range to their bounds.
2656 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2657 (builtin_access::overlap): Avoid setting the size of overlap if it's
2659 (maybe_diag_overlap): Also consider arrays when deciding what values
2660 of offsets to include in diagnostics.
2662 2018-02-14 Martin Sebor <msebor@redhat.com>
2665 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2666 that correspond to the kind of a declaration.
2668 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2671 * config/pa/pa.md: Load address of PIC label using the linkage table
2672 if the label is nonlocal.
2674 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2676 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2677 warning message if user requests -maltivec=be.
2678 * doc/invoke.texi: Document deprecation of -maltivec=be.
2680 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2683 * config/rs6000/rs6000-c.c: Update definitions for
2684 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2685 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2687 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2690 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2691 add _get_ssp intrinsics. Remove argument from
2692 __builtin_ia32_rdssp[d|q].
2693 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2694 * config/i386/i386-builtin.def: Remove argument from
2695 __builtin_ia32_rdssp[d|q].
2696 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2697 ix86_expand_special_args_builtin for _rdssp[d|q].
2698 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2699 Clear register before usage.
2700 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2701 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2703 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2705 PR tree-optimization/84357
2706 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2707 operand 1 of an ARRAY_REF too.
2709 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2712 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2713 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2715 (set_of_reg): New struct.
2716 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2717 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2718 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2720 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2721 Split into bitclr, bitset, bitinvert patterns if appropriate.
2722 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2723 use rx_fuse_in_memory_bitop.
2724 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2725 to named insn, correct maximum insn length.
2727 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2730 * machmode.def: Define a complex mode for PARTIAL_INT.
2731 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2733 * doc/rtl.texi: Document CSPImode.
2734 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2736 (msp430_hard_regno_nregs_with_padding): Likewise.
2738 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2741 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2743 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2745 PR rtl-optimization/84169
2746 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2747 we generated a parallel as new i3 and we split that to new i2 and i3
2748 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2749 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2750 those to i2, not i1. Partially rewrite this scan code.
2752 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2755 * stor-layout.c (place_field): For variable length fields, adjust
2756 offset_align afterwards not just based on the field's alignment,
2757 but also on the size.
2760 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2761 of exps and logs in the use_exp2 case.
2763 2018-02-13 Jeff Law <law@redhat.com>
2765 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2768 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2771 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2775 * common.opt (gas-loc-support, gas-locview-support): New.
2776 (ginline-points, ginternal-reset-location-views): New.
2777 * doc/invoke.texi: Document them. Use @itemx where intended.
2778 (gvariable-location-views): Adjust.
2779 * target.def (reset_location_view): New.
2780 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2781 (TARGET_RESET_LOCATION_VIEW): New.
2782 * doc/tm.texi: Rebuilt.
2783 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2784 (dwarf2out_default_as_locview_support): New.
2785 (output_asm_line_debug_info): Use option variables.
2786 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2787 (output_loc_list): Likewise.
2788 (add_high_low_attributes): Check option variables.
2789 Don't output entry view attribute in strict mode.
2790 (gen_inlined_subroutine_die): Check option variables.
2791 (dwarf2out_inline_entry): Likewise.
2792 (init_sections_and_labels): Likewise.
2793 (dwarf2out_early_finish): Likewise.
2794 (maybe_reset_location_view): New, from...
2795 (dwarf2out_var_location): ... here. Call it.
2796 * debug.h (dwarf2out_default_as_loc_support): Declare.
2797 (dwarf2out_default_as_locview_support): Declare.
2798 * hooks.c (hook_int_rtx_insn_0): New.
2799 * hooks.h (hook_int_rtx_insn_0): Declare.
2800 * toplev.c (process_options): Take -gas-loc-support and
2801 -gas-locview-support from dwarf2out. Enable
2802 -gvariable-location-views by default only with locview
2803 assembler support. Enable -ginternal-reset-location-views by
2804 default only if the target defines the corresponding hook.
2805 Enable -ginline-points by default if location views are
2806 enabled; force it disabled if statement frontiers are
2808 * tree-inline.c (expand_call_inline): Check option variables.
2809 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2811 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2813 PR tree-optimization/84321
2814 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2815 handling. Also check whether the anti-range contains any values
2816 that satisfy the mask; switch to a VR_RANGE if not.
2818 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2821 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2823 2018-02-13 Martin Jambor <mjambor@suse.cz>
2826 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2827 of call statements, also set location of a load to a temporary.
2829 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2831 * config/rl78/rl78.c (add_vector_labels): New function.
2832 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2833 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2834 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2835 which checks that no arguments are passed.
2836 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2837 * doc/extend.texi: Documentation for the new attribute.
2839 2018-02-13 Andreas Schwab <schwab@suse.de>
2841 * config/riscv/linux.h (CPP_SPEC): Define.
2843 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2846 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2847 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2848 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2849 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2850 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2851 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2852 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2853 temporarily for AES and PCLMUL builtins.
2855 PR tree-optimization/84339
2856 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2857 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2861 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2862 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2863 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2865 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2867 * omp-simd-clone.h: New file.
2868 * omp-simd-clone.c: Include omp-simd-clone.h.
2869 (expand_simd_clones): No longer static.
2870 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2871 cgraph.h and omp-simd-clone.h.
2872 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2873 (vect_recog_widen_shift_pattern): Formatting fix.
2874 (vect_pattern_recog_1): Don't check optab for calls.
2877 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2878 operands[2] into a REG before using gen_lowpart on it.
2880 2018-02-12 Jeff Law <law@redhat.com>
2883 * config/sh/sh.c (find_barrier): Consider a sibling call
2886 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2887 successfully back substituting a reg.
2889 2018-02-12 Richard Biener <rguenther@suse.de>
2891 PR tree-optimization/84037
2892 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2893 parameter, move visited init to caller.
2894 (vect_slp_analyze_operations): Separate cost from validity
2895 check, initialize visited once for all instances.
2896 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2898 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2900 (vect_model_promotion_demotion_cost): Likewise.
2901 (vectorizable_bswap): Guard cost modeling with !slp_node
2902 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2904 (vectorizable_call): Likewise.
2905 (vectorizable_conversion): Likewise.
2906 (vectorizable_assignment): Likewise.
2907 (vectorizable_shift): Likewise.
2908 (vectorizable_operation): Likewise.
2909 (vectorizable_store): Likewise.
2910 (vectorizable_load): Likewise.
2911 (vectorizable_condition): Likewise.
2912 (vectorizable_comparison): Likewise.
2914 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2917 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2918 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2919 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2921 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2924 * params.def (inline-min-speedup): Increase from 8 to 15.
2925 (max-inline-insns-auto): Decrease from 40 to 30.
2926 * ipa-split.c (consider_split): Add some buffer for function to
2927 be considered inlining candidate.
2928 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2931 2018-02-12 Richard Biener <rguenther@suse.de>
2933 PR tree-optimization/84037
2934 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2935 matched stmts if we cannot swap the non-matched ones.
2937 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2939 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2940 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2941 _mm_maskz_scalef_round_ss): New intrinsics.
2942 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2943 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2944 __builtin_ia32_scalefss_round): Remove.
2945 (__builtin_ia32_scalefsd_mask_round,
2946 __builtin_ia32_scalefss_mask_round): New intrinsics.
2947 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2948 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2949 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2950 "<round_constraint>")): Changed to ...
2951 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2952 "<round_scalar_constraint>")): ... this.
2953 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2954 %0, %1, %2<round_op3>}"): Changed to ...
2955 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2956 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2957 %2<round_scalar_mask_op3>}"): ... this.
2958 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2960 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2962 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2963 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2964 (_mm_maskz_sqrt_round_ss): New intrinsics.
2965 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2966 (__builtin_ia32_sqrtsd_mask_round)
2967 (__builtin_ia32_sqrtss_mask_round): New builtins.
2968 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2969 (__builtin_ia32_sqrtss_round): Remove.
2970 (__builtin_ia32_sqrtsd_mask_round)
2971 (__builtin_ia32_sqrtss_mask_round): New builtins.
2972 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2973 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2974 ((match_operand:VF_128 1 "vector_operand"
2975 "xBm,<round_constraint>")): Changed to ...
2976 ((match_operand:VF_128 1 "vector_operand"
2977 "xBm,<round_scalar_constraint>")): ... this.
2978 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2979 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2980 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2981 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2982 %<iptr>1<round_scalar_mask_op3>}): ... this.
2983 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2984 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2986 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2989 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2990 Cast vec_cmpeq result to correct type.
2991 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2992 Cast vec_cmpgt result to correct type.
2994 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2996 * final.c (final_scan_insn_1): Renamed from...
2997 (final_scan_insn): ... this. New wrapper, to recover
2998 seen from the outermost call in recursive ones.
2999 * config/sparc/sparc.c (output_return): Drop seen from call.
3000 (output_sibcall): Likewise.
3001 * config/visium/visium.c (output_branch): Likewise.
3003 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3005 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3008 2018-02-10 Alan Modra <amodra@gmail.com>
3011 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3012 Specify LR as an input.
3014 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3017 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3018 remove_member_access_dummy_vars): New functions.
3019 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3020 lower_omp_1, execute_lower_omp): Use them.
3022 PR rtl-optimization/84308
3023 * shrink-wrap.c (spread_components): Release todo vector.
3025 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3027 PR rtl-optimization/57193
3028 * ira-color.c (struct allocno_color_data): Add member
3029 conflict_allocno_hard_prefs.
3030 (update_conflict_allocno_hard_prefs): New.
3031 (bucket_allocno_compare_func): Add a preference based on
3032 conflict_allocno_hard_prefs.
3033 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3034 (color_allocnos): Remove a dead code. Initiate
3035 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3037 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3040 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3041 constraint from =wa to wa. Avoid a subreg on the output operand,
3042 instead use a pseudo and subreg it in a move.
3043 (p9_xxbrd_<mode>): Changed to ...
3044 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3045 (p9_xxbrd_v2df): New expander.
3046 (p9_xxbrw_<mode>): Changed to ...
3047 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3048 (p9_xxbrw_v4sf): New expander.
3050 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3052 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3054 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3057 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3058 multiply in 32-bit mode.
3059 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3060 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3063 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3065 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3066 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3067 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3068 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3070 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3073 * dwarf2out.c (is_trivial_indirect_ref): New function.
3074 (dwarf2out_late_global_decl): Do not generate a location
3075 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3076 and that are not defined in the current unit.
3078 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3080 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3081 instead of a libcall for UNORDERED.
3083 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3086 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3087 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3089 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3092 * config/s390/s390.c (s390_set_current_function): Invoke
3093 s390_indirect_branch_settings also if fndecl didn't change.
3095 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3097 * config/rs6000/rs6000.md (blockage): Set length to zero.
3099 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3101 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3103 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3106 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3107 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3111 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3112 PARALLEL incoming that failed vt_get_decl_and_offset check.
3115 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3116 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3118 (get_variable_section): For decls in named .bss* sections pass true as
3119 second argument to bss_initializer_p.
3121 2018-02-09 Marek Polacek <polacek@redhat.com>
3122 Jakub Jelinek <jakub@redhat.com>
3125 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3126 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3127 Sync some changes from cxx_fold_indirect_ref.
3129 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3131 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3133 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3134 (BLOCK_INLINE_ENTRY_LABEL): New.
3135 (dwarf2out_var_location): Disregard inline entry markers.
3136 (inline_entry_data): New struct.
3137 (inline_entry_data_hasher): New hashtable type.
3138 (inline_entry_data_hasher::hash): New.
3139 (inline_entry_data_hasher::equal): New.
3140 (inline_entry_data_table): New variable.
3141 (add_high_low_attributes): Add DW_AT_entry_pc and
3142 DW_AT_GNU_entry_view attributes if a pending entry is found
3143 in inline_entry_data_table. Add old entry_pc attribute only
3144 if debug nonbinding markers are disabled.
3145 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3146 markers are enabled.
3147 (block_within_block_p, dwarf2out_inline_entry): New.
3148 (dwarf2out_finish): Check that no entries remained in
3149 inline_entry_data_table.
3150 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3151 (final_scan_insn, notice_source_line): Likewise.
3152 (rest_of_clean_state): Skip inline entry markers.
3153 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3155 * gimple.c (gimple_build_debug_inline_entry): New.
3156 * gimple.h (enum gimple_debug_subcode): Add
3157 GIMPLE_DEBUG_INLINE_ENTRY.
3158 (gimple_build_debug_inline_entry): Declare.
3159 (gimple_debug_inline_entry_p): New.
3160 (gimple_debug_nonbind_marker_p): Adjust.
3161 * insn-notes.def (INLINE_ENTRY): New.
3162 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3163 inline entry marker notes.
3164 (print_insn): Likewise.
3165 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3166 (INSN_DEBUG_MARKER_KIND): Likewise.
3167 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3168 * tree-inline.c (expand_call_inline): Build and insert
3169 debug_inline_entry stmt.
3170 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3171 inline entry blocks early, if nonbind markers are enabled.
3172 (dump_scope_block): Dump fragment info.
3173 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3174 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3175 (gimple_build_debug_inline_entry): New.
3176 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3177 Enable/disable inline entry points too.
3178 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3179 (DEBUG_INSN): Describe inline entry markers.
3181 * common.opt (gvariable-location-views): New.
3182 (gvariable-location-views=incompat5): New.
3183 * config.in: Rebuilt.
3184 * configure: Rebuilt.
3185 * configure.ac: Test assembler for view support.
3186 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3187 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3188 * dwarf2out.c (var_loc_view): New typedef.
3189 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3190 (dwarf2out_locviews_in_attribute): New.
3191 (dwarf2out_locviews_in_loclist): New.
3192 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3193 (enum dw_line_info_opcode): Add LI_adv_address.
3194 (struct dw_line_info_table): Add view.
3195 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3196 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3197 (zero_view_p): New variable.
3198 (ZERO_VIEW_P): New macro.
3199 (output_asm_line_debug_info): New.
3200 (struct var_loc_node): Add view.
3201 (add_AT_view_list, AT_loc_list): New.
3202 (add_var_loc_to_decl): Add view param. Test it against last.
3203 (new_loc_list): Add view params. Record them.
3204 (AT_loc_list_ptr): Handle loc and view lists.
3205 (view_list_to_loc_list_val_node): New.
3206 (print_dw_val): Handle dw_val_class_view_list.
3207 (size_of_die): Likewise.
3208 (value_format): Likewise.
3209 (loc_list_has_views): New.
3210 (gen_llsym): Set vl_symbol too.
3211 (maybe_gen_llsym, skip_loc_list_entry): New.
3212 (dwarf2out_maybe_output_loclist_view_pair): New.
3213 (output_loc_list): Output view list or entries too.
3214 (output_view_list_offset): New.
3215 (output_die): Handle dw_val_class_view_list.
3216 (output_dwarf_version): New.
3217 (output_compilation_unit_header): Use it.
3218 (output_skeleton_debug_sections): Likewise.
3219 (output_rnglists, output_line_info): Likewise.
3220 (output_pubnames, output_aranges): Update version comments.
3221 (output_one_line_info_table): Output view numbers in asm comments.
3222 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3223 Call maybe_gen_llsym.
3224 (loc_list_from_tree_1): Adjust.
3225 (add_AT_location_description): Create view list attribute if
3226 needed, check it's absent otherwise.
3227 (convert_cfa_to_fb_loc_list): Adjust.
3228 (maybe_emit_file): Call output_asm_line_debug_info for test.
3229 (dwarf2out_var_location): Reset views as needed. Precompute
3230 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3231 attribute. Set view.
3232 (new_line_info_table): Reset next view.
3233 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3234 (dwarf2out_source_line): Likewise. Output view resets and labels to
3235 the assembler, or select appropriate line info opcodes.
3236 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3237 (optimize_string_length): Catch it. Adjust.
3238 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3239 dw_val_class_view_list, and remove it if no longer needed.
3240 (hash_loc_list): Hash view numbers.
3241 (loc_list_hasher::equal): Compare them.
3242 (optimize_location_lists): Check whether a view list symbol is
3243 needed, and whether the locview attribute is present, and
3244 whether they match. Remove the locview attribute if no longer
3246 (index_location_lists): Call skip_loc_list_entry for test.
3247 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3248 Use output_dwarf_version.
3249 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3250 (struct dw_val_node): Add val_view_list.
3251 * final.c (SEEN_NEXT_VIEW): New.
3252 (set_next_view_needed): New.
3253 (clear_next_view_needed): New.
3254 (maybe_output_next_view): New.
3255 (final_start_function): Rename to...
3256 (final_start_function_1): ... this. Take pointer to FIRST,
3257 add SEEN parameter. Emit param bindings in the initial view.
3258 (final_start_function): Reintroduce SEEN-less interface.
3259 (final): Rename to...
3260 (final_1): ... this. Take SEEN parameter. Output final pending
3261 next view at the end.
3262 (final): Reintroduce seen-less interface.
3263 (final_scan_insn): Output pending next view before switching
3264 sections or ending a block. Mark the next view as needed when
3265 outputting variable locations. Notify debug backend of section
3266 changes, and of location view changes.
3267 (rest_of_handle_final): Adjust.
3268 * toplev.c (process_options): Autodetect value for debug variable
3269 location views option. Warn on incompat5 without -gdwarf-5.
3270 * doc/invoke.texi (gvariable-location-views): New.
3271 (gvariable-location-views=incompat5): New.
3272 (gno-variable-location-views): New.
3274 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3276 PR tree-optimization/84136
3277 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3278 that the result of find_edge is non-NULL.
3280 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3283 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3284 storing integer register in SImode. Fix cost of 256 and 512
3285 byte aligned SSE register store.
3287 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3289 * config/i386/i386.c (ix86_multiplication_cost): Fix
3290 multiplication cost for TARGET_AVX512DQ.
3292 2018-02-08 Marek Polacek <polacek@redhat.com>
3294 PR tree-optimization/84238
3295 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3298 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3300 PR tree-optimization/84265
3301 * tree-vect-stmts.c (vectorizable_store): Don't treat
3302 VMAT_CONTIGUOUS accesses as grouped.
3303 (vectorizable_load): Likewise.
3305 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3307 PR tree-optimization/81635
3308 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3309 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3310 (test_round_for_mask): New functions.
3311 (wide_int_cc_tests): Call test_round_for_mask.
3312 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3313 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3314 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3315 range returned by get_range_info.
3317 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3320 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3321 * symtab.c: Include builtins.h
3322 (symtab_node::output_to_lto_symbol_table_p): Move here
3323 from lto-streamer-out.c:output_symbol_p.
3324 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3325 (output_symbol_p): Move all logic to symtab.c
3326 (produce_symtab): Update.
3328 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3330 * config/s390/s390-opts.h (enum indirect_branch): Define.
3331 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3332 (s390_indirect_branch_via_thunk)
3333 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3334 (enum s390_indirect_branch_type): Define.
3335 * config/s390/s390.c (struct s390_frame_layout, struct
3336 machine_function): Remove.
3337 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3338 (indirect_branch_table_label_no, indirect_branch_table_name):
3340 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3341 (enum s390_indirect_branch_option): Define.
3342 (s390_return_addr_from_memory): New function.
3343 (s390_handle_string_attribute): New function.
3344 (s390_attribute_table): Add new attribute handler.
3345 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3346 (s390_indirect_branch_via_thunk): New function.
3347 (s390_indirect_branch_via_inline_thunk): New function.
3348 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3349 sibling call optimization for non z10 compiles.
3350 (s390_emit_call): Force indirect branch target to be a single
3351 register. Add r1 clobber for non-z10 compiles.
3352 (s390_emit_epilogue): Emit return jump via return_use expander.
3353 (s390_reorg): Handle JUMP_INSNs as execute targets.
3354 (s390_option_override_internal): Perform validity checks for the
3355 new command line options.
3356 (s390_indirect_branch_attrvalue): New function.
3357 (s390_indirect_branch_settings): New function.
3358 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3359 (s390_output_indirect_thunk_function): New function.
3360 (s390_code_end): Implement target hook.
3361 (s390_case_values_threshold): Implement target hook.
3362 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3364 * config/s390/s390.h (struct s390_frame_layout)
3365 (struct machine_function): Move here from s390.c.
3366 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3367 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3368 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3369 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3370 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3371 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3372 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3373 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3374 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3375 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3376 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3377 (mnemonic attribute): Add values which aren't recognized
3379 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3380 pattern for branch conversion. Fix mnemonic attribute.
3381 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3382 indirect branch via thunk if requested.
3383 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3384 ("*indirect_jump"): Disable for branch conversion using out of
3386 ("indirect_jump_via_thunk<mode>_z10")
3387 ("indirect_jump_via_thunk<mode>")
3388 ("indirect_jump_via_inlinethunk<mode>_z10")
3389 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3390 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3391 ("casesi_jump_via_inlinethunk<mode>_z10")
3392 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3393 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3394 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3395 ("*indirect2_jump"): Disable for branch conversion.
3396 ("casesi_jump"): Turn into expander and expand patterns for branch
3398 ("return_use"): New expander.
3399 ("*return"): Emit return via thunk and rename it to ...
3400 ("*return<mode>"): ... this one.
3401 * config/s390/s390.opt: Add new options and and enum for the
3404 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3406 * lra-constraints.c (match_reload): Unconditionally use
3407 gen_lowpart_SUBREG, rather than selecting between that
3408 and equivalent gen_rtx_SUBREG code.
3410 2018-02-08 Richard Biener <rguenther@suse.de>
3412 PR tree-optimization/84233
3413 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3414 changed flag instead of boguously re-using phi_inserted.
3416 2018-02-08 Martin Jambor <mjambor@suse.cz>
3418 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3419 static local variables.
3421 2018-02-08 Richard Biener <rguenther@suse.de>
3423 PR tree-optimization/84278
3424 * tree-vect-stmts.c (vectorizable_store): When looking for
3425 smaller vector types to perform grouped strided loads/stores
3426 make sure the mode is supported by the target.
3427 (vectorizable_load): Likewise.
3429 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3431 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3432 Increase LDP/STP opportunities by adding adjacent callee-saves.
3434 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3436 PR rtl-optimization/84068
3437 PR rtl-optimization/83459
3438 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3440 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3442 PR tree-optimization/84224
3443 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3444 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3447 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3450 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3451 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3452 count, remove one USE.
3454 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3456 * doc/install.texi (Configuration): Document the
3457 --with-long-double-format={ibm,ieee} PowerPC configuration
3461 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3462 Convert from define_expand to be define_insn_and_split. Rework
3463 float/double/_Float128 conversions to QI/HI/SImode to work with
3464 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3465 conversions to QI/HImode types did a store and then a load to
3466 truncate the value. For conversions to VSX registers, don't split
3467 the insn, instead emit the code directly. Use the code iterator
3468 any_fix to combine signed and unsigned conversions.
3469 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3470 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3471 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3472 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3473 (fix_<mode>di2_hw): Likewise.
3474 (fixuns_<mode>di2_hw): Likewise.
3475 (fix_<mode>si2_hw): Likewise.
3476 (fixuns_<mode>si2_hw): Likewise.
3477 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3478 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3479 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3480 fix<uns>_trunc<SFDF:mode>si2_p8.
3481 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3483 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3484 (fix<uns>_<mode>_mem): Likewise.
3485 (fctiw<u>z_<mode>_mem): Likewise.
3486 (fix<uns>_<mode>_mem): Likewise.
3487 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3488 the register allocator from doing a direct move to the GPRs to do
3489 a store, and instead use the ISA 3.0 store byte/half-word from
3490 vector register instruction. For IEEE 128-bit floating point,
3491 also optimize stores of 32-bit ints.
3492 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3494 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3496 * genextract.c (push_pathstr_operand): New function to support
3498 (walk_rtx): Call push_pathstr_operand.
3499 (print_path): Support [a-zA-Z].
3501 2018-02-07 Richard Biener <rguenther@suse.de>
3503 PR tree-optimization/84037
3504 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3505 (cse_and_gimplify_to_preheader): Declare.
3506 (vect_get_place_in_interleaving_chain): Likewise.
3507 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3509 (_loop_vec_info::~_loop_vec_info): Delete it.
3510 (cse_and_gimplify_to_preheader): New function.
3511 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3512 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3513 (vectorizable_load): Likewise. For grouped stores always base
3514 the IV on the first element.
3515 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3516 condition before gimplifying.
3518 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3520 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3521 *DIV_EXPR and *MOD_EXPR.
3523 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3526 * config/i386/i386.c (ix86_option_override_internal): Mask out
3527 the CF_SET bit when checking -fcf-protection.
3529 2018-02-07 Tom de Vries <tom@codesourcery.com>
3532 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3535 2018-02-07 Richard Biener <rguenther@suse.de>
3537 PR tree-optimization/84204
3538 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3541 PR tree-optimization/84205
3542 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3543 special-case isl_ast_op_zdiv_r.
