1 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
15 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
17 for RS6000_BTM_POWERPC64.
18 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
19 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
25 2018-03-28 Carl Love <cel@us.ibm.com>
28 2017-09-27 Carl Love <cel@us.ibm.com>
30 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
31 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
32 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
35 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
37 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
38 instead of __vector bool.
39 (_mm_max_pu8): Likewise.
40 (_mm_min_pi16): Likewise.
42 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
45 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
49 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
51 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
52 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
53 (div_extend): Likewise.
54 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
56 (__builtin_divweuo): Likewise.
57 (__builtin_divdeo): Likewise.
58 (__builtin_divdeuo): Likewise.
60 2018-03-28 Jakub Jelinek <jakub@redhat.com>
63 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
64 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
66 PR tree-optimization/82004
67 * gimple-match-head.c (optimize_pow_to_exp): New function.
68 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
69 Don't fold to exp if optimize_pow_to_exp is false.
71 2018-03-28 Martin Liska <mliska@suse.cz>
74 * calls.c (initialize_argument_information): Fix trailing space.
75 * common.opt: Fix typo and provide better explanation for
76 -fsanitize-coverage option.
77 * config/i386/i386.opt: Fix typo.
79 2018-03-28 Jakub Jelinek <jakub@redhat.com>
80 Martin Liska <mliska@suse.cz>
83 * gimplify.c (asan_poison_variable): Don't do the check for
84 gimplify_omp_ctxp here.
85 (gimplify_decl_expr): Do it here.
86 (gimplify_target_expr): Likewise.
88 2018-03-28 Martin Liska <mliska@suse.cz>
91 * config/i386/i386.c (ix86_function_arg_advance): Do not call
92 chkp_type_bounds_count if MPX is not enabled.
94 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
96 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
98 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
101 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
102 function to create the function decl for complex long double
103 multiply and divide for -mabi=ieeelongdouble.
104 (init_float128_ieee): Call it.
106 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
109 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
110 -fcf-protection=branch -mibt.
111 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
113 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
116 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
118 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
121 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
122 extern array declarations.
124 2018-03-27 Richard Biener <rguenther@suse.de>
127 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
128 explicit single_use checks.
130 2018-03-27 Richard Biener <rguenther@suse.de>
132 PR tree-optimization/85082
133 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
136 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
138 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
139 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
140 Turn on fasynchronous-unwind-tables and funwind-tables.
142 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
145 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
146 (*bmi_blsr_<mode>_ccz): Ditto.
148 2018-03-26 Tom de Vries <tom@codesourcery.com>
150 PR tree-optimization/85063
151 * omp-general.c (offloading_function_p): New function. Factor out
153 * omp-offload.c (pass_omp_target_link::gate): ... here.
154 * omp-general.h (offloading_function_p): Declare.
155 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
156 with attribute omp declare target for offloading functions.
158 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
160 PR tree-optimization/84005
161 * tree-data-ref.h (get_base_for_alignment): Declare.
162 * tree-data-ref.c (get_base_for_alignment_1): New function.
163 (get_base_for_alignment): Likewise.
164 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
165 get_base_for_alignment to find a suitable base object, instead
166 of always using drb->base_address.
168 2018-03-23 Jakub Jelinek <jakub@redhat.com>
171 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
172 known size by default.
174 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
177 * lra-constraints.c (process_alt_operands): Don't match BLKmode
178 and non BLKmode operands.
180 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
183 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
186 2018-03-23 Richard Biener <rguenther@suse.de>
189 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
190 we are going to emit early debug for LTO.
192 2018-03-23 Jakub Jelinek <jakub@redhat.com>
195 * function.c (match_asm_constraints_1): Don't optimize if input
196 doesn't satisfy general_operand predicate for output's mode.
199 * alias.c (write_dependence_p): Don't require for x_canonicalized
200 non-VOIDmode if x has VOIDmode.
203 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
204 just don't try to optimize it rather than assert it never happens.
206 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
208 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
209 macro expansions for definition of ST_INTERNAL_<mode> and
210 LD_INTERNAL_<mode> builtins.
211 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
213 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
215 (altivec_expand_st_builtin): Likewise.
216 (altivec_expand_builtin): Remove calls to deleted functions.
217 (rs6000_address_for_altivec): Delete this function.
218 * config/rs6000/vector.md: Remove expands for
219 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
221 2018-03-22 Sudakshina Das <sudi.das@arm.com>
224 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
225 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
226 re-computing once computed.
227 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
228 (arm_init_machine_status): Initialize
229 machine->static_chain_stack_bytes.
231 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
234 * doc/extend.texi: Add four new prototypes for vec_ld.
235 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
236 definitions for more logical presentation.
237 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
238 entries for V1TI variants of __builtin_altivec_ld builtin.
239 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
240 handling of V1TI variant of LVX icode pattern.
241 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
242 (rs6000_gimple_fold_builtin): Likewise.
243 (altivec_init_builtins): Add code to define
244 __builtin_altivec_lvx_v1ti function.
246 2018-03-22 Jakub Jelinek <jakub@redhat.com>
249 * function.c (match_asm_constraints_1): Don't do the optimization
250 if input isn't a REG, SUBREG, MEM or constant.
252 2018-03-22 Tom de Vries <tom@codesourcery.com>
254 PR tree-optimization/84956
255 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
256 bb_has_abnormal_pred.
258 2018-03-22 Jakub Jelinek <jakub@redhat.com>
261 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
262 DECL_INITIAL (decl) to decl at the end.
263 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
266 2018-03-21 Joseph Myers <joseph@codesourcery.com>
268 * doc/extend.texi (__builtin_tgmath): Document when complex
269 integer types are treated as _Complex _Float64.
271 2018-03-21 Tom de Vries <tom@codesourcery.com>
273 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
275 2018-03-21 Jakub Jelinek <jakub@redhat.com>
277 PR tree-optimization/84960
278 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
279 if it is ENTRY block, move them into single succ of ENTRY in that case.
281 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
283 PR tree-optimization/84811
284 * poly-int.h (poly_span_traits): Remove the T3 parameter and
285 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
286 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
287 (known_subrange_p): Update accordingly. Cast each value involved
288 in the size comparison, rather than casting the result of the
291 2018-03-21 Jakub Jelinek <jakub@redhat.com>
293 PR tree-optimization/84982
294 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
295 by flipping the least significant bit rather than all bits from
296 bitpos to bitpos + bitsize - 1.
298 2018-03-21 Nathan Sidwell <nathan@acm.org>
300 * doc/extend.texi (Deprecated Features): Remove mention of
301 long-deleted deprecations.
303 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
306 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
307 * configure: Regenerate.
309 2018-03-21 Tom de Vries <tom@codesourcery.com>
311 PR tree-optimization/83126
312 * tree-parloops.c (num_phis): New function.
313 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
315 2018-03-21 Nathan Sidwell <nathan@acm.org>
317 * doc/extend.texi (Deprecated Features): Update deprecated flags,
318 mention anon-struct/union members and trailing attributes.
320 2018-03-21 Bin Cheng <bin.cheng@arm.com>
322 PR tree-optimization/84969
323 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
324 builtin memset partitions if they set different rhs values.
326 2018-03-21 Jakub Jelinek <jakub@redhat.com>
328 PR rtl-optimization/84989
329 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
330 VEC_DUPLICATE with scalar result mode.
332 2018-03-21 Martin Liska <mliska@suse.cz>
335 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
336 not intended return statement.
338 2018-03-21 Martin Liska <mliska@suse.cz>
341 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
342 (chkp_find_bound_slots_1): Limit number of iterations.
344 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
347 * Minor grammar fixes for x86 options.
349 2018-03-20 Jakub Jelinek <jakub@redhat.com>
352 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
353 holding REG_CFA_RESTORE notes, instead turn them into a USE.
355 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
358 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
359 (altivec_lvx_<mode>_1op): Likewise.
360 (altivec_stvx_<mode>_2op): Likewise.
361 (altivec_stvx_<mode>_1op): Likewise.
362 (altivec_lvx_<VM2:mode>): New define_expand.
363 (altivec_stvx_<VM2:mode>): Likewise.
364 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
365 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
366 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
367 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
368 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
369 (rs6000_gen_lvx): Likewise.
370 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
371 (altivec_expand_stv_builtin): Likewise.
372 (altivec_expand_builtin): Likewise.
373 * config/rs6000/vector.md: Likewise.
375 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
378 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
381 2018-03-20 Richard Biener <rguenther@suse.de>
384 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
385 sign-conversions as zero, fall back to standard scalar_stmt
388 2018-03-20 Martin Liska <mliska@suse.cz>
391 * predict.c (rebuild_frequencies): Handle case when we have
392 PROFILE_ABSENT, but flag_guess_branch_prob is false.
394 2018-03-20 Jakub Jelinek <jakub@redhat.com>
397 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
398 flag_section_anchors.
399 * varasm.c (use_blocks_for_decl_p): Remove hack for
403 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
405 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
406 be created, use lowpart_subreg of operands[0] rather than operands[0]
408 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
409 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
410 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
411 and n constraint instead of aarch64_shift_imm_di and Usd.
412 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
413 (*aarch64_<optab>_reg_minus<mode>3): ... this.
415 2018-03-20 Sudakshina Das <sudi.das@arm.com>
418 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
419 to favor GPR over NEON registers.
420 (<shift>di3_neon): Likewise.
422 2018-03-20 Tom de Vries <tom@codesourcery.com>
425 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
426 (nvptx_process_pars): Emit bar.sync asap and alap.
428 2018-03-20 Tom de Vries <tom@codesourcery.com>
431 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
432 seen_label if seen_label is already set.
434 2018-03-20 Jakub Jelinek <jakub@redhat.com>
437 * config/i386/i386.c (fold_builtin_cpu): For features above 31
438 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
439 Use 1U instead of 1. Formatting fixes.
442 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
443 instead of TREE_TYPE (s1) for the return value.
445 2018-03-19 Jakub Jelinek <jakub@redhat.com>
447 PR tree-optimization/84946
448 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
449 bitsize + bitsize in poly_uint64 rather than poly_int64.
452 * dwarf2asm.c: Include fold-const.c.
453 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
454 of decl rather than decl itself.
456 PR rtl-optimization/84643
457 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
459 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
462 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
463 calling assemble_variable.
465 2018-03-19 Sudakshina Das <sudi.das@arm.com>
468 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
469 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
471 2018-03-19 Jim Wilson <jimw@sifive.com>
474 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
475 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
476 (riscv_first_stack_step): Likewise.
477 (riscv_option_override): Use STACK_BOUNDARY instead of
479 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
481 (BIGGEST_ALIGNMENT): Set to 128.
482 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
483 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
486 2018-03-19 Richard Biener <rguenther@suse.de>
488 PR tree-optimization/84933
489 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
490 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
492 2018-03-19 Richard Biener <rguenther@suse.de>
494 PR tree-optimization/84859
495 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
496 (cond_if_else_store_replacement): Perform sinking operation on
497 single-store BBs regardless of MAX_STORES_TO_SINK setting.
498 Generalize what a BB with a single eligible store is.
500 2018-03-19 Richard Biener <rguenther@suse.de>
502 PR tree-optimization/84929
503 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
504 chrec_is_positive against non-chrec arg.
506 2018-03-19 Tamar Christina <tamar.christina@arm.com>
509 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
511 2018-03-18 Martin Liska <mliska@suse.cz>
513 PR rtl-optimization/84635
514 * regrename.c (build_def_use): Use matches_mode only when
517 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
519 PR tree-optimization/84913
520 * tree-vect-loop.c (vectorizable_reduction): Don't try to
521 vectorize chains of COND_EXPRs.
523 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
525 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
527 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
529 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
531 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
533 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
535 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
536 Kito Cheng <kito.cheng@gmail.com>
538 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
539 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
540 (nds32_adjust_reg_alloc_order): New function.
541 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
543 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
545 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
546 nds32_print_operand, nds32_print_operand_address): Use
547 HOST_WIDE_INT_PRINT_DEC instead.
549 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
551 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
553 2018-03-17 Jakub Jelinek <jakub@redhat.com>
556 * config/i386/i386.c (initial_ix86_tune_features,
557 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
559 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
560 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
561 rather than 1u << ix86_tune. Formatting fix.
562 (ix86_option_override_internal): Change ix86_arch_mask from
563 unsigned int to unsigned HOST_WIDE_INT, initialize to
564 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
565 (ix86_function_specific_restore): Likewise.
567 2018-03-16 Jakub Jelinek <jakub@redhat.com>
570 * postreload.c (reload_combine_recognize_pattern): Perform
571 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
572 truncate_int_for_mode the result for the destination's mode.
575 * hsa-gen.c (mem_type_for_type): Fix comment typo.
576 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
578 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
581 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
584 * lra-assigns.c (lra_split_hard_reg_for): Don't use
585 regno_allocno_class_array and sorted_pseudos.
586 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
587 insns where regno is used.
589 2018-03-16 Martin Liska <mliska@suse.cz>
592 * multiple_target.c (create_dispatcher_calls): Redirect
593 reference in the symbol table.
595 2018-03-16 Martin Liska <mliska@suse.cz>
598 * multiple_target.c (create_dispatcher_calls): Redirect also
601 2018-03-16 Jakub Jelinek <jakub@redhat.com>
605 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
606 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
607 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
609 2018-03-16 Julia Koval <julia.koval@intel.com>
611 * doc/invoke.texi (Skylake Server): Add CLWB.
612 Cannonlake): Remove CLWB.
614 2018-03-16 Jakub Jelinek <jakub@redhat.com>
616 PR tree-optimization/84841
617 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
619 (FLOAT_ONE_CONST_TYPE): Define.
620 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
621 (sort_by_operand_rank): Put entries with higher constant_type last
622 rather than first to match comments.
624 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
626 * config/nios2/nios2.md (movsi_internal): Fix thinko in
629 2018-03-15 Jakub Jelinek <jakub@redhat.com>
632 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
633 check and use address of target always.
635 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
638 * config/i386/i386.c (indirect_thunk_needed): Update comments.
639 (indirect_thunk_bnd_needed): Likewise.
640 (indirect_thunks_used): Likewise.
641 (indirect_thunks_bnd_used): Likewise.
642 (indirect_return_needed): New.
643 (indirect_return_bnd_needed): Likewise.
644 (output_indirect_thunk_function): Add a bool argument for
646 (output_indirect_thunk_function): Don't generate alias for
647 function return thunk.
648 (ix86_code_end): Call output_indirect_thunk_function to generate
649 function return thunks.
650 (ix86_output_function_return): Set indirect_return_bnd_needed
651 and indirect_return_needed instead of indirect_thunk_bnd_needed
652 and indirect_thunk_needed.
654 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
656 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
657 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
658 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
660 2018-03-15 David Malcolm <dmalcolm@redhat.com>
661 Paul Hua <paul.hua.gm@gmail.com>
664 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
666 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
668 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
669 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
672 2018-03-15 Tamar Christina <tamar.christina@arm.com>
675 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
676 instead of GET_MODE_SIZE when comparing Units.
678 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
681 * varasm.c (hash_section): Return an unchangeble hash value
682 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
683 Return !aarch64_can_use_per_function_literal_pools_p ().
685 2018-03-15 Jakub Jelinek <jakub@redhat.com>
688 * optabs.c (emit_conditional_move): Pass address of cmode's copy
689 rather than address of cmode as last argument to prepare_cmp_insn.
691 2018-03-15 Julia Koval <julia.koval@intel.com>
693 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
694 F_AVX512VNNI, F_AVX512BITALG): New.
696 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
699 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
700 insn for floating-point loads and stores.
702 2018-03-14 Carl Love <cel@us.ibm.com>
704 * config/rs6000/rs6000-c.c: Add macro definitions for
705 ALTIVEC_BUILTIN_VEC_PERMXOR.
706 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
707 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
708 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
709 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
711 * config/doc/extend.texi: Add prototypes for vec_permxor.
713 2018-03-14 David Malcolm <dmalcolm@redhat.com>
716 * diagnostic-show-locus.c (class layout_point): Convert m_line
717 from int to linenum_type.
718 (line_span::comparator): Use linenum "compare" function when
719 comparing line numbers.
720 (test_line_span): New function.
721 (layout_range::contains_point): Convert param "row" from int to
723 (layout_range::intersects_line_p): Likewise.
724 (layout::will_show_line_p): Likewise.
725 (layout::print_source_line): Likewise.
726 (layout::should_print_annotation_line_p): Likewise.
727 (layout::print_annotation_line): Likewise.
728 (layout::print_leading_fixits): Likewise.
729 (layout::annotation_line_showed_range_p): Likewise.
730 (struct line_corrections): Likewise for field m_row.
731 (line_corrections::line_corrections): Likewise for param "row".
732 (layout::print_trailing_fixits): Likewise.
733 (layout::get_state_at_point): Likewise.
734 (layout::get_x_bound_for_row): Likewise.
735 (layout::print_line): Likewise.
736 (diagnostic_show_locus): Likewise for locals "last_line" and
738 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
739 * input.c (selftest::test_linenum_comparisons): New function.
740 (selftest::input_c_tests): Call it.
741 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
742 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
743 * selftest.h (ASSERT_GT): New macro.
744 (ASSERT_GT_AT): New macro.
745 (ASSERT_LT): New macro.
746 (ASSERT_LT_AT): New macro.
748 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
750 PR rtl-optimization/84780
751 * combine.c (distribute_links): Don't make a link based on pc_rtx.
753 2018-03-14 Martin Liska <mliska@suse.cz>
755 * tree.c (record_node_allocation_statistics): Use
757 (get_stats_node_kind): New function extracted from
758 record_node_allocation_statistics.
759 (free_node): Use get_stats_node_kind.
761 2018-03-14 Richard Biener <rguenther@suse.de>
763 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
764 that the value-set of ANTIC_IN doesn't grow.
767 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
769 (BB_VISITED_WITH_VISITED_SUCCS): New define.
770 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
772 2018-03-14 Julia Koval <julia.koval@intel.com>
774 * config.gcc (icelake-client, icelake-server): New.
776 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
777 (initial_ix86_arch_features): Ditto.
778 (PTA_SKYLAKE): Add SGX.
779 (PTA_ICELAKE): Remove.
780 (PTA_ICELAKE_CLIENT): New.
781 (PTA_ICELAKE_SERVER): New.
782 (ix86_option_override_internal): Split up icelake on icelake client and
784 (get_builtin_code_for_version): Ditto.
785 (fold_builtin_cpu): Ditto.
786 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
787 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
788 * config/i386/i386.h (processor_type): Ditto.
789 * doc/invoke.texi: Ditto.
791 2018-03-14 Jakub Jelinek <jakub@redhat.com>
794 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
795 INTEGER_CST offset, add it together with bitpos / 8 and
796 sign extend based on POINTER_SIZE.
800 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
803 * config/i386/constraints.md (Yc): New register constraint.
804 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
805 Use Yc constraint for alternative 2 of operand 0. Remove
806 preferred_for_speed attribute.
808 2018-03-14 Richard Biener <rguenther@suse.de>
810 PR tree-optimization/84830
811 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
812 with the old one to avoid oscillations.
814 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
817 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
819 (assign_by_spills): Return a flag of reload assignment failure.
820 Do not process the reload assignment failures. Do not spill other
821 reload pseudos if they has the same reg class. Update n if
823 (lra_assign): Add a return arg. Set up from the result of
824 assign_by_spills call.
825 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
826 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
827 usage_insns if it is not NULL.
828 (spill_hard_reg_in_range): New function.
829 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
830 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
832 (lra_assign): Change prototype.
833 * lra.c (lra): Add code to deal with fails by splitting hard reg
836 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
838 * config/riscv/riscv.opt (mrelax): New option.
839 * config/riscv/riscv.c (riscv_file_start): Emit ".option
840 "norelax" when riscv_mrelax is disabled.
841 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
843 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
846 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
847 reassociation for int modes.
849 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
851 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
852 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
854 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
855 * config/aarch64/aarch64-sve.md
856 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
857 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
858 (*extend<mode><Vwide>2): Rename to...
859 (aarch64_sve_extend<mode><Vwide>2): ...this.
860 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
861 renaming the old pattern to...
862 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
864 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
865 define_expand, renaming the old pattern to...
866 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
867 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
868 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
869 account when deciding which SVE instruction the optab should use.
870 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
872 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
874 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
875 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
876 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
877 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
878 (tlsdesc_small_<mode>): Turn a define_expand and use
879 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
880 (tlsdesc_small_advsimd_<mode>): ...this.
881 (tlsdesc_small_sve_<mode>): New pattern.
883 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
885 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
886 (UNSPEC_UMUL_HIGHPART): New constants.
887 (MUL_HIGHPART): New int iteraor.
888 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
889 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
891 (*<su>mul<mode>3_highpart): New define_insn.
893 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
896 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
899 2018-03-13 Martin Liska <mliska@suse.cz>
902 * (sem_item_optimizer::sem_item_optimizer): Initialize new
904 (sem_item_optimizer::~sem_item_optimizer): Release it.
905 (sem_item_optimizer::merge_classes): Register variable aliases.
906 (sem_item_optimizer::fixup_pt_set): New function.
907 (sem_item_optimizer::fixup_points_to_sets): Likewise.
908 * ipa-icf.h: Declare new variables and functions.
910 2018-03-13 Jakub Jelinek <jakub@redhat.com>
913 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
914 integer_pow2p@2 and test integer_pow2p in condition.
915 (A < 0 ? C : 0): Similarly for @1.
918 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
919 characters starting at p contain '\0' character, don't look beyond
923 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
924 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
927 * reg-stack.c (change_stack): Change update_end var from int to
928 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
929 also call set_block_for_insn on the newly added insns and rescan.
932 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
936 * tree.c (stabilize_reference_1): Return save_expr (e) for
937 STATEMENT_LIST even if it doesn't have side-effects.
939 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
941 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
943 2018-03-12 Renlin Li <renlin.li@arm.com>
945 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
946 aarch64_output_scalar_simd_mov_immediate.
948 2018-03-12 Martin Sebor <msebor@redhat.com>
950 PR tree-optimization/83456
951 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
952 for perfectly overlapping calls to memcpy.
953 (gimple_fold_builtin_memory_chk): Same.
954 (gimple_fold_builtin_strcpy): Handle no-warning.
955 (gimple_fold_builtin_stxcpy_chk): Same.
956 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
958 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
960 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
961 parameter. Use it for SFmode.
962 (rs6000_function_arg_advance_1): Adjust.
963 (rs6000_function_arg): Adjust.
964 (rs6000_gimplify_va_arg): Pass false for that new parameter.
966 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
968 PR rtl-optimization/84169
969 PR rtl-optimization/84780
970 * combine.c (can_combine_p): Check for a 2-insn combination whether
971 the destination register is used between the two insns, too.
973 2018-03-12 Richard Biener <rguenther@suse.de>
975 PR tree-optimization/84803
976 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
977 for refs DR analysis didn't process.
979 2018-03-12 Richard Biener <rguenther@suse.de>
981 PR tree-optimization/84777
982 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
983 force-vectorize loops ignore whether we are optimizing for size.
985 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
987 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
988 (TARGET_MD_ASM_ADJUST): Define.
990 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
991 Kito Cheng <kito.cheng@gmail.com>
992 Chung-Ju Wu <jasonwucj@gmail.com>
994 * config/nds32/nds32.c (nds32_compute_stack_frame,
995 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
996 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
997 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
998 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
999 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1000 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1001 * config/nds32/nds32.md (prologue, epilogue): Use macro
1002 NDS32_V3PUSH_AVAILABLE_P to do checking.
1004 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1007 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1008 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1009 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1010 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1011 addition of most attributes on !orig_type_die or the attribute not
1012 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1014 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1015 Chung-Ju Wu <jasonwucj@gmail.com>
1017 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1019 * config/nds32/nds32.opt (mvh): New option.
1021 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1022 Chung-Ju Wu <jasonwucj@gmail.com>
1024 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1026 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1027 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1030 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1031 Chung-Ju Wu <jasonwucj@gmail.com>
1033 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1035 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1036 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1038 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1039 Kito Cheng <kito.cheng@gmail.com>
1040 Chung-Ju Wu <jasonwucj@gmail.com>
1042 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1043 UNSPEC_FFMISM and UNSPEC_FLMISM.
1044 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1045 for ffb, ffmism and flmism.
1046 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1047 (unspec_ffmism): Ditto.
1048 (unspec_flmism): Ditto.
1049 (nds32_expand_builtin_impl): Check if string extension is available.
1050 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1051 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1053 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1056 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1059 * lra-assigns.c (assign_by_spills): Return a flag of reload
1060 assignment failure. Do not process the reload assignment
1061 failures. Do not spill other reload pseudos if they has the same
1063 (lra_assign): Add a return arg. Set up from the result of
1064 assign_by_spills call.
1065 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1066 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1067 usage_insns if it is not NULL.
1068 (spill_hard_reg_in_range): New function.
1069 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1070 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1071 function prototypes.
1072 (lra_assign): Change prototype.
1073 * lra.c (lra): Add code to deal with fails by splitting hard reg
1076 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1079 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1081 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1084 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1085 (dw_val_node): Add val_symbolic_view.
1086 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1087 (symview_upper_bound): New.
1088 (new_line_info_table): Initialize symviews_since_reset.
1089 (dwarf2out_source_line): Count symviews_since_reset and set
1090 symview_upper_bound.
1091 (dw_val_equal_p): Handle symview.
1092 (add_AT_symview): New.
1093 (print_dw_val): Handle symview.
1094 (attr_checksum, attr_checksum_ordered): Likewise.
1095 (same_dw_val_p, size_of_die): Likewise.
1096 (value_format, output_die): Likewise.
1097 (add_high_low_attributes): Use add_AT_symview for entry_view.
1098 (dwarf2out_finish): Reset symview_upper_bound, clear
1101 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1104 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1105 Add strict argument and use it.
1106 (rs6000_split_multireg_move): Update for new strict argument.
1107 (mem_operand_gpr): Disallow all non-offsettable addresses.
1108 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1110 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1113 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1114 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1115 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1118 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1119 decl, use remap_type if we want to use the type.
1121 2018-03-09 Martin Sebor <msebor@redhat.com>
1123 PR tree-optimization/84526
1124 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1126 (builtin_access::generic_overlap): Be prepared to handle non-array
1129 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1131 PR rtl-optimization/84682
1132 * lra-constraints.c (process_address_1): Check is_address flag
1133 for address constraints.
1134 (process_alt_operands): Likewise.
1135 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1136 preprocess_constraints.
1137 * recog.h (preprocess_constraints): Add oploc parameter.
1139 * recog.c (preprocess_constraints): Test address_operand for
1140 CT_ADDRESS constraints.
1142 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1145 * lra-assigns.c (assign_by_spills): Return a flag of reload
1146 assignment failure. Do not process the reload assignment
1147 failures. Do not spill other reload pseudos if they has the same
1149 (lra_assign): Add a return arg. Set up from the result of
1150 assign_by_spills call.
1151 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1152 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1153 usage_insns if it is not NULL.
1154 (spill_hard_reg_in_range): New function.
1155 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1156 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1157 function prototypes.
1158 (lra_assign): Change prototype.
1159 * lra.c (lra): Add code to deal with fails by splitting hard reg
1162 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1165 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1166 Accept complain bool parameter. Only emit errors if it is true.
1167 (arm_parse_cpu_option_name): Likewise.
1168 (arm_target_thumb_only): Adjust callers of the above.
1169 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1170 prototype to take a default true bool parameter.
1171 (arm_parse_arch_option_name): Likewise.
1173 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1174 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1178 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1179 * configure: Regenerate.
1180 * configure.ac ("linker --version-script option"): New.
1181 ("linker soname option"): New.
1183 2018-03-09 Richard Biener <rguenther@suse.de>
1185 PR tree-optimization/84775
1186 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1187 immediate uses of predicate stmts and mark them modified.
1190 PR tree-optimization/84178
1191 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1193 (version_loop_for_if_conversion): Delay update_ssa call.
1194 (tree_if_conversion): Delay update_ssa until after predicate
1197 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1200 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1201 when the function accesses prior frames.
1203 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1206 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1207 gen_llsym, otherwise call maybe_gen_llsym.
1210 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1211 has ',' character inside of it.
1213 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1216 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1217 as clobbering CC_REGNUM.
1219 2018-03-08 Richard Biener <rguenther@suse.de>
1222 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1223 (follow_copies_to_constant): Do not follow SSA names registered
1226 2018-03-08 Richard Biener <rguenther@suse.de>
1228 PR tree-optimization/84178
1229 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1231 (version_loop_for_if_conversion): Delay update_ssa call.
1232 (tree_if_conversion): Delay update_ssa until after predicate
1235 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1237 PR tree-optimization/84178
1238 * tree-if-conv.c (release_bb_predicate): Remove the
1239 the assertion that the stmts have NULL use_ops.
1240 Discard the statements, asserting that they haven't
1241 yet been added to a BB.
1243 2018-03-08 Richard Biener <rguenther@suse.de>
1245 PR tree-optimization/84746
1246 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1247 (phi_translate): Pass in destination ANTIC_OUT set.
1248 (phi_translate_1): Likewise. For a simplified result lookup
1249 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1250 (phi_translate_set): Adjust.
1251 (do_pre_regular_insertion): Likewise.
1252 (do_pre_partial_partial_insertion): Likewise.
1254 2018-03-08 Martin Liska <mliska@suse.cz>
1256 PR gcov-profile/84735
1257 * doc/gcov.texi: Document usage of profile files.
1258 * gcov-io.h: Document changes in the format.
1260 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1264 * dwarf2out.c (struct dw_line_info_table): Update comments for
1266 (FORCE_RESET_NEXT_VIEW): New.
1267 (FORCE_RESETTING_VIEW_P): New.
1268 (RESETTING_VIEW_P): Check for -1 too.
1269 (ZERO_VIEW_P): Likewise.
1270 (new_line_info_table): Force-reset next view.
1271 (dwarf2out_begin_function): Likewise.
1272 (dwarf2out_source_line): Simplify zero_view_p initialization.
1273 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1274 view directly. Omit view when omitting .loc at line 0.
1276 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1278 PR tree-optimization/84740
1279 * tree-switch-conversion.c (process_switch): Call build_constructors
1280 only if info.phi_count is non-zero.
1282 PR tree-optimization/84739
1283 * tree-tailcall.c (find_tail_calls): Check call arguments against
1284 DECL_ARGUMENTS (current_function_decl) rather than
1285 DECL_ARGUMENTS (func) when checking for tail recursion.
1287 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1289 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1290 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1291 Volker Reichelt's entry and add entries for people that perform
1292 GCC fuzzy testing and report numerous bugs.
1294 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1297 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1298 readonly data in sdata, if that is disabled.
1299 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1300 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1301 -mreadonly-in-sdata option.
1303 2018-03-07 Martin Sebor <msebor@redhat.com>
1305 PR tree-optimization/84468
1306 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1307 basic block when looking for nul assignment.
1309 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1312 * except.h (output_function_exception_table): Adjust prototype.
1313 * except.c (output_function_exception_table): Remove FNNAME parameter
1314 and add SECTION parameter. Ouput one part of the table at a time.
1315 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1316 the first part of the exception table and emit unwind directives.
1317 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1318 (i386_pe_seh_cold_init): Likewise.
1319 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1320 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1321 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1322 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1323 * config/i386/winnt.c: Include except.h.
1324 (struct seh_frame_state): Add reg_offset, after_prologue and
1325 in_cold_section fields.
1326 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1327 (i386_pe_seh_cold_init): New function.
1328 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1329 to seh->in_cold_section.
1330 (seh_emit_push): Record the offset of the push.
1331 (seh_emit_save): Record the offet of the save.
1332 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1333 Test seh->after_prologue to disregard the epilogue.
1334 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1335 (i386_pe_end_cold_function): New function.
1337 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1340 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1341 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1344 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1346 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1347 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1350 * multiple_target.c: Include tree-inline.h and intl.h.
1351 (expand_target_clones): Diagnose and fail if node->definition and
1352 !tree_versionable_function_p (node->decl).
1354 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1356 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1358 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1359 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1360 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1362 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1365 * combine.c (try_combine): Use reg_or_subregno instead of handling
1366 just paradoxical SUBREGs and REGs.
1368 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1370 * config/arc/arc.c (arc_finalize_pic): Remove function.
1371 (arc_must_save_register): We use single base PIC register, remove
1372 checks to save/restore the PIC register.
1373 (arc_expand_prologue): Likewise.
