1 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2 Kito Cheng <kito.cheng@gmail.com>
4 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5 nds32_cond_code_str, output_cond_branch,
6 output_cond_branch_compare_zero, nds32_expand_cbranch,
7 nds32_expand_cstore, nds32_expand_movcc,
8 nds32_output_cbranchsi4_equality_zero,
9 nds32_output_cbranchsi4_equality_reg,
10 nds32_output_cbranchsi4_equality_reg_or_const_int,
11 nds32_output_cbranchsi4_greater_less_zero: New functions.
12 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13 nds32_expand_cstore, nds32_expand_movcc,
14 nds32_output_cbranchsi4_equality_zero,
15 nds32_output_cbranchsi4_equality_reg,
16 nds32_output_cbranchsi4_equality_reg_or_const_int,
17 nds32_output_cbranchsi4_greater_less_zero): Declare.
18 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19 nds32_rimm11s_operand): New predicates.
20 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
21 * config/nds32/nds32.md: Rewrite all the branch and conditional move
24 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
27 * config/nds32/nds32.md: Ditto.
28 * config/nds32/pipelines.md: Ditto.
30 2018-04-04 Richard Biener <rguenther@suse.de>
32 PR tree-optimization/85168
33 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
34 propagating abnormals.
36 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
38 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
40 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
41 Kito Cheng <kito.cheng@gmail.com>
43 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
44 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
45 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
46 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
47 * config/nds32/nds32.md (sibcall_internal): New.
48 (sibcall_register): Remove.
49 (sibcall_immediate): Remove.
50 (sibcall_value_internal): New.
51 (sibcall_value_register): Remove.
52 (sibcall_value_immediate): Remove.
53 * config/nds32/predicates.md (nds32_general_register_operand): New.
54 (nds32_call_address_operand): New.
56 2018-04-03 Jakub Jelinek <jakub@redhat.com>
58 PR rtl-optimization/85167
59 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
60 bb_defs if *split_p, instead preinitialize it to NULL.
62 PR tree-optimization/85156
63 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
64 evaluating the argument multiple times.
66 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
68 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
70 (_mm_cvtpd_ps): Likewise.
71 (_mm_cvttpd_epi32): Likewise.
72 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
73 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
74 vector, pixel, and bool following altivec.h include.
76 2018-04-03 Martin Sebor <msebor@redhat.com>
78 * doc/extend.texi (Common Function Attributes): Clarify.
79 (const attribute): Likewise.
80 (pure attribute): Likewise.
82 2018-04-03 Jakub Jelinek <jakub@redhat.com>
85 * config/i386/i386.c (ix86_expand_vector_set): Use
86 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
88 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
90 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
91 instructions when changing rounding bits to preserve precision bits
92 in the x87 control word.
94 2018-04-03 Martin Liska <mliska@suse.cz>
96 PR tree-optimization/82491
97 * rtl.h (strip_offset_and_add): Replace += suboffset with
98 poly_uint64 () + suboffset.
100 2018-03-29 Martin Liska <mliska@suse.cz>
101 Martin Jambor <mjambor@suse.cz>
104 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
105 param_type is not an integral or pointer type.
107 2018-04-03 Richard Biener <rguenther@suse.de>
109 * sese.h (recompute_all_dominators): Remove.
111 2018-04-02 Martin Sebor <msebor@redhat.com>
113 * doc/invoke.texi (-Wrestrict): Fix typos.
115 2018-04-02 Jim Wilson <jimw@sifive.com>
117 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
118 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
119 (<optab>di3, <optab>si3_extend): Likewise.
120 (<optab>si3_mask, <optab>si3_mask_1): New.
121 (<optab>di3_mask, <optab>di3_mask_1): New.
122 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
123 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
124 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
126 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
128 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
131 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
133 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
134 (nds32_canonicalize_comparison): New function.
136 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
137 Kito Cheng <kito.cheng@gmail.com>
138 Kuan-Lin Chen <kuanlinchentw@gmail.com>
140 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
141 * config/nds32/constants.md (unspec_volatile_element): Add
142 UNSPEC_VOLATILE_RELAX_GROUP.
143 * config/nds32/nds32-relax-opt.c: New file.
144 * config/nds32/nds32-predicates.c
145 (nds32_symbol_load_store_p): New function.
146 * config/nds32/nds32-protos.h
147 (nds32_symbol_load_store_p): Declare function.
148 (make_pass_nds32_relax_opt): Declare new rtl pass function.
149 * config/nds32/nds32.c
150 (nds32_register_pass): New function to register pass.