3545 PR tree-optimization/84223
3546 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3547 Only add conditions from within the region.
3548 (gather_bbs::after_dom_children): Adjust.
3550 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3553 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3554 * config/avr/avr.md: Only post-reload split REG-REG moves if
3555 either register is GENERAL_REG_P.
3557 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3559 PR tree-optimization/84235
3560 * tree-ssa-scopedtables.c
3561 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3562 if the subtraction is performed in floating point type where NaNs are
3563 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3564 build 1. Formatting fix.
3566 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3569 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3570 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3571 and skip it regardless of bb boundaries. Use CALL_P macro,
3572 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3573 unnecessarily, formatting fix.
3575 2018-02-06 Michael Collison <michael.collison@arm.com>
3577 * config/arm/thumb2.md:
3578 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3579 (*thumb_mov_notscc): Ditto.
3581 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3584 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3585 unsigned_fix, not "s".
3587 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3589 * configure.ac (gcc_fn_eh_frame_ro): New function.
3590 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3591 correct .eh_frame permissions.
3592 * configure: Regenerate.
3594 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3596 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3599 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3601 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3602 Display warning message for -mno-speculate-indirect-jumps.
3604 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3606 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3608 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3610 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3612 PR tree-optimization/84225
3613 * tree-eh.c (find_trapping_overflow): Only call
3614 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3616 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3619 * config/i386/i386.c: Reimplement the check of possible options
3620 -mibt/-mshstk conbination. Change error messages.
3621 * doc/invoke.texi: Fix a typo: remove extra '='.
3623 2018-02-06 Marek Polacek <polacek@redhat.com>
3625 PR tree-optimization/84228
3626 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3628 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3631 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3632 emitted arch directives.
3633 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3634 __ARM_FEATURE_COPROC before changing architectures.
3636 2018-02-06 Richard Biener <rguenther@suse.de>
3638 * config/i386/i386.c (print_reg): Fix typo.
3639 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3641 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3643 * configure: Regenerate.
3645 2018-02-05 Martin Sebor <msebor@redhat.com>
3647 PR tree-optimization/83369
3648 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3651 2018-02-05 Martin Liska <mliska@suse.cz>
3653 * doc/invoke.texi: Cherry-pick upstream r323995.
3655 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3657 * ira.c (ira_init_register_move_cost): Adjust comment.
3659 2018-02-05 Martin Liska <mliska@suse.cz>
3661 PR gcov-profile/84137
3662 * doc/gcov.texi: Fix typo in documentation.
3664 2018-02-05 Martin Liska <mliska@suse.cz>
3666 PR gcov-profile/83879
3667 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3668 using dlopen functionality.
3670 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3672 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3673 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3674 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3675 _mm_maskz_range_round_ss): New intrinsics.
3676 (__builtin_ia32_rangesd128_round)
3677 (__builtin_ia32_rangess128_round): Remove.
3678 (__builtin_ia32_rangesd128_mask_round,
3679 __builtin_ia32_rangess128_mask_round): New builtins.
3680 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3681 __builtin_ia32_rangess128_round): Remove.
3682 (__builtin_ia32_rangesd128_mask_round,
3683 __builtin_ia32_rangess128_mask_round): New builtins.
3684 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3685 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3686 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3687 "<round_saeonly_constraint>")): Changed to ...
3688 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3689 "<round_saeonly_scalar_constraint>")): ... this.
3690 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3691 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3692 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3693 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3694 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3696 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3698 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3700 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3701 Remove all values except native, 8540 and 8548.
3703 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3705 * config/i386/i386.c (ix86_output_function_return): Pass
3706 INVALID_REGNUM, instead of -1, as invalid register number to
3707 indirect_thunk_name and output_indirect_thunk.
3709 2018-02-02 Julia Koval <julia.koval@intel.com>
3711 * config.gcc: Add -march=icelake.
3712 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3713 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3714 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3715 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3716 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3717 (processor_target_table): Add icelake.
3718 (ix86_option_override_internal): Handle new PTAs.
3719 (get_builtin_code_for_version): Handle icelake.
3720 (M_INTEL_COREI7_ICELAKE): New.
3721 (fold_builtin_cpu): Handle icelake.
3722 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3723 * doc/invoke.texi: Add -march=icelake.
3725 2018-02-02 Julia Koval <julia.koval@intel.com>
3727 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3728 to wide_int_bitmask.
3729 * wide-int-bitmask.h: New.
3731 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3734 * config/i386/i386.md: Replace Pmode with word_mode in
3735 builtin_setjmp_setup and builtin_longjmp to support x32.
3737 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3741 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3743 (rs6000_supported_cpu_names): New static variable.
3744 (linux_cpu_translation_table): Likewise.
3745 (elf_platform) <cpu>: Define new static variable and use it.
3746 Translate kernel AT_PLATFORM name to canonical name if needed.
3747 Error if platform name is unknown.
3749 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3752 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3754 2018-02-01 Jeff Law <law@redhat.com>
3757 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3758 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3759 the scratch if RELEASE_VIA_POP is false.
3760 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3761 If we have to save a temporary register, decrement SIZE appropriately.
3762 Pass new arguments to release_scratch_register_on_entry.
3763 (ix86_adjust_stack_and_probe): Likewise.
3764 (ix86_emit_probe_stack_range): Pass new arguments to
3765 release_scratch_register_on_entry.
3767 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3769 PR rtl-optimization/84157
3770 * combine.c (change_zero_ext): Use REG_P predicate in
3771 front of HARD_REGISTER_P predicate.
3773 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3775 * config/avr/avr.c (avr_option_override): Move disabling of
3776 -fdelete-null-pointer-checks to...
3777 * common/config/avr/avr-common.c (avr_option_optimization_table):
3780 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3782 PR tree-optimization/81635
3783 * tree-data-ref.c (split_constant_offset_1): For types that
3784 wrap on overflow, try to use range info to prove that wrapping
3787 2018-02-01 Renlin Li <renlin.li@arm.com>
3790 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3792 (aarch64_register_move_cost): Likewise.
3793 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3795 (REG_CLASS_NAMES): Likewise.
3796 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3797 TAILCALL_ADDR_REGS. Remove IP registers.
3798 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3800 2018-02-01 Richard Biener <rguenther@suse.de>
3802 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3803 for specifying RPO order and allow NULL for that.
3804 * domwalk.c (dom_walker::dom_walker): Likewise.
3805 (dom_walker::walk): Handle NULL RPO order.
3806 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3808 (rewrite_update_dom_walker): Likewise.
3809 (mark_def_dom_walker): Likewise.
3811 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3813 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3814 (aarch64_maybe_expand_sve_subreg_move): Declare.
3815 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3816 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3818 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3819 that are semantically a reverse operation.
3820 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3821 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3822 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3824 (aarch64_can_change_mode_class): For big-endian, forbid changes
3825 between two SVE modes if they have different element sizes.
3827 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3829 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3830 the TImode handling for big-endian targets.
3832 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3834 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3835 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3837 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3838 Remove BSWAP handing for big-endian targets and use the form of
3839 LD1RQ appropariate for the mode.
3841 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3843 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3844 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3847 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3850 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3851 check for operands that need to go through aarch64_sve_reload_be.
3853 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3855 PR tree-optimization/81661
3856 PR tree-optimization/84117
3857 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3858 * tree-eh.c: Include gimplify.h.
3859 (find_trapping_overflow, replace_trapping_overflow,
3860 rewrite_to_non_trapping_overflow): New functions.
3861 * tree-vect-loop.c: Include tree-eh.h.
3862 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3863 * tree-data-ref.c: Include tree-eh.h.
3864 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3866 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3868 PR rtl-optimization/84123
3869 * combine.c (change_zero_ext): Check if hard register satisfies
3870 can_change_dest_mode before calling gen_lowpart_SUBREG.
3872 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3875 * ira.c (ira_init_register_move_cost): Remove assert.
3877 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3879 PR rtl-optimization/84071
3880 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3881 * doc/tm.texi: Regenerate.
3883 2018-01-31 Richard Biener <rguenther@suse.de>
3885 PR tree-optimization/84132
3886 * tree-data-ref.c (analyze_miv_subscript): Properly
3887 check whether evolution_function_is_affine_multivariate_p
3888 before calling gcd_of_steps_may_divide_p.
3890 2018-01-31 Julia Koval <julia.koval@intel.com>
3893 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3894 * config/i386/i386.md (rdpid_rex64) New.
3895 (rdpid): Make 32bit only.
3897 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3900 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3901 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3903 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3906 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3908 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3910 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3912 PR rtl-optimization/84071
3913 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3914 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3916 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3918 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3919 (arc_attribute_table): Add 'aux' attribute.
3920 (arc_in_small_data_p): Consider aux like variables.
3921 (arc_is_aux_reg_p): New function.
3922 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3923 (arc_get_aux_arg): New function.
3924 (prepare_move_operands): Handle aux-register access.
3925 (arc_handle_aux_attribute): New function.
3926 * doc/extend.texi (ARC Variable attributes): Add subsection.
3928 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3930 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3931 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3932 (arc_attribute_table): Add 'uncached' attribute.
3933 (arc_print_operand): Print '.di' flag for uncached memory
3935 (arc_in_small_data_p): Do not consider for small data the uncached
3937 (arc_is_uncached_mem_p): New function.
3938 * config/arc/predicates.md (compact_store_memory_operand): Check
3939 for uncached memory accesses.
3940 (nonvol_nonimm_operand): Likewise.
3941 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3943 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3946 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3947 falign-loops=): Add Optimization flag.
3949 2018-01-30 Jeff Law <law@redhat.com>
3952 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3953 INT_REGISTERS_SAVED. Check it prior to calling
3954 get_scratch_register_on_entry.
3955 (ix86_adjust_stack_and_probe): Similarly.
3956 (ix86_emit_probe_stack_range): Similarly.
3957 (ix86_expand_prologue): Corresponding changes.
3959 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3962 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3963 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3965 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3968 * lra-constraints.c (curr_insn_transform): Process AND in the
3971 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3973 PR rtl-optimization/83986
3974 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3975 dependence against last_pending_memory_flush in addition to
3978 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3980 PR tree-optimization/81611
3981 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3984 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3987 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3990 2018-01-30 Richard Biener <rguenther@suse.de>
3991 Jakub Jelinek <jakub@redhat.com>
3993 PR tree-optimization/84111
3994 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3995 inner loops added during recursion, as they don't have up-to-date
3998 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4001 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4002 (can_inline_edge_by_limits_p): ... here.
4003 (can_early_inline_edge_p, check_callers,
4004 update_caller_keys, update_callee_keys, recursive_inlining,
4005 add_new_edges_to_heap, speculation_useful_p,
4006 inline_small_functions,
4007 inline_small_functions, flatten_function,
4008 inline_to_all_callers_1): Update.
4010 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4012 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4015 2018-01-30 Richard Biener <rguenther@suse.de>
4017 PR tree-optimization/83008
4018 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4019 invariant and constant vector uses in stmts when they need
4022 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4025 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4026 * configure: Regenerate.
4028 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4030 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4032 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4033 Use gen_rtx_REG rather than gen_lowpart.
4035 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4037 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4038 rather than 0 when creating partial subregs.
4040 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4042 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4045 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4048 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4049 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4050 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4051 flags. This restores the settings used before the 2017-07-24.
4052 Turning off pre increment/decrement/modify allows IVOPTS to
4053 optimize DF/SF loops where the index is an int.
4055 2018-01-29 Richard Biener <rguenther@suse.de>
4056 Kelvin Nilsen <kelvin@gcc.gnu.org>
4059 * tree-vect-stmts.c (vectorizable_call): Don't call
4060 targetm.vectorize_builtin_md_vectorized_function if callee is
4063 2018-01-22 Carl Love <cel@us.ibm.com>
4065 * doc/extend.tex: Fix typo in second arg in
4066 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4068 2018-01-29 Richard Biener <rguenther@suse.de>
4070 PR tree-optimization/84086
4071 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4072 (flush_ssaname_freelist): When SSA names were released reset
4073 the SCEV hash table.
4075 2018-01-29 Richard Biener <rguenther@suse.de>
4077 PR tree-optimization/84057
4078 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4079 removed paths when removing edges.
4081 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4083 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4084 -mfunction-return=@var{choice}.
4086 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4089 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4091 (layout::print_source_line): Likewise.
4092 (test_get_line_width_without_trailing_whitespace): Add test cases.
4094 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4097 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4100 2018-01-26 Jim Wilson <jimw@sifive.com>
4102 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4104 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4107 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4109 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4110 and CMP + SUB-immediate -> SUBS.
4112 2018-01-26 Martin Sebor <msebor@redhat.com>
4114 PR tree-optimization/83896
4115 * tree-ssa-strlen.c (get_string_len): Rename...
4116 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4117 Avoid assuming length is constant.
4118 (handle_char_store): Use HOST_WIDE_INT for string length.
4120 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4123 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4124 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4126 2018-01-26 Richard Biener <rguenther@suse.de>
4128 PR rtl-optimization/84003
4129 * dse.c (record_store): Only record redundant stores when
4130 the earlier store aliases at least all accesses the later one does.
4132 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4134 PR rtl-optimization/83985
4135 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4136 REG_CFA_RESTORE insns.
4137 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4138 REG_CFA_RESTORE insns here.
4141 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4142 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4144 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4146 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4147 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4148 (arc_init): Likewise.
4149 (arc_override_options): Likewise.
4150 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4152 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4154 * config/arc/arc.h (TARGET_DBNZ): Define.
4155 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4156 properly set the tune attribute.
4157 (dbnz): Use TARGET_DBNZ guard.
4158 * config/arc/arc.opt (mtune): Add core3 option.
4160 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4162 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4163 recognize new pic like addresses.
4164 (arc_delegitimize_address): Clean up.
4166 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4168 * config/arc/arc-arches.def: Option mrf16 valid for all
4170 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4171 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4172 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4173 * config/arc/arc-tables.opt: Regenerate.
4174 * config/arc/arc.c (arc_conditional_register_usage): Handle
4175 reduced register file case.
4176 (arc_file_start): Set must have build attributes.
4177 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4179 * config/arc/arc.opt (mrf16): Add new option.
4180 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4181 * config/arc/genmultilib.awk: Handle new mrf16 option.
4182 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4183 * config/arc/t-multilib: Regenerate.
4184 * doc/invoke.texi (ARC Options): Document mrf16 option.
4186 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4188 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4189 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4190 (arc_attribute_table): Add 'secure_call' attribute.
4191 (arc_print_operand): Print secure call operand.
4192 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4194 (arc_is_secure_call_p): New function. * config/arc/arc.md
4195 (call_i): Add support for sjli instruction.
4196 (call_value_i): Likewise.
4197 * config/arc/constraints.md (Csc): New constraint.
4199 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4200 John Eric Martin <John.Martin@emmicro-us.com>
4202 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4203 * config/arc/arc.c (_arc_jli_section): New struct.
4204 (arc_jli_section): New type.
4205 (rc_jli_sections): New static variable.
4206 (arc_handle_jli_attribute): New function.
4207 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4208 (arc_file_end): New function.
4209 (TARGET_ASM_FILE_END): Define.
4210 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4211 (arc_add_jli_section): New function.
4212 (jli_call_scan): Likewise.
4213 (arc_reorg): Call jli_call_scan.
4214 (arc_output_addsi): Remove 'S' from printing asm operand.
4215 (arc_is_jli_call_p): New function.
4216 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4218 (movhi_insn): Likewise.
4219 (movsi_insn): Likewise.
4220 (movsi_set_cc_insn): Likewise.
4221 (loadqi_update): Likewise.
4222 (load_zeroextendqisi_update): Likewise.
4223 (load_signextendqisi_update): Likewise.
4224 (loadhi_update): Likewise.
4225 (load_zeroextendhisi_update): Likewise.
4226 (load_signextendhisi_update): Likewise.
4227 (loadsi_update): Likewise.
4228 (loadsf_update): Likewise.
4229 (movsicc_insn): Likewise.
4230 (bset_insn): Likewise.
4231 (bxor_insn): Likewise.
4232 (bclr_insn): Likewise.
4233 (bmsk_insn): Likewise.
4234 (bicsi3_insn): Likewise.
4235 (cmpsi_cc_c_insn): Likewise.
4236 (movsi_ne): Likewise.
4237 (movsi_cond_exec): Likewise.
4238 (clrsbsi2): Likewise.
4249 (call_i): Remove 'S' asm letter, add jli instruction.
4250 (call_value_i): Likewise.
4251 * config/arc/arc.op (mjli-always): New option.
4252 * config/arc/constraints.md (Cji): New constraint.
4253 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4255 (subsf3_fpx): Likewise.
4256 (mulsf3_fpx): Likewise.
4257 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4259 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4261 * doc/invoke.texi (ARC): Document mjli-always option.
4263 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4265 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4266 avoid addition with 0 and use incw and decw where possible.
4268 2018-01-26 Richard Biener <rguenther@suse.de>
4270 PR tree-optimization/81082
4271 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4272 association if it requires casting to unsigned.
4273 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4274 from fold_plusminus_mult_expr to catch important cases late when
4275 range info is available.
4277 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4279 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4280 * configure.ac (hidden_linkonce): New test.
4281 * configure: Regenerate.
4282 * config.in: Regenerate.
4284 2018-01-26 Julia Koval <julia.koval@intel.com>
4286 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4287 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4288 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4289 _mm_mask_bitshuffle_epi64_mask): Fix type.
4290 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4291 USI_FTYPE_V4DI_V4DI_USI): Remove.
4292 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4293 __builtin_ia32_vpshufbitqmb256_mask,
4294 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4295 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4296 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4298 2018-01-26 Alan Modra <amodra@gmail.com>
4301 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4302 UNSPEC_VBPERMQ. Sort other unspecs.
4304 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4306 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4308 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4311 * predict.c (drop_profile): Do not push/pop cfun; update also
4313 (handle_missing_profiles): Fix logic looking for zero profiles.
4315 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4318 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4319 on functions with #pragma omp declare simd or functions with simd
4321 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4322 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4323 Remove trailing \n from warning_at calls.
4325 2018-01-25 Tom de Vries <tom@codesourcery.com>
4328 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4329 for neutered workers.
4331 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4334 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4335 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4337 2018-01-24 Jeff Law <law@redhat.com>
4340 * i386.c (get_probe_interval): Move to earlier point.
4341 (ix86_compute_frame_layout): If -fstack-clash-protection and
4342 the frame is larger than the probe interval, then use pushes
4343 to save registers rather than reg->mem moves.
4344 (ix86_expand_prologue): Remove conditional for int_registers_saved
4347 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4350 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4351 min/max for never referenced object.
4353 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4356 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4358 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4359 attributes from DECL_ATTRIBUTES (decl) without affecting
4360 DECL_ATTRIBUTES (current_function_decl).
4361 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4362 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4364 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4366 PR tree-optimization/83979
4367 * fold-const.c (fold_comparison): Use constant_boolean_node
4368 instead of boolean_{true,false}_node.
4370 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4372 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4375 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4377 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4378 Simplify the clause that sets the length attribute.
4379 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4380 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4381 clause that sets the length attribute.
4382 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4384 2018-01-24 Tom de Vries <tom@codesourcery.com>
4387 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4388 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4389 Add strict parameter.
4390 (prevent_branch_around_nothing): Insert dummy insn between branch to
4391 label and label with no ptx insn inbetween.
4392 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4394 2018-01-24 Tom de Vries <tom@codesourcery.com>
4397 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4398 for neutered threads in warp.
4399 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4401 2018-01-24 Richard Biener <rguenther@suse.de>
4403 PR tree-optimization/83176
4404 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4407 2018-01-24 Richard Biener <rguenther@suse.de>
4409 PR tree-optimization/82819
4410 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4411 code generating pluses that are no-ops in the target precision.
4413 2018-01-24 Richard Biener <rguenther@suse.de>
4416 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4418 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4420 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4421 to merge probabilities.
4422 * predict.c (probably_never_executed): Also mark as cold functions
4423 with global 0 profile and guessed local profile.
4424 * profile-count.c (profile_probability::combine_with_count): New
4426 * profile-count.h (profile_probability::operator*,
4427 profile_probability::operator*=, profile_probability::operator/,
4428 profile_probability::operator/=): Reduce precision to adjusted
4429 and set value to guessed on contradictory divisions.
4430 (profile_probability::combine_with_freq): Remove.
4431 (profile_probability::combine_wiht_count): Declare.
4432 (profile_count::force_nonzero):: Set to adjusted.
4433 (profile_count::probability_in):: Set quality to adjusted.
4434 * tree-ssa-tail-merge.c (replace_block_by): Use
4437 2018-01-23 Andrew Waterman <andrew@sifive.com>
4438 Jim Wilson <jimw@sifive.com>
4440 * config/riscv/riscv.c (riscv_stack_boundary): New.
4441 (riscv_option_override): Set riscv_stack_boundary. Handle
4442 riscv_preferred_stack_boundary_arg.
4443 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4444 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4445 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4446 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4447 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4448 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4450 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4453 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4454 of struct ix86_frame.
4455 (ix86_expand_epilogue): Likewise. Add a local variable for
4456 the reg_save_offset field in struct ix86_frame.
4458 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4460 PR tree-optimization/82604
4461 * tree-loop-distribution.c (enum partition_kind): New enum item
4462 PKIND_PARTIAL_MEMSET.
4463 (partition_builtin_p): Support above new enum item.
4464 (generate_code_for_partition): Ditto.
4465 (compute_access_range): Differentiate cases that equality can be
4466 proven at all loops, the innermost loops or no loops.
4467 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4468 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4469 (finalize_partitions, distribute_loop): Don't fuse partition of
4470 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4471 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4474 2018-01-23 Martin Liska <mliska@suse.cz>
4476 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4477 order to ignore the predictor.
4478 (PRED_POLYMORPHIC_CALL): Likewise.
4479 (PRED_RECURSIVE_CALL): Likewise.
4481 2018-01-23 Martin Liska <mliska@suse.cz>
4483 * tree-profile.c (tree_profiling): Print function header to
4484 aware reader which function we are working on.
4485 * value-prof.c (gimple_find_values_to_profile): Do not print
4486 not interesting value histograms.
4488 2018-01-23 Martin Liska <mliska@suse.cz>
4490 * profile-count.h (enum profile_quality): Add
4491 profile_uninitialized as the first value. Do not number values
4492 as they are zero based.
4493 (profile_count::verify): Update sanity check.
4494 (profile_probability::verify): Likewise.
4496 2018-01-23 Nathan Sidwell <nathan@acm.org>
4498 * doc/invoke.texi (ffor-scope): Deprecate.
4500 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4502 PR tree-optimization/83510
4503 * domwalk.c (set_all_edges_as_executable): New function.
4504 (dom_walker::dom_walker): Convert bool param
4505 "skip_unreachable_blocks" to enum reachability. Move setup of
4506 edge flags to set_all_edges_as_executable and only do it when
4507 reachability is REACHABLE_BLOCKS.
4508 * domwalk.h (enum dom_walker::reachability): New enum.
4509 (dom_walker::dom_walker): Convert bool param
4510 "skip_unreachable_blocks" to enum reachability.
4511 (set_all_edges_as_executable): New decl.
4512 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4513 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4515 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4516 but converting true to REACHABLE_BLOCKS.
4517 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4519 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4520 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4521 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4523 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4524 if check_all_array_refs will be called.
4526 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4528 * tree.c (selftest::test_location_wrappers): Add more test
4531 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4533 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4534 (selftest::test_bit_in_range): Likewise.
4536 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4539 * doc/sourcebuild.texi (vect_float): Say that the selector
4540 only describes the situation when -funsafe-math-optimizations is on.
4541 (vect_float_strict): Document.
4543 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4545 PR tree-optimization/83965
4546 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4547 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4548 instead of checking only for a reduction.
4549 (vect_recog_widen_sum_pattern): Likewise.
4551 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4553 * predict.c (probably_never_executed): Only use precise profile info.
4554 (compute_function_frequency): Skip after inlining hack since we now
4555 have quality checking.
4557 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4559 * profile-count.h (profile_probability::very_unlikely,
4560 profile_probability::unlikely, profile_probability::even): Set
4561 precision to guessed.
4563 2018-01-23 Richard Biener <rguenther@suse.de>
4565 PR tree-optimization/83963
4566 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4567 Properly terminate dominator walk when crossing the exit edge not
4568 when visiting its source block.
4570 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4573 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4574 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4576 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4578 PR tree-optimization/83957
4579 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4580 semicolon after for body surrounded by braces.
4582 PR tree-optimization/83081
4583 * profile-count.h (profile_probability::split): New method.
4584 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4585 Use profile_probability::split.
4586 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4587 when splitting a single conditional jump into 2.