1374 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1376 (arc_verify_short): Likewise.
1377 (arc_attr_type): Likewise.
1378 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1379 (walk_stores): Likewise.
1380 (arc_address_cost): Make it static.
1381 (arc_verify_short): Likewise.
1382 (branch_dest): Likewise.
1383 (arc_attr_type): Likewise.
1384 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1385 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1386 (arc_final_prescan_insn): Remove inserting the nops due to
1387 hardware hazards. It is done in reorg step.
1388 (insn_length_variant_t): Remove.
1389 (insn_length_parameters_t): Likewise.
1390 (arc_insn_length_parameters): Likewise.
1391 (arc_get_insn_variants): Likewise.
1392 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1394 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1397 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1400 PR tree-optimization/84687
1401 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1403 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1405 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1407 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1408 Rename to ppc_speculation_barrier.
1409 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1410 __builtin_ppc_speculation_barrier.
1412 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1415 * combine.c (combine_simplify_rtx): Don't try to simplify if
1416 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1419 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1421 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1422 to 32 bytes when compiling for POWER9.
1424 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1427 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1428 regparm >= 3 with no arg reg available also for calls with
1429 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1432 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1434 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1436 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1439 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1441 2018-03-05 Richard Biener <rguenther@suse.de>
1443 PR tree-optimization/84486
1444 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1445 When inserting a __builtin_assume_aligned call set the LHS
1446 SSA name alignment info accordingly.
1448 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1450 PR tree-optimization/84114
1451 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1452 Avoid reassociation of FLOAT_MODE addition.
1454 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1456 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1457 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1458 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1459 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1460 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1461 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1462 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1464 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1465 __builtin_ia32_wbinvd): New builtins.
1466 (SPECIAL_ARGS2): New.
1467 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1468 (SPECIAL_ARGS2): New.
1469 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1470 (ix86_valid_target_attribute_inner_p): Ditto.
1471 (ix86_init_mmx_sse_builtins): Add special_args2.
1472 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1473 TARGET_WBNOINVD_P): New.
1474 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1475 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1476 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1477 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1478 * config/i386/pconfigintrin.h: New file.
1479 * config/i386/wbnoinvdintrin.h: Ditto.
1480 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1481 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1483 2018-03-05 Richard Biener <rguenther@suse.de>
1485 PR tree-optimization/84670
1486 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1488 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1489 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1490 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1491 doesn't grow if all successors (recursively) were visited at least
1494 2018-03-05 Richard Biener <rguenther@suse.de>
1496 PR tree-optimization/84650
1497 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1498 if executed in the loop pipeline.
1500 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1502 * doc/configfiles.texi (Configuration Files): Move info about
1503 conditionalizing $target-protos.h to...
1504 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1505 differs from $target-protos.h.
1507 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1508 Chung-Ju Wu <jasonwucj@gmail.com>
1510 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1511 * config/nds32/nds32-multiple.md (setmemsi): Define.
1512 * config/nds32/nds32-memory-manipulation.c
1513 (nds32_gen_dup_4_byte_to_word_value): New.
1514 (emit_setmem_word_loop): New.
1515 (emit_setmem_byte_loop): New.
1516 (nds32_expand_setmem_loop): New.
1517 (nds32_expand_setmem_loop_v3m): New.
1518 (nds32_expand_setmem_unroll): New.
1519 (nds32_expand_setmem): New.
1521 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1522 Chung-Ju Wu <jasonwucj@gmail.com>
1524 * config/nds32/nds32-memory-manipulation.c
1525 (nds32_emit_load_store): New.
1526 (nds32_emit_post_inc_load_store): New.
1527 (nds32_emit_mem_move): New.
1528 (nds32_emit_mem_move_block): New.
1529 (nds32_expand_movmemsi_loop_unknown_size): New.
1530 (nds32_expand_movmemsi_loop_known_size): New.
1531 (nds32_expand_movmemsi_loop): New.
1532 (nds32_expand_movmemsi_unroll): New.
1533 (nds32_expand_movmemqi): Rename ...
1534 (nds32_expand_movmemsi): ... to this.
1535 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1536 (movmemsi): ... to this.
1537 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1538 (nds32_expand_movmemsi): ... to this.
1540 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1541 Monk Chiang <sh.chiang04@gmail.com>
1542 Chung-Ju Wu <jasonwucj@gmail.com>
1544 * config/nds32/nds32-protos.h
1545 (nds32_expand_load_multiple): New arguments.
1546 (nds32_expand_store_multiple): Ditto.
1547 (nds32_valid_multiple_load_store): Rename ...
1548 (nds32_valid_multiple_load_store_p): ... to this.
1549 * config/nds32/nds32-memory-manipulation.c
1550 (nds32_expand_load_multiple): Refine implementation.
1551 (nds32_expand_store_multiple): Ditto.
1552 * config/nds32/nds32-multiple.md
1553 (load_multiple): Update nds32_expand_load_multiple interface.
1554 (store_multiple): Update nds32_expand_store_multiple interface.
1555 * config/nds32/nds32-predicates.c
1556 (nds32_valid_multiple_load_store): Rename ...
1557 (nds32_valid_multiple_load_store_p): ... to this and refine
1559 * config/nds32/predicates.md
1560 (nds32_load_multiple_and_update_address_operation): New predicate.
1561 (nds32_store_multiple_and_update_address_operation): New predicate.
1563 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1564 Chung-Ju Wu <jasonwucj@gmail.com>
1566 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1567 (combo): New attribute.
1568 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1570 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1572 * config/nds32/nds32.opt: Change -mcmodel= default value.
1574 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1575 Monk Chiang <sh.chiang04@gmail.com>
1576 Chung-Ju Wu <jasonwucj@gmail.com>
1578 * config/nds32/constants.md (unspec_element): New enum.
1579 * config/nds32/constraints.md (Umw): New constraint.
1580 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1581 * config/nds32/nds32-intrinsic.md: Likewise.
1582 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1583 (nds32_valid_smw_lwm_base_p): New.
1584 (nds32_output_smw_single_word): New.
1585 (nds32_output_lmw_single_word): New.
1586 (nds32_expand_unaligned_load): New.
1587 (nds32_expand_unaligned_store): New.
1588 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1589 (nds32_output_smw_single_word): Declare.
1590 (nds32_output_lmw_single_word): Declare.
1591 (nds32_expand_unaligned_load): Declare.
1592 (nds32_expand_unaligned_store): Declare.
1593 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1594 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1595 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1596 NDS32_BUILTIN_UASTORE_DW.
1597 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1600 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1601 Kito Cheng <kito.cheng@gmail.com>
1602 Chung-Ju Wu <jasonwucj@gmail.com>
1604 * config/nds32/nds32-intrinsic.c
1605 (nds32_expand_builtin_null_ftype_reg): Delete.
1606 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1607 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1608 (nds32_read_argument): New.
1609 (nds32_legitimize_target): Ditto.
1610 (nds32_legitimize_argument): Ditto.
1611 (nds32_check_constant_argument): Ditto.
1612 (nds32_expand_unop_builtin): Ditto.
1613 (nds32_expand_unopimm_builtin): Ditto.
1614 (nds32_expand_binop_builtin): Ditto.
1615 (nds32_builtin_decl_impl): Ditto.
1616 (builtin_description): Ditto.
1617 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1618 (nds32_init_builtins_impl): Ditto.
1619 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1620 (nds32_builtin_decl): New.
1621 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1622 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1624 2018-03-02 Jeff Law <law@redhat.com>
1626 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1627 (redundant_insn, fill_simple_delay_slots): Likewise.
1628 (fill_slots_from_thread): Likewise.
1629 * resource.c (mark_referenced_resources): Likewise.
1630 (mark_set_resources, find_dead_or_set_registers): Likewise.
1632 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1634 * substring-locations.h (format_warning_va): Formatting fix for
1636 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1638 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1639 * substring-locations.c: Include intl.h.
1640 (format_warning_va): Turned into small wrapper around
1641 format_warning_n_va, renamed to ...
1642 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1643 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1645 (format_warning_at_substring_n): New function.
1646 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1647 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1648 format_warning_at_substring with just a shorter name instead of
1649 const function pointer.
1650 (fmtwarn_n): New function.
1651 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1652 appropriate, get rid of all the fmtstr temporaries, move conditionals
1653 with G_() wrapped string literals directly into fmtwarn arguments,
1654 cast dir.len to (int), formatting fixes.
1656 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1658 * doc/invoke.texi: Remove "Cilk Plus" references.
1660 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1661 Richard Biener <rguenther@suse.de>
1664 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1665 for error or warning attributes if CALL_FROM_THUNK_P is set.
1668 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1671 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1672 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1675 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1676 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1678 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1679 instead of -1U in last predictors element's probability member.
1681 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1684 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1685 arguments if they are comparable.
1687 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1689 PR tree-optimization/84634
1690 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1691 masks and masked_loop_p with a single loop_masks, making sure it's
1692 null for bb vectorization.
1694 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1696 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1697 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1698 loop->force_vectorize to check whether there is no alias.
1700 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1703 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1705 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1707 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1708 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1709 instead of a loop around prev_real_insn.
1710 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1714 * config/i386/i386.c (ix86_print_operand): Use conditional
1715 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1718 2018-03-02 Richard Biener <rguenther@suse.de>
1720 PR tree-optimization/84427
1721 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1722 (bitmap_set_subtract_values): Rewrite to handle multiple
1725 (prune_clobbered_mems): Likewise.
1726 (phi_translate): Take edge instead of pred/phiblock.
1727 (phi_translate_1): Likewise.
1728 (phi_translate_set): Likewise. Insert all translated
1729 exprs for a value into the set, keeping possibly multiple
1730 expressions per value.
1731 (compute_antic_aux): Adjust for phi_translate changes.
1732 When intersecting union the expressions and prune those
1733 not in the final value set, keeping possibly multiple
1734 expressions per value. Do not use value-insertion
1735 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1736 all expressions. Add verification that the value-sets
1737 only shrink during iteration.
1738 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1739 (do_pre_regular_insertion): Likewise.
1740 (do_pre_partial_partial_insertion): Likewise.
1742 2018-03-02 Richard Biener <rguenther@suse.de>
1745 * config/darwin.c (saved_debug_info_level): New static global.
1746 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1747 (darwin_asm_lto_end): Restore debug info generation settings.
1749 2018-03-01 Martin Liska <mliska@suse.cz>
1752 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1755 2018-03-01 Richard Biener <rguenther@suse.de>
1758 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1759 type annotation with LTO when debug was disabled at compile-time.
1761 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1763 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1765 (mips_final_postscan_insn): Likewise.
1767 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1769 PR rtl-optimization/84528
1770 * alias.c (init_alias_target): Add commentary.
1771 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1772 a unique base value if the frame pointer is not eliminated
1773 to the stack pointer.
1775 2018-03-01 Tom de Vries <tom@codesourcery.com>
1777 PR rtl-optimization/83327
1778 * lra-int.h (hard_regs_spilled_into): Declare.
1779 * lra.c (hard_regs_spilled_into): Define.
1780 (init_reg_info): Init hard_regs_spilled_into.
1781 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1782 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1783 (process_bb_lives): Handle hard_regs_spilled_into.
1784 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1785 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1787 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1789 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1790 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1791 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1792 * config/rs6000/aix72.h: New file.
1794 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1796 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1797 instead of warning_at with conditional singular and plural messages
1801 * stor-layout.c (update_alignment_for_field): For
1802 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1803 && !DECL_PACKED (field), do the alignment update, just use
1804 only desired_align instead of MAX (type_align, desired_align)
1806 (place_field): Don't do known_align < desired_align handling
1807 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1808 is non-NULL, instead do it after rli->prev_field handling and
1809 only if not within a bitfield word. For DECL_PACKED (field)
1810 use type_align of BITS_PER_UNIT.
1812 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1814 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1815 superfluous parentheses and trailing spaces.
1817 2018-02-28 Richard Biener <rguenther@suse.de>
1819 PR tree-optimization/84584
1820 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1821 SCoPs with fake exit edge.
1823 2018-02-28 Martin Liska <mliska@suse.cz>
1826 * timevar.c (timer::print): Fix format to properly print 100%
1829 2018-02-28 Richard Biener <rguenther@suse.de>
1832 * genmatch.c (capture_info::walk_match): Do not mark
1833 captured expressions without operands as expr_p given
1834 they act more like predicates and should be subject to
1835 "lost tail" side-effect preserving.
1837 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1839 PR rtl-optimization/81611
1840 * auto-inc-dec.c (attempt_change): Move dead note from
1841 mem_insn if it's the next use of regno
1842 (find_address): Take address use of reg holding
1843 non-incremented value. Add parm to limit search to the named
1845 (merge_in_block): Attempt to use a mem insn that is the next
1846 use of the original regno.
1848 2018-02-27 Martin Sebor <msebor@redhat.com>
1851 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1852 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1854 2018-02-27 Martin Sebor <msebor@redhat.com>
1856 PR translation/84207
1857 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1858 n argument to unsigned HOST_WIDE_INT.
1859 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1860 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1861 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1862 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1864 2018-02-27 Richard Biener <rguenther@suse.de>
1866 PR tree-optimization/84512
1867 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1868 Do not use the estimate returned from record_stmt_cost for
1869 the scalar iteration cost but sum properly using add_stmt_cost.
1871 2018-02-27 Richard Biener <rguenther@suse.de>
1873 PR tree-optimization/84466
1874 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1875 Adjust last change to less strictly validate use operands.
1877 2018-02-27 Martin Liska <mliska@suse.cz>
1879 PR gcov-profile/84548
1880 * gcov.c (process_file): Allow partial overlap and consider it
1881 also as group functions.
1882 (output_lines): Properly calculate range of lines for a group.
1884 2018-02-27 Martin Liska <mliska@suse.cz>
1886 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1887 'ggc' suffixes. Change first column width.
1888 (timer::print): Fix formatting of the column.
1890 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1892 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1893 preserve inline entry blocks for the sake of debug inline
1894 entry point markers alone.
1895 (remove_unused_locals): Suggest in comments a better place to
1896 force the preservation of inline entry blocks that are
1897 otherwise unused, but do not preserve them.
1899 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1901 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1903 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1906 * config/i386/constraints.md (Bs): Replace
1907 ix86_indirect_branch_register with
1908 TARGET_INDIRECT_BRANCH_REGISTER.
1910 * config/i386/i386.md (indirect_jump): Likewise.
1911 (tablejump): Likewise.
1912 (*sibcall_memory): Likewise.
1913 (*sibcall_value_memory): Likewise.
1914 Peepholes of indirect call and jump via memory: Likewise.
1915 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1916 (*sibcall_value_GOT_32): Likewise.
1917 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1918 (GOT_memory_operand): Likewise.
1919 (call_insn_operand): Likewise.
1920 (sibcall_insn_operand): Likewise.
1921 (GOT32_symbol_operand): Likewise.
1922 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1924 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1926 PR rtl-optimization/83496
1927 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1928 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1929 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1930 redundant insn, if any.
1931 (relax_delay_slots): Likewise.
1932 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1934 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1936 PR tree-optimization/83965
1937 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1938 that grouped statements are part of a reduction chain. Return
1939 true if the statement is not marked as a reduction itself but
1941 (vect_recog_dot_prod_pattern): Don't check whether the statement
1942 is part of a group here.
1943 (vect_recog_sad_pattern): Likewise.
1944 (vect_recog_widen_sum_pattern): Likewise.
1946 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1949 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1951 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1954 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1956 (ix86_output_indirect_function_return): New prototype.
1957 (ix86_split_simple_return_pop_internal): Likewise.
1958 * config/i386/i386.c (indirect_return_via_cx): New.
1959 (indirect_return_via_cx_bnd): Likewise.
1960 (indirect_thunk_name): Handle return va CX_REG.
1961 (output_indirect_thunk_function): Create alias for
1962 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1963 (ix86_output_indirect_jmp): Remove the bool argument.
1964 (ix86_output_indirect_function_return): New function.
1965 (ix86_split_simple_return_pop_internal): Likewise.
1966 * config/i386/i386.md (*indirect_jump): Don't pass false
1967 to ix86_output_indirect_jmp.
1968 (*tablejump_1): Likewise.
1969 (simple_return_pop_internal): Change it to define_insn_and_split.
1970 Call ix86_split_simple_return_pop_internal to split it for
1972 (simple_return_indirect_internal): Call
1973 ix86_output_indirect_function_return instead of
1974 ix86_output_indirect_jmp.
1976 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1979 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1980 memset and value initialization afterwards.
1982 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1984 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1986 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1989 * common/config/aarch64/aarch64-common.c
1990 (aarch_option_optimization_table[]): Switch
1991 off fomit-frame-pointer
1993 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1994 Chung-Ju Wu <jasonwucj@gmail.com>
1996 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1998 (store_multiple): Ditto.
2000 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2002 * config.gcc: Add --with-cpu support for nds32 target.
2003 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2004 * config/nds32/nds32.opt: Add -mcpu= option.
2006 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2008 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2009 isel=yes): Warn for these deprecated options.
2011 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2013 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2014 ISA_2_5_MASKS_EMBEDDED.
2016 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2018 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2019 p->max as pointers rather than using iterative_hash_expr.
2021 2018-02-23 Carl Love <cel@us.ibm.com>
2023 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2024 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2026 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2027 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2028 P8V_BUILTIN_VEC_VUNSIGNED2.
2030 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2033 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2034 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2036 * lra-constraints.c (curr_insn_transform): Set up
2037 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2039 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2040 * lra-eliminations.c (spill_pseudos): Ditto.
2041 (process_insn_for_elimination): Ditto.
2042 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2043 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2044 LRA_NON_CLOBBERED_ALT.
2046 2018-02-22 Martin Sebor <msebor@redhat.com>
2048 PR tree-optimization/84480
2049 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2050 to maybe_diag_stxncpy_trunc. Call it.
2051 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2052 from gimple_fold_builtin_strcpy. Print inlining stack.
2053 (handle_builtin_stxncpy): Print inlining stack.
2054 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2056 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2059 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2060 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2061 and -fcheck-pointer-bounds are used together.
2062 (indirect_thunk_prefix): New enum.
2063 (indirect_thunk_need_prefix): New function.
2064 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2065 "_nt" instead of "_bnd" for NOTRACK prefix.
2066 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2067 (output_indirect_thunk_function): Likewise.
2069 (ix86_code_end): Update output_indirect_thunk_function calls.
2070 (ix86_output_indirect_branch_via_reg): Replace
2071 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2072 (ix86_output_indirect_branch_via_push): Likewise.
2073 (ix86_output_function_return): Likewise.
2074 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2075 incompatible with -fcf-protection=branch and
2076 -fcheck-pointer-bounds.
2078 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2081 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2082 Change gcc_assert call to output_operand_lossage.
2084 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2086 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2088 2018-02-22 DJ Delorie <dj@redhat.com>
2089 Sebastian Perta <sebastian.perta@renesas.com>
2090 Oleg Endo <olegendo@gcc.gnu.org>
2092 * config/rx/rx.c (rx_rtx_costs): New function.
2093 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2095 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2097 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2099 2018-02-22 Martin Liska <mliska@suse.cz>
2102 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2103 Add "native" as a possible value.
2105 2018-02-22 Martin Liska <mliska@suse.cz>
2108 * config/i386/i386.c (ix86_option_override_internal):
2109 Add "native" as a possible value for -march and -mtune.
2111 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2114 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2115 to all type variants.
2117 PR tree-optimization/84503
2118 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2119 width as info->bitpos + info->bitsize - start.
2120 (merged_store_group::merge_overlapping): Simplify width computation.
2121 (check_no_overlap): New function.
2122 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2123 start + width and last_order of the group, fail if check_no_overlap
2125 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2126 to group if check_no_overlap fails.
2128 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2130 * config/rs6000/altivec.md: Delete contraint arguments to
2131 define_expand, define_split, and define_peephole2, and in
2132 define_insn_and_split if always unused.
2133 * config/rs6000/darwin.md: Ditto.
2134 * config/rs6000/dfp.md: Ditto.
2135 * config/rs6000/rs6000.md: Ditto.
2136 * config/rs6000/sync.md: Ditto.
2137 * config/rs6000/vector.md: Ditto.
2138 * config/rs6000/vsx.md: Ditto.
2140 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2142 * config/rs6000/altivec.md: Write output control strings as braced
2143 blocks instead of double-quoted strings.
2144 * config/rs6000/darwin.md: Ditto.
2145 * config/rs6000/rs6000.md: Ditto.
2146 * config/rs6000/vector.md: Ditto.
2147 * config/rs6000/vsx.md: Ditto.
2149 2018-02-21 Jason Merrill <jason@redhat.com>
2151 PR c++/84314 - ICE with templates and fastcall attribute.
2152 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2154 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2156 * ipa-cp.c (determine_versionability): Fix comment typos.
2158 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2161 * ipa-cp.c (determine_versionability): Do not version functions caling
2164 2018-02-21 Martin Liska <mliska@suse.cz>
2167 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2168 Add "native" as a possible value.
2169 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2170 the macro when native cpu detection is available.
2172 2018-02-21 Martin Liska <mliska@suse.cz>
2175 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2176 Add "native" as a possible value.
2177 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2178 when native cpu detection is available.
2180 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2181 Martin Sebor <msebor@redhat.com>
2183 PR tree-optimization/84478
2184 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2186 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2187 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2188 support which is conservatively correct, for 2 only stay conservative
2189 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2190 argument to the 2 argument get_range_strlen, adjust 6 arg
2191 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2193 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2194 (gimple_fold_builtin_strlen): Pass true as last argument to
2197 2018-02-20 Martin Sebor <msebor@redhat.com>
2200 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2201 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2202 (builtin_memref::builtin_memref): Factor out parts into
2203 set_base_and_offset and call it.
2205 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2208 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2209 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2210 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2211 search at the associated MODE_INT.
2213 2018-02-20 Jeff Law <law@redhat.com>
2216 PR tree-optimization/81592
2218 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2219 for range data rather than using global data.
2220 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2221 range data rather than using global data.
2222 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2223 pass it to children as needed.
2224 (struct directive::fmtresult): Similarly.
2225 (struct directive::set_width): Similarly.
2226 (struct directive::set_precision): Similarly.
2227 (format_integer, format_directive, parse_directive): Similarly.
2228 (format_none): Accept unnamed vr_values parameter.
2229 (format_percent, format_floating, format_character): Similarly.
2230 (format_string, format_plain): Similarly.
2231 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2232 the EVRP range analyzer for range data rather than using global data.
2233 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2234 gimple-ssa-evrp-analyze.h
2235 (class sprintf_dom_walker): Add after_dom_children member function.
2236 Add evrp_range_analyzer member.
2237 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2238 range analyzer as needed.
2239 (sprintf_dom_walker::after_dom_children): New member function.
2240 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2242 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2243 (evrp_range_analyzer::pop_to_marker): Likewise.
2245 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2247 PR tree-optimization/84419
2248 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2249 with the required type if its current type is compatible but
2252 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2255 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2256 after vectorization.
2258 2018-02-20 Martin Liska <mliska@suse.cz>
2261 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2262 possible values if we don't have a hint.
2264 2018-02-20 Martin Liska <mliska@suse.cz>
2268 * final.c (shorten_branches): Build align_tab array with one
2270 * opts.c (finish_options): Add alignment option limit check.
2271 (MAX_CODE_ALIGN): Likewise.
2272 (MAX_CODE_ALIGN_VALUE): Likewise.
2273 * doc/invoke.texi: Document maximum allowed option value for
2274 all -falign-* options.
2276 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2279 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2280 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2281 * var-tracking.c (emit_note_insn_var_location): Remove all references
2282 to NOTE_INSN_CALL_ARG_LOCATION.
2283 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2284 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2285 Use copy_rtx_if_shared.
2286 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2287 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2288 (dwarf2out_var_location): Remove handling of
2289 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2291 * final.c (final_scan_insn): Remove all references to
2292 NOTE_INSN_CALL_ARG_LOCATION.
2293 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2294 before dumping final insns.
2295 * except.c (emit_note_eh_region_end): Remove all references to
2296 NOTE_INSN_CALL_ARG_LOCATION.
2297 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2298 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2299 * config/arc/arc.c (hwloop_optimize): Likewise.
2300 * config/arm/arm.c (create_fix_barrier): Likewise.
2301 * config/s390/s390.c (s390_chunkify_start): Likewise.
2302 * config/sh/sh.c (find_barrier): Likewise.
2303 * config/i386/i386.c (rest_of_insert_endbranch,
2304 ix86_seh_fixup_eh_fallthru): Likewise.
2305 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2306 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2307 * config/frv/frv.c (frv_function_prologue): Likewise.
2308 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2310 (note_outside_basic_block_p): Remove all references to
2311 NOTE_INSN_CALL_ARG_LOCATION.
2312 * gengtype.c (adjust_field_rtx_def): Likewise.
2313 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2315 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2316 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2319 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2322 PR tree-optimization/84452
2323 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2324 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2327 2018-02-19 Martin Liska <mliska@suse.cz>
2330 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2332 2018-02-19 Martin Liska <mliska@suse.cz>
2333 Richard Sandiford <richard.sandiford@linaro.org>
2335 PR tree-optimization/82491
2336 * gimple-fold.c (get_base_constructor): Make earlier bail out
2339 2018-02-19 Carl Love <cel@us.ibm.com>
2341 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2342 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2344 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2345 P8V_BUILTIN_VEC_NEG.
2347 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2349 * config/rl78/rl78.md (movdf): New define expand.
2351 2018-02-19 Martin Liska <mliska@suse.cz>
2354 * doc/invoke.texi: Fix typo.
2355 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2357 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2359 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2360 handle rs6000_single_float and rs6000_double_float specially for
2363 2018-02-16 Jeff Law <law@redhat.com>
2365 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2368 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2371 * ipa-inline.c (inline_small_functions): Fix a typo.
2373 2018-02-16 Nathan Sidwell <nathan@acm.org>
2375 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2377 2018-02-16 Carl Love <cel@us.ibm.com>
2379 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2380 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2381 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2382 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2383 expansion to P8V_BUILTIN_VEC_FLOAT2.
2385 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2387 PR rtl-optimization/70023
2388 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2389 src_regno into account.
2391 2018-02-16 Carl Love <cel@us.ibm.com>
2393 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2394 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2395 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2396 * config/rs6000/rs6000.c: Remove case statements for
2397 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2398 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2399 and P9V_BUILTIN_VEC_VINSERT4B.
2400 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2401 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2402 * config/rs6000/vsx.md:
2403 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2406 2018-02-16 Carl Love <cel@us.ibm.com>
2408 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2410 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2412 * config/rs6000/rs6000-c.c: Add the definitions for
2413 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2414 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2415 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2416 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2417 definition for insert4b and define insn *insert3b_internal.
2418 * doc/extend.texi: Add documentation for vec_extract4b.
2420 2018-02-16 Nathan Sidwell <nathan@acm.org>
2422 * doc/extend.texi (Backwards Compatibility): Mention friend
2423 injection. Note for-scope is deprecated.
2424 * doc/invoke.texi (-ffriend-injection): Deprecate.
2426 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2428 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2429 that moved to I2, also allow destinations that are a paradoxical
2430 subreg (instead of a normal reg).
2432 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2435 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2438 2018-02-16 Richard Biener <rguenther@suse.de>
2440 PR tree-optimization/84037
2441 PR tree-optimization/84016
2443 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2444 Adjust vec_construct for the fact we need additional higher latency
2445 128bit inserts for AVX256 and AVX512 vector builds.
2446 (ix86_add_stmt_cost): Scale vector construction cost for
2449 2018-02-16 Richard Biener <rguenther@suse.de>
2451 PR tree-optimization/84417
2452 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2453 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2454 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2456 2018-02-16 Martin Liska <mliska@suse.cz>
2459 * internal-fn.def (ASAN_CHECK): Set proper flags.
2460 (ASAN_MARK): Likewise.
2462 2018-02-16 Julia Koval <julia.koval@intel.com>
2464 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2465 from PTA_CANNONLAKE.
2467 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2470 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2471 Use ++iter rather than iter++ for std::list iterators.
2472 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2473 defer deleting them until all nodes in the forest are processed. Do
2474 free even leaf nodes. Change to_process into auto_vec.
2477 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2478 * vec.h (vec_default_construct): Use memset instead of placement new
2479 if BROKEN_VALUE_INITIALIZATION is defined.
2480 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2481 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2484 PR rtl-optimization/83723
2485 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2486 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2487 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2489 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2491 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2493 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2495 PR rtl-optimization/81443
2496 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2497 from inner REGs to paradoxical SUBREGs.
2499 2018-02-16 Richard Biener <rguenther@suse.de>
2501 PR tree-optimization/84399
2502 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2503 For operands we can analyze at their definition make sure we can
2504 analyze them at each use as well.
2506 2018-02-16 Richard Biener <rguenther@suse.de>
2508 PR tree-optimization/84190
2509 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2510 volatile accesses if the decl isn't volatile.
2512 2018-02-15 Jason Merrill <jason@redhat.com>
2514 PR c++/84314 - ICE with templates and fastcall attribute.
2515 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2516 TYPE_CANONICAL on an existing type.
2518 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2520 PR tree-optimization/84383
2521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2522 dstoff nor call operand_equal_p if dstbase is NULL.
2524 PR tree-optimization/84334
2525 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2526 also a CONSTANT_CLASS_P, punt.
2528 2018-02-14 Jim Wilson <jimw@sifive.com>
2530 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2531 first SMALL_OPERAND check. New local min_second_step. Move assert
2532 to where locals are set. Add TARGET_RVC support.
2533 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2535 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2537 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2539 2018-02-14 Martin Sebor <msebor@redhat.com>
2541 PR tree-optimization/83698
2542 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2543 arrays constrain the offset range to their bounds.
2544 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2545 (builtin_access::overlap): Avoid setting the size of overlap if it's
2547 (maybe_diag_overlap): Also consider arrays when deciding what values
2548 of offsets to include in diagnostics.
2550 2018-02-14 Martin Sebor <msebor@redhat.com>
2553 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2554 that correspond to the kind of a declaration.
2556 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2559 * config/pa/pa.md: Load address of PIC label using the linkage table
2560 if the label is nonlocal.
2562 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2564 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2565 warning message if user requests -maltivec=be.
2566 * doc/invoke.texi: Document deprecation of -maltivec=be.
2568 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2571 * config/rs6000/rs6000-c.c: Update definitions for
2572 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2573 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2575 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2578 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2579 add _get_ssp intrinsics. Remove argument from
2580 __builtin_ia32_rdssp[d|q].
2581 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2582 * config/i386/i386-builtin.def: Remove argument from
2583 __builtin_ia32_rdssp[d|q].
2584 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2585 ix86_expand_special_args_builtin for _rdssp[d|q].
2586 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2587 Clear register before usage.
2588 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2589 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2591 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2593 PR tree-optimization/84357
2594 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2595 operand 1 of an ARRAY_REF too.
2597 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2600 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2601 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2603 (set_of_reg): New struct.
2604 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2605 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2606 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2608 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2609 Split into bitclr, bitset, bitinvert patterns if appropriate.
2610 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2611 use rx_fuse_in_memory_bitop.
2612 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2613 to named insn, correct maximum insn length.
2615 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2618 * machmode.def: Define a complex mode for PARTIAL_INT.
2619 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2621 * doc/rtl.texi: Document CSPImode.
2622 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2624 (msp430_hard_regno_nregs_with_padding): Likewise.
2626 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2629 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2631 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2633 PR rtl-optimization/84169
2634 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2635 we generated a parallel as new i3 and we split that to new i2 and i3
2636 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2637 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2638 those to i2, not i1. Partially rewrite this scan code.
2640 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2643 * stor-layout.c (place_field): For variable length fields, adjust
2644 offset_align afterwards not just based on the field's alignment,
2645 but also on the size.
2648 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2649 of exps and logs in the use_exp2 case.
2651 2018-02-13 Jeff Law <law@redhat.com>
2653 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2656 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2659 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2663 * common.opt (gas-loc-support, gas-locview-support): New.
2664 (ginline-points, ginternal-reset-location-views): New.
2665 * doc/invoke.texi: Document them. Use @itemx where intended.
2666 (gvariable-location-views): Adjust.
2667 * target.def (reset_location_view): New.
2668 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2669 (TARGET_RESET_LOCATION_VIEW): New.
2670 * doc/tm.texi: Rebuilt.
2671 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2672 (dwarf2out_default_as_locview_support): New.