151 (nds32_register_passes): New function to register passes.
152 * config/nds32/nds32.md (relax_group): New pattern.
153 * config/nds32/nds32.opt (mrelax-hint): New option.
154 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
156 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
158 * config/nds32/t-nds32: Modify files dependency.
160 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
162 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
163 (PROFILE_HOOK): Define its implementation.
165 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
167 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
168 type and 32-bit size.
170 2018-04-01 Jakub Jelinek <jakub@redhat.com>
173 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
174 (V_128_256): New mode iterator.
175 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
176 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
177 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
179 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
180 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
182 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
185 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
186 NaN inputs correctly.
188 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
191 * config/rs6000/vsx.md (??r): New mode attribute.
192 (*vsx_mov<mode>_64bit): Use it.
193 (*vsx_mov<mode>_32bit): Likewise.
195 2018-03-30 Martin Sebor <msebor@redhat.com>
197 PR tree-optimization/84818
198 * builtins.c (check_access): Use warning_n.
200 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
203 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
205 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
208 2018-03-30 Julia Koval <julia.koval@intel.com>
211 * x86-tune.def (movx, partial_reg_dependency): Enable for
214 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
217 * lra-constraints.c (process_alt_operands): Move setting
218 this_alternative_matches below.
220 2018-03-29 Martin Liska <mliska@suse.cz>
223 * doc/invoke.texi: Document how LTO works with debug info.
224 Describe auto-load support of binutils. Mention 'x86-64'
225 as valid option value of -march option.
227 2018-03-29 Jakub Jelinek <jakub@redhat.com>
229 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
232 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
233 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
234 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
237 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
240 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
241 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
242 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
243 for RS6000_BTM_POWERPC64.
244 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
245 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
246 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
251 2018-03-28 Carl Love <cel@us.ibm.com>
254 2017-09-27 Carl Love <cel@us.ibm.com>
256 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
257 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
258 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
261 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
263 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
264 instead of __vector bool.
265 (_mm_max_pu8): Likewise.
266 (_mm_min_pi16): Likewise.
268 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
271 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
275 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
277 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
278 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
279 (div_extend): Likewise.
280 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
282 (__builtin_divweuo): Likewise.
283 (__builtin_divdeo): Likewise.
284 (__builtin_divdeuo): Likewise.
286 2018-03-28 Jakub Jelinek <jakub@redhat.com>
289 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
290 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
292 PR tree-optimization/82004
293 * gimple-match-head.c (optimize_pow_to_exp): New function.
294 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
295 Don't fold to exp if optimize_pow_to_exp is false.
297 2018-03-28 Martin Liska <mliska@suse.cz>
300 * calls.c (initialize_argument_information): Fix trailing space.
301 * common.opt: Fix typo and provide better explanation for
302 -fsanitize-coverage option.
303 * config/i386/i386.opt: Fix typo.
305 2018-03-28 Jakub Jelinek <jakub@redhat.com>
306 Martin Liska <mliska@suse.cz>
309 * gimplify.c (asan_poison_variable): Don't do the check for
310 gimplify_omp_ctxp here.
311 (gimplify_decl_expr): Do it here.
312 (gimplify_target_expr): Likewise.
314 2018-03-28 Martin Liska <mliska@suse.cz>
317 * config/i386/i386.c (ix86_function_arg_advance): Do not call
318 chkp_type_bounds_count if MPX is not enabled.
320 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
322 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
324 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
327 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
328 function to create the function decl for complex long double
329 multiply and divide for -mabi=ieeelongdouble.
330 (init_float128_ieee): Call it.
332 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
335 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
336 -fcf-protection=branch -mibt.
337 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
339 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
342 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
344 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
347 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
348 extern array declarations.
350 2018-03-27 Richard Biener <rguenther@suse.de>
353 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
354 explicit single_use checks.
356 2018-03-27 Richard Biener <rguenther@suse.de>
358 PR tree-optimization/85082
359 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
362 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
364 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
365 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
366 Turn on fasynchronous-unwind-tables and funwind-tables.
368 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
371 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
372 (*bmi_blsr_<mode>_ccz): Ditto.
374 2018-03-26 Tom de Vries <tom@codesourcery.com>
376 PR tree-optimization/85063
377 * omp-general.c (offloading_function_p): New function. Factor out
379 * omp-offload.c (pass_omp_target_link::gate): ... here.
380 * omp-general.h (offloading_function_p): Declare.
381 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
382 with attribute omp declare target for offloading functions.