4589 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4591 PR tree-optimization/69452
4592 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4595 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4597 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4598 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4599 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4601 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4603 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4604 * config/rl78/rl78.md (movdi): New define_expand.
4605 * config/rl78/rl78.c (rl78_split_movdi): New function.
4607 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4610 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4612 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4613 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4614 128-bit to produce an UNSPEC move to get the double word with the
4615 signbit and then a shift directly to do signbit.
4616 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4617 implementation with a new version that just does either a direct
4618 move or a regular move. Move memory interface to separate insns.
4619 Move insns so they are next to the expander.
4620 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4621 with signbit move. Split big and little endian case.
4622 (signbit<mode>2_dm_mem_le): Likewise.
4623 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4624 (signbit<mode>2_dm2): Likewise.
4626 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4628 * config/rl78/rl78.md (anddi3): New define_expand.
4630 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4632 * config/rl78/rl78.md (umindi3): New define_expand.
4634 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4636 * config/rl78/rl78.md (smindi3): New define_expand.
4638 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4640 * config/rl78/rl78.md (smaxdi3): New define_expand.
4642 2018-01-22 Carl Love <cel@us.ibm.com>
4644 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4645 LVX_V1TI): Add macro expansion.
4646 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4647 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4648 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4649 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4650 Change check to determine if the instruction is a byte reversing
4651 entry. Fix typo in comment.
4652 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4653 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4654 Add def_builtin calls for new builtins.
4655 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4656 Add define_insn expansion.
4658 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4660 * config/rl78/rl78.md (umaxdi3): New define_expand.
4662 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4664 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4665 for non-QImode registers.
4667 2018-01-22 Richard Biener <rguenther@suse.de>
4669 PR tree-optimization/83963
4670 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4671 including the loop exit block.
4672 (scop_detection::merge_sese): Likewise.
4673 (scop_detection::add_scop): Do it here instead.
4675 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4677 * doc/sourcebuild.texi (arm_softfloat): Document.
4679 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4682 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4683 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4684 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4686 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4687 David Edelsohn <dje.gcc@gmail.com>
4690 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4691 Change "crset eq" to "crset 2".
4692 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4693 (*call_indirect_aix<mode>_nospec): Likewise.
4694 (*call_value_indirect_aix<mode>_nospec): Likewise.
4695 (*call_indirect_elfv2<mode>_nospec): Likewise.
4696 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4697 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4698 change assembly output from . to $.
4699 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4700 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4701 (*tablejump<mode>_internal1_nospec): Likewise.
4703 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4706 * config/sh/sh_optimize_sett_clrt.cc:
4707 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4709 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4711 PR tree-optimization/83940
4712 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4713 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4714 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4715 use it to pass back the definition type.
4716 (vect_check_store_rhs): Likewise rhs_dt_out.
4717 (vect_build_gather_load_calls): Add a mask_dt argument and use
4718 it instead of a call to vect_is_simple_use.
4719 (vectorizable_store): Update calls to vect_check_load_store_mask
4720 and vect_check_store_rhs. Use the dt returned by the latter instead
4721 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4722 instead of calls to vect_is_simple_use. Pass the scalar rather
4723 than the vector operand to vect_is_simple_use when handling
4724 second and subsequent copies of an rhs value.
4725 (vectorizable_load): Update calls to vect_check_load_store_mask
4726 and vect_build_gather_load_calls. Use the cached mask_dt and
4727 gs_info.offset_dt instead of calls to vect_is_simple_use.
4729 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4732 * tree-emutls.c: Include gimplify.h.
4733 (lower_emutls_2): New function.
4734 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4735 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4736 it before further processing.
4739 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4740 UINTVAL (trueop1) instead of INTVAL (op1).
4742 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4746 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4747 INCOMING_FRAME_SP_OFFSET if not defined.
4748 (scan_trace): Add ENTRY argument. If true and
4749 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4750 emit a note to adjust the CFA offset.
4751 (create_cfi_notes): Adjust scan_trace callers.
4752 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4753 INCOMING_FRAME_SP_OFFSET in the CIE.
4754 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4755 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4757 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4758 * doc/tm.texi: Regenerated.
4760 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4762 PR rtl-optimization/83147
4763 * lra-constraints.c (remove_inheritance_pseudos): Use
4764 lra_substitute_pseudo_within_insn.
4766 2018-01-19 Tom de Vries <tom@codesourcery.com>
4767 Cesar Philippidis <cesar@codesourcery.com>
4770 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4772 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4775 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4776 spaces for function labels.
4778 2018-01-19 Martin Liska <mliska@suse.cz>
4780 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4781 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4782 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4783 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4784 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4785 (PRED_CONST_RETURN): Change from 69 to 65.
4786 (PRED_NULL_RETURN): Change from 91 to 71.
4787 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4788 (PRED_LOOP_GUARD): Change from 66 to 73.
4790 2018-01-19 Martin Liska <mliska@suse.cz>
4792 * predict.c (predict_insn_def): Add new assert.
4793 (struct branch_predictor): Change type to signed integer.
4794 (test_prediction_value_range): Amend test to cover
4796 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4797 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4798 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4799 (PRED_LOOP_IV_COMPARE): Likewise.
4800 * predict.h (PROB_UNINITIALIZED): Define new constant.
4802 2018-01-19 Martin Liska <mliska@suse.cz>
4804 * predict.c (dump_prediction): Add new format for
4805 analyze_brprob.py script which is enabled with -details
4807 * profile-count.h (precise_p): New function.
4809 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4811 PR tree-optimization/83922
4812 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4813 there are no statements that need masking.
4814 (vect_active_double_reduction_p): New function.
4815 (vect_analyze_loop_operations): Use it when handling phis that
4816 are not in the loop header.
4818 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4820 PR tree-optimization/83914
4821 * tree-vect-loop.c (vectorizable_induction): Don't convert
4822 init_expr or apply the peeling adjustment for inductions
4823 that are nested within the vectorized loop.
4825 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4827 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4830 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4834 * function.h (gimplify_parameters): Add gimple_seq * argument.
4835 * function.c: Include gimple.h and options.h.
4836 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4837 for the added local temporaries if needed.
4838 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4839 if there are any parameter cleanups, wrap whole body into a
4840 try/finally with the cleanups.
4842 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4845 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4846 Use GET_MODE_CLASS for scalar floating point.
4848 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4852 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4853 Fix call of call_cgraph_insertion_hooks.
4855 2018-01-18 Martin Sebor <msebor@redhat.com>
4857 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4859 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4862 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4865 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4868 * common.opt: (-ffile-prefix-map): New option.
4869 * opts.c (common_handle_option): Defer it.
4870 * opts-global.c (handle_common_deferred_options): Handle it.
4871 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4872 * file-prefix-map.h: New file.
4873 (remap_debug_filename, add_debug_prefix_map): ...here.
4874 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4875 * final.c (debug_prefix_map, add_debug_prefix_map
4876 remap_debug_filename): Move to...
4877 * file-prefix-map.c: New file.
4878 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4879 generalize, get rid of alloca(), use strrchr() instead of strchr().
4880 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4881 Implement in terms of add_prefix_map().
4882 (remap_macro_filename, remap_debug_filename): Implement in term of
4884 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4885 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4886 * dbxout.c: Include file-prefix-map.h.
4887 * varasm.c: Likewise.
4888 * vmsdbgout.c: Likewise.
4889 * xcoffout.c: Likewise.
4890 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4891 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4892 * doc/invoke.texi (-ffile-prefix-map): Document.
4893 (-fdebug-prefix-map): Update description.
4895 2018-01-18 Martin Liska <mliska@suse.cz>
4897 * config/i386/i386.c (indirect_thunk_name): Document that also
4899 (output_indirect_thunk): Document why both instructions
4900 (pause and lfence) are generated.
4902 2018-01-18 Richard Biener <rguenther@suse.de>
4904 PR tree-optimization/83887
4905 * graphite-scop-detection.c
4906 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4907 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4908 (scop_detection::merge_sese): Re-implement with a flood-fill
4909 algorithm that properly finds a SESE region if it exists.
4911 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4914 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4915 pointer_diff optimizations use view_convert instead of convert.
4917 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4919 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4920 Generate different code for -mno-speculate-indirect-jumps.
4921 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4922 (*call_indirect_aix<mode>): Disable for
4923 -mno-speculate-indirect-jumps.
4924 (*call_indirect_aix<mode>_nospec): New define_insn.
4925 (*call_value_indirect_aix<mode>): Disable for
4926 -mno-speculate-indirect-jumps.
4927 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4928 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4929 -mno-speculate-indirect-jumps.
4930 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4932 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4934 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4935 long double type, set the flags for noting the default long double
4936 type, even if we don't pass or return a long double type.
4938 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4941 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4944 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4946 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4947 support for merge[hl].
4948 (fold_mergehl_helper): New helper function.
4949 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4950 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4951 (altivec_vmrglw_direct): Add xxmrglw insn.
4953 2018-01-17 Andrew Waterman <andrew@sifive.com>
4955 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4956 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4958 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4961 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4962 call the lto_location_cache before reading the
4963 DECL_SOURCE_LOCATION of the types.
4965 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4966 Richard Sandiford <richard.sandiford@linaro.org>
4968 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4969 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4970 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4971 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4972 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4974 * config/aarch64/constraints.md (aarch64_movti_operand):
4976 * config/aarch64/predicates.md (Uti): Add new constraint.
4978 2018-01-17 Carl Love <cel@us.ibm.com>
4980 * config/rs6000/vsx.md (define_expand xl_len_r,
4981 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4982 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4984 (define_expand, define_insn): Move the shift left from the
4985 define_insn to the define_expand for lxvl and stxvl instructions.
4986 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4987 and XL_LEN_R definitions to PURE.
4989 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4991 * config/i386/i386.c (indirect_thunk_name): Declare regno
4992 as unsigned int. Compare regno with INVALID_REGNUM.
4993 (output_indirect_thunk): Ditto.
4994 (output_indirect_thunk_function): Ditto.
4995 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4996 in the call to output_indirect_thunk_function.
4998 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5001 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5002 rather than the size of inner_type to determine the stack slot size
5003 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5005 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5008 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5011 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5013 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5014 endian Linux systems to optionally enable multilibs for selecting
5015 the long double type if the user configured an explicit type.
5016 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5017 have no long double multilibs if not defined.
5018 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5019 warn if the user used -mabi={ieee,ibm}longdouble and we built
5020 multilibs for long double.
5021 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5022 appropriate multilib option.
5023 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5025 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5026 for building long double multilibs.
5027 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5029 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5031 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5034 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5036 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5039 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5042 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5045 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5047 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5048 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5050 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5052 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5053 different rtl trees depending on TARGET_64BIT.
5054 (rs6000_gen_lvx): Likewise.
5056 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5058 * config/visium/visium.md (nop): Tweak comment.
5059 (hazard_nop): Likewise.
5061 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5063 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5064 -mspeculate-indirect-jumps.
5065 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5066 for -mno-speculate-indirect-jumps.
5067 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5068 (*call_value_indirect_elfv2<mode>): Disable for
5069 -mno-speculate-indirect-jumps.
5070 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5071 (indirect_jump): Emit different RTL for
5072 -mno-speculate-indirect-jumps.
5073 (*indirect_jump<mode>): Disable for
5074 -mno-speculate-indirect-jumps.
5075 (*indirect_jump<mode>_nospec): New define_insn.
5076 (tablejump): Emit different RTL for
5077 -mno-speculate-indirect-jumps.
5078 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5079 (tablejumpsi_nospec): New define_expand.
5080 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5081 (tablejumpdi_nospec): New define_expand.
5082 (*tablejump<mode>_internal1): Disable for
5083 -mno-speculate-indirect-jumps.
5084 (*tablejump<mode>_internal1_nospec): New define_insn.
5085 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5088 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5090 * caller-save.c (insert_save): Drop unnecessary parameter. All
5093 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5094 Richard Biener <rguenth@suse.de>
5097 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5098 return early, inline manually is_gimple_sizepos. Make sure if we
5099 call gimplify_expr we don't end up with a gimple constant.
5100 * tree.c (variably_modified_type_p): Don't return true for
5101 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5102 * gimplify.h (is_gimple_sizepos): Remove.
5104 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5106 PR tree-optimization/83857
5107 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5108 vectorizable_live_operation for pure SLP statements.
5109 (vectorizable_live_operation): Handle PHIs.
5111 2018-01-16 Richard Biener <rguenther@suse.de>
5113 PR tree-optimization/83867
5114 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5115 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5117 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5120 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5121 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5122 If off is not INTEGER_CST, issue a may not be aligned warning
5123 rather than isn't aligned. Use isn%'t rather than isn't.
5124 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5126 <case MULT_EXPR>: Improve the case when bottom and one of the
5127 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5128 operand, in that case check if the other operand is multiple of
5129 bottom divided by the INTEGER_CST operand.
5131 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5134 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5135 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5136 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5137 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5138 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5139 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5140 (pa_function_arg_size): New function.
5142 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5144 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5145 in a separate statement.
5147 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5149 PR tree-optimization/83847
5150 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5151 group gathers and scatters.
5153 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5155 PR rtl-optimization/86620
5156 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5158 PR rtl-optimization/83213
5159 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5160 to last if both are JUMP_INSNs.
5162 PR tree-optimization/83843
5163 * gimple-ssa-store-merging.c
5164 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5165 store_immediate_info for bswap/nop orig_stores.
5167 2018-01-15 Andrew Waterman <andrew@sifive.com>
5169 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5171 <UDIV>: Increase cost if !TARGET_DIV.
5173 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5175 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5176 (define_attr "cr_logical_3op"): New.
5177 (cceq_ior_compare): Adjust.
5178 (cceq_ior_compare_complement): Adjust.
5179 (*cceq_rev_compare): Adjust.
5180 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5181 (is_cracked_insn): Adjust.
5182 (insn_must_be_first_in_group): Adjust.
5183 * config/rs6000/40x.md: Adjust.
5184 * config/rs6000/440.md: Adjust.
5185 * config/rs6000/476.md: Adjust.
5186 * config/rs6000/601.md: Adjust.
5187 * config/rs6000/603.md: Adjust.
5188 * config/rs6000/6xx.md: Adjust.
5189 * config/rs6000/7450.md: Adjust.
5190 * config/rs6000/7xx.md: Adjust.
5191 * config/rs6000/8540.md: Adjust.
5192 * config/rs6000/cell.md: Adjust.
5193 * config/rs6000/e300c2c3.md: Adjust.
5194 * config/rs6000/e500mc.md: Adjust.
5195 * config/rs6000/e500mc64.md: Adjust.
5196 * config/rs6000/e5500.md: Adjust.
5197 * config/rs6000/e6500.md: Adjust.
5198 * config/rs6000/mpc.md: Adjust.
5199 * config/rs6000/power4.md: Adjust.
5200 * config/rs6000/power5.md: Adjust.
5201 * config/rs6000/power6.md: Adjust.
5202 * config/rs6000/power7.md: Adjust.
5203 * config/rs6000/power8.md: Adjust.
5204 * config/rs6000/power9.md: Adjust.
5205 * config/rs6000/rs64.md: Adjust.
5206 * config/rs6000/titan.md: Adjust.
5208 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5210 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5211 ix86_indirect_branch_register logic.
5213 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5215 * config/i386/constraints.md (Bs): Update
5216 ix86_indirect_branch_register check. Don't check
5217 ix86_indirect_branch_register with GOT_memory_operand.
5219 * config/i386/predicates.md (GOT_memory_operand): Don't check
5220 ix86_indirect_branch_register here.
5221 (GOT32_symbol_operand): Likewise.
5223 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5225 * config/i386/predicates.md (constant_call_address_operand):
5226 Rewrite ix86_indirect_branch_register logic.
5227 (sibcall_insn_operand): Likewise.
5229 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5231 * config/i386/constraints.md (Bs): Replace
5232 ix86_indirect_branch_thunk_register with
5233 ix86_indirect_branch_register.
5235 * config/i386/i386.md (indirect_jump): Likewise.
5236 (tablejump): Likewise.
5237 (*sibcall_memory): Likewise.
5238 (*sibcall_value_memory): Likewise.
5239 Peepholes of indirect call and jump via memory: Likewise.
5240 * config/i386/i386.opt: Likewise.
5241 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5242 (GOT_memory_operand): Likewise.
5243 (call_insn_operand): Likewise.
5244 (sibcall_insn_operand): Likewise.
5245 (GOT32_symbol_operand): Likewise.
5247 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5250 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5251 type rather than type addr's type points to.
5252 (expand_omp_atomic_mutex): Likewise.
5253 (expand_omp_atomic): Likewise.
5255 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5258 * config/i386/i386.c (output_indirect_thunk_function): Use
5259 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5260 for __x86_return_thunk.
5262 2018-01-15 Richard Biener <rguenther@suse.de>
5265 * expmed.c (extract_bit_field_1): Fix typo.
5267 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5270 * config/arm/iterators.md (VF): New mode iterator.
5271 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5272 Remove integer-related logic from pattern.
5273 (neon_vabd<mode>_3): Likewise.
5275 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5278 * common.opt (fstrict-overflow): No longer an alias.
5279 (fwrapv-pointer): New option.
5280 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5281 also for pointer types based on flag_wrapv_pointer.
5282 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5283 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5284 opts->x_flag_wrapv got set.
5285 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5286 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5287 POINTER_TYPE_OVERFLOW_UNDEFINED.
5288 * match.pd: Likewise in address comparison pattern.
5289 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5291 2018-01-15 Richard Biener <rguenther@suse.de>
5294 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5295 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5296 Reset type names to their identifier if their TYPE_DECL doesn't
5297 have linkage (and thus is used for ODR and devirt).
5298 (save_debug_info_for_decl): Remove.
5299 (save_debug_info_for_type): Likewise.
5300 (add_tree_to_fld_list): Adjust.
5301 * tree-pretty-print.c (dump_generic_node): Make dumping of
5302 type names more robust.
5304 2018-01-15 Richard Biener <rguenther@suse.de>
5306 * BASE-VER: Bump to 8.0.1.
5308 2018-01-14 Martin Sebor <msebor@redhat.com>
5311 * builtins.c (check_access): Avoid warning when the no-warning bit
5314 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5316 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5317 * ira-color (allocno_hard_regs_compare): Likewise.
5319 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5322 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5323 Use .pushsection/.popsection.
5325 2018-01-14 Martin Sebor <msebor@redhat.com>
5328 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5330 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5332 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5333 entry from extra_headers.
5334 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5335 extra_headers, make the list bitwise identical to the i?86-*-* one.
5337 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5339 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5340 -mcmodel=large with -mindirect-branch=thunk,
5341 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5342 -mfunction-return=thunk-extern.
5343 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5344 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5345 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5347 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5349 * config/i386/i386.c (print_reg): Print the name of the full
5350 integer register without '%'.
5351 (ix86_print_operand): Handle 'V'.
5352 * doc/extend.texi: Document 'V' modifier.
5354 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5356 * config/i386/constraints.md (Bs): Disallow memory operand for
5357 -mindirect-branch-register.
5359 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5360 (GOT_memory_operand): Likewise.
5361 (call_insn_operand): Likewise.
5362 (sibcall_insn_operand): Likewise.
5363 (GOT32_symbol_operand): Likewise.
5364 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5365 for -mindirect-branch-register.
5366 (tablejump): Likewise.
5367 (*sibcall_memory): Likewise.
5368 (*sibcall_value_memory): Likewise.
5369 Disallow peepholes of indirect call and jump via memory for
5370 -mindirect-branch-register.
5371 (*call_pop): Replace m with Bw.
5372 (*call_value_pop): Likewise.
5373 (*sibcall_pop_memory): Replace m with Bs.
5374 * config/i386/i386.opt (mindirect-branch-register): New option.
5375 * doc/invoke.texi: Document -mindirect-branch-register option.
5377 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5379 * config/i386/i386-protos.h (ix86_output_function_return): New.
5380 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5381 set function_return_type.
5382 (indirect_thunk_name): Add ret_p to indicate thunk for function
5384 (output_indirect_thunk_function): Pass false to
5385 indirect_thunk_name.
5386 (ix86_output_indirect_branch_via_reg): Likewise.
5387 (ix86_output_indirect_branch_via_push): Likewise.
5388 (output_indirect_thunk_function): Create alias for function
5389 return thunk if regno < 0.
5390 (ix86_output_function_return): New function.
5391 (ix86_handle_fndecl_attribute): Handle function_return.
5392 (ix86_attribute_table): Add function_return.
5393 * config/i386/i386.h (machine_function): Add
5394 function_return_type.
5395 * config/i386/i386.md (simple_return_internal): Use
5396 ix86_output_function_return.
5397 (simple_return_internal_long): Likewise.
5398 * config/i386/i386.opt (mfunction-return=): New option.
5399 (indirect_branch): Mention -mfunction-return=.
5400 * doc/extend.texi: Document function_return function attribute.
5401 * doc/invoke.texi: Document -mfunction-return= option.
5403 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5405 * config/i386/i386-opts.h (indirect_branch): New.
5406 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5407 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5408 with local indirect jump when converting indirect call and jump.
5409 (ix86_set_indirect_branch_type): New.
5410 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5411 (indirectlabelno): New.
5412 (indirect_thunk_needed): Likewise.
5413 (indirect_thunk_bnd_needed): Likewise.
5414 (indirect_thunks_used): Likewise.
5415 (indirect_thunks_bnd_used): Likewise.
5416 (INDIRECT_LABEL): Likewise.
5417 (indirect_thunk_name): Likewise.
5418 (output_indirect_thunk): Likewise.
5419 (output_indirect_thunk_function): Likewise.
5420 (ix86_output_indirect_branch_via_reg): Likewise.
5421 (ix86_output_indirect_branch_via_push): Likewise.
5422 (ix86_output_indirect_branch): Likewise.
5423 (ix86_output_indirect_jmp): Likewise.
5424 (ix86_code_end): Call output_indirect_thunk_function if needed.
5425 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5427 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5428 (ix86_attribute_table): Add indirect_branch.
5429 * config/i386/i386.h (machine_function): Add indirect_branch_type
5430 and has_local_indirect_jump.
5431 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5433 (tablejump): Likewise.
5434 (*indirect_jump): Use ix86_output_indirect_jmp.
5435 (*tablejump_1): Likewise.
5436 (simple_return_indirect_internal): Likewise.
5437 * config/i386/i386.opt (mindirect-branch=): New option.
5438 (indirect_branch): New.
5441 (thunk-inline): Likewise.
5442 (thunk-extern): Likewise.
5443 * doc/extend.texi: Document indirect_branch function attribute.
5444 * doc/invoke.texi: Document -mindirect-branch= option.
5446 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5449 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5451 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5453 * ipa-inline.c (want_inline_small_function_p): Return false if
5454 inlining has already failed with CIF_FINAL_ERROR.
5455 (update_caller_keys): Call want_inline_small_function_p before
5457 (update_callee_keys): Likewise.
5459 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5461 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5463 (rs6000_quadword_masked_address_p): Likewise.
5464 (quad_aligned_load_p): Likewise.
5465 (quad_aligned_store_p): Likewise.
5466 (const_load_sequence_p): Add comment to describe the outer-most loop.
5467 (mimic_memory_attributes_and_flags): New function.
5468 (rs6000_gen_stvx): Likewise.
5469 (replace_swapped_aligned_store): Likewise.
5470 (rs6000_gen_lvx): Likewise.
5471 (replace_swapped_aligned_load): Likewise.
5472 (replace_swapped_load_constant): Capitalize argument name in
5473 comment describing this function.
5474 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5475 and stores that access quad-word aligned addresses and replace
5476 with stvx or lvx instructions when appropriate.
5477 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5478 New function prototype.
5479 (rs6000_quadword_masked_address_p): Likewise.
5480 (rs6000_gen_lvx): Likewise.
5481 (rs6000_gen_stvx): Likewise.
5482 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5483 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5484 when memory address is aligned.
5485 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5486 this split to select lvx instruction when memory address is aligned.
5487 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5488 instruction when memory address is aligned.
5489 (*vsx_le_perm_load_v16qi): Likewise.
5490 (four unnamed splitters): Modify to select the stvx instruction
5491 when memory is aligned.
5493 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5495 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5496 which appears in the queue multiple times.
5498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5499 Alan Hayward <alan.hayward@arm.com>
5500 David Sherwood <david.sherwood@arm.com>
5502 * tree-vectorizer.h (vec_lower_bound): New structure.
5503 (_loop_vec_info): Add check_nonzero and lower_bounds.
5504 (LOOP_VINFO_CHECK_NONZERO): New macro.
5505 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5506 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5507 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5508 fields. Make seg_len the distance travelled, not including the
5510 (dr_direction_indicator): Declare.
5511 (dr_zero_step_indicator): Likewise.
5512 (dr_known_forward_stride_p): Likewise.