2673 (output_asm_line_debug_info): Use option variables.
2674 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2675 (output_loc_list): Likewise.
2676 (add_high_low_attributes): Check option variables.
2677 Don't output entry view attribute in strict mode.
2678 (gen_inlined_subroutine_die): Check option variables.
2679 (dwarf2out_inline_entry): Likewise.
2680 (init_sections_and_labels): Likewise.
2681 (dwarf2out_early_finish): Likewise.
2682 (maybe_reset_location_view): New, from...
2683 (dwarf2out_var_location): ... here. Call it.
2684 * debug.h (dwarf2out_default_as_loc_support): Declare.
2685 (dwarf2out_default_as_locview_support): Declare.
2686 * hooks.c (hook_int_rtx_insn_0): New.
2687 * hooks.h (hook_int_rtx_insn_0): Declare.
2688 * toplev.c (process_options): Take -gas-loc-support and
2689 -gas-locview-support from dwarf2out. Enable
2690 -gvariable-location-views by default only with locview
2691 assembler support. Enable -ginternal-reset-location-views by
2692 default only if the target defines the corresponding hook.
2693 Enable -ginline-points by default if location views are
2694 enabled; force it disabled if statement frontiers are
2696 * tree-inline.c (expand_call_inline): Check option variables.
2697 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2699 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2701 PR tree-optimization/84321
2702 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2703 handling. Also check whether the anti-range contains any values
2704 that satisfy the mask; switch to a VR_RANGE if not.
2706 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2709 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2711 2018-02-13 Martin Jambor <mjambor@suse.cz>
2714 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2715 of call statements, also set location of a load to a temporary.
2717 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2719 * config/rl78/rl78.c (add_vector_labels): New function.
2720 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2721 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2722 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2723 which checks that no arguments are passed.
2724 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2725 * doc/extend.texi: Documentation for the new attribute.
2727 2018-02-13 Andreas Schwab <schwab@suse.de>
2729 * config/riscv/linux.h (CPP_SPEC): Define.
2731 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2734 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2735 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2736 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2737 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2738 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2739 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2740 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2741 temporarily for AES and PCLMUL builtins.
2743 PR tree-optimization/84339
2744 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2745 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2749 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2750 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2751 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2753 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2755 * omp-simd-clone.h: New file.
2756 * omp-simd-clone.c: Include omp-simd-clone.h.
2757 (expand_simd_clones): No longer static.
2758 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2759 cgraph.h and omp-simd-clone.h.
2760 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2761 (vect_recog_widen_shift_pattern): Formatting fix.
2762 (vect_pattern_recog_1): Don't check optab for calls.
2765 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2766 operands[2] into a REG before using gen_lowpart on it.
2768 2018-02-12 Jeff Law <law@redhat.com>
2771 * config/sh/sh.c (find_barrier): Consider a sibling call
2774 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2775 successfully back substituting a reg.
2777 2018-02-12 Richard Biener <rguenther@suse.de>
2779 PR tree-optimization/84037
2780 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2781 parameter, move visited init to caller.
2782 (vect_slp_analyze_operations): Separate cost from validity
2783 check, initialize visited once for all instances.
2784 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2786 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2788 (vect_model_promotion_demotion_cost): Likewise.
2789 (vectorizable_bswap): Guard cost modeling with !slp_node
2790 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2792 (vectorizable_call): Likewise.
2793 (vectorizable_conversion): Likewise.
2794 (vectorizable_assignment): Likewise.
2795 (vectorizable_shift): Likewise.
2796 (vectorizable_operation): Likewise.
2797 (vectorizable_store): Likewise.
2798 (vectorizable_load): Likewise.
2799 (vectorizable_condition): Likewise.
2800 (vectorizable_comparison): Likewise.
2802 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2805 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2806 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2807 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2809 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2812 * params.def (inline-min-speedup): Increase from 8 to 15.
2813 (max-inline-insns-auto): Decrease from 40 to 30.
2814 * ipa-split.c (consider_split): Add some buffer for function to
2815 be considered inlining candidate.
2816 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2819 2018-02-12 Richard Biener <rguenther@suse.de>
2821 PR tree-optimization/84037
2822 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2823 matched stmts if we cannot swap the non-matched ones.
2825 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2827 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2828 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2829 _mm_maskz_scalef_round_ss): New intrinsics.
2830 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2831 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2832 __builtin_ia32_scalefss_round): Remove.
2833 (__builtin_ia32_scalefsd_mask_round,
2834 __builtin_ia32_scalefss_mask_round): New intrinsics.
2835 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2836 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2837 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2838 "<round_constraint>")): Changed to ...
2839 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2840 "<round_scalar_constraint>")): ... this.
2841 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2842 %0, %1, %2<round_op3>}"): Changed to ...
2843 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2844 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2845 %2<round_scalar_mask_op3>}"): ... this.
2846 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2848 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2850 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2851 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2852 (_mm_maskz_sqrt_round_ss): New intrinsics.
2853 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2854 (__builtin_ia32_sqrtsd_mask_round)
2855 (__builtin_ia32_sqrtss_mask_round): New builtins.
2856 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2857 (__builtin_ia32_sqrtss_round): Remove.
2858 (__builtin_ia32_sqrtsd_mask_round)
2859 (__builtin_ia32_sqrtss_mask_round): New builtins.
2860 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2861 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2862 ((match_operand:VF_128 1 "vector_operand"
2863 "xBm,<round_constraint>")): Changed to ...
2864 ((match_operand:VF_128 1 "vector_operand"
2865 "xBm,<round_scalar_constraint>")): ... this.
2866 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2867 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2868 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2869 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2870 %<iptr>1<round_scalar_mask_op3>}): ... this.
2871 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2872 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2874 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2877 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2878 Cast vec_cmpeq result to correct type.
2879 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2880 Cast vec_cmpgt result to correct type.
2882 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2884 * final.c (final_scan_insn_1): Renamed from...
2885 (final_scan_insn): ... this. New wrapper, to recover
2886 seen from the outermost call in recursive ones.
2887 * config/sparc/sparc.c (output_return): Drop seen from call.
2888 (output_sibcall): Likewise.
2889 * config/visium/visium.c (output_branch): Likewise.
2891 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2893 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2896 2018-02-10 Alan Modra <amodra@gmail.com>
2899 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2900 Specify LR as an input.
2902 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2905 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2906 remove_member_access_dummy_vars): New functions.
2907 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2908 lower_omp_1, execute_lower_omp): Use them.
2910 PR rtl-optimization/84308
2911 * shrink-wrap.c (spread_components): Release todo vector.
2913 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2915 PR rtl-optimization/57193
2916 * ira-color.c (struct allocno_color_data): Add member
2917 conflict_allocno_hard_prefs.
2918 (update_conflict_allocno_hard_prefs): New.
2919 (bucket_allocno_compare_func): Add a preference based on
2920 conflict_allocno_hard_prefs.
2921 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2922 (color_allocnos): Remove a dead code. Initiate
2923 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2925 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2928 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2929 constraint from =wa to wa. Avoid a subreg on the output operand,
2930 instead use a pseudo and subreg it in a move.
2931 (p9_xxbrd_<mode>): Changed to ...
2932 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2933 (p9_xxbrd_v2df): New expander.
2934 (p9_xxbrw_<mode>): Changed to ...
2935 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2936 (p9_xxbrw_v4sf): New expander.
2938 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2940 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2942 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2945 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2946 multiply in 32-bit mode.
2947 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2948 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2951 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2953 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2954 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2955 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2956 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2958 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2961 * dwarf2out.c (is_trivial_indirect_ref): New function.
2962 (dwarf2out_late_global_decl): Do not generate a location
2963 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2964 and that are not defined in the current unit.
2966 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2968 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2969 instead of a libcall for UNORDERED.
2971 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2974 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2975 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2977 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2980 * config/s390/s390.c (s390_set_current_function): Invoke
2981 s390_indirect_branch_settings also if fndecl didn't change.
2983 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2985 * config/rs6000/rs6000.md (blockage): Set length to zero.
2987 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2989 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2991 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2994 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2995 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2999 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3000 PARALLEL incoming that failed vt_get_decl_and_offset check.
3003 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3004 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3006 (get_variable_section): For decls in named .bss* sections pass true as
3007 second argument to bss_initializer_p.
3009 2018-02-09 Marek Polacek <polacek@redhat.com>
3010 Jakub Jelinek <jakub@redhat.com>
3013 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3014 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3015 Sync some changes from cxx_fold_indirect_ref.
3017 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3019 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3021 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3022 (BLOCK_INLINE_ENTRY_LABEL): New.
3023 (dwarf2out_var_location): Disregard inline entry markers.
3024 (inline_entry_data): New struct.
3025 (inline_entry_data_hasher): New hashtable type.
3026 (inline_entry_data_hasher::hash): New.
3027 (inline_entry_data_hasher::equal): New.
3028 (inline_entry_data_table): New variable.
3029 (add_high_low_attributes): Add DW_AT_entry_pc and
3030 DW_AT_GNU_entry_view attributes if a pending entry is found
3031 in inline_entry_data_table. Add old entry_pc attribute only
3032 if debug nonbinding markers are disabled.
3033 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3034 markers are enabled.
3035 (block_within_block_p, dwarf2out_inline_entry): New.
3036 (dwarf2out_finish): Check that no entries remained in
3037 inline_entry_data_table.
3038 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3039 (final_scan_insn, notice_source_line): Likewise.
3040 (rest_of_clean_state): Skip inline entry markers.
3041 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3043 * gimple.c (gimple_build_debug_inline_entry): New.
3044 * gimple.h (enum gimple_debug_subcode): Add
3045 GIMPLE_DEBUG_INLINE_ENTRY.
3046 (gimple_build_debug_inline_entry): Declare.
3047 (gimple_debug_inline_entry_p): New.
3048 (gimple_debug_nonbind_marker_p): Adjust.
3049 * insn-notes.def (INLINE_ENTRY): New.
3050 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3051 inline entry marker notes.
3052 (print_insn): Likewise.
3053 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3054 (INSN_DEBUG_MARKER_KIND): Likewise.
3055 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3056 * tree-inline.c (expand_call_inline): Build and insert
3057 debug_inline_entry stmt.
3058 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3059 inline entry blocks early, if nonbind markers are enabled.
3060 (dump_scope_block): Dump fragment info.
3061 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3062 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3063 (gimple_build_debug_inline_entry): New.
3064 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3065 Enable/disable inline entry points too.
3066 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3067 (DEBUG_INSN): Describe inline entry markers.
3069 * common.opt (gvariable-location-views): New.
3070 (gvariable-location-views=incompat5): New.
3071 * config.in: Rebuilt.
3072 * configure: Rebuilt.
3073 * configure.ac: Test assembler for view support.
3074 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3075 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3076 * dwarf2out.c (var_loc_view): New typedef.
3077 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3078 (dwarf2out_locviews_in_attribute): New.
3079 (dwarf2out_locviews_in_loclist): New.
3080 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3081 (enum dw_line_info_opcode): Add LI_adv_address.
3082 (struct dw_line_info_table): Add view.
3083 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3084 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3085 (zero_view_p): New variable.
3086 (ZERO_VIEW_P): New macro.
3087 (output_asm_line_debug_info): New.
3088 (struct var_loc_node): Add view.
3089 (add_AT_view_list, AT_loc_list): New.
3090 (add_var_loc_to_decl): Add view param. Test it against last.
3091 (new_loc_list): Add view params. Record them.
3092 (AT_loc_list_ptr): Handle loc and view lists.
3093 (view_list_to_loc_list_val_node): New.
3094 (print_dw_val): Handle dw_val_class_view_list.
3095 (size_of_die): Likewise.
3096 (value_format): Likewise.
3097 (loc_list_has_views): New.
3098 (gen_llsym): Set vl_symbol too.
3099 (maybe_gen_llsym, skip_loc_list_entry): New.
3100 (dwarf2out_maybe_output_loclist_view_pair): New.
3101 (output_loc_list): Output view list or entries too.
3102 (output_view_list_offset): New.
3103 (output_die): Handle dw_val_class_view_list.
3104 (output_dwarf_version): New.
3105 (output_compilation_unit_header): Use it.
3106 (output_skeleton_debug_sections): Likewise.
3107 (output_rnglists, output_line_info): Likewise.
3108 (output_pubnames, output_aranges): Update version comments.
3109 (output_one_line_info_table): Output view numbers in asm comments.
3110 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3111 Call maybe_gen_llsym.
3112 (loc_list_from_tree_1): Adjust.
3113 (add_AT_location_description): Create view list attribute if
3114 needed, check it's absent otherwise.
3115 (convert_cfa_to_fb_loc_list): Adjust.
3116 (maybe_emit_file): Call output_asm_line_debug_info for test.
3117 (dwarf2out_var_location): Reset views as needed. Precompute
3118 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3119 attribute. Set view.
3120 (new_line_info_table): Reset next view.
3121 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3122 (dwarf2out_source_line): Likewise. Output view resets and labels to
3123 the assembler, or select appropriate line info opcodes.
3124 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3125 (optimize_string_length): Catch it. Adjust.
3126 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3127 dw_val_class_view_list, and remove it if no longer needed.
3128 (hash_loc_list): Hash view numbers.
3129 (loc_list_hasher::equal): Compare them.
3130 (optimize_location_lists): Check whether a view list symbol is
3131 needed, and whether the locview attribute is present, and
3132 whether they match. Remove the locview attribute if no longer
3134 (index_location_lists): Call skip_loc_list_entry for test.
3135 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3136 Use output_dwarf_version.
3137 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3138 (struct dw_val_node): Add val_view_list.
3139 * final.c (SEEN_NEXT_VIEW): New.
3140 (set_next_view_needed): New.
3141 (clear_next_view_needed): New.
3142 (maybe_output_next_view): New.
3143 (final_start_function): Rename to...
3144 (final_start_function_1): ... this. Take pointer to FIRST,
3145 add SEEN parameter. Emit param bindings in the initial view.
3146 (final_start_function): Reintroduce SEEN-less interface.
3147 (final): Rename to...
3148 (final_1): ... this. Take SEEN parameter. Output final pending
3149 next view at the end.
3150 (final): Reintroduce seen-less interface.
3151 (final_scan_insn): Output pending next view before switching
3152 sections or ending a block. Mark the next view as needed when
3153 outputting variable locations. Notify debug backend of section
3154 changes, and of location view changes.
3155 (rest_of_handle_final): Adjust.
3156 * toplev.c (process_options): Autodetect value for debug variable
3157 location views option. Warn on incompat5 without -gdwarf-5.
3158 * doc/invoke.texi (gvariable-location-views): New.
3159 (gvariable-location-views=incompat5): New.
3160 (gno-variable-location-views): New.
3162 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3164 PR tree-optimization/84136
3165 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3166 that the result of find_edge is non-NULL.
3168 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3171 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3172 storing integer register in SImode. Fix cost of 256 and 512
3173 byte aligned SSE register store.
3175 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3177 * config/i386/i386.c (ix86_multiplication_cost): Fix
3178 multiplication cost for TARGET_AVX512DQ.
3180 2018-02-08 Marek Polacek <polacek@redhat.com>
3182 PR tree-optimization/84238
3183 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3186 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3188 PR tree-optimization/84265
3189 * tree-vect-stmts.c (vectorizable_store): Don't treat
3190 VMAT_CONTIGUOUS accesses as grouped.
3191 (vectorizable_load): Likewise.
3193 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3195 PR tree-optimization/81635
3196 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3197 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3198 (test_round_for_mask): New functions.
3199 (wide_int_cc_tests): Call test_round_for_mask.
3200 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3201 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3202 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3203 range returned by get_range_info.
3205 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3208 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3209 * symtab.c: Include builtins.h
3210 (symtab_node::output_to_lto_symbol_table_p): Move here
3211 from lto-streamer-out.c:output_symbol_p.
3212 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3213 (output_symbol_p): Move all logic to symtab.c
3214 (produce_symtab): Update.
3216 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3218 * config/s390/s390-opts.h (enum indirect_branch): Define.
3219 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3220 (s390_indirect_branch_via_thunk)
3221 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3222 (enum s390_indirect_branch_type): Define.
3223 * config/s390/s390.c (struct s390_frame_layout, struct
3224 machine_function): Remove.
3225 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3226 (indirect_branch_table_label_no, indirect_branch_table_name):
3228 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3229 (enum s390_indirect_branch_option): Define.
3230 (s390_return_addr_from_memory): New function.
3231 (s390_handle_string_attribute): New function.
3232 (s390_attribute_table): Add new attribute handler.
3233 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3234 (s390_indirect_branch_via_thunk): New function.
3235 (s390_indirect_branch_via_inline_thunk): New function.
3236 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3237 sibling call optimization for non z10 compiles.
3238 (s390_emit_call): Force indirect branch target to be a single
3239 register. Add r1 clobber for non-z10 compiles.
3240 (s390_emit_epilogue): Emit return jump via return_use expander.
3241 (s390_reorg): Handle JUMP_INSNs as execute targets.
3242 (s390_option_override_internal): Perform validity checks for the
3243 new command line options.
3244 (s390_indirect_branch_attrvalue): New function.
3245 (s390_indirect_branch_settings): New function.
3246 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3247 (s390_output_indirect_thunk_function): New function.
3248 (s390_code_end): Implement target hook.
3249 (s390_case_values_threshold): Implement target hook.
3250 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3252 * config/s390/s390.h (struct s390_frame_layout)
3253 (struct machine_function): Move here from s390.c.
3254 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3255 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3256 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3257 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3258 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3259 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3260 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3261 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3262 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3263 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3264 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3265 (mnemonic attribute): Add values which aren't recognized
3267 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3268 pattern for branch conversion. Fix mnemonic attribute.
3269 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3270 indirect branch via thunk if requested.
3271 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3272 ("*indirect_jump"): Disable for branch conversion using out of
3274 ("indirect_jump_via_thunk<mode>_z10")
3275 ("indirect_jump_via_thunk<mode>")
3276 ("indirect_jump_via_inlinethunk<mode>_z10")
3277 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3278 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3279 ("casesi_jump_via_inlinethunk<mode>_z10")
3280 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3281 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3282 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3283 ("*indirect2_jump"): Disable for branch conversion.
3284 ("casesi_jump"): Turn into expander and expand patterns for branch
3286 ("return_use"): New expander.
3287 ("*return"): Emit return via thunk and rename it to ...
3288 ("*return<mode>"): ... this one.
3289 * config/s390/s390.opt: Add new options and and enum for the
3292 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3294 * lra-constraints.c (match_reload): Unconditionally use
3295 gen_lowpart_SUBREG, rather than selecting between that
3296 and equivalent gen_rtx_SUBREG code.
3298 2018-02-08 Richard Biener <rguenther@suse.de>
3300 PR tree-optimization/84233
3301 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3302 changed flag instead of boguously re-using phi_inserted.
3304 2018-02-08 Martin Jambor <mjambor@suse.cz>
3306 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3307 static local variables.
3309 2018-02-08 Richard Biener <rguenther@suse.de>
3311 PR tree-optimization/84278
3312 * tree-vect-stmts.c (vectorizable_store): When looking for
3313 smaller vector types to perform grouped strided loads/stores
3314 make sure the mode is supported by the target.
3315 (vectorizable_load): Likewise.
3317 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3319 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3320 Increase LDP/STP opportunities by adding adjacent callee-saves.
3322 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3324 PR rtl-optimization/84068
3325 PR rtl-optimization/83459
3326 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3328 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3330 PR tree-optimization/84224
3331 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3332 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3335 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3338 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3339 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3340 count, remove one USE.
3342 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3344 * doc/install.texi (Configuration): Document the
3345 --with-long-double-format={ibm,ieee} PowerPC configuration
3349 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3350 Convert from define_expand to be define_insn_and_split. Rework
3351 float/double/_Float128 conversions to QI/HI/SImode to work with
3352 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3353 conversions to QI/HImode types did a store and then a load to
3354 truncate the value. For conversions to VSX registers, don't split
3355 the insn, instead emit the code directly. Use the code iterator
3356 any_fix to combine signed and unsigned conversions.
3357 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3358 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3359 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3360 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3361 (fix_<mode>di2_hw): Likewise.
3362 (fixuns_<mode>di2_hw): Likewise.
3363 (fix_<mode>si2_hw): Likewise.
3364 (fixuns_<mode>si2_hw): Likewise.
3365 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3366 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3367 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3368 fix<uns>_trunc<SFDF:mode>si2_p8.
3369 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3371 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3372 (fix<uns>_<mode>_mem): Likewise.
3373 (fctiw<u>z_<mode>_mem): Likewise.
3374 (fix<uns>_<mode>_mem): Likewise.
3375 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3376 the register allocator from doing a direct move to the GPRs to do
3377 a store, and instead use the ISA 3.0 store byte/half-word from
3378 vector register instruction. For IEEE 128-bit floating point,
3379 also optimize stores of 32-bit ints.
3380 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3382 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3384 * genextract.c (push_pathstr_operand): New function to support
3386 (walk_rtx): Call push_pathstr_operand.
3387 (print_path): Support [a-zA-Z].
3389 2018-02-07 Richard Biener <rguenther@suse.de>
3391 PR tree-optimization/84037
3392 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3393 (cse_and_gimplify_to_preheader): Declare.
3394 (vect_get_place_in_interleaving_chain): Likewise.
3395 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3397 (_loop_vec_info::~_loop_vec_info): Delete it.
3398 (cse_and_gimplify_to_preheader): New function.
3399 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3400 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3401 (vectorizable_load): Likewise. For grouped stores always base
3402 the IV on the first element.
3403 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3404 condition before gimplifying.
3406 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3408 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3409 *DIV_EXPR and *MOD_EXPR.
3411 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3414 * config/i386/i386.c (ix86_option_override_internal): Mask out
3415 the CF_SET bit when checking -fcf-protection.
3417 2018-02-07 Tom de Vries <tom@codesourcery.com>
3420 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3423 2018-02-07 Richard Biener <rguenther@suse.de>
3425 PR tree-optimization/84204
3426 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3429 PR tree-optimization/84205
3430 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3431 special-case isl_ast_op_zdiv_r.
3433 PR tree-optimization/84223
3434 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3435 Only add conditions from within the region.
3436 (gather_bbs::after_dom_children): Adjust.
3438 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3441 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3442 * config/avr/avr.md: Only post-reload split REG-REG moves if
3443 either register is GENERAL_REG_P.
3445 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3447 PR tree-optimization/84235
3448 * tree-ssa-scopedtables.c
3449 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3450 if the subtraction is performed in floating point type where NaNs are
3451 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3452 build 1. Formatting fix.
3454 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3457 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3458 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3459 and skip it regardless of bb boundaries. Use CALL_P macro,
3460 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3461 unnecessarily, formatting fix.
3463 2018-02-06 Michael Collison <michael.collison@arm.com>
3465 * config/arm/thumb2.md:
3466 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3467 (*thumb_mov_notscc): Ditto.
3469 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3472 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3473 unsigned_fix, not "s".
3475 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3477 * configure.ac (gcc_fn_eh_frame_ro): New function.
3478 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3479 correct .eh_frame permissions.
3480 * configure: Regenerate.
3482 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3484 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3487 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3489 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3490 Display warning message for -mno-speculate-indirect-jumps.
3492 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3494 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3496 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3498 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3500 PR tree-optimization/84225
3501 * tree-eh.c (find_trapping_overflow): Only call
3502 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3504 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3507 * config/i386/i386.c: Reimplement the check of possible options
3508 -mibt/-mshstk conbination. Change error messages.
3509 * doc/invoke.texi: Fix a typo: remove extra '='.
3511 2018-02-06 Marek Polacek <polacek@redhat.com>
3513 PR tree-optimization/84228
3514 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3516 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3519 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3520 emitted arch directives.
3521 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3522 __ARM_FEATURE_COPROC before changing architectures.
3524 2018-02-06 Richard Biener <rguenther@suse.de>
3526 * config/i386/i386.c (print_reg): Fix typo.
3527 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3529 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3531 * configure: Regenerate.
3533 2018-02-05 Martin Sebor <msebor@redhat.com>
3535 PR tree-optimization/83369
3536 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3539 2018-02-05 Martin Liska <mliska@suse.cz>
3541 * doc/invoke.texi: Cherry-pick upstream r323995.
3543 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3545 * ira.c (ira_init_register_move_cost): Adjust comment.
3547 2018-02-05 Martin Liska <mliska@suse.cz>
3549 PR gcov-profile/84137
3550 * doc/gcov.texi: Fix typo in documentation.
3552 2018-02-05 Martin Liska <mliska@suse.cz>
3554 PR gcov-profile/83879
3555 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3556 using dlopen functionality.
3558 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3560 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3561 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3562 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3563 _mm_maskz_range_round_ss): New intrinsics.
3564 (__builtin_ia32_rangesd128_round)
3565 (__builtin_ia32_rangess128_round): Remove.
3566 (__builtin_ia32_rangesd128_mask_round,
3567 __builtin_ia32_rangess128_mask_round): New builtins.
3568 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3569 __builtin_ia32_rangess128_round): Remove.
3570 (__builtin_ia32_rangesd128_mask_round,
3571 __builtin_ia32_rangess128_mask_round): New builtins.
3572 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3573 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3574 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3575 "<round_saeonly_constraint>")): Changed to ...
3576 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3577 "<round_saeonly_scalar_constraint>")): ... this.
3578 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3579 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3580 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3581 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3582 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3584 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3586 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3588 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3589 Remove all values except native, 8540 and 8548.
3591 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3593 * config/i386/i386.c (ix86_output_function_return): Pass
3594 INVALID_REGNUM, instead of -1, as invalid register number to
3595 indirect_thunk_name and output_indirect_thunk.
3597 2018-02-02 Julia Koval <julia.koval@intel.com>
3599 * config.gcc: Add -march=icelake.
3600 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3601 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3602 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3603 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3604 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3605 (processor_target_table): Add icelake.
3606 (ix86_option_override_internal): Handle new PTAs.
3607 (get_builtin_code_for_version): Handle icelake.
3608 (M_INTEL_COREI7_ICELAKE): New.
3609 (fold_builtin_cpu): Handle icelake.
3610 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3611 * doc/invoke.texi: Add -march=icelake.
3613 2018-02-02 Julia Koval <julia.koval@intel.com>
3615 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3616 to wide_int_bitmask.
3617 * wide-int-bitmask.h: New.
3619 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3622 * config/i386/i386.md: Replace Pmode with word_mode in
3623 builtin_setjmp_setup and builtin_longjmp to support x32.
3625 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3629 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3631 (rs6000_supported_cpu_names): New static variable.
3632 (linux_cpu_translation_table): Likewise.
3633 (elf_platform) <cpu>: Define new static variable and use it.
3634 Translate kernel AT_PLATFORM name to canonical name if needed.
3635 Error if platform name is unknown.
3637 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3640 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3642 2018-02-01 Jeff Law <law@redhat.com>
3645 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3646 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3647 the scratch if RELEASE_VIA_POP is false.
3648 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3649 If we have to save a temporary register, decrement SIZE appropriately.
3650 Pass new arguments to release_scratch_register_on_entry.
3651 (ix86_adjust_stack_and_probe): Likewise.
3652 (ix86_emit_probe_stack_range): Pass new arguments to
3653 release_scratch_register_on_entry.
3655 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3657 PR rtl-optimization/84157
3658 * combine.c (change_zero_ext): Use REG_P predicate in
3659 front of HARD_REGISTER_P predicate.
3661 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3663 * config/avr/avr.c (avr_option_override): Move disabling of
3664 -fdelete-null-pointer-checks to...
3665 * common/config/avr/avr-common.c (avr_option_optimization_table):
3668 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3670 PR tree-optimization/81635
3671 * tree-data-ref.c (split_constant_offset_1): For types that
3672 wrap on overflow, try to use range info to prove that wrapping
3675 2018-02-01 Renlin Li <renlin.li@arm.com>
3678 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3680 (aarch64_register_move_cost): Likewise.
3681 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3683 (REG_CLASS_NAMES): Likewise.
3684 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3685 TAILCALL_ADDR_REGS. Remove IP registers.
3686 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3688 2018-02-01 Richard Biener <rguenther@suse.de>
3690 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3691 for specifying RPO order and allow NULL for that.
3692 * domwalk.c (dom_walker::dom_walker): Likewise.
3693 (dom_walker::walk): Handle NULL RPO order.
3694 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3696 (rewrite_update_dom_walker): Likewise.
3697 (mark_def_dom_walker): Likewise.
3699 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3701 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3702 (aarch64_maybe_expand_sve_subreg_move): Declare.
3703 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3704 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3706 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3707 that are semantically a reverse operation.
3708 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3709 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3710 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3712 (aarch64_can_change_mode_class): For big-endian, forbid changes
3713 between two SVE modes if they have different element sizes.
3715 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3717 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3718 the TImode handling for big-endian targets.
3720 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3722 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3723 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3725 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3726 Remove BSWAP handing for big-endian targets and use the form of
3727 LD1RQ appropariate for the mode.
3729 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3731 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3732 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3735 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3738 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3739 check for operands that need to go through aarch64_sve_reload_be.
3741 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3743 PR tree-optimization/81661
3744 PR tree-optimization/84117
3745 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3746 * tree-eh.c: Include gimplify.h.
3747 (find_trapping_overflow, replace_trapping_overflow,
3748 rewrite_to_non_trapping_overflow): New functions.
3749 * tree-vect-loop.c: Include tree-eh.h.
3750 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3751 * tree-data-ref.c: Include tree-eh.h.
3752 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3754 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3756 PR rtl-optimization/84123
3757 * combine.c (change_zero_ext): Check if hard register satisfies
3758 can_change_dest_mode before calling gen_lowpart_SUBREG.
3760 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3763 * ira.c (ira_init_register_move_cost): Remove assert.
3765 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3767 PR rtl-optimization/84071
3768 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3769 * doc/tm.texi: Regenerate.
3771 2018-01-31 Richard Biener <rguenther@suse.de>
3773 PR tree-optimization/84132
3774 * tree-data-ref.c (analyze_miv_subscript): Properly
3775 check whether evolution_function_is_affine_multivariate_p
3776 before calling gcd_of_steps_may_divide_p.
3778 2018-01-31 Julia Koval <julia.koval@intel.com>
3781 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3782 * config/i386/i386.md (rdpid_rex64) New.
3783 (rdpid): Make 32bit only.
3785 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3788 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3789 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3791 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3794 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3796 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3798 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3800 PR rtl-optimization/84071
3801 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3802 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3804 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3806 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3807 (arc_attribute_table): Add 'aux' attribute.
3808 (arc_in_small_data_p): Consider aux like variables.
3809 (arc_is_aux_reg_p): New function.
3810 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3811 (arc_get_aux_arg): New function.
3812 (prepare_move_operands): Handle aux-register access.
3813 (arc_handle_aux_attribute): New function.
3814 * doc/extend.texi (ARC Variable attributes): Add subsection.
3816 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3818 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3819 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3820 (arc_attribute_table): Add 'uncached' attribute.
3821 (arc_print_operand): Print '.di' flag for uncached memory
3823 (arc_in_small_data_p): Do not consider for small data the uncached
3825 (arc_is_uncached_mem_p): New function.
3826 * config/arc/predicates.md (compact_store_memory_operand): Check
3827 for uncached memory accesses.
3828 (nonvol_nonimm_operand): Likewise.
3829 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3831 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3834 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3835 falign-loops=): Add Optimization flag.
3837 2018-01-30 Jeff Law <law@redhat.com>
3840 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3841 INT_REGISTERS_SAVED. Check it prior to calling
3842 get_scratch_register_on_entry.
3843 (ix86_adjust_stack_and_probe): Similarly.
3844 (ix86_emit_probe_stack_range): Similarly.
3845 (ix86_expand_prologue): Corresponding changes.
3847 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3850 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3851 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3853 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3856 * lra-constraints.c (curr_insn_transform): Process AND in the
3859 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3861 PR rtl-optimization/83986
3862 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3863 dependence against last_pending_memory_flush in addition to
3866 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3868 PR tree-optimization/81611
3869 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3872 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3875 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3878 2018-01-30 Richard Biener <rguenther@suse.de>
3879 Jakub Jelinek <jakub@redhat.com>
3881 PR tree-optimization/84111
3882 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3883 inner loops added during recursion, as they don't have up-to-date
3886 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3889 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3890 (can_inline_edge_by_limits_p): ... here.