384 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
386 PR tree-optimization/84005
387 * tree-data-ref.h (get_base_for_alignment): Declare.
388 * tree-data-ref.c (get_base_for_alignment_1): New function.
389 (get_base_for_alignment): Likewise.
390 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
391 get_base_for_alignment to find a suitable base object, instead
392 of always using drb->base_address.
394 2018-03-23 Jakub Jelinek <jakub@redhat.com>
397 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
398 known size by default.
400 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
403 * lra-constraints.c (process_alt_operands): Don't match BLKmode
404 and non BLKmode operands.
406 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
409 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
412 2018-03-23 Richard Biener <rguenther@suse.de>
415 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
416 we are going to emit early debug for LTO.
418 2018-03-23 Jakub Jelinek <jakub@redhat.com>
421 * function.c (match_asm_constraints_1): Don't optimize if input
422 doesn't satisfy general_operand predicate for output's mode.
425 * alias.c (write_dependence_p): Don't require for x_canonicalized
426 non-VOIDmode if x has VOIDmode.
429 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
430 just don't try to optimize it rather than assert it never happens.
432 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
434 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
435 macro expansions for definition of ST_INTERNAL_<mode> and
436 LD_INTERNAL_<mode> builtins.
437 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
439 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
441 (altivec_expand_st_builtin): Likewise.
442 (altivec_expand_builtin): Remove calls to deleted functions.
443 (rs6000_address_for_altivec): Delete this function.
444 * config/rs6000/vector.md: Remove expands for
445 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
447 2018-03-22 Sudakshina Das <sudi.das@arm.com>
450 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
451 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
452 re-computing once computed.
453 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
454 (arm_init_machine_status): Initialize
455 machine->static_chain_stack_bytes.
457 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
460 * doc/extend.texi: Add four new prototypes for vec_ld.
461 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
462 definitions for more logical presentation.
463 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
464 entries for V1TI variants of __builtin_altivec_ld builtin.
465 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
466 handling of V1TI variant of LVX icode pattern.
467 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
468 (rs6000_gimple_fold_builtin): Likewise.
469 (altivec_init_builtins): Add code to define
470 __builtin_altivec_lvx_v1ti function.
472 2018-03-22 Jakub Jelinek <jakub@redhat.com>
475 * function.c (match_asm_constraints_1): Don't do the optimization
476 if input isn't a REG, SUBREG, MEM or constant.
478 2018-03-22 Tom de Vries <tom@codesourcery.com>
480 PR tree-optimization/84956
481 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
482 bb_has_abnormal_pred.
484 2018-03-22 Jakub Jelinek <jakub@redhat.com>
487 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
488 DECL_INITIAL (decl) to decl at the end.
489 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
492 2018-03-21 Joseph Myers <joseph@codesourcery.com>
494 * doc/extend.texi (__builtin_tgmath): Document when complex
495 integer types are treated as _Complex _Float64.
497 2018-03-21 Tom de Vries <tom@codesourcery.com>
499 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
501 2018-03-21 Jakub Jelinek <jakub@redhat.com>
503 PR tree-optimization/84960
504 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
505 if it is ENTRY block, move them into single succ of ENTRY in that case.
507 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
509 PR tree-optimization/84811
510 * poly-int.h (poly_span_traits): Remove the T3 parameter and
511 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
512 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
513 (known_subrange_p): Update accordingly. Cast each value involved
514 in the size comparison, rather than casting the result of the
517 2018-03-21 Jakub Jelinek <jakub@redhat.com>
519 PR tree-optimization/84982
520 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
521 by flipping the least significant bit rather than all bits from
522 bitpos to bitpos + bitsize - 1.
524 2018-03-21 Nathan Sidwell <nathan@acm.org>
526 * doc/extend.texi (Deprecated Features): Remove mention of
527 long-deleted deprecations.
529 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
532 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
533 * configure: Regenerate.
535 2018-03-21 Tom de Vries <tom@codesourcery.com>
537 PR tree-optimization/83126
538 * tree-parloops.c (num_phis): New function.
539 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
541 2018-03-21 Nathan Sidwell <nathan@acm.org>
543 * doc/extend.texi (Deprecated Features): Update deprecated flags,
544 mention anon-struct/union members and trailing attributes.
546 2018-03-21 Bin Cheng <bin.cheng@arm.com>
548 PR tree-optimization/84969
549 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
550 builtin memset partitions if they set different rhs values.