5513 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5515 (runtime_alias_check_p): Allow runtime alias checks with
5517 (operator ==): Compare access_size and align.
5518 (prune_runtime_alias_test_list): Rework for new distinction between
5519 the access_size and seg_len.
5520 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5522 (get_segment_min_max): New function.
5523 (create_intersect_range_checks): Use it.
5524 (dr_step_indicator): New function.
5525 (dr_direction_indicator): Likewise.
5526 (dr_zero_step_indicator): Likewise.
5527 (dr_known_forward_stride_p): Likewise.
5528 * tree-loop-distribution.c (data_ref_segment_size): Return
5529 DR_STEP * (niters - 1).
5530 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5532 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5533 (vect_preserves_scalar_order_p): New function, split out from...
5534 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5535 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5536 (vect_vfa_access_size): New function.
5537 (vect_vfa_align): Likewise.
5538 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5539 (dump_lower_bound): New function.
5540 (vect_check_lower_bound): Likewise.
5541 (vect_small_gap_p): Likewise.
5542 (vectorizable_with_step_bound_p): Likewise.
5543 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5544 depencies if the vectorization factor is 1. Convert the checks
5545 for nonzero steps into checks on the bounds of DR_STEP. Try using
5546 a bunds check for variable steps if the minimum required step is
5547 relatively small. Update calls to the dr_with_seg_len
5548 constructor and to vect_compile_time_alias.
5549 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5551 (vect_loop_versioning): Call it.
5552 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5554 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5557 Alan Hayward <alan.hayward@arm.com>
5558 David Sherwood <david.sherwood@arm.com>
5560 * doc/sourcebuild.texi (vect_scatter_store): Document.
5561 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5563 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5565 * genopinit.c (main): Add supports_vec_scatter_store and
5566 supports_vec_scatter_store_cached to target_optabs.
5567 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5568 IFN_MASK_SCATTER_STORE.
5569 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5571 * internal-fn.h (internal_store_fn_p): Declare.
5572 (internal_fn_stored_value_index): Likewise.
5573 * internal-fn.c (scatter_store_direct): New macro.
5574 (expand_scatter_store_optab_fn): New function.
5575 (direct_scatter_store_optab_supported_p): New macro.
5576 (internal_store_fn_p): New function.
5577 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5578 IFN_MASK_SCATTER_STORE.
5579 (internal_fn_mask_index): Likewise.
5580 (internal_fn_stored_value_index): New function.
5581 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5583 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5584 * optabs-query.c (supports_vec_scatter_store_p): New function.
5585 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5586 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5587 true for scatter stores.
5588 (vect_gather_scatter_fn_p): Handle scatter stores too.
5589 (vect_check_gather_scatter): Consider using scatter stores if
5590 supports_vec_scatter_store_p.
5591 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5593 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5594 internal_fn_stored_value_index.
5595 (check_load_store_masking): Handle scatter stores too.
5596 (vect_get_store_rhs): Make public.
5597 (vectorizable_call): Use internal_store_fn_p.
5598 (vectorizable_store): Handle scatter store internal functions.
5599 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5600 when deciding whether the end of the group has been reached.
5601 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5602 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5603 (mask_scatter_store<mode>): New insns.
5605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5606 Alan Hayward <alan.hayward@arm.com>
5607 David Sherwood <david.sherwood@arm.com>
5609 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5610 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5611 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5613 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5614 Use vect_truncate_gather_scatter_offset if we can't treat the
5615 operation as a normal gather load or scatter store.
5616 (get_group_load_store_type): Take the gather_scatter_info
5617 as argument. Try using a gather load or scatter store for
5618 single-element groups.
5619 (get_load_store_type): Update calls to get_group_load_store_type
5620 and vect_use_strided_gather_scatters_p.
5622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5623 Alan Hayward <alan.hayward@arm.com>
5624 David Sherwood <david.sherwood@arm.com>
5626 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5627 optional tree argument.
5628 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5630 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5631 but continue to use the current value as a fallback.
5632 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5633 to compare the updates.
5634 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5635 (get_load_store_type): Use it when handling a strided access.
5636 (vect_get_strided_load_store_ops): New function.
5637 (vect_get_data_ptr_increment): Likewise.
5638 (vectorizable_load): Handle strided gather loads. Always pass
5639 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5642 Alan Hayward <alan.hayward@arm.com>
5643 David Sherwood <david.sherwood@arm.com>
5645 * doc/md.texi (gather_load@var{m}): Document.
5646 (mask_gather_load@var{m}): Likewise.
5647 * genopinit.c (main): Add supports_vec_gather_load and
5648 supports_vec_gather_load_cached to target_optabs.
5649 * optabs-tree.c (init_tree_optimization_optabs): Use
5650 ggc_cleared_alloc to allocate target_optabs.
5651 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5652 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5654 * internal-fn.h (internal_load_fn_p): Declare.
5655 (internal_gather_scatter_fn_p): Likewise.
5656 (internal_fn_mask_index): Likewise.
5657 (internal_gather_scatter_fn_supported_p): Likewise.
5658 * internal-fn.c (gather_load_direct): New macro.
5659 (expand_gather_load_optab_fn): New function.
5660 (direct_gather_load_optab_supported_p): New macro.
5661 (direct_internal_fn_optab): New function.
5662 (internal_load_fn_p): Likewise.
5663 (internal_gather_scatter_fn_p): Likewise.
5664 (internal_fn_mask_index): Likewise.
5665 (internal_gather_scatter_fn_supported_p): Likewise.
5666 * optabs-query.c (supports_at_least_one_mode_p): New function.
5667 (supports_vec_gather_load_p): Likewise.
5668 * optabs-query.h (supports_vec_gather_load_p): Declare.
5669 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5670 and memory_type field.
5671 (NUM_PATTERNS): Bump to 15.
5672 * tree-vect-data-refs.c: Include internal-fn.h.
5673 (vect_gather_scatter_fn_p): New function.
5674 (vect_describe_gather_scatter_call): Likewise.
5675 (vect_check_gather_scatter): Try using internal functions for
5676 gather loads. Recognize existing calls to a gather load function.
5677 (vect_analyze_data_refs): Consider using gather loads if
5678 supports_vec_gather_load_p.
5679 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5680 (vect_get_gather_scatter_offset_type): Likewise.
5681 (vect_convert_mask_for_vectype): Likewise.
5682 (vect_add_conversion_to_patterm): Likewise.
5683 (vect_try_gather_scatter_pattern): Likewise.
5684 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5685 (vect_vect_recog_func_ptrs): Add it.
5686 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5687 internal_fn_mask_index and internal_gather_scatter_fn_p.
5688 (check_load_store_masking): Take the gather_scatter_info as an
5689 argument and handle gather loads.
5690 (vect_get_gather_scatter_ops): New function.
5691 (vectorizable_call): Check internal_load_fn_p.
5692 (vectorizable_load): Likewise. Handle gather load internal
5694 (vectorizable_store): Update call to check_load_store_masking.
5695 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5696 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5697 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5698 (aarch64_gather_scale_operand_d): New predicates.
5699 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5700 (mask_gather_load<mode>): New insns.
5702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5703 Alan Hayward <alan.hayward@arm.com>
5704 David Sherwood <david.sherwood@arm.com>
5706 * optabs.def (fold_left_plus_optab): New optab.
5707 * doc/md.texi (fold_left_plus_@var{m}): Document.
5708 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5709 * internal-fn.c (fold_left_direct): Define.
5710 (expand_fold_left_optab_fn): Likewise.
5711 (direct_fold_left_optab_supported_p): Likewise.
5712 * fold-const-call.c (fold_const_fold_left): New function.
5713 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5714 * tree-parloops.c (valid_reduction_p): New function.
5715 (gather_scalar_reductions): Use it.
5716 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5717 (vect_finish_replace_stmt): Declare.
5718 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5719 (needs_fold_left_reduction_p): New function, split out from...
5720 (vect_is_simple_reduction): ...here. Accept reductions that
5721 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5722 (vect_force_simple_reduction): Also store the reduction type in
5723 the assignment's STMT_VINFO_REDUC_TYPE.
5724 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5725 (merge_with_identity): New function.
5726 (vect_expand_fold_left): Likewise.
5727 (vectorize_fold_left_reduction): Likewise.
5728 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5729 scalar phi in place for it. Check for target support and reject
5730 cases that would reassociate the operation. Defer the transform
5731 phase to vectorize_fold_left_reduction.
5732 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5733 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5734 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5738 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5739 call to ifc_temp_var.
5741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5742 Alan Hayward <alan.hayward@arm.com>
5743 David Sherwood <david.sherwood@arm.com>
5745 * target.def (legitimize_address_displacement): Take the original
5746 offset as a poly_int.
5747 * targhooks.h (default_legitimize_address_displacement): Update
5749 * targhooks.c (default_legitimize_address_displacement): Likewise.
5750 * doc/tm.texi: Regenerate.
5751 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5752 as an argument, moving assert of ad->disp == ad->disp_term to...
5753 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5754 Try calling targetm.legitimize_address_displacement before expanding
5755 the address rather than afterwards, and adjust for the new interface.
5756 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5757 Match the new hook interface. Handle SVE addresses.
5758 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5763 * Makefile.in (OBJS): Add early-remat.o.
5764 * target.def (select_early_remat_modes): New hook.
5765 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5766 * doc/tm.texi: Regenerate.
5767 * targhooks.h (default_select_early_remat_modes): Declare.
5768 * targhooks.c (default_select_early_remat_modes): New function.
5769 * timevar.def (TV_EARLY_REMAT): New timevar.
5770 * passes.def (pass_early_remat): New pass.
5771 * tree-pass.h (make_pass_early_remat): Declare.
5772 * early-remat.c: New file.
5773 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5775 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5778 Alan Hayward <alan.hayward@arm.com>
5779 David Sherwood <david.sherwood@arm.com>
5781 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5782 vfm1 with a bound_epilog parameter.
5783 (vect_do_peeling): Update calls accordingly, and move the prologue
5784 call earlier in the function. Treat the base bound_epilog as 0 for
5785 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5786 this base when peeling for gaps.
5787 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5788 with fully-masked loops.
5789 (vect_estimate_min_profitable_iters): Handle the single peeled
5790 iteration in that case.
5792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5793 Alan Hayward <alan.hayward@arm.com>
5794 David Sherwood <david.sherwood@arm.com>
5796 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5797 single-element interleaving even if the size is not a power of 2.
5798 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5799 accesses for single-element interleaving if the group size is
5802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5803 Alan Hayward <alan.hayward@arm.com>
5804 David Sherwood <david.sherwood@arm.com>
5806 * doc/md.texi (fold_extract_last_@var{m}): Document.
5807 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5808 * optabs.def (fold_extract_last_optab): New optab.
5809 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5810 * internal-fn.c (fold_extract_direct): New macro.
5811 (expand_fold_extract_optab_fn): Likewise.
5812 (direct_fold_extract_optab_supported_p): Likewise.
5813 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5814 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5815 EXTRACT_LAST_REDUCTION.
5816 (get_initial_def_for_reduction): Do not create an initial vector
5817 for EXTRACT_LAST_REDUCTION reductions.
5818 (vectorizable_reduction): Leave the scalar phi in place for
5819 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5820 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5821 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5822 transform phase to vectorizable_condition.
5823 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5825 (vect_finish_stmt_generation): ...here.
5826 (vect_finish_replace_stmt): New function.
5827 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5828 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5830 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5833 Alan Hayward <alan.hayward@arm.com>
5834 David Sherwood <david.sherwood@arm.com>
5836 * doc/md.texi (extract_last_@var{m}): Document.
5837 * optabs.def (extract_last_optab): New optab.
5838 * internal-fn.def (EXTRACT_LAST): New internal function.
5839 * internal-fn.c (cond_unary_direct): New macro.
5840 (expand_cond_unary_optab_fn): Likewise.
5841 (direct_cond_unary_optab_supported_p): Likewise.
5842 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5843 loops using EXTRACT_LAST.
5844 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5845 (extract_last_<mode>): ...this optab.
5846 (vec_extract<mode><Vel>): Update accordingly.
5848 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5849 Alan Hayward <alan.hayward@arm.com>
5850 David Sherwood <david.sherwood@arm.com>
5852 * target.def (empty_mask_is_expensive): New hook.
5853 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5854 * doc/tm.texi: Regenerate.
5855 * targhooks.h (default_empty_mask_is_expensive): Declare.
5856 * targhooks.c (default_empty_mask_is_expensive): New function.
5857 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5858 if the target says that empty masks are expensive.
5859 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5861 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5864 Alan Hayward <alan.hayward@arm.com>
5865 David Sherwood <david.sherwood@arm.com>
5867 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5868 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5869 (vect_use_loop_mask_for_alignment_p): New function.
5870 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5871 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5872 niters_skip argument. Make sure that the first niters_skip elements
5873 of the first iteration are inactive.
5874 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5875 Update call to vect_set_loop_masks_directly.
5876 (get_misalign_in_elems): New function, split out from...
5877 (vect_gen_prolog_loop_niters): ...here.
5878 (vect_update_init_of_dr): Take a code argument that specifies whether
5879 the adjustment should be added or subtracted.
5880 (vect_update_init_of_drs): Likewise.
5881 (vect_prepare_for_masked_peels): New function.
5882 (vect_do_peeling): Skip prologue peeling if we're using a mask
5883 instead. Update call to vect_update_inits_of_drs.
5884 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5886 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5887 alignment. Do not include the number of peeled iterations in
5888 the minimum threshold in that case.
5889 (vectorizable_induction): Adjust the start value down by
5890 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5891 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5892 Take the number of skipped iterations into account when calculating
5894 * tree-vect-stmts.c (vect_gen_while_not): New function.
5896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5897 Alan Hayward <alan.hayward@arm.com>
5898 David Sherwood <david.sherwood@arm.com>
5900 * doc/sourcebuild.texi (vect_fully_masked): Document.
5901 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5903 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5905 (vect_analyze_loop_2): ...here. Don't check the vectorization
5906 factor against the number of loop iterations if the loop is
5909 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5910 Alan Hayward <alan.hayward@arm.com>
5911 David Sherwood <david.sherwood@arm.com>
5913 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5914 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5915 (dump_groups): Update accordingly.
5916 (iv_use::mem_type): New member variable.
5917 (address_p): New function.
5918 (record_use): Add a mem_type argument and initialize the new
5920 (record_group_use): Add a mem_type argument. Use address_p.
5921 Remove obsolete null checks of base_object. Update call to record_use.
5922 (find_interesting_uses_op): Update call to record_group_use.
5923 (find_interesting_uses_cond): Likewise.
5924 (find_interesting_uses_address): Likewise.
5925 (get_mem_type_for_internal_fn): New function.
5926 (find_address_like_use): Likewise.
5927 (find_interesting_uses_stmt): Try find_address_like_use before
5928 calling find_interesting_uses_op.
5929 (addr_offset_valid_p): Use the iv mem_type field as the type
5930 of the addressed memory.
5931 (add_autoinc_candidates): Likewise.
5932 (get_address_cost): Likewise.
5933 (split_small_address_groups_p): Use address_p.
5934 (split_address_groups): Likewise.
5935 (add_iv_candidate_for_use): Likewise.
5936 (autoinc_possible_for_pair): Likewise.
5937 (rewrite_groups): Likewise.
5938 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5939 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5940 (get_alias_ptr_type_for_ptr_address): New function.
5941 (rewrite_use_address): Rewrite address uses in calls that were
5942 identified by find_address_like_use.
5944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5945 Alan Hayward <alan.hayward@arm.com>
5946 David Sherwood <david.sherwood@arm.com>
5948 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5950 * gimple-expr.h (is_gimple_addressable: Likewise.
5951 * gimple-expr.c (is_gimple_address): Likewise.
5952 * internal-fn.c (expand_call_mem_ref): New function.
5953 (expand_mask_load_optab_fn): Use it.
5954 (expand_mask_store_optab_fn): Likewise.
5956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5957 Alan Hayward <alan.hayward@arm.com>
5958 David Sherwood <david.sherwood@arm.com>
5960 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5961 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5962 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5963 (cond_umax@var{mode}): Document.
5964 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5965 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5966 (cond_umin_optab, cond_umax_optab): New optabs.
5967 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5968 (COND_IOR, COND_XOR): New internal functions.
5969 * internal-fn.h (get_conditional_internal_fn): Declare.
5970 * internal-fn.c (cond_binary_direct): New macro.
5971 (expand_cond_binary_optab_fn): Likewise.
5972 (direct_cond_binary_optab_supported_p): Likewise.
5973 (get_conditional_internal_fn): New function.
5974 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5975 Cope with reduction statements that are vectorized as calls rather
5977 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5978 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5979 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5980 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5981 (UNSPEC_COND_EOR): New unspecs.
5982 (optab): Add mappings for them.
5983 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5984 (sve_int_op, sve_fp_op): New int attributes.
5986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5987 Alan Hayward <alan.hayward@arm.com>
5988 David Sherwood <david.sherwood@arm.com>
5990 * optabs.def (while_ult_optab): New optab.
5991 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5992 * internal-fn.def (WHILE_ULT): New internal function.
5993 * internal-fn.h (direct_internal_fn_supported_p): New override
5994 that takes two types as argument.
5995 * internal-fn.c (while_direct): New macro.
5996 (expand_while_optab_fn): New function.
5997 (convert_optab_supported_p): Likewise.
5998 (direct_while_optab_supported_p): New macro.
5999 * wide-int.h (wi::udiv_ceil): New function.
6000 * tree-vectorizer.h (rgroup_masks): New structure.
6001 (vec_loop_masks): New typedef.
6002 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6004 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6005 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6006 (vect_max_vf): New function.
6007 (slpeel_make_loop_iterate_ntimes): Delete.
6008 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6009 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6010 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6011 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6012 internal-fn.h, stor-layout.h and optabs-query.h.
6013 (vect_set_loop_mask): New function.
6014 (add_preheader_seq): Likewise.
6015 (add_header_seq): Likewise.
6016 (interleave_supported_p): Likewise.
6017 (vect_maybe_permute_loop_masks): Likewise.
6018 (vect_set_loop_masks_directly): Likewise.
6019 (vect_set_loop_condition_masked): Likewise.
6020 (vect_set_loop_condition_unmasked): New function, split out from
6021 slpeel_make_loop_iterate_ntimes.
6022 (slpeel_make_loop_iterate_ntimes): Rename to..
6023 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6024 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6025 (vect_do_peeling): Update call accordingly.
6026 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6028 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6029 mask_compare_type, can_fully_mask_p and fully_masked_p.
6030 (release_vec_loop_masks): New function.
6031 (_loop_vec_info): Use it to free the loop masks.
6032 (can_produce_all_loop_masks_p): New function.
6033 (vect_get_max_nscalars_per_iter): Likewise.
6034 (vect_verify_full_masking): Likewise.
6035 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6036 retries, and free the mask rgroups before retrying. Check loop-wide
6037 reasons for disallowing fully-masked loops. Make the final decision
6038 about whether use a fully-masked loop or not.
6039 (vect_estimate_min_profitable_iters): Do not assume that peeling
6040 for the number of iterations will be needed for fully-masked loops.
6041 (vectorizable_reduction): Disable fully-masked loops.
6042 (vectorizable_live_operation): Likewise.
6043 (vect_halve_mask_nunits): New function.
6044 (vect_double_mask_nunits): Likewise.
6045 (vect_record_loop_mask): Likewise.
6046 (vect_get_loop_mask): Likewise.
6047 (vect_transform_loop): Handle the case in which the final loop
6048 iteration might handle a partial vector. Call vect_set_loop_condition
6049 instead of slpeel_make_loop_iterate_ntimes.
6050 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6051 (check_load_store_masking): New function.
6052 (prepare_load_store_mask): Likewise.
6053 (vectorizable_store): Handle fully-masked loops.
6054 (vectorizable_load): Likewise.
6055 (supportable_widening_operation): Use vect_halve_mask_nunits for
6057 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6058 (vect_gen_while): New function.
6059 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6060 (aarch64_uqdec<mode>): New insn.
6062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6063 Alan Hayward <alan.hayward@arm.com>
6064 David Sherwood <david.sherwood@arm.com>
6066 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6067 (reduc_xor_scal_optab): New optabs.
6068 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6069 (reduc_xor_scal_@var{m}): Document.
6070 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6071 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6073 * fold-const-call.c (fold_const_call): Handle them.
6074 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6075 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6076 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6077 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6078 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6079 (UNSPEC_XORV): New unspecs.
6080 (optab): Add entries for them.
6081 (BITWISEV): New int iterator.
6082 (bit_reduc_op): New int attributes.
6084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6085 Alan Hayward <alan.hayward@arm.com>
6086 David Sherwood <david.sherwood@arm.com>
6088 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6089 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6090 * optabs.def (vec_shl_insert_optab): New optab.
6091 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6092 (duplicate_and_interleave): Likewise.
6093 * tree-vect-loop.c: Include internal-fn.h.
6094 (neutral_op_for_slp_reduction): New function, split out from
6095 get_initial_defs_for_reduction.
6096 (get_initial_def_for_reduction): Handle option 2 for variable-length
6097 vectors by loading the neutral value into a vector and then shifting
6098 the initial value into element 0.
6099 (get_initial_defs_for_reduction): Replace the code argument with
6100 the neutral value calculated by neutral_op_for_slp_reduction.
6101 Use gimple_build_vector for constant-length vectors.
6102 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6103 but the first group_size elements have a neutral value.
6104 Use duplicate_and_interleave otherwise.
6105 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6106 Update call to get_initial_defs_for_reduction. Handle SLP
6107 reductions for variable-length vectors by creating one vector
6108 result for each scalar result, with the elements associated
6109 with other scalar results stubbed out with the neutral value.
6110 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6111 Require IFN_VEC_SHL_INSERT for double reductions on
6112 variable-length vectors, or SLP reductions that have
6113 a neutral value. Require can_duplicate_and_interleave_p
6114 support for variable-length unchained SLP reductions if there
6115 is no neutral value, such as for MIN/MAX reductions. Also require
6116 the number of vector elements to be a multiple of the number of
6117 SLP statements when doing variable-length unchained SLP reductions.
6118 Update call to vect_create_epilog_for_reduction.
6119 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6120 and remove initial values.
6121 (duplicate_and_interleave): Make public.
6122 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6123 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6126 Alan Hayward <alan.hayward@arm.com>
6127 David Sherwood <david.sherwood@arm.com>
6129 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6130 (can_duplicate_and_interleave_p): New function.
6131 (vect_get_and_check_slp_defs): Take the vector of statements
6132 rather than just the current one. Remove excess parentheses.
6133 Restriction rejectinon of vect_constant_def and vect_external_def
6134 for variable-length vectors to boolean types, or types for which
6135 can_duplicate_and_interleave_p is false.
6136 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6137 (duplicate_and_interleave): New function.
6138 (vect_get_constant_vectors): Use gimple_build_vector for
6139 constant-length vectors and suitable variable-length constant
6140 vectors. Use duplicate_and_interleave for other variable-length
6141 vectors. Don't defer the update when inserting new statements.
6143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6144 Alan Hayward <alan.hayward@arm.com>
6145 David Sherwood <david.sherwood@arm.com>
6147 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6148 min_profitable_iters doesn't go negative.
6150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6151 Alan Hayward <alan.hayward@arm.com>
6152 David Sherwood <david.sherwood@arm.com>
6154 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6155 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6156 * optabs.def (vec_mask_load_lanes_optab): New optab.
6157 (vec_mask_store_lanes_optab): Likewise.
6158 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6159 (MASK_STORE_LANES): Likewise.
6160 * internal-fn.c (mask_load_lanes_direct): New macro.
6161 (mask_store_lanes_direct): Likewise.
6162 (expand_mask_load_optab_fn): Handle masked operations.
6163 (expand_mask_load_lanes_optab_fn): New macro.
6164 (expand_mask_store_optab_fn): Handle masked operations.
6165 (expand_mask_store_lanes_optab_fn): New macro.
6166 (direct_mask_load_lanes_optab_supported_p): Likewise.
6167 (direct_mask_store_lanes_optab_supported_p): Likewise.
6168 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6170 (vect_load_lanes_supported): Likewise.
6171 * tree-vect-data-refs.c (strip_conversion): New function.
6172 (can_group_stmts_p): Likewise.
6173 (vect_analyze_data_ref_accesses): Use it instead of checking
6174 for a pair of assignments.
6175 (vect_store_lanes_supported): Take a masked_p parameter.
6176 (vect_load_lanes_supported): Likewise.
6177 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6178 vect_store_lanes_supported and vect_load_lanes_supported.
6179 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6180 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6181 parameter. Don't allow gaps for masked accesses.
6182 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6183 and vect_load_lanes_supported.
6184 (get_load_store_type): Take a masked_p parameter and update
6185 call to get_group_load_store_type.