3891 (can_early_inline_edge_p, check_callers,
3892 update_caller_keys, update_callee_keys, recursive_inlining,
3893 add_new_edges_to_heap, speculation_useful_p,
3894 inline_small_functions,
3895 inline_small_functions, flatten_function,
3896 inline_to_all_callers_1): Update.
3898 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3900 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3903 2018-01-30 Richard Biener <rguenther@suse.de>
3905 PR tree-optimization/83008
3906 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3907 invariant and constant vector uses in stmts when they need
3910 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3913 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3914 * configure: Regenerate.
3916 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3918 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3920 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3921 Use gen_rtx_REG rather than gen_lowpart.
3923 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3925 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3926 rather than 0 when creating partial subregs.
3928 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3930 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3933 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3936 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3937 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3938 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3939 flags. This restores the settings used before the 2017-07-24.
3940 Turning off pre increment/decrement/modify allows IVOPTS to
3941 optimize DF/SF loops where the index is an int.
3943 2018-01-29 Richard Biener <rguenther@suse.de>
3944 Kelvin Nilsen <kelvin@gcc.gnu.org>
3947 * tree-vect-stmts.c (vectorizable_call): Don't call
3948 targetm.vectorize_builtin_md_vectorized_function if callee is
3951 2018-01-22 Carl Love <cel@us.ibm.com>
3953 * doc/extend.tex: Fix typo in second arg in
3954 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3956 2018-01-29 Richard Biener <rguenther@suse.de>
3958 PR tree-optimization/84086
3959 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3960 (flush_ssaname_freelist): When SSA names were released reset
3961 the SCEV hash table.
3963 2018-01-29 Richard Biener <rguenther@suse.de>
3965 PR tree-optimization/84057
3966 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3967 removed paths when removing edges.
3969 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3971 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3972 -mfunction-return=@var{choice}.
3974 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3977 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3979 (layout::print_source_line): Likewise.
3980 (test_get_line_width_without_trailing_whitespace): Add test cases.
3982 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3985 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3988 2018-01-26 Jim Wilson <jimw@sifive.com>
3990 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3992 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3995 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3997 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3998 and CMP + SUB-immediate -> SUBS.
4000 2018-01-26 Martin Sebor <msebor@redhat.com>
4002 PR tree-optimization/83896
4003 * tree-ssa-strlen.c (get_string_len): Rename...
4004 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4005 Avoid assuming length is constant.
4006 (handle_char_store): Use HOST_WIDE_INT for string length.
4008 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4011 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4012 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4014 2018-01-26 Richard Biener <rguenther@suse.de>
4016 PR rtl-optimization/84003
4017 * dse.c (record_store): Only record redundant stores when
4018 the earlier store aliases at least all accesses the later one does.
4020 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4022 PR rtl-optimization/83985
4023 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4024 REG_CFA_RESTORE insns.
4025 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4026 REG_CFA_RESTORE insns here.
4029 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4030 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4032 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4034 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4035 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4036 (arc_init): Likewise.
4037 (arc_override_options): Likewise.
4038 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4040 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4042 * config/arc/arc.h (TARGET_DBNZ): Define.
4043 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4044 properly set the tune attribute.
4045 (dbnz): Use TARGET_DBNZ guard.
4046 * config/arc/arc.opt (mtune): Add core3 option.
4048 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4050 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4051 recognize new pic like addresses.
4052 (arc_delegitimize_address): Clean up.
4054 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4056 * config/arc/arc-arches.def: Option mrf16 valid for all
4058 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4059 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4060 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4061 * config/arc/arc-tables.opt: Regenerate.
4062 * config/arc/arc.c (arc_conditional_register_usage): Handle
4063 reduced register file case.
4064 (arc_file_start): Set must have build attributes.
4065 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4067 * config/arc/arc.opt (mrf16): Add new option.
4068 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4069 * config/arc/genmultilib.awk: Handle new mrf16 option.
4070 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4071 * config/arc/t-multilib: Regenerate.
4072 * doc/invoke.texi (ARC Options): Document mrf16 option.
4074 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4076 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4077 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4078 (arc_attribute_table): Add 'secure_call' attribute.
4079 (arc_print_operand): Print secure call operand.
4080 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4082 (arc_is_secure_call_p): New function. * config/arc/arc.md
4083 (call_i): Add support for sjli instruction.
4084 (call_value_i): Likewise.
4085 * config/arc/constraints.md (Csc): New constraint.
4087 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4088 John Eric Martin <John.Martin@emmicro-us.com>
4090 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4091 * config/arc/arc.c (_arc_jli_section): New struct.
4092 (arc_jli_section): New type.
4093 (rc_jli_sections): New static variable.
4094 (arc_handle_jli_attribute): New function.
4095 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4096 (arc_file_end): New function.
4097 (TARGET_ASM_FILE_END): Define.
4098 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4099 (arc_add_jli_section): New function.
4100 (jli_call_scan): Likewise.
4101 (arc_reorg): Call jli_call_scan.
4102 (arc_output_addsi): Remove 'S' from printing asm operand.
4103 (arc_is_jli_call_p): New function.
4104 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4106 (movhi_insn): Likewise.
4107 (movsi_insn): Likewise.
4108 (movsi_set_cc_insn): Likewise.
4109 (loadqi_update): Likewise.
4110 (load_zeroextendqisi_update): Likewise.
4111 (load_signextendqisi_update): Likewise.
4112 (loadhi_update): Likewise.
4113 (load_zeroextendhisi_update): Likewise.
4114 (load_signextendhisi_update): Likewise.
4115 (loadsi_update): Likewise.
4116 (loadsf_update): Likewise.
4117 (movsicc_insn): Likewise.
4118 (bset_insn): Likewise.
4119 (bxor_insn): Likewise.
4120 (bclr_insn): Likewise.
4121 (bmsk_insn): Likewise.
4122 (bicsi3_insn): Likewise.
4123 (cmpsi_cc_c_insn): Likewise.
4124 (movsi_ne): Likewise.
4125 (movsi_cond_exec): Likewise.
4126 (clrsbsi2): Likewise.
4137 (call_i): Remove 'S' asm letter, add jli instruction.
4138 (call_value_i): Likewise.
4139 * config/arc/arc.op (mjli-always): New option.
4140 * config/arc/constraints.md (Cji): New constraint.
4141 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4143 (subsf3_fpx): Likewise.
4144 (mulsf3_fpx): Likewise.
4145 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4147 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4149 * doc/invoke.texi (ARC): Document mjli-always option.
4151 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4153 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4154 avoid addition with 0 and use incw and decw where possible.
4156 2018-01-26 Richard Biener <rguenther@suse.de>
4158 PR tree-optimization/81082
4159 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4160 association if it requires casting to unsigned.
4161 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4162 from fold_plusminus_mult_expr to catch important cases late when
4163 range info is available.
4165 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4167 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4168 * configure.ac (hidden_linkonce): New test.
4169 * configure: Regenerate.
4170 * config.in: Regenerate.
4172 2018-01-26 Julia Koval <julia.koval@intel.com>
4174 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4175 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4176 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4177 _mm_mask_bitshuffle_epi64_mask): Fix type.
4178 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4179 USI_FTYPE_V4DI_V4DI_USI): Remove.
4180 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4181 __builtin_ia32_vpshufbitqmb256_mask,
4182 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4183 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4184 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4186 2018-01-26 Alan Modra <amodra@gmail.com>
4189 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4190 UNSPEC_VBPERMQ. Sort other unspecs.
4192 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4194 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4196 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4199 * predict.c (drop_profile): Do not push/pop cfun; update also
4201 (handle_missing_profiles): Fix logic looking for zero profiles.
4203 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4206 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4207 on functions with #pragma omp declare simd or functions with simd
4209 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4210 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4211 Remove trailing \n from warning_at calls.
4213 2018-01-25 Tom de Vries <tom@codesourcery.com>
4216 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4217 for neutered workers.
4219 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4222 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4223 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4225 2018-01-24 Jeff Law <law@redhat.com>
4228 * i386.c (get_probe_interval): Move to earlier point.
4229 (ix86_compute_frame_layout): If -fstack-clash-protection and
4230 the frame is larger than the probe interval, then use pushes
4231 to save registers rather than reg->mem moves.
4232 (ix86_expand_prologue): Remove conditional for int_registers_saved
4235 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4238 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4239 min/max for never referenced object.
4241 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4244 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4246 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4247 attributes from DECL_ATTRIBUTES (decl) without affecting
4248 DECL_ATTRIBUTES (current_function_decl).
4249 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4250 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4252 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4254 PR tree-optimization/83979
4255 * fold-const.c (fold_comparison): Use constant_boolean_node
4256 instead of boolean_{true,false}_node.
4258 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4260 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4263 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4265 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4266 Simplify the clause that sets the length attribute.
4267 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4268 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4269 clause that sets the length attribute.
4270 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4272 2018-01-24 Tom de Vries <tom@codesourcery.com>
4275 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4276 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4277 Add strict parameter.
4278 (prevent_branch_around_nothing): Insert dummy insn between branch to
4279 label and label with no ptx insn inbetween.
4280 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4282 2018-01-24 Tom de Vries <tom@codesourcery.com>
4285 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4286 for neutered threads in warp.
4287 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4289 2018-01-24 Richard Biener <rguenther@suse.de>
4291 PR tree-optimization/83176
4292 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4295 2018-01-24 Richard Biener <rguenther@suse.de>
4297 PR tree-optimization/82819
4298 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4299 code generating pluses that are no-ops in the target precision.
4301 2018-01-24 Richard Biener <rguenther@suse.de>
4304 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4306 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4308 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4309 to merge probabilities.
4310 * predict.c (probably_never_executed): Also mark as cold functions
4311 with global 0 profile and guessed local profile.
4312 * profile-count.c (profile_probability::combine_with_count): New
4314 * profile-count.h (profile_probability::operator*,
4315 profile_probability::operator*=, profile_probability::operator/,
4316 profile_probability::operator/=): Reduce precision to adjusted
4317 and set value to guessed on contradictory divisions.
4318 (profile_probability::combine_with_freq): Remove.
4319 (profile_probability::combine_wiht_count): Declare.
4320 (profile_count::force_nonzero):: Set to adjusted.
4321 (profile_count::probability_in):: Set quality to adjusted.
4322 * tree-ssa-tail-merge.c (replace_block_by): Use
4325 2018-01-23 Andrew Waterman <andrew@sifive.com>
4326 Jim Wilson <jimw@sifive.com>
4328 * config/riscv/riscv.c (riscv_stack_boundary): New.
4329 (riscv_option_override): Set riscv_stack_boundary. Handle
4330 riscv_preferred_stack_boundary_arg.
4331 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4332 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4333 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4334 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4335 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4336 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4338 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4341 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4342 of struct ix86_frame.
4343 (ix86_expand_epilogue): Likewise. Add a local variable for
4344 the reg_save_offset field in struct ix86_frame.
4346 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4348 PR tree-optimization/82604
4349 * tree-loop-distribution.c (enum partition_kind): New enum item
4350 PKIND_PARTIAL_MEMSET.
4351 (partition_builtin_p): Support above new enum item.
4352 (generate_code_for_partition): Ditto.
4353 (compute_access_range): Differentiate cases that equality can be
4354 proven at all loops, the innermost loops or no loops.
4355 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4356 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4357 (finalize_partitions, distribute_loop): Don't fuse partition of
4358 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4359 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4362 2018-01-23 Martin Liska <mliska@suse.cz>
4364 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4365 order to ignore the predictor.
4366 (PRED_POLYMORPHIC_CALL): Likewise.
4367 (PRED_RECURSIVE_CALL): Likewise.
4369 2018-01-23 Martin Liska <mliska@suse.cz>
4371 * tree-profile.c (tree_profiling): Print function header to
4372 aware reader which function we are working on.
4373 * value-prof.c (gimple_find_values_to_profile): Do not print
4374 not interesting value histograms.
4376 2018-01-23 Martin Liska <mliska@suse.cz>
4378 * profile-count.h (enum profile_quality): Add
4379 profile_uninitialized as the first value. Do not number values
4380 as they are zero based.
4381 (profile_count::verify): Update sanity check.
4382 (profile_probability::verify): Likewise.
4384 2018-01-23 Nathan Sidwell <nathan@acm.org>
4386 * doc/invoke.texi (ffor-scope): Deprecate.
4388 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4390 PR tree-optimization/83510
4391 * domwalk.c (set_all_edges_as_executable): New function.
4392 (dom_walker::dom_walker): Convert bool param
4393 "skip_unreachable_blocks" to enum reachability. Move setup of
4394 edge flags to set_all_edges_as_executable and only do it when
4395 reachability is REACHABLE_BLOCKS.
4396 * domwalk.h (enum dom_walker::reachability): New enum.
4397 (dom_walker::dom_walker): Convert bool param
4398 "skip_unreachable_blocks" to enum reachability.
4399 (set_all_edges_as_executable): New decl.
4400 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4401 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4403 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4404 but converting true to REACHABLE_BLOCKS.
4405 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4407 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4408 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4409 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4411 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4412 if check_all_array_refs will be called.
4414 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4416 * tree.c (selftest::test_location_wrappers): Add more test
4419 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4421 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4422 (selftest::test_bit_in_range): Likewise.
4424 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4427 * doc/sourcebuild.texi (vect_float): Say that the selector
4428 only describes the situation when -funsafe-math-optimizations is on.
4429 (vect_float_strict): Document.
4431 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4433 PR tree-optimization/83965
4434 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4435 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4436 instead of checking only for a reduction.
4437 (vect_recog_widen_sum_pattern): Likewise.
4439 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4441 * predict.c (probably_never_executed): Only use precise profile info.
4442 (compute_function_frequency): Skip after inlining hack since we now
4443 have quality checking.
4445 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4447 * profile-count.h (profile_probability::very_unlikely,
4448 profile_probability::unlikely, profile_probability::even): Set
4449 precision to guessed.
4451 2018-01-23 Richard Biener <rguenther@suse.de>
4453 PR tree-optimization/83963
4454 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4455 Properly terminate dominator walk when crossing the exit edge not
4456 when visiting its source block.
4458 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4461 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4462 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4464 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4466 PR tree-optimization/83957
4467 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4468 semicolon after for body surrounded by braces.
4470 PR tree-optimization/83081
4471 * profile-count.h (profile_probability::split): New method.
4472 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4473 Use profile_probability::split.
4474 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4475 when splitting a single conditional jump into 2.
4477 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4479 PR tree-optimization/69452
4480 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4483 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4485 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4486 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4487 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4489 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4491 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4492 * config/rl78/rl78.md (movdi): New define_expand.
4493 * config/rl78/rl78.c (rl78_split_movdi): New function.
4495 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4498 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4500 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4501 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4502 128-bit to produce an UNSPEC move to get the double word with the
4503 signbit and then a shift directly to do signbit.
4504 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4505 implementation with a new version that just does either a direct
4506 move or a regular move. Move memory interface to separate insns.
4507 Move insns so they are next to the expander.
4508 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4509 with signbit move. Split big and little endian case.
4510 (signbit<mode>2_dm_mem_le): Likewise.
4511 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4512 (signbit<mode>2_dm2): Likewise.
4514 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4516 * config/rl78/rl78.md (anddi3): New define_expand.
4518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4520 * config/rl78/rl78.md (umindi3): New define_expand.
4522 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4524 * config/rl78/rl78.md (smindi3): New define_expand.
4526 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4528 * config/rl78/rl78.md (smaxdi3): New define_expand.
4530 2018-01-22 Carl Love <cel@us.ibm.com>
4532 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4533 LVX_V1TI): Add macro expansion.
4534 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4535 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4536 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4537 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4538 Change check to determine if the instruction is a byte reversing
4539 entry. Fix typo in comment.
4540 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4541 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4542 Add def_builtin calls for new builtins.
4543 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4544 Add define_insn expansion.
4546 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4548 * config/rl78/rl78.md (umaxdi3): New define_expand.
4550 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4552 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4553 for non-QImode registers.
4555 2018-01-22 Richard Biener <rguenther@suse.de>
4557 PR tree-optimization/83963
4558 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4559 including the loop exit block.
4560 (scop_detection::merge_sese): Likewise.
4561 (scop_detection::add_scop): Do it here instead.
4563 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4565 * doc/sourcebuild.texi (arm_softfloat): Document.
4567 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4570 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4571 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4572 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4574 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4575 David Edelsohn <dje.gcc@gmail.com>
4578 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4579 Change "crset eq" to "crset 2".
4580 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4581 (*call_indirect_aix<mode>_nospec): Likewise.
4582 (*call_value_indirect_aix<mode>_nospec): Likewise.
4583 (*call_indirect_elfv2<mode>_nospec): Likewise.
4584 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4585 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4586 change assembly output from . to $.
4587 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4588 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4589 (*tablejump<mode>_internal1_nospec): Likewise.
4591 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4594 * config/sh/sh_optimize_sett_clrt.cc:
4595 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4597 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4599 PR tree-optimization/83940
4600 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4601 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4602 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4603 use it to pass back the definition type.
4604 (vect_check_store_rhs): Likewise rhs_dt_out.
4605 (vect_build_gather_load_calls): Add a mask_dt argument and use
4606 it instead of a call to vect_is_simple_use.
4607 (vectorizable_store): Update calls to vect_check_load_store_mask
4608 and vect_check_store_rhs. Use the dt returned by the latter instead
4609 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4610 instead of calls to vect_is_simple_use. Pass the scalar rather
4611 than the vector operand to vect_is_simple_use when handling
4612 second and subsequent copies of an rhs value.
4613 (vectorizable_load): Update calls to vect_check_load_store_mask
4614 and vect_build_gather_load_calls. Use the cached mask_dt and
4615 gs_info.offset_dt instead of calls to vect_is_simple_use.
4617 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4620 * tree-emutls.c: Include gimplify.h.
4621 (lower_emutls_2): New function.
4622 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4623 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4624 it before further processing.
4627 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4628 UINTVAL (trueop1) instead of INTVAL (op1).
4630 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4634 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4635 INCOMING_FRAME_SP_OFFSET if not defined.
4636 (scan_trace): Add ENTRY argument. If true and
4637 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4638 emit a note to adjust the CFA offset.
4639 (create_cfi_notes): Adjust scan_trace callers.
4640 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4641 INCOMING_FRAME_SP_OFFSET in the CIE.
4642 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4643 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4645 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4646 * doc/tm.texi: Regenerated.
4648 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4650 PR rtl-optimization/83147
4651 * lra-constraints.c (remove_inheritance_pseudos): Use
4652 lra_substitute_pseudo_within_insn.
4654 2018-01-19 Tom de Vries <tom@codesourcery.com>
4655 Cesar Philippidis <cesar@codesourcery.com>
4658 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4660 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4663 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4664 spaces for function labels.
4666 2018-01-19 Martin Liska <mliska@suse.cz>
4668 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4669 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4670 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4671 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4672 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4673 (PRED_CONST_RETURN): Change from 69 to 65.
4674 (PRED_NULL_RETURN): Change from 91 to 71.
4675 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4676 (PRED_LOOP_GUARD): Change from 66 to 73.
4678 2018-01-19 Martin Liska <mliska@suse.cz>
4680 * predict.c (predict_insn_def): Add new assert.
4681 (struct branch_predictor): Change type to signed integer.
4682 (test_prediction_value_range): Amend test to cover
4684 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4685 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4686 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4687 (PRED_LOOP_IV_COMPARE): Likewise.
4688 * predict.h (PROB_UNINITIALIZED): Define new constant.
4690 2018-01-19 Martin Liska <mliska@suse.cz>
4692 * predict.c (dump_prediction): Add new format for
4693 analyze_brprob.py script which is enabled with -details
4695 * profile-count.h (precise_p): New function.
4697 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4699 PR tree-optimization/83922
4700 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4701 there are no statements that need masking.
4702 (vect_active_double_reduction_p): New function.
4703 (vect_analyze_loop_operations): Use it when handling phis that
4704 are not in the loop header.
4706 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4708 PR tree-optimization/83914
4709 * tree-vect-loop.c (vectorizable_induction): Don't convert
4710 init_expr or apply the peeling adjustment for inductions
4711 that are nested within the vectorized loop.
4713 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4715 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4718 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4722 * function.h (gimplify_parameters): Add gimple_seq * argument.
4723 * function.c: Include gimple.h and options.h.
4724 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4725 for the added local temporaries if needed.
4726 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4727 if there are any parameter cleanups, wrap whole body into a
4728 try/finally with the cleanups.
4730 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4733 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4734 Use GET_MODE_CLASS for scalar floating point.
4736 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4740 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4741 Fix call of call_cgraph_insertion_hooks.
4743 2018-01-18 Martin Sebor <msebor@redhat.com>
4745 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4747 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4750 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4753 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4756 * common.opt: (-ffile-prefix-map): New option.
4757 * opts.c (common_handle_option): Defer it.
4758 * opts-global.c (handle_common_deferred_options): Handle it.
4759 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4760 * file-prefix-map.h: New file.
4761 (remap_debug_filename, add_debug_prefix_map): ...here.
4762 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4763 * final.c (debug_prefix_map, add_debug_prefix_map
4764 remap_debug_filename): Move to...
4765 * file-prefix-map.c: New file.
4766 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4767 generalize, get rid of alloca(), use strrchr() instead of strchr().
4768 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4769 Implement in terms of add_prefix_map().
4770 (remap_macro_filename, remap_debug_filename): Implement in term of
4772 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4773 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4774 * dbxout.c: Include file-prefix-map.h.
4775 * varasm.c: Likewise.
4776 * vmsdbgout.c: Likewise.
4777 * xcoffout.c: Likewise.
4778 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4779 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4780 * doc/invoke.texi (-ffile-prefix-map): Document.
4781 (-fdebug-prefix-map): Update description.
4783 2018-01-18 Martin Liska <mliska@suse.cz>
4785 * config/i386/i386.c (indirect_thunk_name): Document that also
4787 (output_indirect_thunk): Document why both instructions
4788 (pause and lfence) are generated.
4790 2018-01-18 Richard Biener <rguenther@suse.de>
4792 PR tree-optimization/83887
4793 * graphite-scop-detection.c
4794 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4795 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4796 (scop_detection::merge_sese): Re-implement with a flood-fill
4797 algorithm that properly finds a SESE region if it exists.
4799 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4802 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4803 pointer_diff optimizations use view_convert instead of convert.
4805 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4807 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4808 Generate different code for -mno-speculate-indirect-jumps.
4809 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4810 (*call_indirect_aix<mode>): Disable for
4811 -mno-speculate-indirect-jumps.
4812 (*call_indirect_aix<mode>_nospec): New define_insn.
4813 (*call_value_indirect_aix<mode>): Disable for
4814 -mno-speculate-indirect-jumps.
4815 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4816 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4817 -mno-speculate-indirect-jumps.
4818 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4820 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4822 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4823 long double type, set the flags for noting the default long double
4824 type, even if we don't pass or return a long double type.
4826 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4829 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4832 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4834 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4835 support for merge[hl].
4836 (fold_mergehl_helper): New helper function.
4837 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4838 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4839 (altivec_vmrglw_direct): Add xxmrglw insn.
4841 2018-01-17 Andrew Waterman <andrew@sifive.com>
4843 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4844 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4846 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4849 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4850 call the lto_location_cache before reading the
4851 DECL_SOURCE_LOCATION of the types.
4853 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4854 Richard Sandiford <richard.sandiford@linaro.org>
4856 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4857 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4858 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4859 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4860 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4862 * config/aarch64/constraints.md (aarch64_movti_operand):
4864 * config/aarch64/predicates.md (Uti): Add new constraint.
4866 2018-01-17 Carl Love <cel@us.ibm.com>
4868 * config/rs6000/vsx.md (define_expand xl_len_r,
4869 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4870 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4872 (define_expand, define_insn): Move the shift left from the
4873 define_insn to the define_expand for lxvl and stxvl instructions.
4874 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4875 and XL_LEN_R definitions to PURE.
4877 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4879 * config/i386/i386.c (indirect_thunk_name): Declare regno
4880 as unsigned int. Compare regno with INVALID_REGNUM.
4881 (output_indirect_thunk): Ditto.
4882 (output_indirect_thunk_function): Ditto.
4883 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4884 in the call to output_indirect_thunk_function.
4886 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4889 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4890 rather than the size of inner_type to determine the stack slot size
4891 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4893 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4896 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4899 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4901 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4902 endian Linux systems to optionally enable multilibs for selecting
4903 the long double type if the user configured an explicit type.
4904 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4905 have no long double multilibs if not defined.
4906 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4907 warn if the user used -mabi={ieee,ibm}longdouble and we built
4908 multilibs for long double.
4909 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4910 appropriate multilib option.
4911 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4913 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4914 for building long double multilibs.
4915 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4917 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4919 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4922 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4924 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4927 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4930 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4933 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4935 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4936 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4938 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4940 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4941 different rtl trees depending on TARGET_64BIT.
4942 (rs6000_gen_lvx): Likewise.
4944 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4946 * config/visium/visium.md (nop): Tweak comment.
4947 (hazard_nop): Likewise.
4949 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4951 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4952 -mspeculate-indirect-jumps.
4953 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4954 for -mno-speculate-indirect-jumps.
4955 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4956 (*call_value_indirect_elfv2<mode>): Disable for
4957 -mno-speculate-indirect-jumps.
4958 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4959 (indirect_jump): Emit different RTL for
4960 -mno-speculate-indirect-jumps.
4961 (*indirect_jump<mode>): Disable for
4962 -mno-speculate-indirect-jumps.
4963 (*indirect_jump<mode>_nospec): New define_insn.
4964 (tablejump): Emit different RTL for
4965 -mno-speculate-indirect-jumps.
4966 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4967 (tablejumpsi_nospec): New define_expand.
4968 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4969 (tablejumpdi_nospec): New define_expand.
4970 (*tablejump<mode>_internal1): Disable for
4971 -mno-speculate-indirect-jumps.
4972 (*tablejump<mode>_internal1_nospec): New define_insn.
4973 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4976 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4978 * caller-save.c (insert_save): Drop unnecessary parameter. All
4981 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4982 Richard Biener <rguenth@suse.de>
4985 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4986 return early, inline manually is_gimple_sizepos. Make sure if we
4987 call gimplify_expr we don't end up with a gimple constant.
4988 * tree.c (variably_modified_type_p): Don't return true for
4989 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4990 * gimplify.h (is_gimple_sizepos): Remove.
4992 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4994 PR tree-optimization/83857
4995 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4996 vectorizable_live_operation for pure SLP statements.
4997 (vectorizable_live_operation): Handle PHIs.
4999 2018-01-16 Richard Biener <rguenther@suse.de>
5001 PR tree-optimization/83867
5002 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5003 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5005 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5008 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5009 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5010 If off is not INTEGER_CST, issue a may not be aligned warning
5011 rather than isn't aligned. Use isn%'t rather than isn't.
5012 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5014 <case MULT_EXPR>: Improve the case when bottom and one of the
5015 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5016 operand, in that case check if the other operand is multiple of
5017 bottom divided by the INTEGER_CST operand.
5019 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5022 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5023 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5024 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5025 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5026 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5027 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5028 (pa_function_arg_size): New function.
5030 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5032 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5033 in a separate statement.
5035 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5037 PR tree-optimization/83847
5038 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5039 group gathers and scatters.
5041 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5043 PR rtl-optimization/86620
5044 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5046 PR rtl-optimization/83213
5047 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5048 to last if both are JUMP_INSNs.
5050 PR tree-optimization/83843
5051 * gimple-ssa-store-merging.c
5052 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5053 store_immediate_info for bswap/nop orig_stores.
5055 2018-01-15 Andrew Waterman <andrew@sifive.com>
5057 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5059 <UDIV>: Increase cost if !TARGET_DIV.
5061 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5063 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5064 (define_attr "cr_logical_3op"): New.
5065 (cceq_ior_compare): Adjust.
5066 (cceq_ior_compare_complement): Adjust.
5067 (*cceq_rev_compare): Adjust.
5068 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5069 (is_cracked_insn): Adjust.
5070 (insn_must_be_first_in_group): Adjust.
5071 * config/rs6000/40x.md: Adjust.
5072 * config/rs6000/440.md: Adjust.
5073 * config/rs6000/476.md: Adjust.
5074 * config/rs6000/601.md: Adjust.
5075 * config/rs6000/603.md: Adjust.
5076 * config/rs6000/6xx.md: Adjust.
5077 * config/rs6000/7450.md: Adjust.
5078 * config/rs6000/7xx.md: Adjust.
5079 * config/rs6000/8540.md: Adjust.
5080 * config/rs6000/cell.md: Adjust.
5081 * config/rs6000/e300c2c3.md: Adjust.
5082 * config/rs6000/e500mc.md: Adjust.
5083 * config/rs6000/e500mc64.md: Adjust.
5084 * config/rs6000/e5500.md: Adjust.
5085 * config/rs6000/e6500.md: Adjust.
5086 * config/rs6000/mpc.md: Adjust.
5087 * config/rs6000/power4.md: Adjust.
5088 * config/rs6000/power5.md: Adjust.
5089 * config/rs6000/power6.md: Adjust.
5090 * config/rs6000/power7.md: Adjust.
5091 * config/rs6000/power8.md: Adjust.
5092 * config/rs6000/power9.md: Adjust.
5093 * config/rs6000/rs64.md: Adjust.
5094 * config/rs6000/titan.md: Adjust.
5096 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5098 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5099 ix86_indirect_branch_register logic.
5101 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5103 * config/i386/constraints.md (Bs): Update
5104 ix86_indirect_branch_register check. Don't check
5105 ix86_indirect_branch_register with GOT_memory_operand.
5107 * config/i386/predicates.md (GOT_memory_operand): Don't check
5108 ix86_indirect_branch_register here.
5109 (GOT32_symbol_operand): Likewise.
5111 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5113 * config/i386/predicates.md (constant_call_address_operand):
5114 Rewrite ix86_indirect_branch_register logic.
5115 (sibcall_insn_operand): Likewise.
5117 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5119 * config/i386/constraints.md (Bs): Replace
5120 ix86_indirect_branch_thunk_register with
5121 ix86_indirect_branch_register.
5123 * config/i386/i386.md (indirect_jump): Likewise.
5124 (tablejump): Likewise.
5125 (*sibcall_memory): Likewise.
5126 (*sibcall_value_memory): Likewise.
5127 Peepholes of indirect call and jump via memory: Likewise.
5128 * config/i386/i386.opt: Likewise.
5129 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5130 (GOT_memory_operand): Likewise.
5131 (call_insn_operand): Likewise.
5132 (sibcall_insn_operand): Likewise.
5133 (GOT32_symbol_operand): Likewise.
5135 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5138 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5139 type rather than type addr's type points to.
5140 (expand_omp_atomic_mutex): Likewise.
5141 (expand_omp_atomic): Likewise.
5143 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5146 * config/i386/i386.c (output_indirect_thunk_function): Use
5147 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5148 for __x86_return_thunk.
5150 2018-01-15 Richard Biener <rguenther@suse.de>
5153 * expmed.c (extract_bit_field_1): Fix typo.
5155 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5158 * config/arm/iterators.md (VF): New mode iterator.
5159 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5160 Remove integer-related logic from pattern.
5161 (neon_vabd<mode>_3): Likewise.
5163 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5166 * common.opt (fstrict-overflow): No longer an alias.
5167 (fwrapv-pointer): New option.
5168 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5169 also for pointer types based on flag_wrapv_pointer.
5170 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5171 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5172 opts->x_flag_wrapv got set.
5173 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5174 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5175 POINTER_TYPE_OVERFLOW_UNDEFINED.
5176 * match.pd: Likewise in address comparison pattern.
5177 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5179 2018-01-15 Richard Biener <rguenther@suse.de>
5182 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5183 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5184 Reset type names to their identifier if their TYPE_DECL doesn't
5185 have linkage (and thus is used for ODR and devirt).
5186 (save_debug_info_for_decl): Remove.
5187 (save_debug_info_for_type): Likewise.
5188 (add_tree_to_fld_list): Adjust.
5189 * tree-pretty-print.c (dump_generic_node): Make dumping of
5190 type names more robust.
5192 2018-01-15 Richard Biener <rguenther@suse.de>
5194 * BASE-VER: Bump to 8.0.1.
5196 2018-01-14 Martin Sebor <msebor@redhat.com>
5199 * builtins.c (check_access): Avoid warning when the no-warning bit
5202 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5204 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5205 * ira-color (allocno_hard_regs_compare): Likewise.
5207 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5210 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5211 Use .pushsection/.popsection.