552 2018-03-21 Jakub Jelinek <jakub@redhat.com>
554 PR rtl-optimization/84989
555 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
556 VEC_DUPLICATE with scalar result mode.
558 2018-03-21 Martin Liska <mliska@suse.cz>
561 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
562 not intended return statement.
564 2018-03-21 Martin Liska <mliska@suse.cz>
567 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
568 (chkp_find_bound_slots_1): Limit number of iterations.
570 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
573 * Minor grammar fixes for x86 options.
575 2018-03-20 Jakub Jelinek <jakub@redhat.com>
578 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
579 holding REG_CFA_RESTORE notes, instead turn them into a USE.
581 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
584 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
585 (altivec_lvx_<mode>_1op): Likewise.
586 (altivec_stvx_<mode>_2op): Likewise.
587 (altivec_stvx_<mode>_1op): Likewise.
588 (altivec_lvx_<VM2:mode>): New define_expand.
589 (altivec_stvx_<VM2:mode>): Likewise.
590 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
591 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
592 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
593 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
594 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
595 (rs6000_gen_lvx): Likewise.
596 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
597 (altivec_expand_stv_builtin): Likewise.
598 (altivec_expand_builtin): Likewise.
599 * config/rs6000/vector.md: Likewise.
601 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
604 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
607 2018-03-20 Richard Biener <rguenther@suse.de>
610 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
611 sign-conversions as zero, fall back to standard scalar_stmt
614 2018-03-20 Martin Liska <mliska@suse.cz>
617 * predict.c (rebuild_frequencies): Handle case when we have
618 PROFILE_ABSENT, but flag_guess_branch_prob is false.
620 2018-03-20 Jakub Jelinek <jakub@redhat.com>
623 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
624 flag_section_anchors.
625 * varasm.c (use_blocks_for_decl_p): Remove hack for
629 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
631 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
632 be created, use lowpart_subreg of operands[0] rather than operands[0]
634 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
635 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
636 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
637 and n constraint instead of aarch64_shift_imm_di and Usd.
638 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
639 (*aarch64_<optab>_reg_minus<mode>3): ... this.
641 2018-03-20 Sudakshina Das <sudi.das@arm.com>
644 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
645 to favor GPR over NEON registers.
646 (<shift>di3_neon): Likewise.
648 2018-03-20 Tom de Vries <tom@codesourcery.com>
651 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
652 (nvptx_process_pars): Emit bar.sync asap and alap.
654 2018-03-20 Tom de Vries <tom@codesourcery.com>
657 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
658 seen_label if seen_label is already set.
660 2018-03-20 Jakub Jelinek <jakub@redhat.com>
663 * config/i386/i386.c (fold_builtin_cpu): For features above 31
664 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
665 Use 1U instead of 1. Formatting fixes.
668 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
669 instead of TREE_TYPE (s1) for the return value.
671 2018-03-19 Jakub Jelinek <jakub@redhat.com>
673 PR tree-optimization/84946
674 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
675 bitsize + bitsize in poly_uint64 rather than poly_int64.
678 * dwarf2asm.c: Include fold-const.c.
679 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
680 of decl rather than decl itself.
682 PR rtl-optimization/84643
683 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
685 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
688 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
689 calling assemble_variable.
691 2018-03-19 Sudakshina Das <sudi.das@arm.com>
694 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
695 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
697 2018-03-19 Jim Wilson <jimw@sifive.com>
700 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
701 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
702 (riscv_first_stack_step): Likewise.
703 (riscv_option_override): Use STACK_BOUNDARY instead of
705 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
707 (BIGGEST_ALIGNMENT): Set to 128.
708 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
709 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
712 2018-03-19 Richard Biener <rguenther@suse.de>
714 PR tree-optimization/84933
715 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
716 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
718 2018-03-19 Richard Biener <rguenther@suse.de>
720 PR tree-optimization/84859
721 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
722 (cond_if_else_store_replacement): Perform sinking operation on
723 single-store BBs regardless of MAX_STORES_TO_SINK setting.
724 Generalize what a BB with a single eligible store is.
726 2018-03-19 Richard Biener <rguenther@suse.de>
728 PR tree-optimization/84929
729 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
730 chrec_is_positive against non-chrec arg.
732 2018-03-19 Tamar Christina <tamar.christina@arm.com>
735 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
737 2018-03-18 Martin Liska <mliska@suse.cz>
739 PR rtl-optimization/84635
740 * regrename.c (build_def_use): Use matches_mode only when
743 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
745 PR tree-optimization/84913
746 * tree-vect-loop.c (vectorizable_reduction): Don't try to
747 vectorize chains of COND_EXPRs.