6186 (vectorizable_store): Update call to get_load_store_type.
6187 Handle IFN_MASK_STORE_LANES.
6188 (vectorizable_load): Update call to get_load_store_type.
6189 Handle IFN_MASK_LOAD_LANES.
6191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6192 Alan Hayward <alan.hayward@arm.com>
6193 David Sherwood <david.sherwood@arm.com>
6195 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6197 * config/aarch64/aarch64-protos.h
6198 (aarch64_sve_struct_memory_operand_p): Declare.
6199 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6200 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6201 (VPRED, vpred): Handle SVE structure modes.
6202 * config/aarch64/constraints.md (Utx): New constraint.
6203 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6204 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6205 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6206 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6207 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6208 structure modes. Split into pieces after RA.
6209 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6210 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6212 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6213 SVE structure modes.
6214 (aarch64_classify_address): Likewise.
6215 (sizetochar): Move earlier in file.
6216 (aarch64_print_operand): Handle SVE register lists.
6217 (aarch64_array_mode): New function.
6218 (aarch64_sve_struct_memory_operand_p): Likewise.
6219 (TARGET_ARRAY_MODE): Redefine.
6221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6222 Alan Hayward <alan.hayward@arm.com>
6223 David Sherwood <david.sherwood@arm.com>
6225 * target.def (array_mode): New target hook.
6226 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6227 * doc/tm.texi: Regenerate.
6228 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6229 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6230 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6232 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6235 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6236 Alan Hayward <alan.hayward@arm.com>
6237 David Sherwood <david.sherwood@arm.com>
6239 * fold-const.c (fold_binary_loc): Check the argument types
6240 rather than the result type when testing for a vector operation.
6242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6244 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6245 * doc/tm.texi: Regenerate.
6247 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6248 Alan Hayward <alan.hayward@arm.com>
6249 David Sherwood <david.sherwood@arm.com>
6251 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6252 (sve): Document new AArch64 extension.
6253 * doc/md.texi (w): Extend the description of the AArch64
6254 constraint to include SVE vectors.
6255 (Upl, Upa): Document new AArch64 predicate constraints.
6256 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6258 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6259 (msve-vector-bits=): New option.
6260 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6261 SVE when these are disabled.
6262 (sve): New extension.
6263 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6264 modes. Adjust their number of units based on aarch64_sve_vg.
6265 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6266 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6267 aarch64_addr_query_type.
6268 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6269 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6270 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6271 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6272 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6273 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6274 (aarch64_simd_imm_zero_p): Delete.
6275 (aarch64_check_zero_based_sve_index_immediate): Declare.
6276 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6277 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6278 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6279 (aarch64_sve_float_mul_immediate_p): Likewise.
6280 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6282 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6283 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6284 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6285 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6286 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6287 (aarch64_regmode_natural_size): Likewise.
6288 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6289 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6291 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6292 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6293 for VG and the SVE predicate registers.
6294 (V_ALIASES): Add a "z"-prefixed alias.
6295 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6296 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6297 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6298 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6299 (REG_CLASS_NAMES): Add entries for them.
6300 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6301 and the predicate registers.
6302 (aarch64_sve_vg): Declare.
6303 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6304 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6305 (REGMODE_NATURAL_SIZE): Define.
6306 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6308 * config/aarch64/aarch64.c: Include cfgrtl.h.
6309 (simd_immediate_info): Add a constructor for series vectors,
6310 and an associated step field.
6311 (aarch64_sve_vg): New variable.
6312 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6313 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6314 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6315 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6316 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6317 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6318 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6319 (aarch64_get_mask_mode): New functions.
6320 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6321 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6322 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6323 predicate modes and predicate registers. Explicitly restrict
6324 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6325 to store a vector mode if it is recognized by
6326 aarch64_classify_vector_mode.
6327 (aarch64_regmode_natural_size): New function.
6328 (aarch64_hard_regno_caller_save_mode): Return the original mode
6330 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6331 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6332 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6333 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6335 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6336 does not overlap dest if the function is frame-related. Handle
6338 (aarch64_split_add_offset): New function.
6339 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6340 them aarch64_add_offset.
6341 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6342 and update call to aarch64_sub_sp.
6343 (aarch64_add_cfa_expression): New function.
6344 (aarch64_expand_prologue): Pass extra temporary registers to the
6345 functions above. Handle the case in which we need to emit new
6346 DW_CFA_expressions for registers that were originally saved
6347 relative to the stack pointer, but now have to be expressed
6348 relative to the frame pointer.
6349 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6351 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6352 IP0 and IP1 values for SVE frames.
6353 (aarch64_expand_vec_series): New function.
6354 (aarch64_expand_sve_widened_duplicate): Likewise.
6355 (aarch64_expand_sve_const_vector): Likewise.
6356 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6357 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6358 into the register, rather than emitting a SET directly.
6359 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6360 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6361 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6362 (offset_9bit_signed_scaled_p): New functions.
6363 (aarch64_replicate_bitmask_imm): New function.
6364 (aarch64_bitmask_imm): Use it.
6365 (aarch64_cannot_force_const_mem): Reject expressions involving
6366 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6367 (aarch64_classify_index): Handle SVE indices, by requiring
6368 a plain register index with a scale that matches the element size.
6369 (aarch64_classify_address): Handle SVE addresses. Assert that
6370 the mode of the address is VOIDmode or an integer mode.
6371 Update call to aarch64_classify_symbol.
6372 (aarch64_classify_symbolic_expression): Update call to
6373 aarch64_classify_symbol.
6374 (aarch64_const_vec_all_in_range_p): New function.
6375 (aarch64_print_vector_float_operand): Likewise.
6376 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6377 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6378 and the FP immediates 1.0 and 0.5.
6379 (aarch64_print_address_internal): Handle SVE addresses.
6380 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6381 (aarch64_regno_regclass): Handle predicate registers.
6382 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6384 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6385 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6386 (aarch64_convert_sve_vector_bits): New function.
6387 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6388 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6390 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6391 Handle SVE vector and predicate modes. Accept VL-based constants
6392 that need only one temporary register, and VL offsets that require
6393 no temporary registers.
6394 (aarch64_conditional_register_usage): Mark the predicate registers
6395 as fixed if SVE isn't available.
6396 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6397 Return true for SVE vector and predicate modes.
6398 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6399 rather than an unsigned int. Handle SVE modes.
6400 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6402 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6404 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6405 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6406 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6407 (aarch64_sve_float_mul_immediate_p): New functions.
6408 (aarch64_sve_valid_immediate): New function.
6409 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6410 Explicitly reject structure modes. Check for INDEX constants.
6411 Handle PTRUE and PFALSE constants.
6412 (aarch64_check_zero_based_sve_index_immediate): New function.
6413 (aarch64_simd_imm_zero_p): Delete.
6414 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6415 vector modes. Accept constants in the range of CNT[BHWD].
6416 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6417 ask for an Advanced SIMD mode.
6418 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6419 (aarch64_simd_vector_alignment): Handle SVE predicates.
6420 (aarch64_vectorize_preferred_vector_alignment): New function.
6421 (aarch64_simd_vector_alignment_reachable): Use it instead of
6423 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6424 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6426 (MAX_VECT_LEN): Delete.
6427 (expand_vec_perm_d): Add a vec_flags field.
6428 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6429 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6430 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6432 (aarch64_evpc_rev): Rename to...
6433 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6434 (aarch64_evpc_rev_global): New function.
6435 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6436 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6438 (aarch64_evpc_sve_tbl): New function.
6439 (aarch64_expand_vec_perm_const_1): Update after rename of
6440 aarch64_evpc_rev. Handle SVE permutes too, trying
6441 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6442 than aarch64_evpc_tbl.
6443 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6444 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6445 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6446 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6447 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6448 (aarch64_expand_sve_vcond): New functions.
6449 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6450 of aarch64_vector_mode_p.
6451 (aarch64_dwarf_poly_indeterminate_value): New function.
6452 (aarch64_compute_pressure_classes): Likewise.
6453 (aarch64_can_change_mode_class): Likewise.
6454 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6455 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6456 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6457 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6458 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6459 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6460 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6461 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6463 (Dn, Dl, Dr): Accept const as well as const_vector.
6464 (Dz): Likewise. Compare against CONST0_RTX.
6465 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6466 of "vector" where appropriate.
6467 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6468 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6469 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6470 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6471 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6472 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6473 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6474 (v_int_equiv): Extend to SVE modes.
6475 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6477 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6478 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6479 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6480 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6481 (SVE_COND_FP_CMP): New int iterators.
6482 (perm_hilo): Handle the new unpack unspecs.
6483 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6485 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6486 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6487 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6488 (aarch64_equality_operator, aarch64_constant_vector_operand)
6489 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6490 (aarch64_sve_nonimmediate_operand): Likewise.
6491 (aarch64_sve_general_operand): Likewise.
6492 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6493 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6494 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6495 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6496 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6497 (aarch64_sve_float_arith_immediate): Likewise.
6498 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6499 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6500 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6501 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6502 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6503 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6504 (aarch64_sve_float_arith_operand): Likewise.
6505 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6506 (aarch64_sve_float_mul_operand): Likewise.
6507 (aarch64_sve_vec_perm_operand): Likewise.
6508 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6509 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6510 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6511 as well as const_vector.
6512 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6513 in file. Use CONST0_RTX and CONSTM1_RTX.
6514 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6515 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6516 Use aarch64_simd_imm_zero.
6517 * config/aarch64/aarch64-sve.md: New file.
6518 * config/aarch64/aarch64.md: Include it.
6519 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6520 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6521 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6522 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6523 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6524 (sve): New attribute.
6525 (enabled): Disable instructions with the sve attribute unless
6527 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6528 aarch64_expand_mov_immediate.
6529 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6530 CNT[BHSD] immediates.
6531 (movti): Split CONST_POLY_INT moves into two halves.
6532 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6533 Split additions that need a temporary here if the destination
6534 is the stack pointer.
6535 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6536 (*add<mode>3_poly_1): New instruction.
6537 (set_clobber_cc): New expander.
6539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6541 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6542 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6543 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6544 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6545 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6546 Change innermode from fixed_mode_size to machine_mode.
6547 (simplify_subreg): Update call accordingly. Handle a constant-sized
6548 subreg of a variable-length CONST_VECTOR.
6550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6551 Alan Hayward <alan.hayward@arm.com>
6552 David Sherwood <david.sherwood@arm.com>
6554 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6555 (add_offset_to_base): New function, split out from...
6556 (create_mem_ref): ...here. When handling a scale other than 1,
6557 check first whether the address is valid without the offset.
6558 Add it into the base if so, leaving the index and scale as-is.
6560 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6563 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6564 fold_for_warn before checking if arg2 is INTEGER_CST.
6566 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6568 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6569 (store_multiple_operation): Delete.
6570 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6571 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6572 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6573 guarded by TARGET_STRING.
6574 (rs6000_output_load_multiple): Delete.
6575 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6576 OPTION_MASK_STRING / TARGET_STRING handling.
6577 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6578 (const rs6000_opt_masks) <"string">: Change mask to 0.
6579 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6580 (MASK_STRING): Delete.
6581 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6583 (load_multiple): Delete.
6590 (store_multiple): Delete.
6597 (movmemsi_8reg): Delete.
6598 (corresponding unnamed define_insn): Delete.
6599 (movmemsi_6reg): Delete.
6600 (corresponding unnamed define_insn): Delete.
6601 (movmemsi_4reg): Delete.
6602 (corresponding unnamed define_insn): Delete.
6603 (movmemsi_2reg): Delete.
6604 (corresponding unnamed define_insn): Delete.
6605 (movmemsi_1reg): Delete.
6606 (corresponding unnamed define_insn): Delete.
6607 * config/rs6000/rs6000.opt (mno-string): New.
6608 (mstring): Replace by deprecation warning stub.
6609 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6611 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6613 * regrename.c (regrename_do_replace): If replacing the same
6614 reg multiple times, try to reuse last created gen_raw_REG.
6617 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6618 main to workaround a bug in GDB.
6620 2018-01-12 Tom de Vries <tom@codesourcery.com>
6623 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6625 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6627 PR rtl-optimization/80481
6628 * ira-color.c (get_cap_member): New function.
6629 (allocnos_conflict_by_live_ranges_p): Use it.
6630 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6631 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6633 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6636 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6637 (*saddl_se_1): Ditto.
6639 (*ssubl_se_1): Ditto.
6641 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6643 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6644 rather than wi::to_widest for DR_INITs.
6645 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6646 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6647 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6649 (vect_analyze_group_access_1): Note that here.
6651 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6653 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6654 polynomial type sizes.
6656 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6658 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6659 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6660 (gimple_add_tmp_var): Likewise.
6662 2018-01-12 Martin Liska <mliska@suse.cz>
6664 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6665 (gimple_alloc_sizes): Likewise.
6666 (dump_gimple_statistics): Use PRIu64 in printf format.
6667 * gimple.h: Change uint64_t to int.
6669 2018-01-12 Martin Liska <mliska@suse.cz>
6671 * tree-core.h: Use uint64_t instead of int.
6672 * tree.c (tree_node_counts): Likewise.
6673 (tree_node_sizes): Likewise.
6674 (dump_tree_statistics): Use PRIu64 in printf format.
6676 2018-01-12 Martin Liska <mliska@suse.cz>
6678 * Makefile.in: As qsort_chk is implemented in vec.c, add
6679 vec.o to linkage of gencfn-macros.
6680 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6681 passing the info to record_node_allocation_statistics.
6682 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6684 * ggc-common.c (struct ggc_usage): Add operator== and use
6685 it in operator< and compare function.
6686 * mem-stats.h (struct mem_usage): Likewise.
6687 * vec.c (struct vec_usage): Remove operator< and compare
6688 function. Can be simply inherited.
6690 2018-01-12 Martin Jambor <mjambor@suse.cz>
6693 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6694 * tree-ssa-math-opts.c: Include domwalk.h.
6695 (convert_mult_to_fma_1): New function.
6696 (fma_transformation_info): New type.
6697 (fma_deferring_state): Likewise.
6698 (cancel_fma_deferring): New function.
6699 (result_of_phi): Likewise.
6700 (last_fma_candidate_feeds_initial_phi): Likewise.
6701 (convert_mult_to_fma): Added deferring logic, split actual
6702 transformation to convert_mult_to_fma_1.
6703 (math_opts_dom_walker): New type.
6704 (math_opts_dom_walker::after_dom_children): New method, body moved
6705 here from pass_optimize_widening_mul::execute, added deferring logic
6707 (pass_optimize_widening_mul::execute): Moved most of code to
6708 math_opts_dom_walker::after_dom_children.
6709 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6710 * config/i386/i386.c (ix86_option_override_internal): Added
6711 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6713 2018-01-12 Richard Biener <rguenther@suse.de>
6716 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6717 inline instance vars.
6719 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6722 * config/rx/rx.c (rx_is_restricted_memory_address):
6725 2018-01-12 Richard Biener <rguenther@suse.de>
6727 PR tree-optimization/80846
6728 * target.def (split_reduction): New target hook.
6729 * targhooks.c (default_split_reduction): New function.
6730 * targhooks.h (default_split_reduction): Declare.
6731 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6732 target requests first reduce vectors by combining low and high
6734 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6735 (get_vectype_for_scalar_type_and_size): Export.
6736 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6737 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6738 * doc/tm.texi: Regenerate.
6739 * config/i386/i386.c (ix86_split_reduction): Implement
6740 TARGET_VECTORIZE_SPLIT_REDUCTION.
6742 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6745 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6746 in PIC mode except for TARGET_VXWORKS_RTP.
6747 * config/sparc/sparc.c: Include cfgrtl.h.
6748 (TARGET_INIT_PIC_REG): Define.
6749 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6750 (sparc_pic_register_p): New predicate.
6751 (sparc_legitimate_address_p): Use it.
6752 (sparc_legitimize_pic_address): Likewise.
6753 (sparc_delegitimize_address): Likewise.
6754 (sparc_mode_dependent_address_p): Likewise.
6755 (gen_load_pcrel_sym): Remove 4th parameter.
6756 (load_got_register): Adjust call to above. Remove obsolete stuff.
6757 (sparc_expand_prologue): Do not call load_got_register here.
6758 (sparc_flat_expand_prologue): Likewise.
6759 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6760 (sparc_use_pseudo_pic_reg): New function.
6761 (sparc_init_pic_reg): Likewise.
6762 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6763 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6765 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6767 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6768 Add item for branch_cost.
6770 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6772 PR rtl-optimization/83565
6773 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6774 not extend the result to a larger mode for rotate operations.
6775 (num_sign_bit_copies1): Likewise.
6777 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6780 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6782 Use values-Xc.o for -pedantic.
6783 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6785 2018-01-12 Martin Liska <mliska@suse.cz>
6788 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6790 (possible_polymorphic_call_targets): Use it.
6791 (ipa_devirt): Likewise.
6793 2018-01-12 Martin Liska <mliska@suse.cz>
6795 * profile-count.h (enum profile_quality): Use 0 as invalid
6796 enum value of profile_quality.
6798 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6800 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6801 -mext-string options.
6803 2018-01-12 Richard Biener <rguenther@suse.de>
6805 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6806 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6807 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6809 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6811 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6813 * configure.ac (--with-long-double-format): Add support for the
6814 configuration option to change the default long double format on
6816 * config.gcc (powerpc*-linux*-*): Likewise.
6817 * configure: Regenerate.
6818 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6819 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6820 used without modification.
6822 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6824 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6825 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6826 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6827 MISC_BUILTIN_SPEC_BARRIER.
6828 (rs6000_init_builtins): Likewise.
6829 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6831 (speculation_barrier): New define_insn.
6832 * doc/extend.texi: Document __builtin_speculation_barrier.
6834 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6837 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6838 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6839 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6841 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6842 integral modes instead of "ss" and "sd".
6843 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6844 vectors with 32-bit and 64-bit elements.
6845 (vecdupssescalarmodesuffix): New mode attribute.
6846 (vec_dup<mode>): Use it.
6848 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6851 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6852 frame if argument is passed on stack.
6854 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6857 * ree.c (combine_reaching_defs): Optimize also
6858 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6859 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6861 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6864 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6866 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6869 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6870 after they are computed.
6872 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6874 PR tree-optimization/83695
6875 * gimple-loop-linterchange.cc
6876 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6877 reset cached scev information after interchange.
6878 (pass_linterchange::execute): Remove call to scev_reset_htab.
6880 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6882 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6883 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6884 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6885 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6886 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6887 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6888 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6889 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6890 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6891 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6892 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6893 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6894 (V_lane_reg): Likewise.
6895 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6897 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6898 (vfmal_lane_low<mode>_intrinsic,
6899 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6900 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6901 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6902 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6903 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6904 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6906 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6908 * config/arm/arm-cpus.in (fp16fml): New feature.
6909 (ALL_SIMD): Add fp16fml.
6910 (armv8.2-a): Add fp16fml as an option.
6911 (armv8.3-a): Likewise.
6912 (armv8.4-a): Add fp16fml as part of fp16.
6913 * config/arm/arm.h (TARGET_FP16FML): Define.
6914 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6916 * config/arm/arm-modes.def (V2HF): Define.
6917 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6918 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6919 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6920 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6921 vfmsl_low, vfmsl_high): New set of builtins.
6922 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6923 (vfml_op): New code attribute.
6924 (VFMLHALVES): New int iterator.
6925 (VFML, VFMLSEL): New mode attributes.
6926 (V_reg): Define mapping for V2HF.
6927 (V_hi, V_lo): New mode attributes.
6928 (VF_constraint): Likewise.
6929 (vfml_half, vfml_half_selector): New int attributes.
6930 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6932 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6933 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6935 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6936 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6937 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6938 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6940 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6941 Document new effective target and option set.
6943 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6945 * config/arm/arm-cpus.in (armv8_4): New feature.
6946 (ARMv8_4a): New fgroup.
6947 (armv8.4-a): New arch.
6948 * config/arm/arm-tables.opt: Regenerate.
6949 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6950 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6951 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6952 Add matching rules for -march=armv8.4-a and extensions.
6953 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6955 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6958 * config/rx/rx.md (BW): New mode attribute.
6959 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6961 2018-01-11 Richard Biener <rguenther@suse.de>
6963 PR tree-optimization/83435
6964 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6965 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6966 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6968 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6969 Alan Hayward <alan.hayward@arm.com>
6970 David Sherwood <david.sherwood@arm.com>
6972 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6974 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6975 (aarch64_print_address_internal): Use it to check for a zero offset.
6977 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6978 Alan Hayward <alan.hayward@arm.com>
6979 David Sherwood <david.sherwood@arm.com>
6981 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6982 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6983 Return a poly_int64 rather than a HOST_WIDE_INT.
6984 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6985 rather than a HOST_WIDE_INT.
6986 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6987 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6988 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6989 final_offset from HOST_WIDE_INT to poly_int64.
6990 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6991 to_constant when getting the number of units in an Advanced SIMD
6993 (aarch64_builtin_vectorized_function): Check for a constant number
6995 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6997 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6998 attribute instead of GET_MODE_NUNITS.
6999 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7000 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7001 GET_MODE_SIZE for fixed-size registers.
7002 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7003 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7004 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7005 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7006 (aarch64_print_operand, aarch64_print_address_internal)
7007 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7008 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7009 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7010 Handle polynomial GET_MODE_SIZE.
7011 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7012 wider than SImode without modification.
7013 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7014 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7015 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7016 passing and returning SVE modes.
7017 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7018 rather than GEN_INT.
7019 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7020 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7021 (aarch64_allocate_and_probe_stack_space): Likewise.
7022 (aarch64_layout_frame): Cope with polynomial offsets.
7023 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7024 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7026 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7027 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7028 poly_int64 rather than a HOST_WIDE_INT.
7029 (aarch64_get_separate_components, aarch64_process_components)
7030 (aarch64_expand_prologue, aarch64_expand_epilogue)
7031 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7032 (aarch64_anchor_offset): New function, split out from...
7033 (aarch64_legitimize_address): ...here.
7034 (aarch64_builtin_vectorization_cost): Handle polynomial
7035 TYPE_VECTOR_SUBPARTS.
7036 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7038 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7039 number of elements from the PARALLEL rather than the mode.
7040 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7041 rather than GET_MODE_BITSIZE.
7042 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7043 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7044 (aarch64_expand_vec_perm_const_1): Handle polynomial
7045 d->perm.length () and d->perm elements.
7046 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7047 Apply to_constant to d->perm elements.
7048 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7049 polynomial CONST_VECTOR_NUNITS.
7050 (aarch64_move_pointer): Take amount as a poly_int64 rather
7052 (aarch64_progress_pointer): Avoid temporary variable.
7053 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7054 the mode attribute instead of GET_MODE.
7056 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7057 Alan Hayward <alan.hayward@arm.com>
7058 David Sherwood <david.sherwood@arm.com>
7060 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7061 x exists before using it.
7062 (aarch64_add_constant_internal): Rename to...
7063 (aarch64_add_offset_1): ...this. Replace regnum with separate
7064 src and dest rtxes. Handle the case in which they're different,
7065 including when the offset is zero. Replace scratchreg with an rtx.
7066 Use 2 additions if there is no spare register into which we can
7067 move a 16-bit constant.
7068 (aarch64_add_constant): Delete.
7069 (aarch64_add_offset): Replace reg with separate src and dest
7070 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7071 Use aarch64_add_offset_1.
7072 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7073 an rtx rather than an int. Take the delta as a poly_int64
7074 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7075 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7076 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7077 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7078 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7080 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7081 aarch64_add_constant.
7083 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7085 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7086 Use scalar_float_mode.
7088 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7090 * config/aarch64/aarch64-simd.md
7091 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7092 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7093 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7094 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7095 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7096 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7097 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7098 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7099 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7100 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7102 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7105 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7106 targ_options->x_arm_arch_string is non NULL.
7108 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7110 * config/aarch64/aarch64.h
7111 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7113 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7116 * expmed.c (emit_store_flag_force): Swap if const op0
7117 and change VOIDmode to mode of op0.
7119 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7121 PR rtl-optimization/83761
7122 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7123 than bytes to mode_for_size.
7125 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7128 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7129 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7132 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7135 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7136 when in layout mode.
7137 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7138 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7141 2018-01-10 Michael Collison <michael.collison@arm.com>
7143 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7144 * config/aarch64/aarch64-option-extension.def: Add
7145 AARCH64_OPT_EXTENSION of 'fp16fml'.
7146 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7147 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7148 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7149 * config/aarch64/constraints.md (Ui7): New constraint.
7150 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7151 (VFMLA_SEL_W): Ditto.
7154 (VFMLA16_LOW): New int iterator.
7155 (VFMLA16_HIGH): Ditto.
7156 (UNSPEC_FMLAL): New unspec.