5213 2018-01-14 Martin Sebor <msebor@redhat.com>
5216 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5218 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5220 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5221 entry from extra_headers.
5222 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5223 extra_headers, make the list bitwise identical to the i?86-*-* one.
5225 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5227 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5228 -mcmodel=large with -mindirect-branch=thunk,
5229 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5230 -mfunction-return=thunk-extern.
5231 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5232 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5233 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5235 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5237 * config/i386/i386.c (print_reg): Print the name of the full
5238 integer register without '%'.
5239 (ix86_print_operand): Handle 'V'.
5240 * doc/extend.texi: Document 'V' modifier.
5242 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5244 * config/i386/constraints.md (Bs): Disallow memory operand for
5245 -mindirect-branch-register.
5247 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5248 (GOT_memory_operand): Likewise.
5249 (call_insn_operand): Likewise.
5250 (sibcall_insn_operand): Likewise.
5251 (GOT32_symbol_operand): Likewise.
5252 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5253 for -mindirect-branch-register.
5254 (tablejump): Likewise.
5255 (*sibcall_memory): Likewise.
5256 (*sibcall_value_memory): Likewise.
5257 Disallow peepholes of indirect call and jump via memory for
5258 -mindirect-branch-register.
5259 (*call_pop): Replace m with Bw.
5260 (*call_value_pop): Likewise.
5261 (*sibcall_pop_memory): Replace m with Bs.
5262 * config/i386/i386.opt (mindirect-branch-register): New option.
5263 * doc/invoke.texi: Document -mindirect-branch-register option.
5265 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5267 * config/i386/i386-protos.h (ix86_output_function_return): New.
5268 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5269 set function_return_type.
5270 (indirect_thunk_name): Add ret_p to indicate thunk for function
5272 (output_indirect_thunk_function): Pass false to
5273 indirect_thunk_name.
5274 (ix86_output_indirect_branch_via_reg): Likewise.
5275 (ix86_output_indirect_branch_via_push): Likewise.
5276 (output_indirect_thunk_function): Create alias for function
5277 return thunk if regno < 0.
5278 (ix86_output_function_return): New function.
5279 (ix86_handle_fndecl_attribute): Handle function_return.
5280 (ix86_attribute_table): Add function_return.
5281 * config/i386/i386.h (machine_function): Add
5282 function_return_type.
5283 * config/i386/i386.md (simple_return_internal): Use
5284 ix86_output_function_return.
5285 (simple_return_internal_long): Likewise.
5286 * config/i386/i386.opt (mfunction-return=): New option.
5287 (indirect_branch): Mention -mfunction-return=.
5288 * doc/extend.texi: Document function_return function attribute.
5289 * doc/invoke.texi: Document -mfunction-return= option.
5291 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5293 * config/i386/i386-opts.h (indirect_branch): New.
5294 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5295 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5296 with local indirect jump when converting indirect call and jump.
5297 (ix86_set_indirect_branch_type): New.
5298 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5299 (indirectlabelno): New.
5300 (indirect_thunk_needed): Likewise.
5301 (indirect_thunk_bnd_needed): Likewise.
5302 (indirect_thunks_used): Likewise.
5303 (indirect_thunks_bnd_used): Likewise.
5304 (INDIRECT_LABEL): Likewise.
5305 (indirect_thunk_name): Likewise.
5306 (output_indirect_thunk): Likewise.
5307 (output_indirect_thunk_function): Likewise.
5308 (ix86_output_indirect_branch_via_reg): Likewise.
5309 (ix86_output_indirect_branch_via_push): Likewise.
5310 (ix86_output_indirect_branch): Likewise.
5311 (ix86_output_indirect_jmp): Likewise.
5312 (ix86_code_end): Call output_indirect_thunk_function if needed.
5313 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5315 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5316 (ix86_attribute_table): Add indirect_branch.
5317 * config/i386/i386.h (machine_function): Add indirect_branch_type
5318 and has_local_indirect_jump.
5319 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5321 (tablejump): Likewise.
5322 (*indirect_jump): Use ix86_output_indirect_jmp.
5323 (*tablejump_1): Likewise.
5324 (simple_return_indirect_internal): Likewise.
5325 * config/i386/i386.opt (mindirect-branch=): New option.
5326 (indirect_branch): New.
5329 (thunk-inline): Likewise.
5330 (thunk-extern): Likewise.
5331 * doc/extend.texi: Document indirect_branch function attribute.
5332 * doc/invoke.texi: Document -mindirect-branch= option.
5334 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5337 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5339 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5341 * ipa-inline.c (want_inline_small_function_p): Return false if
5342 inlining has already failed with CIF_FINAL_ERROR.
5343 (update_caller_keys): Call want_inline_small_function_p before
5345 (update_callee_keys): Likewise.
5347 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5349 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5351 (rs6000_quadword_masked_address_p): Likewise.
5352 (quad_aligned_load_p): Likewise.
5353 (quad_aligned_store_p): Likewise.
5354 (const_load_sequence_p): Add comment to describe the outer-most loop.
5355 (mimic_memory_attributes_and_flags): New function.
5356 (rs6000_gen_stvx): Likewise.
5357 (replace_swapped_aligned_store): Likewise.
5358 (rs6000_gen_lvx): Likewise.
5359 (replace_swapped_aligned_load): Likewise.
5360 (replace_swapped_load_constant): Capitalize argument name in
5361 comment describing this function.
5362 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5363 and stores that access quad-word aligned addresses and replace
5364 with stvx or lvx instructions when appropriate.
5365 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5366 New function prototype.
5367 (rs6000_quadword_masked_address_p): Likewise.
5368 (rs6000_gen_lvx): Likewise.
5369 (rs6000_gen_stvx): Likewise.
5370 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5371 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5372 when memory address is aligned.
5373 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5374 this split to select lvx instruction when memory address is aligned.
5375 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5376 instruction when memory address is aligned.
5377 (*vsx_le_perm_load_v16qi): Likewise.
5378 (four unnamed splitters): Modify to select the stvx instruction
5379 when memory is aligned.
5381 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5383 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5384 which appears in the queue multiple times.
5386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5387 Alan Hayward <alan.hayward@arm.com>
5388 David Sherwood <david.sherwood@arm.com>
5390 * tree-vectorizer.h (vec_lower_bound): New structure.
5391 (_loop_vec_info): Add check_nonzero and lower_bounds.
5392 (LOOP_VINFO_CHECK_NONZERO): New macro.
5393 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5394 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5395 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5396 fields. Make seg_len the distance travelled, not including the
5398 (dr_direction_indicator): Declare.
5399 (dr_zero_step_indicator): Likewise.
5400 (dr_known_forward_stride_p): Likewise.
5401 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5403 (runtime_alias_check_p): Allow runtime alias checks with
5405 (operator ==): Compare access_size and align.
5406 (prune_runtime_alias_test_list): Rework for new distinction between
5407 the access_size and seg_len.
5408 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5410 (get_segment_min_max): New function.
5411 (create_intersect_range_checks): Use it.
5412 (dr_step_indicator): New function.
5413 (dr_direction_indicator): Likewise.
5414 (dr_zero_step_indicator): Likewise.
5415 (dr_known_forward_stride_p): Likewise.
5416 * tree-loop-distribution.c (data_ref_segment_size): Return
5417 DR_STEP * (niters - 1).
5418 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5420 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5421 (vect_preserves_scalar_order_p): New function, split out from...
5422 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5423 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5424 (vect_vfa_access_size): New function.
5425 (vect_vfa_align): Likewise.
5426 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5427 (dump_lower_bound): New function.
5428 (vect_check_lower_bound): Likewise.
5429 (vect_small_gap_p): Likewise.
5430 (vectorizable_with_step_bound_p): Likewise.
5431 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5432 depencies if the vectorization factor is 1. Convert the checks
5433 for nonzero steps into checks on the bounds of DR_STEP. Try using
5434 a bunds check for variable steps if the minimum required step is
5435 relatively small. Update calls to the dr_with_seg_len
5436 constructor and to vect_compile_time_alias.
5437 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5439 (vect_loop_versioning): Call it.
5440 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5442 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5445 Alan Hayward <alan.hayward@arm.com>
5446 David Sherwood <david.sherwood@arm.com>
5448 * doc/sourcebuild.texi (vect_scatter_store): Document.
5449 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5451 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5453 * genopinit.c (main): Add supports_vec_scatter_store and
5454 supports_vec_scatter_store_cached to target_optabs.
5455 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5456 IFN_MASK_SCATTER_STORE.
5457 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5459 * internal-fn.h (internal_store_fn_p): Declare.
5460 (internal_fn_stored_value_index): Likewise.
5461 * internal-fn.c (scatter_store_direct): New macro.
5462 (expand_scatter_store_optab_fn): New function.
5463 (direct_scatter_store_optab_supported_p): New macro.
5464 (internal_store_fn_p): New function.
5465 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5466 IFN_MASK_SCATTER_STORE.
5467 (internal_fn_mask_index): Likewise.
5468 (internal_fn_stored_value_index): New function.
5469 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5471 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5472 * optabs-query.c (supports_vec_scatter_store_p): New function.
5473 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5474 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5475 true for scatter stores.
5476 (vect_gather_scatter_fn_p): Handle scatter stores too.
5477 (vect_check_gather_scatter): Consider using scatter stores if
5478 supports_vec_scatter_store_p.
5479 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5481 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5482 internal_fn_stored_value_index.
5483 (check_load_store_masking): Handle scatter stores too.
5484 (vect_get_store_rhs): Make public.
5485 (vectorizable_call): Use internal_store_fn_p.
5486 (vectorizable_store): Handle scatter store internal functions.
5487 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5488 when deciding whether the end of the group has been reached.
5489 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5490 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5491 (mask_scatter_store<mode>): New insns.
5493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5494 Alan Hayward <alan.hayward@arm.com>
5495 David Sherwood <david.sherwood@arm.com>
5497 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5498 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5499 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5501 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5502 Use vect_truncate_gather_scatter_offset if we can't treat the
5503 operation as a normal gather load or scatter store.
5504 (get_group_load_store_type): Take the gather_scatter_info
5505 as argument. Try using a gather load or scatter store for
5506 single-element groups.
5507 (get_load_store_type): Update calls to get_group_load_store_type
5508 and vect_use_strided_gather_scatters_p.
5510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5511 Alan Hayward <alan.hayward@arm.com>
5512 David Sherwood <david.sherwood@arm.com>
5514 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5515 optional tree argument.
5516 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5518 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5519 but continue to use the current value as a fallback.
5520 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5521 to compare the updates.
5522 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5523 (get_load_store_type): Use it when handling a strided access.
5524 (vect_get_strided_load_store_ops): New function.
5525 (vect_get_data_ptr_increment): Likewise.
5526 (vectorizable_load): Handle strided gather loads. Always pass
5527 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5530 Alan Hayward <alan.hayward@arm.com>
5531 David Sherwood <david.sherwood@arm.com>
5533 * doc/md.texi (gather_load@var{m}): Document.
5534 (mask_gather_load@var{m}): Likewise.
5535 * genopinit.c (main): Add supports_vec_gather_load and
5536 supports_vec_gather_load_cached to target_optabs.
5537 * optabs-tree.c (init_tree_optimization_optabs): Use
5538 ggc_cleared_alloc to allocate target_optabs.
5539 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5540 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5542 * internal-fn.h (internal_load_fn_p): Declare.
5543 (internal_gather_scatter_fn_p): Likewise.
5544 (internal_fn_mask_index): Likewise.
5545 (internal_gather_scatter_fn_supported_p): Likewise.
5546 * internal-fn.c (gather_load_direct): New macro.
5547 (expand_gather_load_optab_fn): New function.
5548 (direct_gather_load_optab_supported_p): New macro.
5549 (direct_internal_fn_optab): New function.
5550 (internal_load_fn_p): Likewise.
5551 (internal_gather_scatter_fn_p): Likewise.
5552 (internal_fn_mask_index): Likewise.
5553 (internal_gather_scatter_fn_supported_p): Likewise.
5554 * optabs-query.c (supports_at_least_one_mode_p): New function.
5555 (supports_vec_gather_load_p): Likewise.
5556 * optabs-query.h (supports_vec_gather_load_p): Declare.
5557 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5558 and memory_type field.
5559 (NUM_PATTERNS): Bump to 15.
5560 * tree-vect-data-refs.c: Include internal-fn.h.
5561 (vect_gather_scatter_fn_p): New function.
5562 (vect_describe_gather_scatter_call): Likewise.
5563 (vect_check_gather_scatter): Try using internal functions for
5564 gather loads. Recognize existing calls to a gather load function.
5565 (vect_analyze_data_refs): Consider using gather loads if
5566 supports_vec_gather_load_p.
5567 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5568 (vect_get_gather_scatter_offset_type): Likewise.
5569 (vect_convert_mask_for_vectype): Likewise.
5570 (vect_add_conversion_to_patterm): Likewise.
5571 (vect_try_gather_scatter_pattern): Likewise.
5572 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5573 (vect_vect_recog_func_ptrs): Add it.
5574 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5575 internal_fn_mask_index and internal_gather_scatter_fn_p.
5576 (check_load_store_masking): Take the gather_scatter_info as an
5577 argument and handle gather loads.
5578 (vect_get_gather_scatter_ops): New function.
5579 (vectorizable_call): Check internal_load_fn_p.
5580 (vectorizable_load): Likewise. Handle gather load internal
5582 (vectorizable_store): Update call to check_load_store_masking.
5583 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5584 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5585 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5586 (aarch64_gather_scale_operand_d): New predicates.
5587 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5588 (mask_gather_load<mode>): New insns.
5590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5591 Alan Hayward <alan.hayward@arm.com>
5592 David Sherwood <david.sherwood@arm.com>
5594 * optabs.def (fold_left_plus_optab): New optab.
5595 * doc/md.texi (fold_left_plus_@var{m}): Document.
5596 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5597 * internal-fn.c (fold_left_direct): Define.
5598 (expand_fold_left_optab_fn): Likewise.
5599 (direct_fold_left_optab_supported_p): Likewise.
5600 * fold-const-call.c (fold_const_fold_left): New function.
5601 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5602 * tree-parloops.c (valid_reduction_p): New function.
5603 (gather_scalar_reductions): Use it.
5604 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5605 (vect_finish_replace_stmt): Declare.
5606 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5607 (needs_fold_left_reduction_p): New function, split out from...
5608 (vect_is_simple_reduction): ...here. Accept reductions that
5609 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5610 (vect_force_simple_reduction): Also store the reduction type in
5611 the assignment's STMT_VINFO_REDUC_TYPE.
5612 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5613 (merge_with_identity): New function.
5614 (vect_expand_fold_left): Likewise.
5615 (vectorize_fold_left_reduction): Likewise.
5616 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5617 scalar phi in place for it. Check for target support and reject
5618 cases that would reassociate the operation. Defer the transform
5619 phase to vectorize_fold_left_reduction.
5620 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5621 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5622 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5626 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5627 call to ifc_temp_var.
5629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5630 Alan Hayward <alan.hayward@arm.com>
5631 David Sherwood <david.sherwood@arm.com>
5633 * target.def (legitimize_address_displacement): Take the original
5634 offset as a poly_int.
5635 * targhooks.h (default_legitimize_address_displacement): Update
5637 * targhooks.c (default_legitimize_address_displacement): Likewise.
5638 * doc/tm.texi: Regenerate.
5639 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5640 as an argument, moving assert of ad->disp == ad->disp_term to...
5641 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5642 Try calling targetm.legitimize_address_displacement before expanding
5643 the address rather than afterwards, and adjust for the new interface.
5644 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5645 Match the new hook interface. Handle SVE addresses.
5646 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5651 * Makefile.in (OBJS): Add early-remat.o.
5652 * target.def (select_early_remat_modes): New hook.
5653 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5654 * doc/tm.texi: Regenerate.
5655 * targhooks.h (default_select_early_remat_modes): Declare.
5656 * targhooks.c (default_select_early_remat_modes): New function.
5657 * timevar.def (TV_EARLY_REMAT): New timevar.
5658 * passes.def (pass_early_remat): New pass.
5659 * tree-pass.h (make_pass_early_remat): Declare.
5660 * early-remat.c: New file.
5661 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5663 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5666 Alan Hayward <alan.hayward@arm.com>
5667 David Sherwood <david.sherwood@arm.com>
5669 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5670 vfm1 with a bound_epilog parameter.
5671 (vect_do_peeling): Update calls accordingly, and move the prologue
5672 call earlier in the function. Treat the base bound_epilog as 0 for
5673 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5674 this base when peeling for gaps.
5675 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5676 with fully-masked loops.
5677 (vect_estimate_min_profitable_iters): Handle the single peeled
5678 iteration in that case.
5680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5681 Alan Hayward <alan.hayward@arm.com>
5682 David Sherwood <david.sherwood@arm.com>
5684 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5685 single-element interleaving even if the size is not a power of 2.
5686 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5687 accesses for single-element interleaving if the group size is
5690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5691 Alan Hayward <alan.hayward@arm.com>
5692 David Sherwood <david.sherwood@arm.com>
5694 * doc/md.texi (fold_extract_last_@var{m}): Document.
5695 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5696 * optabs.def (fold_extract_last_optab): New optab.
5697 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5698 * internal-fn.c (fold_extract_direct): New macro.
5699 (expand_fold_extract_optab_fn): Likewise.
5700 (direct_fold_extract_optab_supported_p): Likewise.
5701 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5702 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5703 EXTRACT_LAST_REDUCTION.
5704 (get_initial_def_for_reduction): Do not create an initial vector
5705 for EXTRACT_LAST_REDUCTION reductions.
5706 (vectorizable_reduction): Leave the scalar phi in place for
5707 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5708 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5709 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5710 transform phase to vectorizable_condition.
5711 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5713 (vect_finish_stmt_generation): ...here.
5714 (vect_finish_replace_stmt): New function.
5715 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5716 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5718 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5721 Alan Hayward <alan.hayward@arm.com>
5722 David Sherwood <david.sherwood@arm.com>
5724 * doc/md.texi (extract_last_@var{m}): Document.
5725 * optabs.def (extract_last_optab): New optab.
5726 * internal-fn.def (EXTRACT_LAST): New internal function.
5727 * internal-fn.c (cond_unary_direct): New macro.
5728 (expand_cond_unary_optab_fn): Likewise.
5729 (direct_cond_unary_optab_supported_p): Likewise.
5730 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5731 loops using EXTRACT_LAST.
5732 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5733 (extract_last_<mode>): ...this optab.
5734 (vec_extract<mode><Vel>): Update accordingly.
5736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5737 Alan Hayward <alan.hayward@arm.com>
5738 David Sherwood <david.sherwood@arm.com>
5740 * target.def (empty_mask_is_expensive): New hook.
5741 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5742 * doc/tm.texi: Regenerate.
5743 * targhooks.h (default_empty_mask_is_expensive): Declare.
5744 * targhooks.c (default_empty_mask_is_expensive): New function.
5745 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5746 if the target says that empty masks are expensive.
5747 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5749 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5752 Alan Hayward <alan.hayward@arm.com>
5753 David Sherwood <david.sherwood@arm.com>
5755 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5756 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5757 (vect_use_loop_mask_for_alignment_p): New function.
5758 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5759 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5760 niters_skip argument. Make sure that the first niters_skip elements
5761 of the first iteration are inactive.
5762 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5763 Update call to vect_set_loop_masks_directly.
5764 (get_misalign_in_elems): New function, split out from...
5765 (vect_gen_prolog_loop_niters): ...here.
5766 (vect_update_init_of_dr): Take a code argument that specifies whether
5767 the adjustment should be added or subtracted.
5768 (vect_update_init_of_drs): Likewise.
5769 (vect_prepare_for_masked_peels): New function.
5770 (vect_do_peeling): Skip prologue peeling if we're using a mask
5771 instead. Update call to vect_update_inits_of_drs.
5772 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5774 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5775 alignment. Do not include the number of peeled iterations in
5776 the minimum threshold in that case.
5777 (vectorizable_induction): Adjust the start value down by
5778 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5779 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5780 Take the number of skipped iterations into account when calculating
5782 * tree-vect-stmts.c (vect_gen_while_not): New function.
5784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5785 Alan Hayward <alan.hayward@arm.com>
5786 David Sherwood <david.sherwood@arm.com>
5788 * doc/sourcebuild.texi (vect_fully_masked): Document.
5789 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5791 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5793 (vect_analyze_loop_2): ...here. Don't check the vectorization
5794 factor against the number of loop iterations if the loop is
5797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5798 Alan Hayward <alan.hayward@arm.com>
5799 David Sherwood <david.sherwood@arm.com>
5801 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5802 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5803 (dump_groups): Update accordingly.
5804 (iv_use::mem_type): New member variable.
5805 (address_p): New function.
5806 (record_use): Add a mem_type argument and initialize the new
5808 (record_group_use): Add a mem_type argument. Use address_p.
5809 Remove obsolete null checks of base_object. Update call to record_use.
5810 (find_interesting_uses_op): Update call to record_group_use.
5811 (find_interesting_uses_cond): Likewise.
5812 (find_interesting_uses_address): Likewise.
5813 (get_mem_type_for_internal_fn): New function.
5814 (find_address_like_use): Likewise.
5815 (find_interesting_uses_stmt): Try find_address_like_use before
5816 calling find_interesting_uses_op.
5817 (addr_offset_valid_p): Use the iv mem_type field as the type
5818 of the addressed memory.
5819 (add_autoinc_candidates): Likewise.
5820 (get_address_cost): Likewise.
5821 (split_small_address_groups_p): Use address_p.
5822 (split_address_groups): Likewise.
5823 (add_iv_candidate_for_use): Likewise.
5824 (autoinc_possible_for_pair): Likewise.
5825 (rewrite_groups): Likewise.
5826 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5827 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5828 (get_alias_ptr_type_for_ptr_address): New function.
5829 (rewrite_use_address): Rewrite address uses in calls that were
5830 identified by find_address_like_use.
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 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5838 * gimple-expr.h (is_gimple_addressable: Likewise.
5839 * gimple-expr.c (is_gimple_address): Likewise.
5840 * internal-fn.c (expand_call_mem_ref): New function.
5841 (expand_mask_load_optab_fn): Use it.
5842 (expand_mask_store_optab_fn): Likewise.
5844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5845 Alan Hayward <alan.hayward@arm.com>
5846 David Sherwood <david.sherwood@arm.com>
5848 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5849 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5850 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5851 (cond_umax@var{mode}): Document.
5852 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5853 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5854 (cond_umin_optab, cond_umax_optab): New optabs.
5855 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5856 (COND_IOR, COND_XOR): New internal functions.
5857 * internal-fn.h (get_conditional_internal_fn): Declare.
5858 * internal-fn.c (cond_binary_direct): New macro.
5859 (expand_cond_binary_optab_fn): Likewise.
5860 (direct_cond_binary_optab_supported_p): Likewise.
5861 (get_conditional_internal_fn): New function.
5862 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5863 Cope with reduction statements that are vectorized as calls rather
5865 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5866 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5867 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5868 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5869 (UNSPEC_COND_EOR): New unspecs.
5870 (optab): Add mappings for them.
5871 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5872 (sve_int_op, sve_fp_op): New int attributes.
5874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5875 Alan Hayward <alan.hayward@arm.com>
5876 David Sherwood <david.sherwood@arm.com>
5878 * optabs.def (while_ult_optab): New optab.
5879 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5880 * internal-fn.def (WHILE_ULT): New internal function.
5881 * internal-fn.h (direct_internal_fn_supported_p): New override
5882 that takes two types as argument.
5883 * internal-fn.c (while_direct): New macro.
5884 (expand_while_optab_fn): New function.
5885 (convert_optab_supported_p): Likewise.
5886 (direct_while_optab_supported_p): New macro.
5887 * wide-int.h (wi::udiv_ceil): New function.
5888 * tree-vectorizer.h (rgroup_masks): New structure.
5889 (vec_loop_masks): New typedef.
5890 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5892 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5893 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5894 (vect_max_vf): New function.
5895 (slpeel_make_loop_iterate_ntimes): Delete.
5896 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5897 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5898 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5899 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5900 internal-fn.h, stor-layout.h and optabs-query.h.
5901 (vect_set_loop_mask): New function.
5902 (add_preheader_seq): Likewise.
5903 (add_header_seq): Likewise.
5904 (interleave_supported_p): Likewise.
5905 (vect_maybe_permute_loop_masks): Likewise.
5906 (vect_set_loop_masks_directly): Likewise.
5907 (vect_set_loop_condition_masked): Likewise.
5908 (vect_set_loop_condition_unmasked): New function, split out from
5909 slpeel_make_loop_iterate_ntimes.
5910 (slpeel_make_loop_iterate_ntimes): Rename to..
5911 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5912 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5913 (vect_do_peeling): Update call accordingly.
5914 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5916 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5917 mask_compare_type, can_fully_mask_p and fully_masked_p.
5918 (release_vec_loop_masks): New function.
5919 (_loop_vec_info): Use it to free the loop masks.
5920 (can_produce_all_loop_masks_p): New function.
5921 (vect_get_max_nscalars_per_iter): Likewise.
5922 (vect_verify_full_masking): Likewise.
5923 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5924 retries, and free the mask rgroups before retrying. Check loop-wide
5925 reasons for disallowing fully-masked loops. Make the final decision
5926 about whether use a fully-masked loop or not.
5927 (vect_estimate_min_profitable_iters): Do not assume that peeling
5928 for the number of iterations will be needed for fully-masked loops.
5929 (vectorizable_reduction): Disable fully-masked loops.
5930 (vectorizable_live_operation): Likewise.
5931 (vect_halve_mask_nunits): New function.
5932 (vect_double_mask_nunits): Likewise.
5933 (vect_record_loop_mask): Likewise.
5934 (vect_get_loop_mask): Likewise.
5935 (vect_transform_loop): Handle the case in which the final loop
5936 iteration might handle a partial vector. Call vect_set_loop_condition
5937 instead of slpeel_make_loop_iterate_ntimes.
5938 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5939 (check_load_store_masking): New function.
5940 (prepare_load_store_mask): Likewise.
5941 (vectorizable_store): Handle fully-masked loops.
5942 (vectorizable_load): Likewise.
5943 (supportable_widening_operation): Use vect_halve_mask_nunits for
5945 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5946 (vect_gen_while): New function.
5947 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5948 (aarch64_uqdec<mode>): New insn.
5950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5951 Alan Hayward <alan.hayward@arm.com>
5952 David Sherwood <david.sherwood@arm.com>
5954 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5955 (reduc_xor_scal_optab): New optabs.
5956 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5957 (reduc_xor_scal_@var{m}): Document.
5958 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5959 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5961 * fold-const-call.c (fold_const_call): Handle them.
5962 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5963 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5964 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5965 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5966 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5967 (UNSPEC_XORV): New unspecs.
5968 (optab): Add entries for them.
5969 (BITWISEV): New int iterator.
5970 (bit_reduc_op): New int attributes.
5972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5973 Alan Hayward <alan.hayward@arm.com>
5974 David Sherwood <david.sherwood@arm.com>
5976 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5977 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5978 * optabs.def (vec_shl_insert_optab): New optab.
5979 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5980 (duplicate_and_interleave): Likewise.
5981 * tree-vect-loop.c: Include internal-fn.h.
5982 (neutral_op_for_slp_reduction): New function, split out from
5983 get_initial_defs_for_reduction.
5984 (get_initial_def_for_reduction): Handle option 2 for variable-length
5985 vectors by loading the neutral value into a vector and then shifting
5986 the initial value into element 0.
5987 (get_initial_defs_for_reduction): Replace the code argument with
5988 the neutral value calculated by neutral_op_for_slp_reduction.
5989 Use gimple_build_vector for constant-length vectors.
5990 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5991 but the first group_size elements have a neutral value.
5992 Use duplicate_and_interleave otherwise.
5993 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5994 Update call to get_initial_defs_for_reduction. Handle SLP
5995 reductions for variable-length vectors by creating one vector
5996 result for each scalar result, with the elements associated
5997 with other scalar results stubbed out with the neutral value.
5998 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5999 Require IFN_VEC_SHL_INSERT for double reductions on
6000 variable-length vectors, or SLP reductions that have
6001 a neutral value. Require can_duplicate_and_interleave_p
6002 support for variable-length unchained SLP reductions if there
6003 is no neutral value, such as for MIN/MAX reductions. Also require
6004 the number of vector elements to be a multiple of the number of
6005 SLP statements when doing variable-length unchained SLP reductions.
6006 Update call to vect_create_epilog_for_reduction.
6007 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6008 and remove initial values.
6009 (duplicate_and_interleave): Make public.
6010 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6011 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6014 Alan Hayward <alan.hayward@arm.com>
6015 David Sherwood <david.sherwood@arm.com>
6017 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6018 (can_duplicate_and_interleave_p): New function.
6019 (vect_get_and_check_slp_defs): Take the vector of statements
6020 rather than just the current one. Remove excess parentheses.
6021 Restriction rejectinon of vect_constant_def and vect_external_def
6022 for variable-length vectors to boolean types, or types for which
6023 can_duplicate_and_interleave_p is false.
6024 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6025 (duplicate_and_interleave): New function.
6026 (vect_get_constant_vectors): Use gimple_build_vector for
6027 constant-length vectors and suitable variable-length constant
6028 vectors. Use duplicate_and_interleave for other variable-length
6029 vectors. Don't defer the update when inserting new statements.
6031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6032 Alan Hayward <alan.hayward@arm.com>
6033 David Sherwood <david.sherwood@arm.com>
6035 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6036 min_profitable_iters doesn't go negative.
6038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6039 Alan Hayward <alan.hayward@arm.com>
6040 David Sherwood <david.sherwood@arm.com>
6042 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6043 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6044 * optabs.def (vec_mask_load_lanes_optab): New optab.
6045 (vec_mask_store_lanes_optab): Likewise.
6046 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6047 (MASK_STORE_LANES): Likewise.
6048 * internal-fn.c (mask_load_lanes_direct): New macro.
6049 (mask_store_lanes_direct): Likewise.
6050 (expand_mask_load_optab_fn): Handle masked operations.
6051 (expand_mask_load_lanes_optab_fn): New macro.
6052 (expand_mask_store_optab_fn): Handle masked operations.
6053 (expand_mask_store_lanes_optab_fn): New macro.
6054 (direct_mask_load_lanes_optab_supported_p): Likewise.
6055 (direct_mask_store_lanes_optab_supported_p): Likewise.
6056 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6058 (vect_load_lanes_supported): Likewise.
6059 * tree-vect-data-refs.c (strip_conversion): New function.
6060 (can_group_stmts_p): Likewise.
6061 (vect_analyze_data_ref_accesses): Use it instead of checking
6062 for a pair of assignments.
6063 (vect_store_lanes_supported): Take a masked_p parameter.
6064 (vect_load_lanes_supported): Likewise.
6065 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6066 vect_store_lanes_supported and vect_load_lanes_supported.
6067 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6068 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6069 parameter. Don't allow gaps for masked accesses.
6070 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6071 and vect_load_lanes_supported.
6072 (get_load_store_type): Take a masked_p parameter and update
6073 call to get_group_load_store_type.
6074 (vectorizable_store): Update call to get_load_store_type.
6075 Handle IFN_MASK_STORE_LANES.
6076 (vectorizable_load): Update call to get_load_store_type.
6077 Handle IFN_MASK_LOAD_LANES.
6079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6080 Alan Hayward <alan.hayward@arm.com>
6081 David Sherwood <david.sherwood@arm.com>
6083 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6085 * config/aarch64/aarch64-protos.h
6086 (aarch64_sve_struct_memory_operand_p): Declare.
6087 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6088 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6089 (VPRED, vpred): Handle SVE structure modes.
6090 * config/aarch64/constraints.md (Utx): New constraint.
6091 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6092 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6093 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6094 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6095 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6096 structure modes. Split into pieces after RA.
6097 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6098 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6100 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6101 SVE structure modes.
6102 (aarch64_classify_address): Likewise.
6103 (sizetochar): Move earlier in file.
6104 (aarch64_print_operand): Handle SVE register lists.
6105 (aarch64_array_mode): New function.
6106 (aarch64_sve_struct_memory_operand_p): Likewise.
6107 (TARGET_ARRAY_MODE): Redefine.
6109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6110 Alan Hayward <alan.hayward@arm.com>
6111 David Sherwood <david.sherwood@arm.com>
6113 * target.def (array_mode): New target hook.