7157 (UNSPEC_FMLSL): Ditto.
7158 (UNSPEC_FMLAL2): Ditto.
7159 (UNSPEC_FMLSL2): Ditto.
7160 (f16mac): New code attribute.
7161 * config/aarch64/aarch64-simd-builtins.def
7162 (aarch64_fmlal_lowv2sf): Ditto.
7163 (aarch64_fmlsl_lowv2sf): Ditto.
7164 (aarch64_fmlalq_lowv4sf): Ditto.
7165 (aarch64_fmlslq_lowv4sf): Ditto.
7166 (aarch64_fmlal_highv2sf): Ditto.
7167 (aarch64_fmlsl_highv2sf): Ditto.
7168 (aarch64_fmlalq_highv4sf): Ditto.
7169 (aarch64_fmlslq_highv4sf): Ditto.
7170 (aarch64_fmlal_lane_lowv2sf): Ditto.
7171 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7172 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7173 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7174 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7175 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7176 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7177 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7178 (aarch64_fmlal_lane_highv2sf): Ditto.
7179 (aarch64_fmlsl_lane_highv2sf): Ditto.
7180 (aarch64_fmlal_laneq_highv2sf): Ditto.
7181 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7182 (aarch64_fmlalq_lane_highv4sf): Ditto.
7183 (aarch64_fmlsl_lane_highv4sf): Ditto.
7184 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7185 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7186 * config/aarch64/aarch64-simd.md:
7187 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7188 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7189 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7190 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7191 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7192 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7193 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7194 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7195 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7196 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7197 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7198 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7199 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7200 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7201 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7202 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7203 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7204 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7205 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7206 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7207 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7208 (vfmlsl_low_u32): Ditto.
7209 (vfmlalq_low_u32): Ditto.
7210 (vfmlslq_low_u32): Ditto.
7211 (vfmlal_high_u32): Ditto.
7212 (vfmlsl_high_u32): Ditto.
7213 (vfmlalq_high_u32): Ditto.
7214 (vfmlslq_high_u32): Ditto.
7215 (vfmlal_lane_low_u32): Ditto.
7216 (vfmlsl_lane_low_u32): Ditto.
7217 (vfmlal_laneq_low_u32): Ditto.
7218 (vfmlsl_laneq_low_u32): Ditto.
7219 (vfmlalq_lane_low_u32): Ditto.
7220 (vfmlslq_lane_low_u32): Ditto.
7221 (vfmlalq_laneq_low_u32): Ditto.
7222 (vfmlslq_laneq_low_u32): Ditto.
7223 (vfmlal_lane_high_u32): Ditto.
7224 (vfmlsl_lane_high_u32): Ditto.
7225 (vfmlal_laneq_high_u32): Ditto.
7226 (vfmlsl_laneq_high_u32): Ditto.
7227 (vfmlalq_lane_high_u32): Ditto.
7228 (vfmlslq_lane_high_u32): Ditto.
7229 (vfmlalq_laneq_high_u32): Ditto.
7230 (vfmlslq_laneq_high_u32): Ditto.
7231 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7232 (AARCH64_FL_FOR_ARCH8_4): New.
7233 (AARCH64_ISA_F16FML): New ISA flag.
7234 (TARGET_F16FML): New feature flag for fp16fml.
7235 (doc/invoke.texi): Document new fp16fml option.
7237 2018-01-10 Michael Collison <michael.collison@arm.com>
7239 * config/aarch64/aarch64-builtins.c:
7240 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7241 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7242 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7243 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7244 (AARCH64_ISA_SHA3): New ISA flag.
7245 (TARGET_SHA3): New feature flag for sha3.
7246 * config/aarch64/iterators.md (sha512_op): New int attribute.
7247 (CRYPTO_SHA512): New int iterator.
7248 (UNSPEC_SHA512H): New unspec.
7249 (UNSPEC_SHA512H2): Ditto.
7250 (UNSPEC_SHA512SU0): Ditto.
7251 (UNSPEC_SHA512SU1): Ditto.
7252 * config/aarch64/aarch64-simd-builtins.def
7253 (aarch64_crypto_sha512hqv2di): New builtin.
7254 (aarch64_crypto_sha512h2qv2di): Ditto.
7255 (aarch64_crypto_sha512su0qv2di): Ditto.
7256 (aarch64_crypto_sha512su1qv2di): Ditto.
7257 (aarch64_eor3qv8hi): Ditto.
7258 (aarch64_rax1qv2di): Ditto.
7259 (aarch64_xarqv2di): Ditto.
7260 (aarch64_bcaxqv8hi): Ditto.
7261 * config/aarch64/aarch64-simd.md:
7262 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7263 (aarch64_crypto_sha512su0qv2di): Ditto.
7264 (aarch64_crypto_sha512su1qv2di): Ditto.
7265 (aarch64_eor3qv8hi): Ditto.
7266 (aarch64_rax1qv2di): Ditto.
7267 (aarch64_xarqv2di): Ditto.
7268 (aarch64_bcaxqv8hi): Ditto.
7269 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7270 (vsha512h2q_u64): Ditto.
7271 (vsha512su0q_u64): Ditto.
7272 (vsha512su1q_u64): Ditto.
7273 (veor3q_u16): Ditto.
7274 (vrax1q_u64): Ditto.
7276 (vbcaxq_u16): Ditto.
7277 * config/arm/types.md (crypto_sha512): New type attribute.
7278 (crypto_sha3): Ditto.
7279 (doc/invoke.texi): Document new sha3 option.
7281 2018-01-10 Michael Collison <michael.collison@arm.com>
7283 * config/aarch64/aarch64-builtins.c:
7284 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7285 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7286 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7287 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7288 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7289 (AARCH64_ISA_SM4): New ISA flag.
7290 (TARGET_SM4): New feature flag for sm4.
7291 * config/aarch64/aarch64-simd-builtins.def
7292 (aarch64_sm3ss1qv4si): Ditto.
7293 (aarch64_sm3tt1aq4si): Ditto.
7294 (aarch64_sm3tt1bq4si): Ditto.
7295 (aarch64_sm3tt2aq4si): Ditto.
7296 (aarch64_sm3tt2bq4si): Ditto.
7297 (aarch64_sm3partw1qv4si): Ditto.
7298 (aarch64_sm3partw2qv4si): Ditto.
7299 (aarch64_sm4eqv4si): Ditto.
7300 (aarch64_sm4ekeyqv4si): Ditto.
7301 * config/aarch64/aarch64-simd.md:
7302 (aarch64_sm3ss1qv4si): Ditto.
7303 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7304 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7305 (aarch64_sm4eqv4si): Ditto.
7306 (aarch64_sm4ekeyqv4si): Ditto.
7307 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7308 (sm3part_op): Ditto.
7309 (CRYPTO_SM3TT): Ditto.
7310 (CRYPTO_SM3PART): Ditto.
7311 (UNSPEC_SM3SS1): New unspec.
7312 (UNSPEC_SM3TT1A): Ditto.
7313 (UNSPEC_SM3TT1B): Ditto.
7314 (UNSPEC_SM3TT2A): Ditto.
7315 (UNSPEC_SM3TT2B): Ditto.
7316 (UNSPEC_SM3PARTW1): Ditto.
7317 (UNSPEC_SM3PARTW2): Ditto.
7318 (UNSPEC_SM4E): Ditto.
7319 (UNSPEC_SM4EKEY): Ditto.
7320 * config/aarch64/constraints.md (Ui2): New constraint.
7321 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7322 * config/arm/types.md (crypto_sm3): New type attribute.
7323 (crypto_sm4): Ditto.
7324 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7325 (vsm3tt1aq_u32): Ditto.
7326 (vsm3tt1bq_u32): Ditto.
7327 (vsm3tt2aq_u32): Ditto.
7328 (vsm3tt2bq_u32): Ditto.
7329 (vsm3partw1q_u32): Ditto.
7330 (vsm3partw2q_u32): Ditto.
7331 (vsm4eq_u32): Ditto.
7332 (vsm4ekeyq_u32): Ditto.
7333 (doc/invoke.texi): Document new sm4 option.
7335 2018-01-10 Michael Collison <michael.collison@arm.com>
7337 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7338 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7339 (AARCH64_FL_FOR_ARCH8_4): New.
7340 (AARCH64_FL_V8_4): New flag.
7341 (doc/invoke.texi): Document new armv8.4-a option.
7343 2018-01-10 Michael Collison <michael.collison@arm.com>
7345 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7346 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7347 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7348 * config/aarch64/aarch64-option-extension.def: Add
7349 AARCH64_OPT_EXTENSION of 'sha2'.
7350 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7351 (crypto): Disable sha2 and aes if crypto disabled.
7352 (crypto): Enable aes and sha2 if enabled.
7353 (simd): Disable sha2 and aes if simd disabled.
7354 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7356 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7357 (TARGET_SHA2): New feature flag for sha2.
7358 (TARGET_AES): New feature flag for aes.
7359 * config/aarch64/aarch64-simd.md:
7360 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7361 conditional on TARGET_AES.
7362 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7363 (aarch64_crypto_sha1hsi): Make pattern conditional
7365 (aarch64_crypto_sha1hv4si): Ditto.
7366 (aarch64_be_crypto_sha1hv4si): Ditto.
7367 (aarch64_crypto_sha1su1v4si): Ditto.
7368 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7369 (aarch64_crypto_sha1su0v4si): Ditto.
7370 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7371 (aarch64_crypto_sha256su0v4si): Ditto.
7372 (aarch64_crypto_sha256su1v4si): Ditto.
7373 (doc/invoke.texi): Document new aes and sha2 options.
7375 2018-01-10 Martin Sebor <msebor@redhat.com>
7377 PR tree-optimization/83781
7378 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7381 2018-01-11 Martin Sebor <msebor@gmail.com>
7382 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7384 PR tree-optimization/83501
7385 PR tree-optimization/81703
7387 * tree-ssa-strlen.c (get_string_cst): Rename...
7388 (get_string_len): ...to this. Handle global constants.
7389 (handle_char_store): Adjust.
7391 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7392 Jim Wilson <jimw@sifive.com>
7394 * config/riscv/riscv-protos.h (riscv_output_return): New.
7395 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7396 (riscv_attribute_table, riscv_output_return),
7397 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7398 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7399 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7401 (riscv_expand_prologue): Add early return for naked function.
7402 (riscv_expand_epilogue): Likewise.
7403 (riscv_function_ok_for_sibcall): Return false for naked function.
7404 (riscv_set_current_function): New.
7405 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7406 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7407 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7408 * doc/extend.texi (RISC-V Function Attributes): New.
7410 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7412 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7413 check for 128-bit long double before checking TCmode.
7414 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7415 128-bit long doubles before checking TFmode or TCmode.
7416 (FLOAT128_IBM_P): Likewise.
7418 2018-01-10 Martin Sebor <msebor@redhat.com>
7420 PR tree-optimization/83671
7421 * builtins.c (c_strlen): Unconditionally return zero for the empty
7423 Use -Warray-bounds for warnings.
7424 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7425 for non-constant array indices with COMPONENT_REF, arrays of
7426 arrays, and pointers to arrays.
7427 (gimple_fold_builtin_strlen): Determine and set length range for
7428 non-constant character arrays.
7430 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7433 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7436 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7438 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7440 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7443 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7444 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7445 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7446 indexed_or_indirect_operand predicate.
7447 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7448 (*vsx_le_perm_load_v8hi): Likewise.
7449 (*vsx_le_perm_load_v16qi): Likewise.
7450 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7451 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7452 (*vsx_le_perm_store_v8hi): Likewise.
7453 (*vsx_le_perm_store_v16qi): Likewise.
7454 (eight unnamed splitters): Likewise.
7456 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7458 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7459 * config/rs6000/emmintrin.h: Likewise.
7460 * config/rs6000/mmintrin.h: Likewise.
7461 * config/rs6000/xmmintrin.h: Likewise.
7463 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7466 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7468 * tree.c (tree_nop_conversion): Return true for location wrapper
7470 (maybe_wrap_with_location): New function.
7471 (selftest::check_strip_nops): New function.
7472 (selftest::test_location_wrappers): New function.
7473 (selftest::tree_c_tests): Call it.
7474 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7475 (maybe_wrap_with_location): New decl.
7476 (EXPR_LOCATION_WRAPPER_P): New macro.
7477 (location_wrapper_p): New inline function.
7478 (tree_strip_any_location_wrapper): New inline function.
7480 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7483 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7484 stack_realign_offset for the largest alignment of stack slot
7486 (ix86_find_max_used_stack_alignment): New function.
7487 (ix86_finalize_stack_frame_flags): Use it. Set
7488 max_used_stack_alignment if we don't realign stack.
7489 * config/i386/i386.h (machine_function): Add
7490 max_used_stack_alignment.
7492 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7494 * config/arm/arm.opt (-mbranch-cost): New option.
7495 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7498 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7501 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7502 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7504 2018-01-10 Richard Biener <rguenther@suse.de>
7507 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7508 early out so it also covers the case where we have a non-NULL
7511 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7513 PR tree-optimization/83753
7514 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7515 for non-strided grouped accesses if the number of elements is 1.
7517 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7520 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7521 * i386.h (TARGET_USE_GATHER): Define.
7522 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7524 2018-01-10 Martin Liska <mliska@suse.cz>
7527 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7528 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7530 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7531 CLEANUP_NO_PARTITIONING is not set.
7533 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7535 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7536 for vectors, as a partial revert of r254296.
7537 * rtl.h (const_vec_p): Delete.
7538 (const_vec_duplicate_p): Don't test for vector CONSTs.
7539 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7540 * expmed.c (make_tree): Likewise.
7543 * common.md (E, F): Use CONSTANT_P instead of checking for
7545 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7546 checking for CONST_VECTOR.
7548 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7551 * predict.c (force_edge_cold): Handle in more sane way edges
7554 2018-01-09 Carl Love <cel@us.ibm.com>
7556 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7558 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7559 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7560 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7561 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7562 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7563 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7564 * config/rs6000/rs6000-protos.h: Add extern defition for
7565 rs6000_generate_float2_double_code.
7566 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7568 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7569 (float2_v2df): Add define_expand.
7571 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7574 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7575 op_mode in the force_to_mode call.
7577 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7579 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7580 instead of checking each element individually.
7581 (aarch64_evpc_uzp): Likewise.
7582 (aarch64_evpc_zip): Likewise.
7583 (aarch64_evpc_ext): Likewise.
7584 (aarch64_evpc_rev): Likewise.
7585 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7586 instead of checking each element individually. Return true without
7588 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7589 whether all selected elements come from the same input, instead of
7590 checking each element individually. Remove calls to gen_rtx_REG,
7591 start_sequence and end_sequence and instead assert that no rtl is
7594 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7596 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7597 order of HIGH and CONST checks.
7599 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7601 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7602 if the destination isn't an SSA_NAME.
7604 2018-01-09 Richard Biener <rguenther@suse.de>
7606 PR tree-optimization/83668
7607 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7609 (canonicalize_loop_form): ... here, renamed from ...
7610 (canonicalize_loop_closed_ssa_form): ... this and amended to
7611 swap successor edges for loop exit blocks to make us use
7612 the RPO order we need for initial schedule generation.
7614 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7616 PR tree-optimization/64811
7617 * match.pd: When optimizing comparisons with Inf, avoid
7618 introducing or losing exceptions from comparisons with NaN.
7620 2018-01-09 Martin Liska <mliska@suse.cz>
7623 * asan.c (shadow_mem_size): Add gcc_assert.
7625 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7627 Don't save registers in main().
7630 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7631 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7632 * config/avr/avr.c (avr_set_current_function): Don't error if
7633 naked, OS_task or OS_main are specified at the same time.
7634 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7636 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7638 * common/config/avr/avr-common.c (avr_option_optimization_table):
7639 Switch on -mmain-is-OS_task for optimizing compilations.
7641 2018-01-09 Richard Biener <rguenther@suse.de>
7643 PR tree-optimization/83572
7644 * graphite.c: Include cfganal.h.
7645 (graphite_transform_loops): Connect infinite loops to exit
7646 and remove fake edges at the end.
7648 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7650 * ipa-inline.c (edge_badness): Revert accidental checkin.
7652 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7655 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7656 symbols; not inline clones.
7658 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7661 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7662 hard registers. Formatting fixes.
7664 PR preprocessor/83722
7665 * gcc.c (try_generate_repro): Pass
7666 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7667 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7670 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7671 Kito Cheng <kito.cheng@gmail.com>
7673 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7674 (riscv_leaf_function_p): Delete.
7675 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7677 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7679 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7681 (do_ifelse): New function.
7682 (do_isel): New function.
7683 (do_sub3): New function.
7684 (do_add3): New function.
7685 (do_load_mask_compare): New function.
7686 (do_overlap_load_compare): New function.
7687 (expand_compare_loop): New function.
7688 (expand_block_compare): Call expand_compare_loop() when appropriate.
7689 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7691 (-mblock-compare-inline-loop-limit): New option.
7693 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7696 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7697 Reverse order of second and third operands in first alternative.
7698 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7699 of first and second elements in UNSPEC_VPERMR vector.
7700 (altivec_expand_vec_perm_le): Likewise.
7702 2018-01-08 Jeff Law <law@redhat.com>
7704 PR rtl-optimizatin/81308
7705 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7706 (process_switch): If group_case_labels makes a change, then set
7708 (pass_convert_switch::execute): If a switch is converted, then
7709 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7711 PR rtl-optimization/81308
7712 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7715 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7717 PR target/83663 - Revert r255946
7718 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7719 generation for cases where splatting a value is not useful.
7720 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7721 across a vec_duplicate and a paradoxical subreg forming a vector
7722 mode to a vec_concat.
7724 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7726 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7727 -march=armv8.3-a variants.
7728 * config/arm/t-multilib: Likewise.
7729 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7731 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7733 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7735 (cceq_ior_compare_complement): Give it a name so I can use it, and
7736 change boolean_or_operator predicate to boolean_operator so it can
7737 be used to generate a crand.
7738 (eqne): New code iterator.
7739 (bd/bd_neg): New code_attrs.
7740 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7741 a single define_insn.
7742 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7743 decrement (bdnzt/bdnzf/bdzt/bdzf).
7744 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7745 with the new names of the branch decrement patterns, and added the
7746 names of the branch decrement conditional patterns.
7748 2018-01-08 Richard Biener <rguenther@suse.de>
7750 PR tree-optimization/83563
7751 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7754 2018-01-08 Richard Biener <rguenther@suse.de>
7757 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7759 2018-01-08 Richard Biener <rguenther@suse.de>
7761 PR tree-optimization/83685
7762 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7763 references to abnormals.
7765 2018-01-08 Richard Biener <rguenther@suse.de>
7768 * dwarf2out.c (output_indirect_strings): Handle empty
7769 skeleton_debug_str_hash.
7770 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7772 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7774 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7775 (emit_store_direct): Likewise.
7776 (arc_trampoline_adjust_address): Likewise.
7777 (arc_asm_trampoline_template): New function.
7778 (arc_initialize_trampoline): Use asm_trampoline_template.
7779 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7780 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7781 * config/arc/arc.md (flush_icache): Delete pattern.
7783 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7785 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7786 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7789 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7792 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7793 by not USED_FOR_TARGET.
7794 (make_pass_resolve_sw_modes): Likewise.
7796 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7798 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7801 2018-01-08 Richard Biener <rguenther@suse.de>
7804 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7806 2018-01-08 Richard Biener <rguenther@suse.de>
7809 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7811 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7814 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7815 basic blocks with a small number of successors.
7816 (convert_control_dep_chain_into_preds): Improve handling of
7818 (dump_predicates): Split apart into...
7819 (dump_pred_chain): ...here...
7820 (dump_pred_info): ...and here.
7821 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7822 (can_chain_union_be_invalidated_p): Improve check for invalidation
7824 (uninit_uses_cannot_happen): Avoid unnecessary if
7825 convert_control_dep_chain_into_preds yielded nothing.
7827 2018-01-06 Martin Sebor <msebor@redhat.com>
7829 PR tree-optimization/83640
7830 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7831 subtracting negative offset from size.
7832 (builtin_access::overlap): Adjust offset bounds of the access to fall
7833 within the size of the object if possible.
7835 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7837 PR rtl-optimization/83699
7838 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7839 extract_bit_field_as_subreg to cases in which the extracted
7840 value is also a vector.
7842 * lra-constraints.c (process_alt_operands): Test for the equivalence
7843 substitutions when detecting a possible reload cycle.
7845 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7848 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7849 by default if flag_selective_schedling{,2}. Formatting fixes.
7851 PR rtl-optimization/83682
7852 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7853 if it has non-VECTOR_MODE element mode.
7854 (vec_duplicate_p): Likewise.
7857 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7858 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7860 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7863 * config/i386/i386-builtin.def
7864 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7865 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7866 Require also OPTION_MASK_ISA_AVX512F in addition to
7867 OPTION_MASK_ISA_GFNI.
7868 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7869 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7870 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7871 to OPTION_MASK_ISA_GFNI.
7872 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7873 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7874 OPTION_MASK_ISA_AVX512BW.
7875 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7876 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7877 addition to OPTION_MASK_ISA_GFNI.
7878 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7879 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7880 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7881 to OPTION_MASK_ISA_GFNI.
7882 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7883 a requirement for all ISAs rather than any of them with a few
7885 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7887 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7888 bitmasks to be enabled with 3 exceptions, instead of requiring any
7889 enabled ISA with lots of exceptions.
7890 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7891 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7892 Change avx512bw in isa attribute to avx512f.
7893 * config/i386/sgxintrin.h: Add license boilerplate.
7894 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7895 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7896 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7897 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7899 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7900 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7901 temporarily sse2 rather than sse if not enabled already.
7904 * config/i386/sse.md (VI248_VLBW): Rename to ...
7905 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7906 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7907 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7908 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7909 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7910 mode iterator instead of VI248_VLBW.
7912 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7914 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7915 (record_modified): Skip clobbers; add debug output.
7916 (param_change_prob): Use sreal frequencies.
7918 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7920 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7921 punt for user-aligned variables.
7923 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7925 * tree-chrec.c (chrec_contains_symbols): Return true for
7928 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7931 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7932 of (x|y) == x for BICS pattern.
7934 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7936 PR tree-optimization/83605
7937 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7938 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7941 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7943 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7944 * config/epiphany/rtems.h: New file.
7946 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7947 Uros Bizjak <ubizjak@gmail.com>
7950 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7951 QIreg_operand instead of register_operand predicate.
7952 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7953 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7954 comments instead of -fmitigate[-_]rop.
7956 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7959 * cgraphunit.c (symbol_table::compile): Switch to text_section
7960 before calling assembly_start debug hook.
7961 * run-rtl-passes.c (run_rtl_passes): Likewise.
7964 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7966 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7967 range_int_cst_p rather than !symbolic_range_p before calling
7968 extract_range_from_multiplicative_op_1.
7970 2018-01-04 Jeff Law <law@redhat.com>
7972 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7973 redundant test in assertion.
7975 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7977 * doc/rtl.texi: Document machine_mode wrapper classes.
7979 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7981 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7984 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7986 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7987 the VEC_PERM_EXPR fold to fail.
7989 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7992 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7993 to switched_sections.
7995 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7998 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8001 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8004 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8005 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8007 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8010 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8011 is BLKmode and bitpos not zero or mode change is needed.
8013 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8016 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8019 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8022 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8023 instead of MULT rtx. Update all corresponding splitters.
8025 (*ssub<modesuffix>): Ditto.
8027 (*cmp_sadd_di): Update split patterns.
8028 (*cmp_sadd_si): Ditto.
8029 (*cmp_sadd_sidi): Ditto.
8030 (*cmp_ssub_di): Ditto.
8031 (*cmp_ssub_si): Ditto.
8032 (*cmp_ssub_sidi): Ditto.
8033 * config/alpha/predicates.md (const23_operand): New predicate.
8034 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8035 Look for ASHIFT, not MULT inner operand.
8036 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8038 2018-01-04 Martin Liska <mliska@suse.cz>
8040 PR gcov-profile/83669
8041 * gcov.c (output_intermediate_file): Add version to intermediate
8043 * doc/gcov.texi: Document new field 'version' in intermediate
8044 file format. Fix location of '-k' option of gcov command.
8046 2018-01-04 Martin Liska <mliska@suse.cz>
8049 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8051 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8053 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8055 2018-01-03 Martin Sebor <msebor@redhat.com>
8057 PR tree-optimization/83655
8058 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8059 checking calls with invalid arguments.
8061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8063 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8064 (vectorizable_mask_load_store): Delete.
8065 (vectorizable_call): Return false for masked loads and stores.
8066 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8067 instead of gimple_assign_rhs1.
8068 (vectorizable_load): Handle IFN_MASK_LOAD.
8069 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8073 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8075 (vectorizable_mask_load_store): ...here.