6114 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6115 * doc/tm.texi: Regenerate.
6116 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6117 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6118 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6120 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6124 Alan Hayward <alan.hayward@arm.com>
6125 David Sherwood <david.sherwood@arm.com>
6127 * fold-const.c (fold_binary_loc): Check the argument types
6128 rather than the result type when testing for a vector operation.
6130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6132 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6133 * doc/tm.texi: Regenerate.
6135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6136 Alan Hayward <alan.hayward@arm.com>
6137 David Sherwood <david.sherwood@arm.com>
6139 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6140 (sve): Document new AArch64 extension.
6141 * doc/md.texi (w): Extend the description of the AArch64
6142 constraint to include SVE vectors.
6143 (Upl, Upa): Document new AArch64 predicate constraints.
6144 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6146 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6147 (msve-vector-bits=): New option.
6148 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6149 SVE when these are disabled.
6150 (sve): New extension.
6151 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6152 modes. Adjust their number of units based on aarch64_sve_vg.
6153 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6154 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6155 aarch64_addr_query_type.
6156 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6157 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6158 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6159 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6160 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6161 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6162 (aarch64_simd_imm_zero_p): Delete.
6163 (aarch64_check_zero_based_sve_index_immediate): Declare.
6164 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6165 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6166 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6167 (aarch64_sve_float_mul_immediate_p): Likewise.
6168 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6170 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6171 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6172 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6173 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6174 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6175 (aarch64_regmode_natural_size): Likewise.
6176 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6177 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6179 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6180 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6181 for VG and the SVE predicate registers.
6182 (V_ALIASES): Add a "z"-prefixed alias.
6183 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6184 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6185 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6186 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6187 (REG_CLASS_NAMES): Add entries for them.
6188 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6189 and the predicate registers.
6190 (aarch64_sve_vg): Declare.
6191 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6192 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6193 (REGMODE_NATURAL_SIZE): Define.
6194 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6196 * config/aarch64/aarch64.c: Include cfgrtl.h.
6197 (simd_immediate_info): Add a constructor for series vectors,
6198 and an associated step field.
6199 (aarch64_sve_vg): New variable.
6200 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6201 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6202 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6203 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6204 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6205 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6206 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6207 (aarch64_get_mask_mode): New functions.
6208 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6209 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6210 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6211 predicate modes and predicate registers. Explicitly restrict
6212 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6213 to store a vector mode if it is recognized by
6214 aarch64_classify_vector_mode.
6215 (aarch64_regmode_natural_size): New function.
6216 (aarch64_hard_regno_caller_save_mode): Return the original mode
6218 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6219 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6220 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6221 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6223 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6224 does not overlap dest if the function is frame-related. Handle
6226 (aarch64_split_add_offset): New function.
6227 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6228 them aarch64_add_offset.
6229 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6230 and update call to aarch64_sub_sp.
6231 (aarch64_add_cfa_expression): New function.
6232 (aarch64_expand_prologue): Pass extra temporary registers to the
6233 functions above. Handle the case in which we need to emit new
6234 DW_CFA_expressions for registers that were originally saved
6235 relative to the stack pointer, but now have to be expressed
6236 relative to the frame pointer.
6237 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6239 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6240 IP0 and IP1 values for SVE frames.
6241 (aarch64_expand_vec_series): New function.
6242 (aarch64_expand_sve_widened_duplicate): Likewise.
6243 (aarch64_expand_sve_const_vector): Likewise.
6244 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6245 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6246 into the register, rather than emitting a SET directly.
6247 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6248 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6249 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6250 (offset_9bit_signed_scaled_p): New functions.
6251 (aarch64_replicate_bitmask_imm): New function.
6252 (aarch64_bitmask_imm): Use it.
6253 (aarch64_cannot_force_const_mem): Reject expressions involving
6254 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6255 (aarch64_classify_index): Handle SVE indices, by requiring
6256 a plain register index with a scale that matches the element size.
6257 (aarch64_classify_address): Handle SVE addresses. Assert that
6258 the mode of the address is VOIDmode or an integer mode.
6259 Update call to aarch64_classify_symbol.
6260 (aarch64_classify_symbolic_expression): Update call to
6261 aarch64_classify_symbol.
6262 (aarch64_const_vec_all_in_range_p): New function.
6263 (aarch64_print_vector_float_operand): Likewise.
6264 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6265 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6266 and the FP immediates 1.0 and 0.5.
6267 (aarch64_print_address_internal): Handle SVE addresses.
6268 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6269 (aarch64_regno_regclass): Handle predicate registers.
6270 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6272 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6273 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6274 (aarch64_convert_sve_vector_bits): New function.
6275 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6276 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6278 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6279 Handle SVE vector and predicate modes. Accept VL-based constants
6280 that need only one temporary register, and VL offsets that require
6281 no temporary registers.
6282 (aarch64_conditional_register_usage): Mark the predicate registers
6283 as fixed if SVE isn't available.
6284 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6285 Return true for SVE vector and predicate modes.
6286 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6287 rather than an unsigned int. Handle SVE modes.
6288 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6290 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6292 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6293 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6294 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6295 (aarch64_sve_float_mul_immediate_p): New functions.
6296 (aarch64_sve_valid_immediate): New function.
6297 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6298 Explicitly reject structure modes. Check for INDEX constants.
6299 Handle PTRUE and PFALSE constants.
6300 (aarch64_check_zero_based_sve_index_immediate): New function.
6301 (aarch64_simd_imm_zero_p): Delete.
6302 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6303 vector modes. Accept constants in the range of CNT[BHWD].
6304 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6305 ask for an Advanced SIMD mode.
6306 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6307 (aarch64_simd_vector_alignment): Handle SVE predicates.
6308 (aarch64_vectorize_preferred_vector_alignment): New function.
6309 (aarch64_simd_vector_alignment_reachable): Use it instead of
6311 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6312 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6314 (MAX_VECT_LEN): Delete.
6315 (expand_vec_perm_d): Add a vec_flags field.
6316 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6317 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6318 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6320 (aarch64_evpc_rev): Rename to...
6321 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6322 (aarch64_evpc_rev_global): New function.
6323 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6324 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6326 (aarch64_evpc_sve_tbl): New function.
6327 (aarch64_expand_vec_perm_const_1): Update after rename of
6328 aarch64_evpc_rev. Handle SVE permutes too, trying
6329 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6330 than aarch64_evpc_tbl.
6331 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6332 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6333 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6334 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6335 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6336 (aarch64_expand_sve_vcond): New functions.
6337 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6338 of aarch64_vector_mode_p.
6339 (aarch64_dwarf_poly_indeterminate_value): New function.
6340 (aarch64_compute_pressure_classes): Likewise.
6341 (aarch64_can_change_mode_class): Likewise.
6342 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6343 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6344 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6345 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6346 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6347 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6348 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6349 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6351 (Dn, Dl, Dr): Accept const as well as const_vector.
6352 (Dz): Likewise. Compare against CONST0_RTX.
6353 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6354 of "vector" where appropriate.
6355 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6356 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6357 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6358 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6359 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6360 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6361 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6362 (v_int_equiv): Extend to SVE modes.
6363 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6365 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6366 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6367 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6368 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6369 (SVE_COND_FP_CMP): New int iterators.
6370 (perm_hilo): Handle the new unpack unspecs.
6371 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6373 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6374 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6375 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6376 (aarch64_equality_operator, aarch64_constant_vector_operand)
6377 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6378 (aarch64_sve_nonimmediate_operand): Likewise.
6379 (aarch64_sve_general_operand): Likewise.
6380 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6381 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6382 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6383 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6384 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6385 (aarch64_sve_float_arith_immediate): Likewise.
6386 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6387 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6388 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6389 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6390 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6391 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6392 (aarch64_sve_float_arith_operand): Likewise.
6393 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6394 (aarch64_sve_float_mul_operand): Likewise.
6395 (aarch64_sve_vec_perm_operand): Likewise.
6396 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6397 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6398 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6399 as well as const_vector.
6400 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6401 in file. Use CONST0_RTX and CONSTM1_RTX.
6402 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6403 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6404 Use aarch64_simd_imm_zero.
6405 * config/aarch64/aarch64-sve.md: New file.
6406 * config/aarch64/aarch64.md: Include it.
6407 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6408 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6409 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6410 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6411 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6412 (sve): New attribute.
6413 (enabled): Disable instructions with the sve attribute unless
6415 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6416 aarch64_expand_mov_immediate.
6417 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6418 CNT[BHSD] immediates.
6419 (movti): Split CONST_POLY_INT moves into two halves.
6420 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6421 Split additions that need a temporary here if the destination
6422 is the stack pointer.
6423 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6424 (*add<mode>3_poly_1): New instruction.
6425 (set_clobber_cc): New expander.
6427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6429 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6430 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6431 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6432 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6433 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6434 Change innermode from fixed_mode_size to machine_mode.
6435 (simplify_subreg): Update call accordingly. Handle a constant-sized
6436 subreg of a variable-length CONST_VECTOR.
6438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6439 Alan Hayward <alan.hayward@arm.com>
6440 David Sherwood <david.sherwood@arm.com>
6442 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6443 (add_offset_to_base): New function, split out from...
6444 (create_mem_ref): ...here. When handling a scale other than 1,
6445 check first whether the address is valid without the offset.
6446 Add it into the base if so, leaving the index and scale as-is.
6448 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6451 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6452 fold_for_warn before checking if arg2 is INTEGER_CST.
6454 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6456 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6457 (store_multiple_operation): Delete.
6458 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6459 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6460 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6461 guarded by TARGET_STRING.
6462 (rs6000_output_load_multiple): Delete.
6463 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6464 OPTION_MASK_STRING / TARGET_STRING handling.
6465 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6466 (const rs6000_opt_masks) <"string">: Change mask to 0.
6467 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6468 (MASK_STRING): Delete.
6469 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6471 (load_multiple): Delete.
6478 (store_multiple): Delete.
6485 (movmemsi_8reg): Delete.
6486 (corresponding unnamed define_insn): Delete.
6487 (movmemsi_6reg): Delete.
6488 (corresponding unnamed define_insn): Delete.
6489 (movmemsi_4reg): Delete.
6490 (corresponding unnamed define_insn): Delete.
6491 (movmemsi_2reg): Delete.
6492 (corresponding unnamed define_insn): Delete.
6493 (movmemsi_1reg): Delete.
6494 (corresponding unnamed define_insn): Delete.
6495 * config/rs6000/rs6000.opt (mno-string): New.
6496 (mstring): Replace by deprecation warning stub.
6497 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6499 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6501 * regrename.c (regrename_do_replace): If replacing the same
6502 reg multiple times, try to reuse last created gen_raw_REG.
6505 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6506 main to workaround a bug in GDB.
6508 2018-01-12 Tom de Vries <tom@codesourcery.com>
6511 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6513 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6515 PR rtl-optimization/80481
6516 * ira-color.c (get_cap_member): New function.
6517 (allocnos_conflict_by_live_ranges_p): Use it.
6518 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6519 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6521 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6524 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6525 (*saddl_se_1): Ditto.
6527 (*ssubl_se_1): Ditto.
6529 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6531 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6532 rather than wi::to_widest for DR_INITs.
6533 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6534 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6535 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6537 (vect_analyze_group_access_1): Note that here.
6539 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6541 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6542 polynomial type sizes.
6544 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6546 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6547 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6548 (gimple_add_tmp_var): Likewise.
6550 2018-01-12 Martin Liska <mliska@suse.cz>
6552 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6553 (gimple_alloc_sizes): Likewise.
6554 (dump_gimple_statistics): Use PRIu64 in printf format.
6555 * gimple.h: Change uint64_t to int.
6557 2018-01-12 Martin Liska <mliska@suse.cz>
6559 * tree-core.h: Use uint64_t instead of int.
6560 * tree.c (tree_node_counts): Likewise.
6561 (tree_node_sizes): Likewise.
6562 (dump_tree_statistics): Use PRIu64 in printf format.
6564 2018-01-12 Martin Liska <mliska@suse.cz>
6566 * Makefile.in: As qsort_chk is implemented in vec.c, add
6567 vec.o to linkage of gencfn-macros.
6568 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6569 passing the info to record_node_allocation_statistics.
6570 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6572 * ggc-common.c (struct ggc_usage): Add operator== and use
6573 it in operator< and compare function.
6574 * mem-stats.h (struct mem_usage): Likewise.
6575 * vec.c (struct vec_usage): Remove operator< and compare
6576 function. Can be simply inherited.
6578 2018-01-12 Martin Jambor <mjambor@suse.cz>
6581 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6582 * tree-ssa-math-opts.c: Include domwalk.h.
6583 (convert_mult_to_fma_1): New function.
6584 (fma_transformation_info): New type.
6585 (fma_deferring_state): Likewise.
6586 (cancel_fma_deferring): New function.
6587 (result_of_phi): Likewise.
6588 (last_fma_candidate_feeds_initial_phi): Likewise.
6589 (convert_mult_to_fma): Added deferring logic, split actual
6590 transformation to convert_mult_to_fma_1.
6591 (math_opts_dom_walker): New type.
6592 (math_opts_dom_walker::after_dom_children): New method, body moved
6593 here from pass_optimize_widening_mul::execute, added deferring logic
6595 (pass_optimize_widening_mul::execute): Moved most of code to
6596 math_opts_dom_walker::after_dom_children.
6597 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6598 * config/i386/i386.c (ix86_option_override_internal): Added
6599 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6601 2018-01-12 Richard Biener <rguenther@suse.de>
6604 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6605 inline instance vars.
6607 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6610 * config/rx/rx.c (rx_is_restricted_memory_address):
6613 2018-01-12 Richard Biener <rguenther@suse.de>
6615 PR tree-optimization/80846
6616 * target.def (split_reduction): New target hook.
6617 * targhooks.c (default_split_reduction): New function.
6618 * targhooks.h (default_split_reduction): Declare.
6619 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6620 target requests first reduce vectors by combining low and high
6622 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6623 (get_vectype_for_scalar_type_and_size): Export.
6624 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6625 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6626 * doc/tm.texi: Regenerate.
6627 * config/i386/i386.c (ix86_split_reduction): Implement
6628 TARGET_VECTORIZE_SPLIT_REDUCTION.
6630 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6633 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6634 in PIC mode except for TARGET_VXWORKS_RTP.
6635 * config/sparc/sparc.c: Include cfgrtl.h.
6636 (TARGET_INIT_PIC_REG): Define.
6637 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6638 (sparc_pic_register_p): New predicate.
6639 (sparc_legitimate_address_p): Use it.
6640 (sparc_legitimize_pic_address): Likewise.
6641 (sparc_delegitimize_address): Likewise.
6642 (sparc_mode_dependent_address_p): Likewise.
6643 (gen_load_pcrel_sym): Remove 4th parameter.
6644 (load_got_register): Adjust call to above. Remove obsolete stuff.
6645 (sparc_expand_prologue): Do not call load_got_register here.
6646 (sparc_flat_expand_prologue): Likewise.
6647 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6648 (sparc_use_pseudo_pic_reg): New function.
6649 (sparc_init_pic_reg): Likewise.
6650 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6651 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6653 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6655 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6656 Add item for branch_cost.
6658 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6660 PR rtl-optimization/83565
6661 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6662 not extend the result to a larger mode for rotate operations.
6663 (num_sign_bit_copies1): Likewise.
6665 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6668 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6670 Use values-Xc.o for -pedantic.
6671 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6673 2018-01-12 Martin Liska <mliska@suse.cz>
6676 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6678 (possible_polymorphic_call_targets): Use it.
6679 (ipa_devirt): Likewise.
6681 2018-01-12 Martin Liska <mliska@suse.cz>
6683 * profile-count.h (enum profile_quality): Use 0 as invalid
6684 enum value of profile_quality.
6686 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6688 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6689 -mext-string options.
6691 2018-01-12 Richard Biener <rguenther@suse.de>
6693 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6694 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6695 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6697 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6699 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6701 * configure.ac (--with-long-double-format): Add support for the
6702 configuration option to change the default long double format on
6704 * config.gcc (powerpc*-linux*-*): Likewise.
6705 * configure: Regenerate.
6706 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6707 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6708 used without modification.
6710 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6712 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6713 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6714 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6715 MISC_BUILTIN_SPEC_BARRIER.
6716 (rs6000_init_builtins): Likewise.
6717 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6719 (speculation_barrier): New define_insn.
6720 * doc/extend.texi: Document __builtin_speculation_barrier.
6722 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6725 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6726 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6727 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6729 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6730 integral modes instead of "ss" and "sd".
6731 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6732 vectors with 32-bit and 64-bit elements.
6733 (vecdupssescalarmodesuffix): New mode attribute.
6734 (vec_dup<mode>): Use it.
6736 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6739 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6740 frame if argument is passed on stack.
6742 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6745 * ree.c (combine_reaching_defs): Optimize also
6746 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6747 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6749 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6752 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6754 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6757 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6758 after they are computed.
6760 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6762 PR tree-optimization/83695
6763 * gimple-loop-linterchange.cc
6764 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6765 reset cached scev information after interchange.
6766 (pass_linterchange::execute): Remove call to scev_reset_htab.
6768 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6770 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6771 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6772 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6773 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6774 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6775 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6776 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6777 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6778 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6779 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6780 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6781 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6782 (V_lane_reg): Likewise.
6783 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6785 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6786 (vfmal_lane_low<mode>_intrinsic,
6787 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6788 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6789 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6790 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6791 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6792 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6794 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6796 * config/arm/arm-cpus.in (fp16fml): New feature.
6797 (ALL_SIMD): Add fp16fml.
6798 (armv8.2-a): Add fp16fml as an option.
6799 (armv8.3-a): Likewise.
6800 (armv8.4-a): Add fp16fml as part of fp16.
6801 * config/arm/arm.h (TARGET_FP16FML): Define.
6802 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6804 * config/arm/arm-modes.def (V2HF): Define.
6805 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6806 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6807 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6808 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6809 vfmsl_low, vfmsl_high): New set of builtins.
6810 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6811 (vfml_op): New code attribute.
6812 (VFMLHALVES): New int iterator.
6813 (VFML, VFMLSEL): New mode attributes.
6814 (V_reg): Define mapping for V2HF.
6815 (V_hi, V_lo): New mode attributes.
6816 (VF_constraint): Likewise.
6817 (vfml_half, vfml_half_selector): New int attributes.
6818 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6820 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6821 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6823 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6824 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6825 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6826 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6828 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6829 Document new effective target and option set.
6831 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6833 * config/arm/arm-cpus.in (armv8_4): New feature.
6834 (ARMv8_4a): New fgroup.
6835 (armv8.4-a): New arch.
6836 * config/arm/arm-tables.opt: Regenerate.
6837 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6838 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6839 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6840 Add matching rules for -march=armv8.4-a and extensions.
6841 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6843 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6846 * config/rx/rx.md (BW): New mode attribute.
6847 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6849 2018-01-11 Richard Biener <rguenther@suse.de>
6851 PR tree-optimization/83435
6852 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6853 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6854 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6856 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6857 Alan Hayward <alan.hayward@arm.com>
6858 David Sherwood <david.sherwood@arm.com>
6860 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6862 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6863 (aarch64_print_address_internal): Use it to check for a zero offset.
6865 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6866 Alan Hayward <alan.hayward@arm.com>
6867 David Sherwood <david.sherwood@arm.com>
6869 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6870 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6871 Return a poly_int64 rather than a HOST_WIDE_INT.
6872 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6873 rather than a HOST_WIDE_INT.
6874 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6875 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6876 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6877 final_offset from HOST_WIDE_INT to poly_int64.
6878 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6879 to_constant when getting the number of units in an Advanced SIMD
6881 (aarch64_builtin_vectorized_function): Check for a constant number
6883 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6885 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6886 attribute instead of GET_MODE_NUNITS.
6887 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6888 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6889 GET_MODE_SIZE for fixed-size registers.
6890 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6891 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6892 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6893 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6894 (aarch64_print_operand, aarch64_print_address_internal)
6895 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6896 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6897 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6898 Handle polynomial GET_MODE_SIZE.
6899 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6900 wider than SImode without modification.
6901 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6902 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6903 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6904 passing and returning SVE modes.
6905 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6906 rather than GEN_INT.
6907 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6908 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6909 (aarch64_allocate_and_probe_stack_space): Likewise.
6910 (aarch64_layout_frame): Cope with polynomial offsets.
6911 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6912 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6914 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6915 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6916 poly_int64 rather than a HOST_WIDE_INT.
6917 (aarch64_get_separate_components, aarch64_process_components)
6918 (aarch64_expand_prologue, aarch64_expand_epilogue)
6919 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6920 (aarch64_anchor_offset): New function, split out from...
6921 (aarch64_legitimize_address): ...here.
6922 (aarch64_builtin_vectorization_cost): Handle polynomial
6923 TYPE_VECTOR_SUBPARTS.
6924 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6926 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6927 number of elements from the PARALLEL rather than the mode.
6928 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6929 rather than GET_MODE_BITSIZE.
6930 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6931 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6932 (aarch64_expand_vec_perm_const_1): Handle polynomial
6933 d->perm.length () and d->perm elements.
6934 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6935 Apply to_constant to d->perm elements.
6936 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6937 polynomial CONST_VECTOR_NUNITS.
6938 (aarch64_move_pointer): Take amount as a poly_int64 rather
6940 (aarch64_progress_pointer): Avoid temporary variable.
6941 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6942 the mode attribute instead of GET_MODE.
6944 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6945 Alan Hayward <alan.hayward@arm.com>
6946 David Sherwood <david.sherwood@arm.com>
6948 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6949 x exists before using it.
6950 (aarch64_add_constant_internal): Rename to...
6951 (aarch64_add_offset_1): ...this. Replace regnum with separate
6952 src and dest rtxes. Handle the case in which they're different,
6953 including when the offset is zero. Replace scratchreg with an rtx.
6954 Use 2 additions if there is no spare register into which we can
6955 move a 16-bit constant.
6956 (aarch64_add_constant): Delete.
6957 (aarch64_add_offset): Replace reg with separate src and dest
6958 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6959 Use aarch64_add_offset_1.
6960 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6961 an rtx rather than an int. Take the delta as a poly_int64
6962 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6963 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6964 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6965 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6966 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6968 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6969 aarch64_add_constant.
6971 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6973 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6974 Use scalar_float_mode.
6976 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6978 * config/aarch64/aarch64-simd.md
6979 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6980 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6981 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6982 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6983 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6984 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6985 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6986 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6987 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6988 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6990 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6993 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6994 targ_options->x_arm_arch_string is non NULL.
6996 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6998 * config/aarch64/aarch64.h
6999 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7001 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7004 * expmed.c (emit_store_flag_force): Swap if const op0
7005 and change VOIDmode to mode of op0.
7007 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7009 PR rtl-optimization/83761
7010 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7011 than bytes to mode_for_size.
7013 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7016 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7017 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7020 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7023 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7024 when in layout mode.
7025 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7026 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7029 2018-01-10 Michael Collison <michael.collison@arm.com>
7031 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7032 * config/aarch64/aarch64-option-extension.def: Add
7033 AARCH64_OPT_EXTENSION of 'fp16fml'.
7034 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7035 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7036 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7037 * config/aarch64/constraints.md (Ui7): New constraint.
7038 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7039 (VFMLA_SEL_W): Ditto.
7042 (VFMLA16_LOW): New int iterator.
7043 (VFMLA16_HIGH): Ditto.
7044 (UNSPEC_FMLAL): New unspec.
7045 (UNSPEC_FMLSL): Ditto.
7046 (UNSPEC_FMLAL2): Ditto.
7047 (UNSPEC_FMLSL2): Ditto.
7048 (f16mac): New code attribute.
7049 * config/aarch64/aarch64-simd-builtins.def
7050 (aarch64_fmlal_lowv2sf): Ditto.
7051 (aarch64_fmlsl_lowv2sf): Ditto.
7052 (aarch64_fmlalq_lowv4sf): Ditto.
7053 (aarch64_fmlslq_lowv4sf): Ditto.
7054 (aarch64_fmlal_highv2sf): Ditto.
7055 (aarch64_fmlsl_highv2sf): Ditto.
7056 (aarch64_fmlalq_highv4sf): Ditto.
7057 (aarch64_fmlslq_highv4sf): Ditto.
7058 (aarch64_fmlal_lane_lowv2sf): Ditto.
7059 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7060 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7061 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7062 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7063 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7064 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7065 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7066 (aarch64_fmlal_lane_highv2sf): Ditto.
7067 (aarch64_fmlsl_lane_highv2sf): Ditto.
7068 (aarch64_fmlal_laneq_highv2sf): Ditto.
7069 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7070 (aarch64_fmlalq_lane_highv4sf): Ditto.
7071 (aarch64_fmlsl_lane_highv4sf): Ditto.
7072 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7073 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7074 * config/aarch64/aarch64-simd.md:
7075 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7076 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7077 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7078 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7079 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7080 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7081 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7082 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7083 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7084 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7085 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7086 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7087 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7088 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7089 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7090 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7091 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7092 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7093 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7094 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7095 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7096 (vfmlsl_low_u32): Ditto.
7097 (vfmlalq_low_u32): Ditto.
7098 (vfmlslq_low_u32): Ditto.
7099 (vfmlal_high_u32): Ditto.
7100 (vfmlsl_high_u32): Ditto.
7101 (vfmlalq_high_u32): Ditto.
7102 (vfmlslq_high_u32): Ditto.
7103 (vfmlal_lane_low_u32): Ditto.
7104 (vfmlsl_lane_low_u32): Ditto.
7105 (vfmlal_laneq_low_u32): Ditto.
7106 (vfmlsl_laneq_low_u32): Ditto.
7107 (vfmlalq_lane_low_u32): Ditto.
7108 (vfmlslq_lane_low_u32): Ditto.
7109 (vfmlalq_laneq_low_u32): Ditto.
7110 (vfmlslq_laneq_low_u32): Ditto.
7111 (vfmlal_lane_high_u32): Ditto.
7112 (vfmlsl_lane_high_u32): Ditto.
7113 (vfmlal_laneq_high_u32): Ditto.
7114 (vfmlsl_laneq_high_u32): Ditto.
7115 (vfmlalq_lane_high_u32): Ditto.
7116 (vfmlslq_lane_high_u32): Ditto.
7117 (vfmlalq_laneq_high_u32): Ditto.
7118 (vfmlslq_laneq_high_u32): Ditto.
7119 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7120 (AARCH64_FL_FOR_ARCH8_4): New.
7121 (AARCH64_ISA_F16FML): New ISA flag.
7122 (TARGET_F16FML): New feature flag for fp16fml.
7123 (doc/invoke.texi): Document new fp16fml option.
7125 2018-01-10 Michael Collison <michael.collison@arm.com>
7127 * config/aarch64/aarch64-builtins.c:
7128 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7129 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7130 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7131 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7132 (AARCH64_ISA_SHA3): New ISA flag.
7133 (TARGET_SHA3): New feature flag for sha3.
7134 * config/aarch64/iterators.md (sha512_op): New int attribute.
7135 (CRYPTO_SHA512): New int iterator.
7136 (UNSPEC_SHA512H): New unspec.
7137 (UNSPEC_SHA512H2): Ditto.
7138 (UNSPEC_SHA512SU0): Ditto.
7139 (UNSPEC_SHA512SU1): Ditto.
7140 * config/aarch64/aarch64-simd-builtins.def
7141 (aarch64_crypto_sha512hqv2di): New builtin.
7142 (aarch64_crypto_sha512h2qv2di): Ditto.
7143 (aarch64_crypto_sha512su0qv2di): Ditto.
7144 (aarch64_crypto_sha512su1qv2di): Ditto.
7145 (aarch64_eor3qv8hi): Ditto.
7146 (aarch64_rax1qv2di): Ditto.
7147 (aarch64_xarqv2di): Ditto.
7148 (aarch64_bcaxqv8hi): Ditto.
7149 * config/aarch64/aarch64-simd.md:
7150 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7151 (aarch64_crypto_sha512su0qv2di): Ditto.
7152 (aarch64_crypto_sha512su1qv2di): Ditto.
7153 (aarch64_eor3qv8hi): Ditto.
7154 (aarch64_rax1qv2di): Ditto.
7155 (aarch64_xarqv2di): Ditto.
7156 (aarch64_bcaxqv8hi): Ditto.
7157 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7158 (vsha512h2q_u64): Ditto.
7159 (vsha512su0q_u64): Ditto.
7160 (vsha512su1q_u64): Ditto.
7161 (veor3q_u16): Ditto.
7162 (vrax1q_u64): Ditto.
7164 (vbcaxq_u16): Ditto.
7165 * config/arm/types.md (crypto_sha512): New type attribute.
7166 (crypto_sha3): Ditto.
7167 (doc/invoke.texi): Document new sha3 option.
7169 2018-01-10 Michael Collison <michael.collison@arm.com>
7171 * config/aarch64/aarch64-builtins.c:
7172 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7173 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7174 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7175 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7176 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7177 (AARCH64_ISA_SM4): New ISA flag.
7178 (TARGET_SM4): New feature flag for sm4.
7179 * config/aarch64/aarch64-simd-builtins.def
7180 (aarch64_sm3ss1qv4si): Ditto.
7181 (aarch64_sm3tt1aq4si): Ditto.
7182 (aarch64_sm3tt1bq4si): Ditto.
7183 (aarch64_sm3tt2aq4si): Ditto.
7184 (aarch64_sm3tt2bq4si): Ditto.
7185 (aarch64_sm3partw1qv4si): Ditto.
7186 (aarch64_sm3partw2qv4si): Ditto.
7187 (aarch64_sm4eqv4si): Ditto.
7188 (aarch64_sm4ekeyqv4si): Ditto.
7189 * config/aarch64/aarch64-simd.md:
7190 (aarch64_sm3ss1qv4si): Ditto.
7191 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7192 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7193 (aarch64_sm4eqv4si): Ditto.
7194 (aarch64_sm4ekeyqv4si): Ditto.
7195 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7196 (sm3part_op): Ditto.
7197 (CRYPTO_SM3TT): Ditto.
7198 (CRYPTO_SM3PART): Ditto.
7199 (UNSPEC_SM3SS1): New unspec.
7200 (UNSPEC_SM3TT1A): Ditto.
7201 (UNSPEC_SM3TT1B): Ditto.
7202 (UNSPEC_SM3TT2A): Ditto.
7203 (UNSPEC_SM3TT2B): Ditto.
7204 (UNSPEC_SM3PARTW1): Ditto.
7205 (UNSPEC_SM3PARTW2): Ditto.
7206 (UNSPEC_SM4E): Ditto.
7207 (UNSPEC_SM4EKEY): Ditto.
7208 * config/aarch64/constraints.md (Ui2): New constraint.
7209 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7210 * config/arm/types.md (crypto_sm3): New type attribute.
7211 (crypto_sm4): Ditto.
7212 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7213 (vsm3tt1aq_u32): Ditto.
7214 (vsm3tt1bq_u32): Ditto.
7215 (vsm3tt2aq_u32): Ditto.
7216 (vsm3tt2bq_u32): Ditto.
7217 (vsm3partw1q_u32): Ditto.
7218 (vsm3partw2q_u32): Ditto.
7219 (vsm4eq_u32): Ditto.
7220 (vsm4ekeyq_u32): Ditto.
7221 (doc/invoke.texi): Document new sm4 option.
7223 2018-01-10 Michael Collison <michael.collison@arm.com>
7225 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7226 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7227 (AARCH64_FL_FOR_ARCH8_4): New.
7228 (AARCH64_FL_V8_4): New flag.
7229 (doc/invoke.texi): Document new armv8.4-a option.
7231 2018-01-10 Michael Collison <michael.collison@arm.com>
7233 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7234 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7235 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7236 * config/aarch64/aarch64-option-extension.def: Add
7237 AARCH64_OPT_EXTENSION of 'sha2'.
7238 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7239 (crypto): Disable sha2 and aes if crypto disabled.
7240 (crypto): Enable aes and sha2 if enabled.
7241 (simd): Disable sha2 and aes if simd disabled.
7242 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7244 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7245 (TARGET_SHA2): New feature flag for sha2.
7246 (TARGET_AES): New feature flag for aes.
7247 * config/aarch64/aarch64-simd.md:
7248 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7249 conditional on TARGET_AES.
7250 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7251 (aarch64_crypto_sha1hsi): Make pattern conditional
7253 (aarch64_crypto_sha1hv4si): Ditto.
7254 (aarch64_be_crypto_sha1hv4si): Ditto.
7255 (aarch64_crypto_sha1su1v4si): Ditto.
7256 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7257 (aarch64_crypto_sha1su0v4si): Ditto.
7258 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7259 (aarch64_crypto_sha256su0v4si): Ditto.
7260 (aarch64_crypto_sha256su1v4si): Ditto.
7261 (doc/invoke.texi): Document new aes and sha2 options.
7263 2018-01-10 Martin Sebor <msebor@redhat.com>
7265 PR tree-optimization/83781
7266 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7269 2018-01-11 Martin Sebor <msebor@gmail.com>
7270 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7272 PR tree-optimization/83501
7273 PR tree-optimization/81703
7275 * tree-ssa-strlen.c (get_string_cst): Rename...
7276 (get_string_len): ...to this. Handle global constants.
7277 (handle_char_store): Adjust.
7279 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7280 Jim Wilson <jimw@sifive.com>
7282 * config/riscv/riscv-protos.h (riscv_output_return): New.
7283 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7284 (riscv_attribute_table, riscv_output_return),
7285 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7286 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7287 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7289 (riscv_expand_prologue): Add early return for naked function.
7290 (riscv_expand_epilogue): Likewise.
7291 (riscv_function_ok_for_sibcall): Return false for naked function.
7292 (riscv_set_current_function): New.
7293 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7294 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7295 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7296 * doc/extend.texi (RISC-V Function Attributes): New.
7298 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7300 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7301 check for 128-bit long double before checking TCmode.
7302 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7303 128-bit long doubles before checking TFmode or TCmode.
7304 (FLOAT128_IBM_P): Likewise.
7306 2018-01-10 Martin Sebor <msebor@redhat.com>
7308 PR tree-optimization/83671
7309 * builtins.c (c_strlen): Unconditionally return zero for the empty
7311 Use -Warray-bounds for warnings.
7312 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7313 for non-constant array indices with COMPONENT_REF, arrays of
7314 arrays, and pointers to arrays.
7315 (gimple_fold_builtin_strlen): Determine and set length range for
7316 non-constant character arrays.
7318 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7321 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7324 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7326 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7328 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7331 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7332 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7333 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7334 indexed_or_indirect_operand predicate.
7335 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7336 (*vsx_le_perm_load_v8hi): Likewise.
7337 (*vsx_le_perm_load_v16qi): Likewise.
7338 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7339 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7340 (*vsx_le_perm_store_v8hi): Likewise.
7341 (*vsx_le_perm_store_v16qi): Likewise.
7342 (eight unnamed splitters): Likewise.
7344 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7346 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7347 * config/rs6000/emmintrin.h: Likewise.
7348 * config/rs6000/mmintrin.h: Likewise.
7349 * config/rs6000/xmmintrin.h: Likewise.
7351 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7354 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7356 * tree.c (tree_nop_conversion): Return true for location wrapper
7358 (maybe_wrap_with_location): New function.
7359 (selftest::check_strip_nops): New function.
7360 (selftest::test_location_wrappers): New function.
7361 (selftest::tree_c_tests): Call it.
7362 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7363 (maybe_wrap_with_location): New decl.
7364 (EXPR_LOCATION_WRAPPER_P): New macro.
7365 (location_wrapper_p): New inline function.
7366 (tree_strip_any_location_wrapper): New inline function.
7368 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7371 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7372 stack_realign_offset for the largest alignment of stack slot
7374 (ix86_find_max_used_stack_alignment): New function.
7375 (ix86_finalize_stack_frame_flags): Use it. Set
7376 max_used_stack_alignment if we don't realign stack.
7377 * config/i386/i386.h (machine_function): Add
7378 max_used_stack_alignment.
7380 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7382 * config/arm/arm.opt (-mbranch-cost): New option.
7383 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7386 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7389 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7390 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7392 2018-01-10 Richard Biener <rguenther@suse.de>
7395 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7396 early out so it also covers the case where we have a non-NULL
7399 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7401 PR tree-optimization/83753
7402 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7403 for non-strided grouped accesses if the number of elements is 1.
7405 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7408 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7409 * i386.h (TARGET_USE_GATHER): Define.
7410 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7412 2018-01-10 Martin Liska <mliska@suse.cz>
7415 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7416 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7418 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7419 CLEANUP_NO_PARTITIONING is not set.
7421 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7423 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7424 for vectors, as a partial revert of r254296.
7425 * rtl.h (const_vec_p): Delete.
7426 (const_vec_duplicate_p): Don't test for vector CONSTs.
7427 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7428 * expmed.c (make_tree): Likewise.
7431 * common.md (E, F): Use CONSTANT_P instead of checking for
7433 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7434 checking for CONST_VECTOR.
7436 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7439 * predict.c (force_edge_cold): Handle in more sane way edges
7442 2018-01-09 Carl Love <cel@us.ibm.com>
7444 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7446 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7447 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7448 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7449 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7450 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7451 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7452 * config/rs6000/rs6000-protos.h: Add extern defition for
7453 rs6000_generate_float2_double_code.
7454 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7456 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7457 (float2_v2df): Add define_expand.
7459 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7462 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7463 op_mode in the force_to_mode call.
7465 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7467 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7468 instead of checking each element individually.
7469 (aarch64_evpc_uzp): Likewise.
7470 (aarch64_evpc_zip): Likewise.
7471 (aarch64_evpc_ext): Likewise.
7472 (aarch64_evpc_rev): Likewise.
7473 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7474 instead of checking each element individually. Return true without
7476 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7477 whether all selected elements come from the same input, instead of
7478 checking each element individually. Remove calls to gen_rtx_REG,
7479 start_sequence and end_sequence and instead assert that no rtl is
7482 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7484 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7485 order of HIGH and CONST checks.
7487 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7489 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7490 if the destination isn't an SSA_NAME.
7492 2018-01-09 Richard Biener <rguenther@suse.de>
7494 PR tree-optimization/83668
7495 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7497 (canonicalize_loop_form): ... here, renamed from ...
7498 (canonicalize_loop_closed_ssa_form): ... this and amended to
7499 swap successor edges for loop exit blocks to make us use
7500 the RPO order we need for initial schedule generation.
7502 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7504 PR tree-optimization/64811
7505 * match.pd: When optimizing comparisons with Inf, avoid
7506 introducing or losing exceptions from comparisons with NaN.
7508 2018-01-09 Martin Liska <mliska@suse.cz>
7511 * asan.c (shadow_mem_size): Add gcc_assert.
7513 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7515 Don't save registers in main().
7518 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7519 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7520 * config/avr/avr.c (avr_set_current_function): Don't error if
7521 naked, OS_task or OS_main are specified at the same time.
7522 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7524 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7526 * common/config/avr/avr-common.c (avr_option_optimization_table):
7527 Switch on -mmain-is-OS_task for optimizing compilations.
7529 2018-01-09 Richard Biener <rguenther@suse.de>
7531 PR tree-optimization/83572
7532 * graphite.c: Include cfganal.h.
7533 (graphite_transform_loops): Connect infinite loops to exit
7534 and remove fake edges at the end.
7536 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7538 * ipa-inline.c (edge_badness): Revert accidental checkin.
7540 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7543 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7544 symbols; not inline clones.
7546 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7549 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7550 hard registers. Formatting fixes.
7552 PR preprocessor/83722
7553 * gcc.c (try_generate_repro): Pass
7554 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7555 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7558 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7559 Kito Cheng <kito.cheng@gmail.com>
7561 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7562 (riscv_leaf_function_p): Delete.
7563 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7565 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7567 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7569 (do_ifelse): New function.
7570 (do_isel): New function.
7571 (do_sub3): New function.
7572 (do_add3): New function.
7573 (do_load_mask_compare): New function.
7574 (do_overlap_load_compare): New function.
7575 (expand_compare_loop): New function.
7576 (expand_block_compare): Call expand_compare_loop() when appropriate.
7577 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7579 (-mblock-compare-inline-loop-limit): New option.
7581 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7584 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7585 Reverse order of second and third operands in first alternative.
7586 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7587 of first and second elements in UNSPEC_VPERMR vector.
7588 (altivec_expand_vec_perm_le): Likewise.
7590 2018-01-08 Jeff Law <law@redhat.com>
7592 PR rtl-optimizatin/81308
7593 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7594 (process_switch): If group_case_labels makes a change, then set
7596 (pass_convert_switch::execute): If a switch is converted, then
7597 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7599 PR rtl-optimization/81308
7600 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7603 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7605 PR target/83663 - Revert r255946
7606 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7607 generation for cases where splatting a value is not useful.
7608 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7609 across a vec_duplicate and a paradoxical subreg forming a vector
7610 mode to a vec_concat.
7612 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7614 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7615 -march=armv8.3-a variants.
7616 * config/arm/t-multilib: Likewise.
7617 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7619 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7621 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7623 (cceq_ior_compare_complement): Give it a name so I can use it, and
7624 change boolean_or_operator predicate to boolean_operator so it can
7625 be used to generate a crand.
7626 (eqne): New code iterator.
7627 (bd/bd_neg): New code_attrs.
7628 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7629 a single define_insn.
7630 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7631 decrement (bdnzt/bdnzf/bdzt/bdzf).
7632 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7633 with the new names of the branch decrement patterns, and added the
7634 names of the branch decrement conditional patterns.
7636 2018-01-08 Richard Biener <rguenther@suse.de>
7638 PR tree-optimization/83563
7639 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7642 2018-01-08 Richard Biener <rguenther@suse.de>
7645 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7647 2018-01-08 Richard Biener <rguenther@suse.de>
7649 PR tree-optimization/83685
7650 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7651 references to abnormals.
7653 2018-01-08 Richard Biener <rguenther@suse.de>
7656 * dwarf2out.c (output_indirect_strings): Handle empty
7657 skeleton_debug_str_hash.
7658 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7660 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7662 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7663 (emit_store_direct): Likewise.
7664 (arc_trampoline_adjust_address): Likewise.
7665 (arc_asm_trampoline_template): New function.
7666 (arc_initialize_trampoline): Use asm_trampoline_template.
7667 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7668 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7669 * config/arc/arc.md (flush_icache): Delete pattern.
7671 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7673 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7674 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7677 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7680 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7681 by not USED_FOR_TARGET.
7682 (make_pass_resolve_sw_modes): Likewise.
7684 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7686 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7689 2018-01-08 Richard Biener <rguenther@suse.de>
7692 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7694 2018-01-08 Richard Biener <rguenther@suse.de>
7697 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7699 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7702 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7703 basic blocks with a small number of successors.
7704 (convert_control_dep_chain_into_preds): Improve handling of
7706 (dump_predicates): Split apart into...
7707 (dump_pred_chain): ...here...
7708 (dump_pred_info): ...and here.
7709 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7710 (can_chain_union_be_invalidated_p): Improve check for invalidation
7712 (uninit_uses_cannot_happen): Avoid unnecessary if
7713 convert_control_dep_chain_into_preds yielded nothing.
7715 2018-01-06 Martin Sebor <msebor@redhat.com>
7717 PR tree-optimization/83640
7718 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7719 subtracting negative offset from size.
7720 (builtin_access::overlap): Adjust offset bounds of the access to fall
7721 within the size of the object if possible.
7723 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7725 PR rtl-optimization/83699
7726 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7727 extract_bit_field_as_subreg to cases in which the extracted
7728 value is also a vector.
7730 * lra-constraints.c (process_alt_operands): Test for the equivalence
7731 substitutions when detecting a possible reload cycle.
7733 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7736 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7737 by default if flag_selective_schedling{,2}. Formatting fixes.
7739 PR rtl-optimization/83682
7740 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7741 if it has non-VECTOR_MODE element mode.
7742 (vec_duplicate_p): Likewise.
7745 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7746 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7748 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7751 * config/i386/i386-builtin.def
7752 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7753 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7754 Require also OPTION_MASK_ISA_AVX512F in addition to
7755 OPTION_MASK_ISA_GFNI.
7756 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7757 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7758 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7759 to OPTION_MASK_ISA_GFNI.
7760 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7761 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7762 OPTION_MASK_ISA_AVX512BW.
7763 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7764 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7765 addition to OPTION_MASK_ISA_GFNI.
7766 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7767 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7768 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7769 to OPTION_MASK_ISA_GFNI.
7770 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7771 a requirement for all ISAs rather than any of them with a few
7773 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7775 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7776 bitmasks to be enabled with 3 exceptions, instead of requiring any
7777 enabled ISA with lots of exceptions.
7778 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7779 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7780 Change avx512bw in isa attribute to avx512f.
7781 * config/i386/sgxintrin.h: Add license boilerplate.
7782 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7783 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7784 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7785 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7787 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7788 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7789 temporarily sse2 rather than sse if not enabled already.
7792 * config/i386/sse.md (VI248_VLBW): Rename to ...
7793 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7794 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7795 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7796 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7797 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7798 mode iterator instead of VI248_VLBW.
7800 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7802 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7803 (record_modified): Skip clobbers; add debug output.
7804 (param_change_prob): Use sreal frequencies.
7806 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7808 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7809 punt for user-aligned variables.
7811 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7813 * tree-chrec.c (chrec_contains_symbols): Return true for
7816 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7819 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7820 of (x|y) == x for BICS pattern.
7822 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7824 PR tree-optimization/83605
7825 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7826 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7829 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7831 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7832 * config/epiphany/rtems.h: New file.
7834 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7835 Uros Bizjak <ubizjak@gmail.com>
7838 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7839 QIreg_operand instead of register_operand predicate.
7840 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7841 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7842 comments instead of -fmitigate[-_]rop.
7844 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7847 * cgraphunit.c (symbol_table::compile): Switch to text_section
7848 before calling assembly_start debug hook.
7849 * run-rtl-passes.c (run_rtl_passes): Likewise.
7852 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7854 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7855 range_int_cst_p rather than !symbolic_range_p before calling
7856 extract_range_from_multiplicative_op_1.
7858 2018-01-04 Jeff Law <law@redhat.com>
7860 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7861 redundant test in assertion.
7863 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7865 * doc/rtl.texi: Document machine_mode wrapper classes.
7867 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7869 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7872 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7874 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7875 the VEC_PERM_EXPR fold to fail.
7877 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7880 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7881 to switched_sections.
7883 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7886 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7889 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7892 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7893 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7895 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7898 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7899 is BLKmode and bitpos not zero or mode change is needed.
7901 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7904 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7907 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7910 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7911 instead of MULT rtx. Update all corresponding splitters.
7913 (*ssub<modesuffix>): Ditto.
7915 (*cmp_sadd_di): Update split patterns.
7916 (*cmp_sadd_si): Ditto.
7917 (*cmp_sadd_sidi): Ditto.
7918 (*cmp_ssub_di): Ditto.
7919 (*cmp_ssub_si): Ditto.
7920 (*cmp_ssub_sidi): Ditto.
7921 * config/alpha/predicates.md (const23_operand): New predicate.
7922 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7923 Look for ASHIFT, not MULT inner operand.
7924 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7926 2018-01-04 Martin Liska <mliska@suse.cz>
7928 PR gcov-profile/83669
7929 * gcov.c (output_intermediate_file): Add version to intermediate
7931 * doc/gcov.texi: Document new field 'version' in intermediate
7932 file format. Fix location of '-k' option of gcov command.
7934 2018-01-04 Martin Liska <mliska@suse.cz>
7937 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7939 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7941 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7943 2018-01-03 Martin Sebor <msebor@redhat.com>
7945 PR tree-optimization/83655
7946 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7947 checking calls with invalid arguments.
7949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7951 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7952 (vectorizable_mask_load_store): Delete.
7953 (vectorizable_call): Return false for masked loads and stores.
7954 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7955 instead of gimple_assign_rhs1.
7956 (vectorizable_load): Handle IFN_MASK_LOAD.
7957 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7961 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7963 (vectorizable_mask_load_store): ...here.
7964 (vectorizable_load): ...and here.
7966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7968 * tree-vect-stmts.c (vect_build_all_ones_mask)
7969 (vect_build_zero_merge_argument): New functions, split out from...
7970 (vectorizable_load): ...here.
7972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7974 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7976 (vectorizable_mask_load_store): ...here.
7977 (vectorizable_store): ...and here.
7979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7981 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7983 (vectorizable_mask_load_store): ...here.
7985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7988 (vect_model_store_cost): Take a vec_load_store_type instead of a
7990 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7991 (vect_model_store_cost): Take a vec_load_store_type instead of a
7993 (vectorizable_mask_load_store): Update accordingly.
7994 (vectorizable_store): Likewise.
7995 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7999 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8000 IFN_MASK_LOAD calls here rather than...
8001 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8004 Alan Hayward <alan.hayward@arm.com>
8005 David Sherwood <david.sherwood@arm.com>
8007 * expmed.c (extract_bit_field_1): For vector extracts,
8008 fall back to extract_bit_field_as_subreg if vec_extract
8011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8012 Alan Hayward <alan.hayward@arm.com>
8013 David Sherwood <david.sherwood@arm.com>
8015 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8016 they are variable or constant sized.
8017 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8018 slots for constant-sized data.
8020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8021 Alan Hayward <alan.hayward@arm.com>
8022 David Sherwood <david.sherwood@arm.com>
8024 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8025 handling COND_EXPRs with boolean comparisons, try to find a better
8026 basis for the mask type than the boolean itself.
8028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8030 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8031 is calculated and how it can be overridden.
8032 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8033 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8035 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8039 Alan Hayward <alan.hayward@arm.com>
8040 David Sherwood <david.sherwood@arm.com>
8042 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8043 Remove the mode argument.
8044 (aarch64_simd_valid_immediate): Remove the mode and inverse
8046 * config/aarch64/iterators.md (bitsize): New iterator.
8047 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8048 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8049 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8050 aarch64_simd_valid_immediate.
8051 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8052 (aarch64_reg_or_bic_imm): Likewise.
8053 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8054 with an insn_type enum and msl with a modifier_type enum.
8055 Replace element_width with a scalar_mode. Change the shift
8056 to unsigned int. Add constructors for scalar_float_mode and
8057 scalar_int_mode elements.
8058 (aarch64_vect_float_const_representable_p): Delete.
8059 (aarch64_can_const_movi_rtx_p)
8060 (aarch64_simd_scalar_immediate_valid_for_move)
8061 (aarch64_simd_make_constant): Update call to
8062 aarch64_simd_valid_immediate.
8063 (aarch64_advsimd_valid_immediate_hs): New function.
8064 (aarch64_advsimd_valid_immediate): Likewise.
8065 (aarch64_simd_valid_immediate): Remove mode and inverse
8066 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8067 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8068 and aarch64_float_const_representable_p on the result.
8069 (aarch64_output_simd_mov_immediate): Remove mode argument.
8070 Update call to aarch64_simd_valid_immediate and use of
8071 simd_immediate_info.
8072 (aarch64_output_scalar_simd_mov_immediate): Update call
8075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8076 Alan Hayward <alan.hayward@arm.com>
8077 David Sherwood <david.sherwood@arm.com>
8079 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8080 (mode_nunits): Likewise CONST_MODE_NUNITS.
8081 * machmode.def (ADJUST_NUNITS): Document.
8082 * genmodes.c (mode_data::need_nunits_adj): New field.
8083 (blank_mode): Update accordingly.
8084 (adj_nunits): New variable.
8085 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8087 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8088 listed in adj_nunits.
8089 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8090 listed in adj_nunits. Don't emit case statements for such modes.
8091 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8092 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8093 nothing if adj_nunits is nonnull.
8094 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8095 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8096 (emit_mode_fbit): Update use of print_maybe_const_decl.
8097 (emit_move_size): Likewise. Treat the array as non-const
8099 (emit_mode_adjustments): Handle adj_nunits.
8101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8103 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8104 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8105 (VECTOR_MODES): Use it.
8106 (make_vector_modes): Take the prefix as an argument.
8108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8109 Alan Hayward <alan.hayward@arm.com>
8110 David Sherwood <david.sherwood@arm.com>
8112 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8113 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8114 for MODE_VECTOR_BOOL.
8115 * machmode.def (VECTOR_BOOL_MODE): Document.
8116 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8117 (make_vector_bool_mode): New function.
8118 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8120 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8121 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8123 * stor-layout.c (int_mode_for_mode): Likewise.
8124 * tree.c (build_vector_type_for_mode): Likewise.
8125 * varasm.c (output_constant_pool_2): Likewise.
8126 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8127 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8128 for MODE_VECTOR_BOOL.
8129 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8130 of mode class checks.
8131 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8132 instead of a list of mode class checks.
8133 (expand_vector_scalar_condition): Likewise.
8134 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8137 Alan Hayward <alan.hayward@arm.com>
8138 David Sherwood <david.sherwood@arm.com>
8140 * machmode.h (mode_size): Change from unsigned short to
8142 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8143 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8144 or if measurement_type is not polynomial.
8145 (fixed_size_mode::includes_p): Check for constant-sized modes.
8146 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8147 return a poly_uint16 rather than an unsigned short.
8148 (emit_mode_size): Change the type of mode_size from unsigned short
8149 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8150 (emit_mode_adjustments): Cope with polynomial vector sizes.
8151 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8153 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8155 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8156 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8157 * caller-save.c (setup_save_areas): Likewise.
8158 (replace_reg_with_saved_mem): Likewise.
8159 * calls.c (emit_library_call_value_1): Likewise.
8160 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8161 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8162 (gen_lowpart_for_combine): Likewise.
8163 * convert.c (convert_to_integer_1): Likewise.
8164 * cse.c (equiv_constant, cse_insn): Likewise.
8165 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8166 (cselib_subst_to_values): Likewise.
8167 * dce.c (word_dce_process_block): Likewise.
8168 * df-problems.c (df_word_lr_mark_ref): Likewise.
8169 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8170 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8171 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8172 (rtl_for_decl_location): Likewise.
8173 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8174 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8175 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8176 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8177 (expand_expr_real_1): Likewise.
8178 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8179 (pad_below): Likewise.
8180 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8181 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8182 * ira.c (get_subreg_tracking_sizes): Likewise.
8183 * ira-build.c (ira_create_allocno_objects): Likewise.
8184 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8185 (ira_sort_regnos_for_alter_reg): Likewise.
8186 * ira-costs.c (record_operand_costs): Likewise.
8187 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8188 (resolve_simple_move): Likewise.
8189 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8190 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8191 (lra_constraints): Likewise.
8192 (CONST_POOL_OK_P): Reject variable-sized modes.
8193 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8194 (add_pseudo_to_slot, lra_spill): Likewise.
8195 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8196 * optabs-query.c (get_best_extraction_insn): Likewise.
8197 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8198 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8199 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8200 * recog.c (offsettable_address_addr_space_p): Likewise.
8201 * regcprop.c (maybe_mode_change): Likewise.
8202 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8203 * regrename.c (build_def_use): Likewise.
8204 * regstat.c (dump_reg_info): Likewise.
8205 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8206 (find_reloads, find_reloads_subreg_address): Likewise.
8207 * reload1.c (eliminate_regs_1): Likewise.
8208 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8209 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8210 (simplify_binary_operation_1, simplify_subreg): Likewise.
8211 * targhooks.c (default_function_arg_padding): Likewise.
8212 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8213 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8214 (verify_gimple_assign_ternary): Likewise.
8215 * tree-inline.c (estimate_move_cost): Likewise.
8216 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8217 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8218 (get_address_cost_ainc): Likewise.
8219 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8220 (vect_supportable_dr_alignment): Likewise.
8221 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8222 (vectorizable_reduction): Likewise.
8223 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8224 (vectorizable_operation, vectorizable_load): Likewise.
8225 * tree.c (build_same_sized_truth_vector_type): Likewise.
8226 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8227 * var-tracking.c (emit_note_insn_var_location): Likewise.
8228 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8229 (ADDR_VEC_ALIGN): Likewise.
8231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8232 Alan Hayward <alan.hayward@arm.com>
8233 David Sherwood <david.sherwood@arm.com>
8235 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8237 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8238 or if measurement_type is polynomial.
8239 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8240 * combine.c (make_extraction): Likewise.
8241 * dse.c (find_shift_sequence): Likewise.
8242 * dwarf2out.c (mem_loc_descriptor): Likewise.
8243 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8244 (extract_bit_field, extract_low_bits): Likewise.
8245 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8246 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8247 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8248 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8249 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8250 * reload.c (find_reloads): Likewise.
8251 * reload1.c (alter_reg): Likewise.
8252 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8253 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8254 * tree-if-conv.c (predicate_mem_writes): Likewise.
8255 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8256 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8257 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8258 * valtrack.c (dead_debug_insert_temp): Likewise.
8259 * varasm.c (mergeable_constant_section): Likewise.
8260 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8263 Alan Hayward <alan.hayward@arm.com>
8264 David Sherwood <david.sherwood@arm.com>
8266 * expr.c (expand_assignment): Cope with polynomial mode sizes
8267 when assigning to a CONCAT.
8269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8270 Alan Hayward <alan.hayward@arm.com>
8271 David Sherwood <david.sherwood@arm.com>
8273 * machmode.h (mode_precision): Change from unsigned short to
8275 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8277 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8278 or if measurement_type is not polynomial.
8279 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8280 in which the mode is already known to be a scalar_int_mode.
8281 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8282 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8284 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8285 for GET_MODE_PRECISION.
8286 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8287 for GET_MODE_PRECISION.
8288 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8290 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8291 (expand_field_assignment, make_extraction): Likewise.
8292 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8293 (get_last_value): Likewise.
8294 * convert.c (convert_to_integer_1): Likewise.
8295 * cse.c (cse_insn): Likewise.
8296 * expr.c (expand_expr_real_1): Likewise.
8297 * lra-constraints.c (simplify_operand_subreg): Likewise.
8298 * optabs-query.c (can_atomic_load_p): Likewise.
8299 * optabs.c (expand_atomic_load): Likewise.
8300 (expand_atomic_store): Likewise.
8301 * ree.c (combine_reaching_defs): Likewise.
8302 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8303 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8304 * tree.h (type_has_mode_precision_p): Likewise.
8305 * ubsan.c (instrument_si_overflow): Likewise.
8307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8308 Alan Hayward <alan.hayward@arm.com>
8309 David Sherwood <david.sherwood@arm.com>
8311 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8312 polynomial numbers of units.
8313 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8314 (valid_vector_subparts_p): New function.
8315 (build_vector_type): Remove temporary shim and take the number
8316 of units as a poly_uint64 rather than an int.
8317 (build_opaque_vector_type): Take the number of units as a
8318 poly_uint64 rather than an int.
8319 * tree.c (build_vector_from_ctor): Handle polynomial
8320 TYPE_VECTOR_SUBPARTS.
8321 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8322 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8323 (build_vector_from_val): If the number of units is variable,
8324 use build_vec_duplicate_cst for constant operands and
8325 VEC_DUPLICATE_EXPR otherwise.
8326 (make_vector_type): Remove temporary is_constant ().
8327 (build_vector_type, build_opaque_vector_type): Take the number of
8328 units as a poly_uint64 rather than an int.
8329 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8331 * cfgexpand.c (expand_debug_expr): Likewise.
8332 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8333 (store_constructor, expand_expr_real_1): Likewise.
8334 (const_scalar_mask_from_tree): Likewise.
8335 * fold-const-call.c (fold_const_reduction): Likewise.
8336 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8337 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8338 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8339 (fold_relational_const): Likewise.
8340 (native_interpret_vector): Likewise. Change the size from an
8341 int to an unsigned int.
8342 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8343 TYPE_VECTOR_SUBPARTS.
8344 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8345 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8346 duplicating a non-constant operand into a variable-length vector.
8347 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8348 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8349 * ipa-icf.c (sem_variable::equals): Likewise.
8350 * match.pd: Likewise.
8351 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8352 * print-tree.c (print_node): Likewise.
8353 * stor-layout.c (layout_type): Likewise.
8354 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8355 * tree-cfg.c (verify_gimple_comparison): Likewise.
8356 (verify_gimple_assign_binary): Likewise.
8357 (verify_gimple_assign_ternary): Likewise.
8358 (verify_gimple_assign_single): Likewise.
8359 * tree-pretty-print.c (dump_generic_node): Likewise.
8360 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8361 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8362 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8363 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8364 (vect_shift_permute_load_chain): Likewise.
8365 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8366 (expand_vector_condition, optimize_vector_constructor): Likewise.
8367 (lower_vec_perm, get_compute_type): Likewise.
8368 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8369 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8370 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8371 (vect_recog_mask_conversion_pattern): Likewise.
8372 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8373 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8374 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8375 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8376 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8377 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8378 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8379 (supportable_widening_operation): Likewise.
8380 (supportable_narrowing_operation): Likewise.
8381 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8383 * varasm.c (output_constant): Likewise.
8385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8386 Alan Hayward <alan.hayward@arm.com>
8387 David Sherwood <david.sherwood@arm.com>
8389 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8390 so that both the length == 3 and length != 3 cases set up their
8391 own permute vectors. Add comments explaining why we know the
8392 number of elements is constant.
8393 (vect_permute_load_chain): Likewise.
8395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8396 Alan Hayward <alan.hayward@arm.com>
8397 David Sherwood <david.sherwood@arm.com>
8399 * machmode.h (mode_nunits): Change from unsigned char to
8401 (ONLY_FIXED_SIZE_MODES): New macro.
8402 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8403 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8404 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8406 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8407 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8408 or if measurement_type is not polynomial.
8409 * genmodes.c (ZERO_COEFFS): New macro.
8410 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8412 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8413 Use ZERO_COEFFS when emitting initializers.
8414 * data-streamer.h (bp_pack_poly_value): New function.
8415 (bp_unpack_poly_value): Likewise.
8416 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8417 for GET_MODE_NUNITS.
8418 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8419 for GET_MODE_NUNITS.
8420 * tree.c (make_vector_type): Remove temporary shim and make
8421 the real function take the number of units as a poly_uint64
8423 (build_vector_type_for_mode): Handle polynomial nunits.
8424 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8425 * emit-rtl.c (const_vec_series_p_1): Likewise.
8426 (gen_rtx_CONST_VECTOR): Likewise.
8427 * fold-const.c (test_vec_duplicate_folding): Likewise.
8428 * genrecog.c (validate_pattern): Likewise.
8429 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8430 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8431 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8432 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8433 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8434 * rtlanal.c (subreg_get_info): Likewise.
8435 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8436 (vect_grouped_load_supported): Likewise.
8437 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8438 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8439 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8440 (simplify_const_unary_operation, simplify_binary_operation_1)
8441 (simplify_const_binary_operation, simplify_ternary_operation)
8442 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8443 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8444 instead of CONST_VECTOR_NUNITS.
8445 * varasm.c (output_constant_pool_2): Likewise.
8446 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8447 explicit-encoded elements in the XVEC for variable-length vectors.
8449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8451 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8454 Alan Hayward <alan.hayward@arm.com>
8455 David Sherwood <david.sherwood@arm.com>
8457 * coretypes.h (fixed_size_mode): Declare.
8458 (fixed_size_mode_pod): New typedef.
8459 * builtins.h (target_builtins::x_apply_args_mode)
8460 (target_builtins::x_apply_result_mode): Change type to
8461 fixed_size_mode_pod.
8462 * builtins.c (apply_args_size, apply_result_size, result_vector)
8463 (expand_builtin_apply_args_1, expand_builtin_apply)
8464 (expand_builtin_return): Update accordingly.
8466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8468 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8469 * cselib.c (cselib_hash_rtx): Likewise.
8470 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8471 CONST_VECTOR encoding.
8473 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8474 Jeff Law <law@redhat.com>
8477 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8478 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8479 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8480 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8483 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8484 explicitly probe *sp in a noreturn function if there were any callee
8485 register saves or frame pointer is needed.
8487 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8490 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8491 BLKmode for ternary, binary or unary expressions.
8494 * var-tracking.c (delete_vta_debug_insn): New inline function.
8495 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8496 insns from get_insns () to NULL instead of each bb separately.
8497 Use delete_vta_debug_insn. No longer static.
8498 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8499 delete_vta_debug_insns callers.
8500 * rtl.h (delete_vta_debug_insns): Declare.
8501 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8502 instead of variable_tracking_main.
8504 2018-01-03 Martin Sebor <msebor@redhat.com>
8506 PR tree-optimization/83603
8507 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8508 arguments past the endof the argument list in functions declared
8509 without a prototype.
8510 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8511 Avoid checking when arguments are null.
8513 2018-01-03 Martin Sebor <msebor@redhat.com>
8516 * doc/extend.texi (attribute const): Fix a typo.