8076 (vectorizable_load): ...and here.
8078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8080 * tree-vect-stmts.c (vect_build_all_ones_mask)
8081 (vect_build_zero_merge_argument): New functions, split out from...
8082 (vectorizable_load): ...here.
8084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8086 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8088 (vectorizable_mask_load_store): ...here.
8089 (vectorizable_store): ...and here.
8091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8093 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8095 (vectorizable_mask_load_store): ...here.
8097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8099 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8100 (vect_model_store_cost): Take a vec_load_store_type instead of a
8102 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8103 (vect_model_store_cost): Take a vec_load_store_type instead of a
8105 (vectorizable_mask_load_store): Update accordingly.
8106 (vectorizable_store): Likewise.
8107 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8111 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8112 IFN_MASK_LOAD calls here rather than...
8113 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8116 Alan Hayward <alan.hayward@arm.com>
8117 David Sherwood <david.sherwood@arm.com>
8119 * expmed.c (extract_bit_field_1): For vector extracts,
8120 fall back to extract_bit_field_as_subreg if vec_extract
8123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8124 Alan Hayward <alan.hayward@arm.com>
8125 David Sherwood <david.sherwood@arm.com>
8127 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8128 they are variable or constant sized.
8129 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8130 slots for constant-sized data.
8132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8133 Alan Hayward <alan.hayward@arm.com>
8134 David Sherwood <david.sherwood@arm.com>
8136 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8137 handling COND_EXPRs with boolean comparisons, try to find a better
8138 basis for the mask type than the boolean itself.
8140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8142 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8143 is calculated and how it can be overridden.
8144 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8145 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8147 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8151 Alan Hayward <alan.hayward@arm.com>
8152 David Sherwood <david.sherwood@arm.com>
8154 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8155 Remove the mode argument.
8156 (aarch64_simd_valid_immediate): Remove the mode and inverse
8158 * config/aarch64/iterators.md (bitsize): New iterator.
8159 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8160 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8161 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8162 aarch64_simd_valid_immediate.
8163 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8164 (aarch64_reg_or_bic_imm): Likewise.
8165 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8166 with an insn_type enum and msl with a modifier_type enum.
8167 Replace element_width with a scalar_mode. Change the shift
8168 to unsigned int. Add constructors for scalar_float_mode and
8169 scalar_int_mode elements.
8170 (aarch64_vect_float_const_representable_p): Delete.
8171 (aarch64_can_const_movi_rtx_p)
8172 (aarch64_simd_scalar_immediate_valid_for_move)
8173 (aarch64_simd_make_constant): Update call to
8174 aarch64_simd_valid_immediate.
8175 (aarch64_advsimd_valid_immediate_hs): New function.
8176 (aarch64_advsimd_valid_immediate): Likewise.
8177 (aarch64_simd_valid_immediate): Remove mode and inverse
8178 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8179 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8180 and aarch64_float_const_representable_p on the result.
8181 (aarch64_output_simd_mov_immediate): Remove mode argument.
8182 Update call to aarch64_simd_valid_immediate and use of
8183 simd_immediate_info.
8184 (aarch64_output_scalar_simd_mov_immediate): Update call
8187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8188 Alan Hayward <alan.hayward@arm.com>
8189 David Sherwood <david.sherwood@arm.com>
8191 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8192 (mode_nunits): Likewise CONST_MODE_NUNITS.
8193 * machmode.def (ADJUST_NUNITS): Document.
8194 * genmodes.c (mode_data::need_nunits_adj): New field.
8195 (blank_mode): Update accordingly.
8196 (adj_nunits): New variable.
8197 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8199 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8200 listed in adj_nunits.
8201 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8202 listed in adj_nunits. Don't emit case statements for such modes.
8203 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8204 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8205 nothing if adj_nunits is nonnull.
8206 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8207 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8208 (emit_mode_fbit): Update use of print_maybe_const_decl.
8209 (emit_move_size): Likewise. Treat the array as non-const
8211 (emit_mode_adjustments): Handle adj_nunits.
8213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8215 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8216 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8217 (VECTOR_MODES): Use it.
8218 (make_vector_modes): Take the prefix as an argument.
8220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8221 Alan Hayward <alan.hayward@arm.com>
8222 David Sherwood <david.sherwood@arm.com>
8224 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8225 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8226 for MODE_VECTOR_BOOL.
8227 * machmode.def (VECTOR_BOOL_MODE): Document.
8228 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8229 (make_vector_bool_mode): New function.
8230 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8232 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8233 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8235 * stor-layout.c (int_mode_for_mode): Likewise.
8236 * tree.c (build_vector_type_for_mode): Likewise.
8237 * varasm.c (output_constant_pool_2): Likewise.
8238 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8239 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8240 for MODE_VECTOR_BOOL.
8241 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8242 of mode class checks.
8243 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8244 instead of a list of mode class checks.
8245 (expand_vector_scalar_condition): Likewise.
8246 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8249 Alan Hayward <alan.hayward@arm.com>
8250 David Sherwood <david.sherwood@arm.com>
8252 * machmode.h (mode_size): Change from unsigned short to
8254 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8255 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8256 or if measurement_type is not polynomial.
8257 (fixed_size_mode::includes_p): Check for constant-sized modes.
8258 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8259 return a poly_uint16 rather than an unsigned short.
8260 (emit_mode_size): Change the type of mode_size from unsigned short
8261 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8262 (emit_mode_adjustments): Cope with polynomial vector sizes.
8263 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8265 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8267 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8268 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8269 * caller-save.c (setup_save_areas): Likewise.
8270 (replace_reg_with_saved_mem): Likewise.
8271 * calls.c (emit_library_call_value_1): Likewise.
8272 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8273 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8274 (gen_lowpart_for_combine): Likewise.
8275 * convert.c (convert_to_integer_1): Likewise.
8276 * cse.c (equiv_constant, cse_insn): Likewise.
8277 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8278 (cselib_subst_to_values): Likewise.
8279 * dce.c (word_dce_process_block): Likewise.
8280 * df-problems.c (df_word_lr_mark_ref): Likewise.
8281 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8282 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8283 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8284 (rtl_for_decl_location): Likewise.
8285 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8286 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8287 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8288 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8289 (expand_expr_real_1): Likewise.
8290 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8291 (pad_below): Likewise.
8292 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8293 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8294 * ira.c (get_subreg_tracking_sizes): Likewise.
8295 * ira-build.c (ira_create_allocno_objects): Likewise.
8296 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8297 (ira_sort_regnos_for_alter_reg): Likewise.
8298 * ira-costs.c (record_operand_costs): Likewise.
8299 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8300 (resolve_simple_move): Likewise.
8301 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8302 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8303 (lra_constraints): Likewise.
8304 (CONST_POOL_OK_P): Reject variable-sized modes.
8305 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8306 (add_pseudo_to_slot, lra_spill): Likewise.
8307 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8308 * optabs-query.c (get_best_extraction_insn): Likewise.
8309 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8310 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8311 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8312 * recog.c (offsettable_address_addr_space_p): Likewise.
8313 * regcprop.c (maybe_mode_change): Likewise.
8314 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8315 * regrename.c (build_def_use): Likewise.
8316 * regstat.c (dump_reg_info): Likewise.
8317 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8318 (find_reloads, find_reloads_subreg_address): Likewise.
8319 * reload1.c (eliminate_regs_1): Likewise.
8320 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8321 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8322 (simplify_binary_operation_1, simplify_subreg): Likewise.
8323 * targhooks.c (default_function_arg_padding): Likewise.
8324 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8325 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8326 (verify_gimple_assign_ternary): Likewise.
8327 * tree-inline.c (estimate_move_cost): Likewise.
8328 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8329 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8330 (get_address_cost_ainc): Likewise.
8331 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8332 (vect_supportable_dr_alignment): Likewise.
8333 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8334 (vectorizable_reduction): Likewise.
8335 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8336 (vectorizable_operation, vectorizable_load): Likewise.
8337 * tree.c (build_same_sized_truth_vector_type): Likewise.
8338 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8339 * var-tracking.c (emit_note_insn_var_location): Likewise.
8340 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8341 (ADDR_VEC_ALIGN): Likewise.
8343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8344 Alan Hayward <alan.hayward@arm.com>
8345 David Sherwood <david.sherwood@arm.com>
8347 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8349 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8350 or if measurement_type is polynomial.
8351 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8352 * combine.c (make_extraction): Likewise.
8353 * dse.c (find_shift_sequence): Likewise.
8354 * dwarf2out.c (mem_loc_descriptor): Likewise.
8355 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8356 (extract_bit_field, extract_low_bits): Likewise.
8357 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8358 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8359 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8360 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8361 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8362 * reload.c (find_reloads): Likewise.
8363 * reload1.c (alter_reg): Likewise.
8364 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8365 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8366 * tree-if-conv.c (predicate_mem_writes): Likewise.
8367 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8368 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8369 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8370 * valtrack.c (dead_debug_insert_temp): Likewise.
8371 * varasm.c (mergeable_constant_section): Likewise.
8372 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8375 Alan Hayward <alan.hayward@arm.com>
8376 David Sherwood <david.sherwood@arm.com>
8378 * expr.c (expand_assignment): Cope with polynomial mode sizes
8379 when assigning to a CONCAT.
8381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8382 Alan Hayward <alan.hayward@arm.com>
8383 David Sherwood <david.sherwood@arm.com>
8385 * machmode.h (mode_precision): Change from unsigned short to
8387 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8389 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8390 or if measurement_type is not polynomial.
8391 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8392 in which the mode is already known to be a scalar_int_mode.
8393 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8394 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8396 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8397 for GET_MODE_PRECISION.
8398 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8399 for GET_MODE_PRECISION.
8400 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8402 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8403 (expand_field_assignment, make_extraction): Likewise.
8404 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8405 (get_last_value): Likewise.
8406 * convert.c (convert_to_integer_1): Likewise.
8407 * cse.c (cse_insn): Likewise.
8408 * expr.c (expand_expr_real_1): Likewise.
8409 * lra-constraints.c (simplify_operand_subreg): Likewise.
8410 * optabs-query.c (can_atomic_load_p): Likewise.
8411 * optabs.c (expand_atomic_load): Likewise.
8412 (expand_atomic_store): Likewise.
8413 * ree.c (combine_reaching_defs): Likewise.
8414 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8415 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8416 * tree.h (type_has_mode_precision_p): Likewise.
8417 * ubsan.c (instrument_si_overflow): Likewise.
8419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8420 Alan Hayward <alan.hayward@arm.com>
8421 David Sherwood <david.sherwood@arm.com>
8423 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8424 polynomial numbers of units.
8425 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8426 (valid_vector_subparts_p): New function.
8427 (build_vector_type): Remove temporary shim and take the number
8428 of units as a poly_uint64 rather than an int.
8429 (build_opaque_vector_type): Take the number of units as a
8430 poly_uint64 rather than an int.
8431 * tree.c (build_vector_from_ctor): Handle polynomial
8432 TYPE_VECTOR_SUBPARTS.
8433 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8434 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8435 (build_vector_from_val): If the number of units is variable,
8436 use build_vec_duplicate_cst for constant operands and
8437 VEC_DUPLICATE_EXPR otherwise.
8438 (make_vector_type): Remove temporary is_constant ().
8439 (build_vector_type, build_opaque_vector_type): Take the number of
8440 units as a poly_uint64 rather than an int.
8441 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8443 * cfgexpand.c (expand_debug_expr): Likewise.
8444 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8445 (store_constructor, expand_expr_real_1): Likewise.
8446 (const_scalar_mask_from_tree): Likewise.
8447 * fold-const-call.c (fold_const_reduction): Likewise.
8448 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8449 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8450 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8451 (fold_relational_const): Likewise.
8452 (native_interpret_vector): Likewise. Change the size from an
8453 int to an unsigned int.
8454 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8455 TYPE_VECTOR_SUBPARTS.
8456 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8457 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8458 duplicating a non-constant operand into a variable-length vector.
8459 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8460 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8461 * ipa-icf.c (sem_variable::equals): Likewise.
8462 * match.pd: Likewise.
8463 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8464 * print-tree.c (print_node): Likewise.
8465 * stor-layout.c (layout_type): Likewise.
8466 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8467 * tree-cfg.c (verify_gimple_comparison): Likewise.
8468 (verify_gimple_assign_binary): Likewise.
8469 (verify_gimple_assign_ternary): Likewise.
8470 (verify_gimple_assign_single): Likewise.
8471 * tree-pretty-print.c (dump_generic_node): Likewise.
8472 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8473 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8474 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8475 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8476 (vect_shift_permute_load_chain): Likewise.
8477 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8478 (expand_vector_condition, optimize_vector_constructor): Likewise.
8479 (lower_vec_perm, get_compute_type): Likewise.
8480 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8481 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8482 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8483 (vect_recog_mask_conversion_pattern): Likewise.
8484 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8485 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8486 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8487 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8488 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8489 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8490 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8491 (supportable_widening_operation): Likewise.
8492 (supportable_narrowing_operation): Likewise.
8493 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8495 * varasm.c (output_constant): Likewise.
8497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8498 Alan Hayward <alan.hayward@arm.com>
8499 David Sherwood <david.sherwood@arm.com>
8501 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8502 so that both the length == 3 and length != 3 cases set up their
8503 own permute vectors. Add comments explaining why we know the
8504 number of elements is constant.
8505 (vect_permute_load_chain): Likewise.
8507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8508 Alan Hayward <alan.hayward@arm.com>
8509 David Sherwood <david.sherwood@arm.com>
8511 * machmode.h (mode_nunits): Change from unsigned char to
8513 (ONLY_FIXED_SIZE_MODES): New macro.
8514 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8515 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8516 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8518 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8519 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8520 or if measurement_type is not polynomial.
8521 * genmodes.c (ZERO_COEFFS): New macro.
8522 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8524 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8525 Use ZERO_COEFFS when emitting initializers.
8526 * data-streamer.h (bp_pack_poly_value): New function.
8527 (bp_unpack_poly_value): Likewise.
8528 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8529 for GET_MODE_NUNITS.
8530 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8531 for GET_MODE_NUNITS.
8532 * tree.c (make_vector_type): Remove temporary shim and make
8533 the real function take the number of units as a poly_uint64
8535 (build_vector_type_for_mode): Handle polynomial nunits.
8536 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8537 * emit-rtl.c (const_vec_series_p_1): Likewise.
8538 (gen_rtx_CONST_VECTOR): Likewise.
8539 * fold-const.c (test_vec_duplicate_folding): Likewise.
8540 * genrecog.c (validate_pattern): Likewise.
8541 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8542 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8543 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8544 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8545 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8546 * rtlanal.c (subreg_get_info): Likewise.
8547 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8548 (vect_grouped_load_supported): Likewise.
8549 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8550 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8551 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8552 (simplify_const_unary_operation, simplify_binary_operation_1)
8553 (simplify_const_binary_operation, simplify_ternary_operation)
8554 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8555 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8556 instead of CONST_VECTOR_NUNITS.
8557 * varasm.c (output_constant_pool_2): Likewise.
8558 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8559 explicit-encoded elements in the XVEC for variable-length vectors.
8561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8563 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8566 Alan Hayward <alan.hayward@arm.com>
8567 David Sherwood <david.sherwood@arm.com>
8569 * coretypes.h (fixed_size_mode): Declare.
8570 (fixed_size_mode_pod): New typedef.
8571 * builtins.h (target_builtins::x_apply_args_mode)
8572 (target_builtins::x_apply_result_mode): Change type to
8573 fixed_size_mode_pod.
8574 * builtins.c (apply_args_size, apply_result_size, result_vector)
8575 (expand_builtin_apply_args_1, expand_builtin_apply)
8576 (expand_builtin_return): Update accordingly.
8578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8580 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8581 * cselib.c (cselib_hash_rtx): Likewise.
8582 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8583 CONST_VECTOR encoding.
8585 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8586 Jeff Law <law@redhat.com>
8589 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8590 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8591 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8592 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8595 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8596 explicitly probe *sp in a noreturn function if there were any callee
8597 register saves or frame pointer is needed.
8599 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8602 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8603 BLKmode for ternary, binary or unary expressions.
8606 * var-tracking.c (delete_vta_debug_insn): New inline function.
8607 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8608 insns from get_insns () to NULL instead of each bb separately.
8609 Use delete_vta_debug_insn. No longer static.
8610 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8611 delete_vta_debug_insns callers.
8612 * rtl.h (delete_vta_debug_insns): Declare.
8613 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8614 instead of variable_tracking_main.
8616 2018-01-03 Martin Sebor <msebor@redhat.com>
8618 PR tree-optimization/83603
8619 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8620 arguments past the endof the argument list in functions declared
8621 without a prototype.
8622 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8623 Avoid checking when arguments are null.
8625 2018-01-03 Martin Sebor <msebor@redhat.com>
8628 * doc/extend.texi (attribute const): Fix a typo.
8629 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8630 issuing -Wsuggest-attribute for void functions.
8632 2018-01-03 Martin Sebor <msebor@redhat.com>
8634 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8635 offset_int::from instead of wide_int::to_shwi.
8636 (maybe_diag_overlap): Remove assertion.
8637 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8638 * gimple-ssa-sprintf.c (format_directive): Same.
8639 (parse_directive): Same.
8640 (sprintf_dom_walker::compute_format_length): Same.
8641 (try_substitute_return_value): Same.
8643 2018-01-03 Jeff Law <law@redhat.com>
8646 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8647 non-constant residual for zero at runtime and avoid probing in
8648 that case. Reorganize code for trailing problem to mirror handling
8651 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8653 PR tree-optimization/83501
8654 * tree-ssa-strlen.c (get_string_cst): New.
8655 (handle_char_store): Call get_string_cst.
8657 2018-01-03 Martin Liska <mliska@suse.cz>
8659 PR tree-optimization/83593
8660 * tree-ssa-strlen.c: Include tree-cfg.h.
8661 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8662 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8663 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8665 (strlen_dom_walker::before_dom_children): Call
8666 gimple_purge_dead_eh_edges. Dump tranformation with details
8668 (strlen_dom_walker::before_dom_children): Update call by adding
8669 new argument cleanup_eh.
8670 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8672 2018-01-03 Martin Liska <mliska@suse.cz>
8675 * cif-code.def (VARIADIC_THUNK): New enum value.
8676 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8679 2018-01-03 Jan Beulich <jbeulich@suse.com>
8681 * sse.md (mov<mode>_internal): Tighten condition for when to use
8682 vmovdqu<ssescalarsize> for TI and OI modes.
8684 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8686 Update copyright years.
8688 2018-01-03 Martin Liska <mliska@suse.cz>
8691 * ipa-visibility.c (function_and_variable_visibility): Skip
8692 functions with noipa attribure.
8694 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8696 * gcc.c (process_command): Update copyright notice dates.
8697 * gcov-dump.c (print_version): Ditto.
8698 * gcov.c (print_version): Ditto.
8699 * gcov-tool.c (print_version): Ditto.
8700 * gengtype.c (create_file): Ditto.
8701 * doc/cpp.texi: Bump @copying's copyright year.
8702 * doc/cppinternals.texi: Ditto.
8703 * doc/gcc.texi: Ditto.
8704 * doc/gccint.texi: Ditto.
8705 * doc/gcov.texi: Ditto.
8706 * doc/install.texi: Ditto.
8707 * doc/invoke.texi: Ditto.
8709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8711 * vector-builder.h (vector_builder::m_full_nelts): Change from
8712 unsigned int to poly_uint64.
8713 (vector_builder::full_nelts): Update prototype accordingly.
8714 (vector_builder::new_vector): Likewise.
8715 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8716 (vector_builder::operator ==): Likewise.
8717 (vector_builder::finalize): Likewise.
8718 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8719 Take the number of elements as a poly_uint64 rather than an
8721 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8722 from unsigned int to poly_uint64.
8723 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8724 (vec_perm_indices::new_vector): Likewise.
8725 (vec_perm_indices::length): Likewise.
8726 (vec_perm_indices::nelts_per_input): Likewise.
8727 (vec_perm_indices::input_nelts): Likewise.
8728 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8729 number of elements per input as a poly_uint64 rather than an
8730 unsigned int. Use the original encoding for variable-length
8731 vectors, rather than clamping each individual element.
8732 For the second and subsequent elements in each pattern,
8733 clamp the step and base before clamping their sum.
8734 (vec_perm_indices::series_p): Handle polynomial element counts.
8735 (vec_perm_indices::all_in_range_p): Likewise.
8736 (vec_perm_indices_to_tree): Likewise.
8737 (vec_perm_indices_to_rtx): Likewise.
8738 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8739 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8740 (tree_vector_builder::new_binary_operation): Handle polynomial
8741 element counts. Return false if we need to know the number
8742 of elements at compile time.
8743 * fold-const.c (fold_vec_perm): Punt if the number of elements
8744 isn't known at compile time.
8746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8748 * vec-perm-indices.h (vec_perm_builder): Change element type
8749 from HOST_WIDE_INT to poly_int64.
8750 (vec_perm_indices::element_type): Update accordingly.
8751 (vec_perm_indices::clamp): Handle polynomial element_types.
8752 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8753 (vec_perm_indices::all_in_range_p): Likewise.
8754 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8756 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8757 polynomial vec_perm_indices element types.
8758 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8759 * fold-const.c (fold_vec_perm): Likewise.
8760 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8761 * tree-vect-generic.c (lower_vec_perm): Likewise.
8762 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8763 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8764 element type to HOST_WIDE_INT.
8766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8767 Alan Hayward <alan.hayward@arm.com>
8768 David Sherwood <david.sherwood@arm.com>
8770 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8771 rather than an int. Use plus_constant.
8772 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8773 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8776 Alan Hayward <alan.hayward@arm.com>
8777 David Sherwood <david.sherwood@arm.com>
8779 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8780 a HOST_WIDE_INT to a poly_int64.
8782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8783 Alan Hayward <alan.hayward@arm.com>
8784 David Sherwood <david.sherwood@arm.com>
8786 * calls.c (load_register_parameters): Cope with polynomial
8787 mode sizes. Require a constant size for BLKmode parameters
8788 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8789 forces a parameter to be padded at the lsb end in order to
8790 fill a complete number of words, require the parameter size
8791 to be ordered wrt UNITS_PER_WORD.
8793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8794 Alan Hayward <alan.hayward@arm.com>
8795 David Sherwood <david.sherwood@arm.com>
8797 * reload1.c (spill_stack_slot_width): Change element type
8798 from unsigned int to poly_uint64_pod.
8799 (alter_reg): Treat mode sizes as polynomial.
8801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8802 Alan Hayward <alan.hayward@arm.com>
8803 David Sherwood <david.sherwood@arm.com>
8805 * reload.c (complex_word_subreg_p): New function.
8806 (reload_inner_reg_of_subreg, push_reload): Use it.
8808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8809 Alan Hayward <alan.hayward@arm.com>
8810 David Sherwood <david.sherwood@arm.com>
8812 * lra-constraints.c (process_alt_operands): Reject matched
8813 operands whose sizes aren't ordered.
8814 (match_reload): Refer to this check here.
8816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8817 Alan Hayward <alan.hayward@arm.com>
8818 David Sherwood <david.sherwood@arm.com>
8820 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8821 that the mode size is in the set {1, 2, 4, 8, 16}.
8823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8824 Alan Hayward <alan.hayward@arm.com>
8825 David Sherwood <david.sherwood@arm.com>
8827 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8828 Use plus_constant instead of gen_rtx_PLUS.
8830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8831 Alan Hayward <alan.hayward@arm.com>
8832 David Sherwood <david.sherwood@arm.com>
8834 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8835 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8836 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8837 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8838 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8839 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8840 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8841 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8842 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8843 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8845 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8846 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8847 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8848 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8849 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8850 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8851 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8852 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8853 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8854 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8856 * expr.c (emit_move_resolve_push): Treat the input and result
8857 of PUSH_ROUNDING as a poly_int64.
8858 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8859 (emit_push_insn): Likewise.
8860 * lra-eliminations.c (mark_not_eliminable): Likewise.
8861 * recog.c (push_operand): Likewise.
8862 * reload1.c (elimination_effects): Likewise.
8863 * rtlanal.c (nonzero_bits1): Likewise.
8864 * calls.c (store_one_arg): Likewise. Require the padding to be
8865 known at compile time.
8867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8868 Alan Hayward <alan.hayward@arm.com>
8869 David Sherwood <david.sherwood@arm.com>
8871 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8872 Use plus_constant instead of gen_rtx_PLUS.
8874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8875 Alan Hayward <alan.hayward@arm.com>
8876 David Sherwood <david.sherwood@arm.com>
8878 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8882 Alan Hayward <alan.hayward@arm.com>
8883 David Sherwood <david.sherwood@arm.com>
8885 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8886 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8887 via stack temporaries. Treat the mode size as polynomial too.