8517 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8518 issuing -Wsuggest-attribute for void functions.
8520 2018-01-03 Martin Sebor <msebor@redhat.com>
8522 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8523 offset_int::from instead of wide_int::to_shwi.
8524 (maybe_diag_overlap): Remove assertion.
8525 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8526 * gimple-ssa-sprintf.c (format_directive): Same.
8527 (parse_directive): Same.
8528 (sprintf_dom_walker::compute_format_length): Same.
8529 (try_substitute_return_value): Same.
8531 2018-01-03 Jeff Law <law@redhat.com>
8534 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8535 non-constant residual for zero at runtime and avoid probing in
8536 that case. Reorganize code for trailing problem to mirror handling
8539 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8541 PR tree-optimization/83501
8542 * tree-ssa-strlen.c (get_string_cst): New.
8543 (handle_char_store): Call get_string_cst.
8545 2018-01-03 Martin Liska <mliska@suse.cz>
8547 PR tree-optimization/83593
8548 * tree-ssa-strlen.c: Include tree-cfg.h.
8549 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8550 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8551 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8553 (strlen_dom_walker::before_dom_children): Call
8554 gimple_purge_dead_eh_edges. Dump tranformation with details
8556 (strlen_dom_walker::before_dom_children): Update call by adding
8557 new argument cleanup_eh.
8558 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8560 2018-01-03 Martin Liska <mliska@suse.cz>
8563 * cif-code.def (VARIADIC_THUNK): New enum value.
8564 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8567 2018-01-03 Jan Beulich <jbeulich@suse.com>
8569 * sse.md (mov<mode>_internal): Tighten condition for when to use
8570 vmovdqu<ssescalarsize> for TI and OI modes.
8572 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8574 Update copyright years.
8576 2018-01-03 Martin Liska <mliska@suse.cz>
8579 * ipa-visibility.c (function_and_variable_visibility): Skip
8580 functions with noipa attribure.
8582 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8584 * gcc.c (process_command): Update copyright notice dates.
8585 * gcov-dump.c (print_version): Ditto.
8586 * gcov.c (print_version): Ditto.
8587 * gcov-tool.c (print_version): Ditto.
8588 * gengtype.c (create_file): Ditto.
8589 * doc/cpp.texi: Bump @copying's copyright year.
8590 * doc/cppinternals.texi: Ditto.
8591 * doc/gcc.texi: Ditto.
8592 * doc/gccint.texi: Ditto.
8593 * doc/gcov.texi: Ditto.
8594 * doc/install.texi: Ditto.
8595 * doc/invoke.texi: Ditto.
8597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8599 * vector-builder.h (vector_builder::m_full_nelts): Change from
8600 unsigned int to poly_uint64.
8601 (vector_builder::full_nelts): Update prototype accordingly.
8602 (vector_builder::new_vector): Likewise.
8603 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8604 (vector_builder::operator ==): Likewise.
8605 (vector_builder::finalize): Likewise.
8606 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8607 Take the number of elements as a poly_uint64 rather than an
8609 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8610 from unsigned int to poly_uint64.
8611 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8612 (vec_perm_indices::new_vector): Likewise.
8613 (vec_perm_indices::length): Likewise.
8614 (vec_perm_indices::nelts_per_input): Likewise.
8615 (vec_perm_indices::input_nelts): Likewise.
8616 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8617 number of elements per input as a poly_uint64 rather than an
8618 unsigned int. Use the original encoding for variable-length
8619 vectors, rather than clamping each individual element.
8620 For the second and subsequent elements in each pattern,
8621 clamp the step and base before clamping their sum.
8622 (vec_perm_indices::series_p): Handle polynomial element counts.
8623 (vec_perm_indices::all_in_range_p): Likewise.
8624 (vec_perm_indices_to_tree): Likewise.
8625 (vec_perm_indices_to_rtx): Likewise.
8626 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8627 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8628 (tree_vector_builder::new_binary_operation): Handle polynomial
8629 element counts. Return false if we need to know the number
8630 of elements at compile time.
8631 * fold-const.c (fold_vec_perm): Punt if the number of elements
8632 isn't known at compile time.
8634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8636 * vec-perm-indices.h (vec_perm_builder): Change element type
8637 from HOST_WIDE_INT to poly_int64.
8638 (vec_perm_indices::element_type): Update accordingly.
8639 (vec_perm_indices::clamp): Handle polynomial element_types.
8640 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8641 (vec_perm_indices::all_in_range_p): Likewise.
8642 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8644 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8645 polynomial vec_perm_indices element types.
8646 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8647 * fold-const.c (fold_vec_perm): Likewise.
8648 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8649 * tree-vect-generic.c (lower_vec_perm): Likewise.
8650 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8651 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8652 element type to HOST_WIDE_INT.
8654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8655 Alan Hayward <alan.hayward@arm.com>
8656 David Sherwood <david.sherwood@arm.com>
8658 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8659 rather than an int. Use plus_constant.
8660 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8661 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8664 Alan Hayward <alan.hayward@arm.com>
8665 David Sherwood <david.sherwood@arm.com>
8667 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8668 a HOST_WIDE_INT to a poly_int64.
8670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8671 Alan Hayward <alan.hayward@arm.com>
8672 David Sherwood <david.sherwood@arm.com>
8674 * calls.c (load_register_parameters): Cope with polynomial
8675 mode sizes. Require a constant size for BLKmode parameters
8676 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8677 forces a parameter to be padded at the lsb end in order to
8678 fill a complete number of words, require the parameter size
8679 to be ordered wrt UNITS_PER_WORD.
8681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8682 Alan Hayward <alan.hayward@arm.com>
8683 David Sherwood <david.sherwood@arm.com>
8685 * reload1.c (spill_stack_slot_width): Change element type
8686 from unsigned int to poly_uint64_pod.
8687 (alter_reg): Treat mode sizes as polynomial.
8689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8690 Alan Hayward <alan.hayward@arm.com>
8691 David Sherwood <david.sherwood@arm.com>
8693 * reload.c (complex_word_subreg_p): New function.
8694 (reload_inner_reg_of_subreg, push_reload): Use it.
8696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8697 Alan Hayward <alan.hayward@arm.com>
8698 David Sherwood <david.sherwood@arm.com>
8700 * lra-constraints.c (process_alt_operands): Reject matched
8701 operands whose sizes aren't ordered.
8702 (match_reload): Refer to this check here.
8704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8705 Alan Hayward <alan.hayward@arm.com>
8706 David Sherwood <david.sherwood@arm.com>
8708 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8709 that the mode size is in the set {1, 2, 4, 8, 16}.
8711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8712 Alan Hayward <alan.hayward@arm.com>
8713 David Sherwood <david.sherwood@arm.com>
8715 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8716 Use plus_constant instead of gen_rtx_PLUS.
8718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8719 Alan Hayward <alan.hayward@arm.com>
8720 David Sherwood <david.sherwood@arm.com>
8722 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8723 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8724 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8725 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8726 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8727 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8728 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8729 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8730 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8731 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8733 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8734 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8735 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8736 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8737 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8738 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8739 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8740 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8741 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8742 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8744 * expr.c (emit_move_resolve_push): Treat the input and result
8745 of PUSH_ROUNDING as a poly_int64.
8746 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8747 (emit_push_insn): Likewise.
8748 * lra-eliminations.c (mark_not_eliminable): Likewise.
8749 * recog.c (push_operand): Likewise.
8750 * reload1.c (elimination_effects): Likewise.
8751 * rtlanal.c (nonzero_bits1): Likewise.
8752 * calls.c (store_one_arg): Likewise. Require the padding to be
8753 known at compile time.
8755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8756 Alan Hayward <alan.hayward@arm.com>
8757 David Sherwood <david.sherwood@arm.com>
8759 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8760 Use plus_constant instead of gen_rtx_PLUS.
8762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8763 Alan Hayward <alan.hayward@arm.com>
8764 David Sherwood <david.sherwood@arm.com>
8766 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8770 Alan Hayward <alan.hayward@arm.com>
8771 David Sherwood <david.sherwood@arm.com>
8773 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8774 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8775 via stack temporaries. Treat the mode size as polynomial too.
8777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8778 Alan Hayward <alan.hayward@arm.com>
8779 David Sherwood <david.sherwood@arm.com>
8781 * expr.c (expand_expr_real_2): When handling conversions involving
8782 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8783 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8784 as a poly_uint64 too.
8786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8787 Alan Hayward <alan.hayward@arm.com>
8788 David Sherwood <david.sherwood@arm.com>
8790 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8793 Alan Hayward <alan.hayward@arm.com>
8794 David Sherwood <david.sherwood@arm.com>
8796 * combine.c (can_change_dest_mode): Handle polynomial
8797 REGMODE_NATURAL_SIZE.
8798 * expmed.c (store_bit_field_1): Likewise.
8799 * expr.c (store_constructor): Likewise.
8800 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8801 and polynomial REGMODE_NATURAL_SIZE.
8802 (gen_lowpart_common): Likewise.
8803 * reginfo.c (record_subregs_of_mode): Likewise.
8804 * rtlanal.c (read_modify_subreg_p): Likewise.
8806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8807 Alan Hayward <alan.hayward@arm.com>
8808 David Sherwood <david.sherwood@arm.com>
8810 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8811 numbers of elements.
8813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8814 Alan Hayward <alan.hayward@arm.com>
8815 David Sherwood <david.sherwood@arm.com>
8817 * match.pd: Cope with polynomial numbers of vector elements.
8819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8820 Alan Hayward <alan.hayward@arm.com>
8821 David Sherwood <david.sherwood@arm.com>
8823 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8824 in a POINTER_PLUS_EXPR.
8826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8827 Alan Hayward <alan.hayward@arm.com>
8828 David Sherwood <david.sherwood@arm.com>
8830 * omp-simd-clone.c (simd_clone_subparts): New function.
8831 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8832 (ipa_simd_modify_function_body): Likewise.
8834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8835 Alan Hayward <alan.hayward@arm.com>
8836 David Sherwood <david.sherwood@arm.com>
8838 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8839 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8840 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8841 (expand_vector_condition, vector_element): Likewise.
8842 (subparts_gt): New function.
8843 (get_compute_type): Use subparts_gt.
8844 (count_type_subparts): Delete.
8845 (expand_vector_operations_1): Use subparts_gt instead of
8846 count_type_subparts.
8848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8849 Alan Hayward <alan.hayward@arm.com>
8850 David Sherwood <david.sherwood@arm.com>
8852 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8853 (vect_compile_time_alias): ...this new function. Do the calculation
8854 on poly_ints rather than trees.
8855 (vect_prune_runtime_alias_test_list): Update call accordingly.
8857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8858 Alan Hayward <alan.hayward@arm.com>
8859 David Sherwood <david.sherwood@arm.com>
8861 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8863 (vect_schedule_slp_instance): Likewise.
8865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8866 Alan Hayward <alan.hayward@arm.com>
8867 David Sherwood <david.sherwood@arm.com>
8869 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8870 constant and extern definitions for variable-length vectors.
8871 (vect_get_constant_vectors): Note that the number of units
8872 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8879 of units as polynomial. Choose between WIDE and NARROW based
8882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8883 Alan Hayward <alan.hayward@arm.com>
8884 David Sherwood <david.sherwood@arm.com>
8886 * tree-vect-stmts.c (simd_clone_subparts): New function.
8887 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8894 vectors as polynomial. Use build_index_vector for
8897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8898 Alan Hayward <alan.hayward@arm.com>
8899 David Sherwood <david.sherwood@arm.com>
8901 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8902 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8903 for variable-length vectors.
8904 (vectorizable_mask_load_store): Treat the number of units as
8905 polynomial, asserting that it is constant if the condition has
8906 already been enforced.
8907 (vectorizable_store, vectorizable_load): Likewise.
8909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8910 Alan Hayward <alan.hayward@arm.com>
8911 David Sherwood <david.sherwood@arm.com>
8913 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8914 of units as polynomial. Punt if we can't tell at compile time
8915 which vector contains the final result.
8917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8918 Alan Hayward <alan.hayward@arm.com>
8919 David Sherwood <david.sherwood@arm.com>
8921 * tree-vect-loop.c (vectorizable_induction): Treat the number
8922 of units as polynomial. Punt on SLP inductions. Use an integer
8923 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8924 cast of such a series for variable-length floating-point
8927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8928 Alan Hayward <alan.hayward@arm.com>
8929 David Sherwood <david.sherwood@arm.com>
8931 * tree.h (build_index_vector): Declare.
8932 * tree.c (build_index_vector): New function.
8933 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8934 of units as polynomial, forcibly converting it to a constant if
8935 vectorizable_reduction has already enforced the condition.
8936 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8937 to create a {1,2,3,...} vector.
8938 (vectorizable_reduction): Treat the number of units as polynomial.
8939 Choose vectype_in based on the largest scalar element size rather
8940 than the smallest number of units. Enforce the restrictions
8943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8944 Alan Hayward <alan.hayward@arm.com>
8945 David Sherwood <david.sherwood@arm.com>
8947 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8948 number of units as polynomial.
8950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8951 Alan Hayward <alan.hayward@arm.com>
8952 David Sherwood <david.sherwood@arm.com>
8954 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8955 * target.def (autovectorize_vector_sizes): Return the vector sizes
8956 by pointer, using vector_sizes rather than a bitmask.
8957 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8958 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8959 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8961 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8962 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8963 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8964 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8965 * omp-general.c (omp_max_vf): Likewise.
8966 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8967 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8968 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8969 * tree-vect-slp.c (vect_slp_bb): Likewise.
8970 * doc/tm.texi: Regenerate.
8971 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8973 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8974 the vector size as a poly_uint64 rather than an unsigned int.
8975 (current_vector_size): Change from an unsigned int to a poly_uint64.
8976 (get_vectype_for_scalar_type): Update accordingly.
8977 * tree.h (build_truth_vector_type): Take the size and number of
8978 units as a poly_uint64 rather than an unsigned int.
8979 (build_vector_type): Add a temporary overload that takes
8980 the number of units as a poly_uint64 rather than an unsigned int.
8981 * tree.c (make_vector_type): Likewise.
8982 (build_truth_vector_type): Take the number of units as a poly_uint64
8983 rather than an unsigned int.
8985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8986 Alan Hayward <alan.hayward@arm.com>
8987 David Sherwood <david.sherwood@arm.com>
8989 * target.def (get_mask_mode): Take the number of units and length
8990 as poly_uint64s rather than unsigned ints.
8991 * targhooks.h (default_get_mask_mode): Update accordingly.
8992 * targhooks.c (default_get_mask_mode): Likewise.
8993 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8994 * doc/tm.texi: Regenerate.
8996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8997 Alan Hayward <alan.hayward@arm.com>
8998 David Sherwood <david.sherwood@arm.com>
9000 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9001 * omp-general.c (omp_max_vf): Likewise.
9002 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9003 (expand_omp_simd): Handle polynomial safelen.
9004 * omp-low.c (omplow_simd_context): Add a default constructor.
9005 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9006 (lower_rec_simd_input_clauses): Update accordingly.
9007 (lower_rec_input_clauses): Likewise.
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-vectorizer.h (vect_nunits_for_cost): New function.
9014 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9015 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9016 (vect_analyze_slp_cost): Likewise.
9017 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9018 (vect_model_load_cost): Likewise.
9020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9021 Alan Hayward <alan.hayward@arm.com>
9022 David Sherwood <david.sherwood@arm.com>
9024 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9025 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9026 from an unsigned int * to a poly_uint64_pod *.
9027 (calculate_unrolling_factor): New function.
9028 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9031 Alan Hayward <alan.hayward@arm.com>
9032 David Sherwood <david.sherwood@arm.com>
9034 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9035 from an unsigned int to a poly_uint64.
9036 (_loop_vec_info::slp_unrolling_factor): Likewise.
9037 (_loop_vec_info::vectorization_factor): Change from an int
9039 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9040 (vect_get_num_vectors): New function.
9041 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9042 (vect_get_num_copies): Use vect_get_num_vectors.
9043 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9044 to an unsigned int *.
9045 (vect_analyze_data_refs): Change min_vf from an int * to a
9047 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9048 than an unsigned HOST_WIDE_INT.
9049 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9050 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9051 to an unsigned int *.
9052 (vect_analyze_data_ref_dependences): Likewise.
9053 (vect_compute_data_ref_alignment): Handle polynomial vf.
9054 (vect_enhance_data_refs_alignment): Likewise.
9055 (vect_prune_runtime_alias_test_list): Likewise.
9056 (vect_shift_permute_load_chain): Likewise.
9057 (vect_supportable_dr_alignment): Likewise.
9058 (dependence_distance_ge_vf): Take the vectorization factor as a
9059 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9060 (vect_analyze_data_refs): Change min_vf from an int * to a
9062 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9063 vfm1 as a poly_uint64 rather than an int. Make the same change
9064 for the returned bound_scalar.
9065 (vect_gen_vector_loop_niters): Handle polynomial vf.
9066 (vect_do_peeling): Likewise. Update call to
9067 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9068 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9070 * tree-vect-loop.c (vect_determine_vectorization_factor)
9071 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9072 (vect_get_known_peeling_cost): Likewise.
9073 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9074 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9075 (vect_transform_loop): Likewise. Use the lowest possible VF when
9076 updating the upper bounds of the loop.
9077 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9079 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9080 polynomial unroll factors.
9081 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9082 (vect_make_slp_decision): Likewise.
9083 (vect_supported_load_permutation_p): Likewise, and polynomial
9085 (vect_analyze_slp_cost): Handle polynomial vf.
9086 (vect_slp_analyze_node_operations): Likewise.
9087 (vect_slp_analyze_bb_1): Likewise.
9088 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9089 than an unsigned HOST_WIDE_INT.
9090 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9091 (vectorizable_load): Handle polynomial vf.
9092 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9094 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9097 Alan Hayward <alan.hayward@arm.com>
9098 David Sherwood <david.sherwood@arm.com>
9100 * match.pd: Handle bit operations involving three constants
9101 and try to fold one pair.
9103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9105 * tree-vect-loop-manip.c: Include gimple-fold.h.
9106 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9107 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9108 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9109 Add a path that uses a step of VF instead of 1, but disable it
9111 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9112 and niters_no_overflow parameters. Update calls to
9113 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9114 Create a new SSA name if the latter choses to use a ste other
9115 than zero, and return it via niters_vector_mult_vf_var.
9116 * tree-vect-loop.c (vect_transform_loop): Update calls to
9117 vect_do_peeling, vect_gen_vector_loop_niters and
9118 slpeel_make_loop_iterate_ntimes.
9119 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9120 (vect_gen_vector_loop_niters): Update declarations after above changes.
9122 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9124 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9125 128-bit round to integer instructions.
9126 (ceil<mode>2): Likewise.
9127 (btrunc<mode>2): Likewise.
9128 (round<mode>2): Likewise.
9130 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9132 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9133 unaligned VSX load/store on P8/P9.
9134 (expand_block_clear): Allow the use of unaligned VSX
9135 load/store on P8/P9.
9137 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9139 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9141 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9142 swap associated with both a load and a store.
9144 2018-01-02 Andrew Waterman <andrew@sifive.com>
9146 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9147 * config/riscv/riscv.md (clear_cache): Use it.
9149 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9151 * web.c: Remove out-of-date comment.
9153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9155 * expr.c (fixup_args_size_notes): Check that any existing
9156 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9157 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9158 (emit_single_push_insn): ...here.
9160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9162 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9163 (const_vector_encoded_nelts): New function.
9164 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9165 (const_vector_int_elt, const_vector_elt): Declare.
9166 * emit-rtl.c (const_vector_int_elt_1): New function.
9167 (const_vector_elt): Likewise.
9168 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9169 of CONST_VECTOR_ELT.
9171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9173 * expr.c: Include rtx-vector-builder.h.
9174 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9175 directly on the tree encoding.
9176 (const_vector_from_tree): Likewise.
9177 * optabs.c: Include rtx-vector-builder.h.
9178 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9179 sequence of "u" values.
9180 * vec-perm-indices.c: Include rtx-vector-builder.h.
9181 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9182 directly on the vec_perm_indices encoding.
9184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9186 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9187 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9188 * rtx-vector-builder.h: New file.
9189 * rtx-vector-builder.c: Likewise.
9190 * rtl.h (rtx_def::u2): Add a const_vector field.
9191 (CONST_VECTOR_NPATTERNS): New macro.
9192 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9193 (CONST_VECTOR_DUPLICATE_P): Likewise.
9194 (CONST_VECTOR_STEPPED_P): Likewise.
9195 (CONST_VECTOR_ENCODED_ELT): Likewise.
9196 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9197 (unwrap_const_vec_duplicate): Likewise.
9198 (const_vec_series_p): Check for a non-duplicated vector encoding.
9199 Say that the function only returns true for integer vectors.
9200 * emit-rtl.c: Include rtx-vector-builder.h.
9201 (gen_const_vec_duplicate_1): Delete.
9202 (gen_const_vector): Call gen_const_vec_duplicate instead of
9203 gen_const_vec_duplicate_1.
9204 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9205 (gen_const_vec_duplicate): Use rtx_vector_builder.
9206 (gen_const_vec_series): Likewise.
9207 (gen_rtx_CONST_VECTOR): Likewise.
9208 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9209 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9210 Build a new vector rather than modifying a CONST_VECTOR in-place.
9211 (handle_special_swappables): Update call accordingly.
9212 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9213 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9214 Build a new vector rather than modifying a CONST_VECTOR in-place.
9215 (handle_special_swappables): Update call accordingly.
9217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9219 * simplify-rtx.c (simplify_const_binary_operation): Use
9220 CONST_VECTOR_ELT instead of XVECEXP.
9222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9224 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9225 the selector elements to be different from the data elements
9226 if the selector is a VECTOR_CST.
9227 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9228 ssizetype for the selector.
9230 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9232 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9233 before testing each element individually.
9234 * tree-vect-generic.c (lower_vec_perm): Likewise.
9236 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9238 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9239 * selftest-run-tests.c (selftest::run_tests): Call it.
9240 * vector-builder.h (vector_builder::operator ==): New function.
9241 (vector_builder::operator !=): Likewise.
9242 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9243 (vec_perm_indices::all_from_input_p): New function.
9244 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9245 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9246 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9247 instead of reading the VECTOR_CST directly. Detect whether both
9248 vector inputs are the same before constructing the vec_perm_indices,
9249 and update the number of inputs argument accordingly. Use the
9250 utility functions added above. Only construct sel2 if we need to.
9252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9254 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9255 the broadcast of the low byte.
9256 (expand_mult_highpart): Use an explicit encoding for the permutes.
9257 * optabs-query.c (can_mult_highpart_p): Likewise.
9258 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9259 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9260 (vectorizable_bswap): Likewise.
9261 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9262 explicit encoding for the power-of-2 permutes.
9263 (vect_permute_store_chain): Likewise.
9264 (vect_grouped_load_supported): Likewise.
9265 (vect_permute_load_chain): Likewise.
9267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9269 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9270 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9271 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9272 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9273 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9274 (vect_gen_perm_mask_any): Likewise.
9276 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9278 * int-vector-builder.h: New file.
9279 * vec-perm-indices.h: Include int-vector-builder.h.
9280 (vec_perm_indices): Redefine as an int_vector_builder.
9281 (auto_vec_perm_indices): Delete.
9282 (vec_perm_builder): Redefine as a stand-alone class.
9283 (vec_perm_indices::vec_perm_indices): New function.
9284 (vec_perm_indices::clamp): Likewise.
9285 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9286 (vec_perm_indices::new_vector): New function.
9287 (vec_perm_indices::new_expanded_vector): Update for new
9288 vec_perm_indices class.
9289 (vec_perm_indices::rotate_inputs): New function.
9290 (vec_perm_indices::all_in_range_p): Operate directly on the
9291 encoded form, without computing elided elements.
9292 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9293 encoding. Update for new vec_perm_indices class.
9294 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9295 the given vec_perm_builder.
9296 (expand_vec_perm_var): Update vec_perm_builder constructor.
9297 (expand_mult_highpart): Use vec_perm_builder instead of
9298 auto_vec_perm_indices.
9299 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9300 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9301 or double series encoding as appropriate.
9302 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9303 vec_perm_indices instead of auto_vec_perm_indices.
9304 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9305 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9306 (vect_permute_store_chain): Likewise.
9307 (vect_grouped_load_supported): Likewise.
9308 (vect_permute_load_chain): Likewise.
9309 (vect_shift_permute_load_chain): Likewise.
9310 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9311 (vect_transform_slp_perm_load): Likewise.
9312 (vect_schedule_slp_instance): Likewise.
9313 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9314 (vectorizable_mask_load_store): Likewise.
9315 (vectorizable_bswap): Likewise.
9316 (vectorizable_store): Likewise.
9317 (vectorizable_load): Likewise.
9318 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9319 vec_perm_indices instead of auto_vec_perm_indices. Use
9320 tree_to_vec_perm_builder to read the vector from a tree.
9321 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9322 vec_perm_builder instead of a vec_perm_indices.
9323 (have_whole_vector_shift): Use vec_perm_builder and
9324 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9325 truncation to calc_vec_perm_mask_for_shift.
9326 (vect_create_epilog_for_reduction): Likewise.
9327 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9328 from auto_vec_perm_indices to vec_perm_indices.
9329 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9330 instead of changing individual elements.
9331 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9332 the vector in d.perm.
9333 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9334 from auto_vec_perm_indices to vec_perm_indices.
9335 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9336 instead of changing individual elements.
9337 (arm_vectorize_vec_perm_const): Use new_vector to install
9338 the vector in d.perm.
9339 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9340 Update vec_perm_builder constructor.
9341 (rs6000_expand_interleave): Likewise.
9342 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9343 (rs6000_expand_interleave): Likewise.
9345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9347 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9348 to qimode could truncate the indices.
9349 * optabs.c (expand_vec_perm_var): Likewise.
9351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9353 * Makefile.in (OBJS): Add vec-perm-indices.o.
9354 * vec-perm-indices.h: New file.
9355 * vec-perm-indices.c: Likewise.
9356 * target.h (vec_perm_indices): Replace with a forward class
9358 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9359 * optabs.h: Include vec-perm-indices.h.
9360 (expand_vec_perm): Delete.
9361 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9362 (expand_vec_perm_const): Declare.
9363 * target.def (vec_perm_const_ok): Replace with...
9364 (vec_perm_const): ...this new hook.
9365 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9366 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9367 * doc/tm.texi: Regenerate.
9368 * optabs.def (vec_perm_const): Delete.
9369 * doc/md.texi (vec_perm_const): Likewise.
9370 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9371 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9372 expand_vec_perm for constant permutation vectors. Assert that
9373 the mode of variable permutation vectors is the integer equivalent
9374 of the mode that is being permuted.
9375 * optabs-query.h (selector_fits_mode_p): Declare.
9376 * optabs-query.c: Include vec-perm-indices.h.
9377 (selector_fits_mode_p): New function.
9378 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9379 is defined, instead of checking whether the vec_perm_const_optab
9380 exists. Use targetm.vectorize.vec_perm_const instead of
9381 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9382 fit in the vector mode before using a variable permute.
9383 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9384 vec_perm_indices instead of an rtx.
9385 (expand_vec_perm): Replace with...
9386 (expand_vec_perm_const): ...this new function. Take the selector
9387 as a vec_perm_indices rather than an rtx. Also take the mode of
9388 the selector. Update call to shift_amt_for_vec_perm_mask.
9389 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9390 Use vec_perm_indices::new_expanded_vector to expand the original
9391 selector into bytes. Check whether the indices fit in the vector
9392 mode before using a variable permute.
9393 (expand_vec_perm_var): Make global.
9394 (expand_mult_highpart): Use expand_vec_perm_const.
9395 * fold-const.c: Includes vec-perm-indices.h.
9396 * tree-ssa-forwprop.c: Likewise.
9397 * tree-vect-data-refs.c: Likewise.
9398 * tree-vect-generic.c: Likewise.
9399 * tree-vect-loop.c: Likewise.
9400 * tree-vect-slp.c: Likewise.
9401 * tree-vect-stmts.c: Likewise.
9402 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9404 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9405 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9406 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9407 (aarch64_vectorize_vec_perm_const): ...this new function.
9408 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9409 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9410 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9411 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9412 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9413 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9414 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9416 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9417 check for NEON modes.
9418 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9419 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9420 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9421 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9423 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9424 the old VEC_PERM_CONST conditions.
9425 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9426 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9427 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9428 (ia64_vectorize_vec_perm_const_ok): Merge into...
9429 (ia64_vectorize_vec_perm_const): ...this new function.
9430 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9431 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9432 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9433 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9434 * config/mips/mips.c (mips_expand_vec_perm_const)
9435 (mips_vectorize_vec_perm_const_ok): Merge into...
9436 (mips_vectorize_vec_perm_const): ...this new function.
9437 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9438 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9439 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9440 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9441 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9442 (rs6000_expand_vec_perm_const): Delete.
9443 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9445 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9446 (altivec_expand_vec_perm_const_le): Take each operand individually.
9447 Operate on constant selectors rather than rtxes.
9448 (altivec_expand_vec_perm_const): Likewise. Update call to
9449 altivec_expand_vec_perm_const_le.
9450 (rs6000_expand_vec_perm_const): Delete.
9451 (rs6000_vectorize_vec_perm_const_ok): Delete.
9452 (rs6000_vectorize_vec_perm_const): New function.
9453 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9454 an element count and rtx array.
9455 (rs6000_expand_extract_even): Update call accordingly.
9456 (rs6000_expand_interleave): Likewise.
9457 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9458 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9459 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9460 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9461 (rs6000_expand_vec_perm_const): Delete.
9462 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9463 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9464 (altivec_expand_vec_perm_const_le): Take each operand individually.
9465 Operate on constant selectors rather than rtxes.
9466 (altivec_expand_vec_perm_const): Likewise. Update call to
9467 altivec_expand_vec_perm_const_le.
9468 (rs6000_expand_vec_perm_const): Delete.
9469 (rs6000_vectorize_vec_perm_const_ok): Delete.
9470 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9471 reference to the SPE evmerge intructions.
9472 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9473 an element count and rtx array.
9474 (rs6000_expand_extract_even): Update call accordingly.
9475 (rs6000_expand_interleave): Likewise.
9476 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9477 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9479 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9481 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9483 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9484 vector mode and that that mode matches the mode of the data
9486 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9487 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9488 directly using expand_vec_perm_1 when forcing selectors into
9490 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9494 * optabs-query.h (can_vec_perm_p): Delete.
9495 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9496 * optabs-query.c (can_vec_perm_p): Split into...
9497 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9498 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9499 particular selector is valid.
9500 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9501 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9502 (vect_grouped_load_supported): Likewise.
9503 (vect_shift_permute_load_chain): Likewise.
9504 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9505 (vect_transform_slp_perm_load): Likewise.
9506 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9507 (vectorizable_bswap): Likewise.
9508 (vect_gen_perm_mask_checked): Likewise.
9509 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9510 implementations of variable permutation vectors into account
9511 when deciding which selector to use.
9512 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9513 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9514 with a false third argument.
9515 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9516 to test whether the constant selector is valid and can_vec_perm_var_p
9517 to test whether a variable selector is valid.
9519 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9521 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9522 * optabs-query.c (can_vec_perm_p): Likewise.
9523 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9524 instead of vec_perm_indices.
9525 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9526 (vect_gen_perm_mask_checked): Likewise,
9527 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9528 (vect_gen_perm_mask_checked): Likewise,
9530 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9532 * optabs-query.h (qimode_for_vec_perm): Declare.
9533 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9534 (qimode_for_vec_perm): ...this new function.
9535 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9537 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9539 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9540 does not have a conditional at the top.
9542 2018-01-02 Richard Biener <rguenther@suse.de>
9544 * ipa-inline.c (big_speedup_p): Fix expression.
9546 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9549 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9552 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9556 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9557 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9558 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9559 cond_taken_branch_cost 3->4.
9561 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9563 PR tree-optimization/83581
9564 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9565 TODO_cleanup_cfg if any changes have been made.
9568 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9569 convert_modes if target mode has the right side, but different mode
9573 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9574 last argument when extracting from CONCAT. If either from_real or
9575 from_imag is NULL, use expansion through memory. If result is not
9576 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9577 the parts directly to inner mode, if even that fails, use expansion
9581 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9582 check for bswap in mode rather than HImode and use that in expand_unop
9585 Copyright (C) 2018 Free Software Foundation, Inc.
9587 Copying and distribu