8889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8890 Alan Hayward <alan.hayward@arm.com>
8891 David Sherwood <david.sherwood@arm.com>
8893 * expr.c (expand_expr_real_2): When handling conversions involving
8894 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8895 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8896 as a poly_uint64 too.
8898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8899 Alan Hayward <alan.hayward@arm.com>
8900 David Sherwood <david.sherwood@arm.com>
8902 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8905 Alan Hayward <alan.hayward@arm.com>
8906 David Sherwood <david.sherwood@arm.com>
8908 * combine.c (can_change_dest_mode): Handle polynomial
8909 REGMODE_NATURAL_SIZE.
8910 * expmed.c (store_bit_field_1): Likewise.
8911 * expr.c (store_constructor): Likewise.
8912 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8913 and polynomial REGMODE_NATURAL_SIZE.
8914 (gen_lowpart_common): Likewise.
8915 * reginfo.c (record_subregs_of_mode): Likewise.
8916 * rtlanal.c (read_modify_subreg_p): Likewise.
8918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8919 Alan Hayward <alan.hayward@arm.com>
8920 David Sherwood <david.sherwood@arm.com>
8922 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8923 numbers of elements.
8925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8926 Alan Hayward <alan.hayward@arm.com>
8927 David Sherwood <david.sherwood@arm.com>
8929 * match.pd: Cope with polynomial numbers of vector elements.
8931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8932 Alan Hayward <alan.hayward@arm.com>
8933 David Sherwood <david.sherwood@arm.com>
8935 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8936 in a POINTER_PLUS_EXPR.
8938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8939 Alan Hayward <alan.hayward@arm.com>
8940 David Sherwood <david.sherwood@arm.com>
8942 * omp-simd-clone.c (simd_clone_subparts): New function.
8943 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8944 (ipa_simd_modify_function_body): Likewise.
8946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8947 Alan Hayward <alan.hayward@arm.com>
8948 David Sherwood <david.sherwood@arm.com>
8950 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8951 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8952 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8953 (expand_vector_condition, vector_element): Likewise.
8954 (subparts_gt): New function.
8955 (get_compute_type): Use subparts_gt.
8956 (count_type_subparts): Delete.
8957 (expand_vector_operations_1): Use subparts_gt instead of
8958 count_type_subparts.
8960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8961 Alan Hayward <alan.hayward@arm.com>
8962 David Sherwood <david.sherwood@arm.com>
8964 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8965 (vect_compile_time_alias): ...this new function. Do the calculation
8966 on poly_ints rather than trees.
8967 (vect_prune_runtime_alias_test_list): Update call accordingly.
8969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8970 Alan Hayward <alan.hayward@arm.com>
8971 David Sherwood <david.sherwood@arm.com>
8973 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8975 (vect_schedule_slp_instance): Likewise.
8977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8978 Alan Hayward <alan.hayward@arm.com>
8979 David Sherwood <david.sherwood@arm.com>
8981 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8982 constant and extern definitions for variable-length vectors.
8983 (vect_get_constant_vectors): Note that the number of units
8984 is known to be constant.
8986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8987 Alan Hayward <alan.hayward@arm.com>
8988 David Sherwood <david.sherwood@arm.com>
8990 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8991 of units as polynomial. Choose between WIDE and NARROW based
8994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8995 Alan Hayward <alan.hayward@arm.com>
8996 David Sherwood <david.sherwood@arm.com>
8998 * tree-vect-stmts.c (simd_clone_subparts): New function.
8999 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9002 Alan Hayward <alan.hayward@arm.com>
9003 David Sherwood <david.sherwood@arm.com>
9005 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9006 vectors as polynomial. Use build_index_vector for
9009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9014 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9015 for variable-length vectors.
9016 (vectorizable_mask_load_store): Treat the number of units as
9017 polynomial, asserting that it is constant if the condition has
9018 already been enforced.
9019 (vectorizable_store, vectorizable_load): Likewise.
9021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9022 Alan Hayward <alan.hayward@arm.com>
9023 David Sherwood <david.sherwood@arm.com>
9025 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9026 of units as polynomial. Punt if we can't tell at compile time
9027 which vector contains the final result.
9029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9030 Alan Hayward <alan.hayward@arm.com>
9031 David Sherwood <david.sherwood@arm.com>
9033 * tree-vect-loop.c (vectorizable_induction): Treat the number
9034 of units as polynomial. Punt on SLP inductions. Use an integer
9035 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9036 cast of such a series for variable-length floating-point
9039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9040 Alan Hayward <alan.hayward@arm.com>
9041 David Sherwood <david.sherwood@arm.com>
9043 * tree.h (build_index_vector): Declare.
9044 * tree.c (build_index_vector): New function.
9045 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9046 of units as polynomial, forcibly converting it to a constant if
9047 vectorizable_reduction has already enforced the condition.
9048 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9049 to create a {1,2,3,...} vector.
9050 (vectorizable_reduction): Treat the number of units as polynomial.
9051 Choose vectype_in based on the largest scalar element size rather
9052 than the smallest number of units. Enforce the restrictions
9055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9056 Alan Hayward <alan.hayward@arm.com>
9057 David Sherwood <david.sherwood@arm.com>
9059 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9060 number of units as polynomial.
9062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9063 Alan Hayward <alan.hayward@arm.com>
9064 David Sherwood <david.sherwood@arm.com>
9066 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9067 * target.def (autovectorize_vector_sizes): Return the vector sizes
9068 by pointer, using vector_sizes rather than a bitmask.
9069 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9070 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9071 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9073 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9074 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9075 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9076 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9077 * omp-general.c (omp_max_vf): Likewise.
9078 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9079 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9080 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9081 * tree-vect-slp.c (vect_slp_bb): Likewise.
9082 * doc/tm.texi: Regenerate.
9083 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9085 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9086 the vector size as a poly_uint64 rather than an unsigned int.
9087 (current_vector_size): Change from an unsigned int to a poly_uint64.
9088 (get_vectype_for_scalar_type): Update accordingly.
9089 * tree.h (build_truth_vector_type): Take the size and number of
9090 units as a poly_uint64 rather than an unsigned int.
9091 (build_vector_type): Add a temporary overload that takes
9092 the number of units as a poly_uint64 rather than an unsigned int.
9093 * tree.c (make_vector_type): Likewise.
9094 (build_truth_vector_type): Take the number of units as a poly_uint64
9095 rather than an unsigned int.
9097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9098 Alan Hayward <alan.hayward@arm.com>
9099 David Sherwood <david.sherwood@arm.com>
9101 * target.def (get_mask_mode): Take the number of units and length
9102 as poly_uint64s rather than unsigned ints.
9103 * targhooks.h (default_get_mask_mode): Update accordingly.
9104 * targhooks.c (default_get_mask_mode): Likewise.
9105 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9106 * doc/tm.texi: Regenerate.
9108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9109 Alan Hayward <alan.hayward@arm.com>
9110 David Sherwood <david.sherwood@arm.com>
9112 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9113 * omp-general.c (omp_max_vf): Likewise.
9114 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9115 (expand_omp_simd): Handle polynomial safelen.
9116 * omp-low.c (omplow_simd_context): Add a default constructor.
9117 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9118 (lower_rec_simd_input_clauses): Update accordingly.
9119 (lower_rec_input_clauses): Likewise.
9121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9122 Alan Hayward <alan.hayward@arm.com>
9123 David Sherwood <david.sherwood@arm.com>
9125 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9126 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9127 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9128 (vect_analyze_slp_cost): Likewise.
9129 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9130 (vect_model_load_cost): Likewise.
9132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9133 Alan Hayward <alan.hayward@arm.com>
9134 David Sherwood <david.sherwood@arm.com>
9136 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9137 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9138 from an unsigned int * to a poly_uint64_pod *.
9139 (calculate_unrolling_factor): New function.
9140 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9143 Alan Hayward <alan.hayward@arm.com>
9144 David Sherwood <david.sherwood@arm.com>
9146 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9147 from an unsigned int to a poly_uint64.
9148 (_loop_vec_info::slp_unrolling_factor): Likewise.
9149 (_loop_vec_info::vectorization_factor): Change from an int
9151 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9152 (vect_get_num_vectors): New function.
9153 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9154 (vect_get_num_copies): Use vect_get_num_vectors.
9155 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9156 to an unsigned int *.
9157 (vect_analyze_data_refs): Change min_vf from an int * to a
9159 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9160 than an unsigned HOST_WIDE_INT.
9161 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9162 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9163 to an unsigned int *.
9164 (vect_analyze_data_ref_dependences): Likewise.
9165 (vect_compute_data_ref_alignment): Handle polynomial vf.
9166 (vect_enhance_data_refs_alignment): Likewise.
9167 (vect_prune_runtime_alias_test_list): Likewise.
9168 (vect_shift_permute_load_chain): Likewise.
9169 (vect_supportable_dr_alignment): Likewise.
9170 (dependence_distance_ge_vf): Take the vectorization factor as a
9171 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9172 (vect_analyze_data_refs): Change min_vf from an int * to a
9174 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9175 vfm1 as a poly_uint64 rather than an int. Make the same change
9176 for the returned bound_scalar.
9177 (vect_gen_vector_loop_niters): Handle polynomial vf.
9178 (vect_do_peeling): Likewise. Update call to
9179 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9180 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9182 * tree-vect-loop.c (vect_determine_vectorization_factor)
9183 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9184 (vect_get_known_peeling_cost): Likewise.
9185 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9186 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9187 (vect_transform_loop): Likewise. Use the lowest possible VF when
9188 updating the upper bounds of the loop.
9189 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9191 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9192 polynomial unroll factors.
9193 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9194 (vect_make_slp_decision): Likewise.
9195 (vect_supported_load_permutation_p): Likewise, and polynomial
9197 (vect_analyze_slp_cost): Handle polynomial vf.
9198 (vect_slp_analyze_node_operations): Likewise.
9199 (vect_slp_analyze_bb_1): Likewise.
9200 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9201 than an unsigned HOST_WIDE_INT.
9202 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9203 (vectorizable_load): Handle polynomial vf.
9204 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9206 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9209 Alan Hayward <alan.hayward@arm.com>
9210 David Sherwood <david.sherwood@arm.com>
9212 * match.pd: Handle bit operations involving three constants
9213 and try to fold one pair.
9215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9217 * tree-vect-loop-manip.c: Include gimple-fold.h.
9218 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9219 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9220 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9221 Add a path that uses a step of VF instead of 1, but disable it
9223 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9224 and niters_no_overflow parameters. Update calls to
9225 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9226 Create a new SSA name if the latter choses to use a ste other
9227 than zero, and return it via niters_vector_mult_vf_var.
9228 * tree-vect-loop.c (vect_transform_loop): Update calls to
9229 vect_do_peeling, vect_gen_vector_loop_niters and
9230 slpeel_make_loop_iterate_ntimes.
9231 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9232 (vect_gen_vector_loop_niters): Update declarations after above changes.
9234 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9236 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9237 128-bit round to integer instructions.
9238 (ceil<mode>2): Likewise.
9239 (btrunc<mode>2): Likewise.
9240 (round<mode>2): Likewise.
9242 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9244 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9245 unaligned VSX load/store on P8/P9.
9246 (expand_block_clear): Allow the use of unaligned VSX
9247 load/store on P8/P9.
9249 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9251 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9253 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9254 swap associated with both a load and a store.
9256 2018-01-02 Andrew Waterman <andrew@sifive.com>
9258 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9259 * config/riscv/riscv.md (clear_cache): Use it.
9261 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9263 * web.c: Remove out-of-date comment.
9265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9267 * expr.c (fixup_args_size_notes): Check that any existing
9268 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9269 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9270 (emit_single_push_insn): ...here.
9272 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9274 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9275 (const_vector_encoded_nelts): New function.
9276 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9277 (const_vector_int_elt, const_vector_elt): Declare.
9278 * emit-rtl.c (const_vector_int_elt_1): New function.
9279 (const_vector_elt): Likewise.
9280 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9281 of CONST_VECTOR_ELT.
9283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9285 * expr.c: Include rtx-vector-builder.h.
9286 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9287 directly on the tree encoding.
9288 (const_vector_from_tree): Likewise.
9289 * optabs.c: Include rtx-vector-builder.h.
9290 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9291 sequence of "u" values.
9292 * vec-perm-indices.c: Include rtx-vector-builder.h.
9293 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9294 directly on the vec_perm_indices encoding.
9296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9298 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9299 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9300 * rtx-vector-builder.h: New file.
9301 * rtx-vector-builder.c: Likewise.
9302 * rtl.h (rtx_def::u2): Add a const_vector field.
9303 (CONST_VECTOR_NPATTERNS): New macro.
9304 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9305 (CONST_VECTOR_DUPLICATE_P): Likewise.
9306 (CONST_VECTOR_STEPPED_P): Likewise.
9307 (CONST_VECTOR_ENCODED_ELT): Likewise.
9308 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9309 (unwrap_const_vec_duplicate): Likewise.
9310 (const_vec_series_p): Check for a non-duplicated vector encoding.
9311 Say that the function only returns true for integer vectors.
9312 * emit-rtl.c: Include rtx-vector-builder.h.
9313 (gen_const_vec_duplicate_1): Delete.
9314 (gen_const_vector): Call gen_const_vec_duplicate instead of
9315 gen_const_vec_duplicate_1.
9316 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9317 (gen_const_vec_duplicate): Use rtx_vector_builder.
9318 (gen_const_vec_series): Likewise.
9319 (gen_rtx_CONST_VECTOR): Likewise.
9320 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9321 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9322 Build a new vector rather than modifying a CONST_VECTOR in-place.
9323 (handle_special_swappables): Update call accordingly.
9324 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9325 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9326 Build a new vector rather than modifying a CONST_VECTOR in-place.
9327 (handle_special_swappables): Update call accordingly.
9329 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9331 * simplify-rtx.c (simplify_const_binary_operation): Use
9332 CONST_VECTOR_ELT instead of XVECEXP.
9334 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9336 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9337 the selector elements to be different from the data elements
9338 if the selector is a VECTOR_CST.
9339 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9340 ssizetype for the selector.
9342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9344 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9345 before testing each element individually.
9346 * tree-vect-generic.c (lower_vec_perm): Likewise.
9348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9350 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9351 * selftest-run-tests.c (selftest::run_tests): Call it.
9352 * vector-builder.h (vector_builder::operator ==): New function.
9353 (vector_builder::operator !=): Likewise.
9354 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9355 (vec_perm_indices::all_from_input_p): New function.
9356 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9357 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9358 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9359 instead of reading the VECTOR_CST directly. Detect whether both
9360 vector inputs are the same before constructing the vec_perm_indices,
9361 and update the number of inputs argument accordingly. Use the
9362 utility functions added above. Only construct sel2 if we need to.
9364 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9366 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9367 the broadcast of the low byte.
9368 (expand_mult_highpart): Use an explicit encoding for the permutes.
9369 * optabs-query.c (can_mult_highpart_p): Likewise.
9370 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9371 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9372 (vectorizable_bswap): Likewise.
9373 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9374 explicit encoding for the power-of-2 permutes.
9375 (vect_permute_store_chain): Likewise.
9376 (vect_grouped_load_supported): Likewise.
9377 (vect_permute_load_chain): Likewise.
9379 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9381 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9382 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9383 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9384 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9385 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9386 (vect_gen_perm_mask_any): Likewise.
9388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9390 * int-vector-builder.h: New file.
9391 * vec-perm-indices.h: Include int-vector-builder.h.
9392 (vec_perm_indices): Redefine as an int_vector_builder.
9393 (auto_vec_perm_indices): Delete.
9394 (vec_perm_builder): Redefine as a stand-alone class.
9395 (vec_perm_indices::vec_perm_indices): New function.
9396 (vec_perm_indices::clamp): Likewise.
9397 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9398 (vec_perm_indices::new_vector): New function.
9399 (vec_perm_indices::new_expanded_vector): Update for new
9400 vec_perm_indices class.
9401 (vec_perm_indices::rotate_inputs): New function.
9402 (vec_perm_indices::all_in_range_p): Operate directly on the
9403 encoded form, without computing elided elements.
9404 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9405 encoding. Update for new vec_perm_indices class.
9406 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9407 the given vec_perm_builder.
9408 (expand_vec_perm_var): Update vec_perm_builder constructor.
9409 (expand_mult_highpart): Use vec_perm_builder instead of
9410 auto_vec_perm_indices.
9411 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9412 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9413 or double series encoding as appropriate.
9414 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9415 vec_perm_indices instead of auto_vec_perm_indices.
9416 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9417 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9418 (vect_permute_store_chain): Likewise.
9419 (vect_grouped_load_supported): Likewise.
9420 (vect_permute_load_chain): Likewise.
9421 (vect_shift_permute_load_chain): Likewise.
9422 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9423 (vect_transform_slp_perm_load): Likewise.
9424 (vect_schedule_slp_instance): Likewise.
9425 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9426 (vectorizable_mask_load_store): Likewise.
9427 (vectorizable_bswap): Likewise.
9428 (vectorizable_store): Likewise.
9429 (vectorizable_load): Likewise.
9430 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9431 vec_perm_indices instead of auto_vec_perm_indices. Use
9432 tree_to_vec_perm_builder to read the vector from a tree.
9433 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9434 vec_perm_builder instead of a vec_perm_indices.
9435 (have_whole_vector_shift): Use vec_perm_builder and
9436 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9437 truncation to calc_vec_perm_mask_for_shift.
9438 (vect_create_epilog_for_reduction): Likewise.
9439 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9440 from auto_vec_perm_indices to vec_perm_indices.
9441 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9442 instead of changing individual elements.
9443 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9444 the vector in d.perm.
9445 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9446 from auto_vec_perm_indices to vec_perm_indices.
9447 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9448 instead of changing individual elements.
9449 (arm_vectorize_vec_perm_const): Use new_vector to install
9450 the vector in d.perm.
9451 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9452 Update vec_perm_builder constructor.
9453 (rs6000_expand_interleave): Likewise.
9454 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9455 (rs6000_expand_interleave): Likewise.
9457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9459 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9460 to qimode could truncate the indices.
9461 * optabs.c (expand_vec_perm_var): Likewise.
9463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9465 * Makefile.in (OBJS): Add vec-perm-indices.o.
9466 * vec-perm-indices.h: New file.
9467 * vec-perm-indices.c: Likewise.
9468 * target.h (vec_perm_indices): Replace with a forward class
9470 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9471 * optabs.h: Include vec-perm-indices.h.
9472 (expand_vec_perm): Delete.
9473 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9474 (expand_vec_perm_const): Declare.
9475 * target.def (vec_perm_const_ok): Replace with...
9476 (vec_perm_const): ...this new hook.
9477 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9478 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9479 * doc/tm.texi: Regenerate.
9480 * optabs.def (vec_perm_const): Delete.
9481 * doc/md.texi (vec_perm_const): Likewise.
9482 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9483 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9484 expand_vec_perm for constant permutation vectors. Assert that
9485 the mode of variable permutation vectors is the integer equivalent
9486 of the mode that is being permuted.
9487 * optabs-query.h (selector_fits_mode_p): Declare.
9488 * optabs-query.c: Include vec-perm-indices.h.
9489 (selector_fits_mode_p): New function.
9490 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9491 is defined, instead of checking whether the vec_perm_const_optab
9492 exists. Use targetm.vectorize.vec_perm_const instead of
9493 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9494 fit in the vector mode before using a variable permute.
9495 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9496 vec_perm_indices instead of an rtx.
9497 (expand_vec_perm): Replace with...
9498 (expand_vec_perm_const): ...this new function. Take the selector
9499 as a vec_perm_indices rather than an rtx. Also take the mode of
9500 the selector. Update call to shift_amt_for_vec_perm_mask.
9501 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9502 Use vec_perm_indices::new_expanded_vector to expand the original
9503 selector into bytes. Check whether the indices fit in the vector
9504 mode before using a variable permute.
9505 (expand_vec_perm_var): Make global.
9506 (expand_mult_highpart): Use expand_vec_perm_const.
9507 * fold-const.c: Includes vec-perm-indices.h.
9508 * tree-ssa-forwprop.c: Likewise.
9509 * tree-vect-data-refs.c: Likewise.
9510 * tree-vect-generic.c: Likewise.
9511 * tree-vect-loop.c: Likewise.
9512 * tree-vect-slp.c: Likewise.
9513 * tree-vect-stmts.c: Likewise.
9514 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9516 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9517 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9518 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9519 (aarch64_vectorize_vec_perm_const): ...this new function.
9520 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9521 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9522 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9523 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9524 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9525 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9526 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9528 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9529 check for NEON modes.
9530 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9531 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9532 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9533 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9535 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9536 the old VEC_PERM_CONST conditions.
9537 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9538 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9539 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9540 (ia64_vectorize_vec_perm_const_ok): Merge into...
9541 (ia64_vectorize_vec_perm_const): ...this new function.
9542 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9543 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9544 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9545 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9546 * config/mips/mips.c (mips_expand_vec_perm_const)
9547 (mips_vectorize_vec_perm_const_ok): Merge into...
9548 (mips_vectorize_vec_perm_const): ...this new function.
9549 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9550 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9551 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9552 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9553 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9554 (rs6000_expand_vec_perm_const): Delete.
9555 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9557 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9558 (altivec_expand_vec_perm_const_le): Take each operand individually.
9559 Operate on constant selectors rather than rtxes.
9560 (altivec_expand_vec_perm_const): Likewise. Update call to
9561 altivec_expand_vec_perm_const_le.
9562 (rs6000_expand_vec_perm_const): Delete.
9563 (rs6000_vectorize_vec_perm_const_ok): Delete.
9564 (rs6000_vectorize_vec_perm_const): New function.
9565 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9566 an element count and rtx array.
9567 (rs6000_expand_extract_even): Update call accordingly.
9568 (rs6000_expand_interleave): Likewise.
9569 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9570 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9571 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9572 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9573 (rs6000_expand_vec_perm_const): Delete.
9574 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9575 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9576 (altivec_expand_vec_perm_const_le): Take each operand individually.
9577 Operate on constant selectors rather than rtxes.
9578 (altivec_expand_vec_perm_const): Likewise. Update call to
9579 altivec_expand_vec_perm_const_le.
9580 (rs6000_expand_vec_perm_const): Delete.
9581 (rs6000_vectorize_vec_perm_const_ok): Delete.
9582 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9583 reference to the SPE evmerge intructions.
9584 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9585 an element count and rtx array.
9586 (rs6000_expand_extract_even): Update call accordingly.
9587 (rs6000_expand_interleave): Likewise.
9588 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9589 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9591 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9595 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9596 vector mode and that that mode matches the mode of the data
9598 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9599 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9600 directly using expand_vec_perm_1 when forcing selectors into
9602 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9606 * optabs-query.h (can_vec_perm_p): Delete.
9607 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9608 * optabs-query.c (can_vec_perm_p): Split into...
9609 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9610 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9611 particular selector is valid.
9612 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9613 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9614 (vect_grouped_load_supported): Likewise.
9615 (vect_shift_permute_load_chain): Likewise.
9616 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9617 (vect_transform_slp_perm_load): Likewise.
9618 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9619 (vectorizable_bswap): Likewise.
9620 (vect_gen_perm_mask_checked): Likewise.
9621 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9622 implementations of variable permutation vectors into account
9623 when deciding which selector to use.
9624 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9625 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9626 with a false third argument.
9627 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9628 to test whether the constant selector is valid and can_vec_perm_var_p
9629 to test whether a variable selector is valid.
9631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9633 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9634 * optabs-query.c (can_vec_perm_p): Likewise.
9635 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9636 instead of vec_perm_indices.
9637 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9638 (vect_gen_perm_mask_checked): Likewise,
9639 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9640 (vect_gen_perm_mask_checked): Likewise,
9642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9644 * optabs-query.h (qimode_for_vec_perm): Declare.
9645 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9646 (qimode_for_vec_perm): ...this new function.
9647 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9649 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9651 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9652 does not have a conditional at the top.
9654 2018-01-02 Richard Biener <rguenther@suse.de>
9656 * ipa-inline.c (big_speedup_p): Fix expression.
9658 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9661 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9664 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9668 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9669 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9670 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9671 cond_taken_branch_cost 3->4.
9673 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9675 PR tree-optimization/83581
9676 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9677 TODO_cleanup_cfg if any changes have been made.
9680 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9681 convert_modes if target mode has the right side, but different mode
9685 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9686 last argument when extracting from CONCAT. If either from_real or
9687 from_imag is NULL, use expansion through memory. If result is not
9688 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9689 the parts directly to inner mode, if even that fails, use expansion
9693 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9694 check for bswap in mode rather than HImode and use that in expand_unop
9697 Copyright (C) 2018 Free Software Foundation, Inc.
9699 Copying and distribu