749 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
751 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
753 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
755 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
757 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
759 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
761 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
762 Kito Cheng <kito.cheng@gmail.com>
764 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
765 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
766 (nds32_adjust_reg_alloc_order): New function.
767 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
769 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
771 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
772 nds32_print_operand, nds32_print_operand_address): Use
773 HOST_WIDE_INT_PRINT_DEC instead.
775 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
777 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
779 2018-03-17 Jakub Jelinek <jakub@redhat.com>
782 * config/i386/i386.c (initial_ix86_tune_features,
783 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
785 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
786 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
787 rather than 1u << ix86_tune. Formatting fix.
788 (ix86_option_override_internal): Change ix86_arch_mask from
789 unsigned int to unsigned HOST_WIDE_INT, initialize to
790 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
791 (ix86_function_specific_restore): Likewise.
793 2018-03-16 Jakub Jelinek <jakub@redhat.com>
796 * postreload.c (reload_combine_recognize_pattern): Perform
797 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
798 truncate_int_for_mode the result for the destination's mode.
801 * hsa-gen.c (mem_type_for_type): Fix comment typo.
802 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
804 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
807 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
810 * lra-assigns.c (lra_split_hard_reg_for): Don't use
811 regno_allocno_class_array and sorted_pseudos.
812 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
813 insns where regno is used.
815 2018-03-16 Martin Liska <mliska@suse.cz>
818 * multiple_target.c (create_dispatcher_calls): Redirect
819 reference in the symbol table.
821 2018-03-16 Martin Liska <mliska@suse.cz>
824 * multiple_target.c (create_dispatcher_calls): Redirect also
827 2018-03-16 Jakub Jelinek <jakub@redhat.com>
831 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
832 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
833 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
835 2018-03-16 Julia Koval <julia.koval@intel.com>
837 * doc/invoke.texi (Skylake Server): Add CLWB.
838 Cannonlake): Remove CLWB.
840 2018-03-16 Jakub Jelinek <jakub@redhat.com>
842 PR tree-optimization/84841
843 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
845 (FLOAT_ONE_CONST_TYPE): Define.
846 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
847 (sort_by_operand_rank): Put entries with higher constant_type last
848 rather than first to match comments.
850 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
852 * config/nios2/nios2.md (movsi_internal): Fix thinko in
855 2018-03-15 Jakub Jelinek <jakub@redhat.com>
858 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
859 check and use address of target always.
861 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
864 * config/i386/i386.c (indirect_thunk_needed): Update comments.
865 (indirect_thunk_bnd_needed): Likewise.
866 (indirect_thunks_used): Likewise.
867 (indirect_thunks_bnd_used): Likewise.
868 (indirect_return_needed): New.
869 (indirect_return_bnd_needed): Likewise.
870 (output_indirect_thunk_function): Add a bool argument for
872 (output_indirect_thunk_function): Don't generate alias for
873 function return thunk.
874 (ix86_code_end): Call output_indirect_thunk_function to generate
875 function return thunks.
876 (ix86_output_function_return): Set indirect_return_bnd_needed
877 and indirect_return_needed instead of indirect_thunk_bnd_needed
878 and indirect_thunk_needed.
880 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
882 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
883 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
884 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
886 2018-03-15 David Malcolm <dmalcolm@redhat.com>
887 Paul Hua <paul.hua.gm@gmail.com>
890 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
892 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
894 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
895 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
898 2018-03-15 Tamar Christina <tamar.christina@arm.com>
901 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
902 instead of GET_MODE_SIZE when comparing Units.
904 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
907 * varasm.c (hash_section): Return an unchangeble hash value
908 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
909 Return !aarch64_can_use_per_function_literal_pools_p ().
911 2018-03-15 Jakub Jelinek <jakub@redhat.com>
914 * optabs.c (emit_conditional_move): Pass address of cmode's copy
915 rather than address of cmode as last argument to prepare_cmp_insn.
917 2018-03-15 Julia Koval <julia.koval@intel.com>
919 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
920 F_AVX512VNNI, F_AVX512BITALG): New.
922 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
925 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
926 insn for floating-point loads and stores.
928 2018-03-14 Carl Love <cel@us.ibm.com>
930 * config/rs6000/rs6000-c.c: Add macro definitions for
931 ALTIVEC_BUILTIN_VEC_PERMXOR.
932 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
933 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
934 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
935 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
937 * config/doc/extend.texi: Add prototypes for vec_permxor.
939 2018-03-14 David Malcolm <dmalcolm@redhat.com>
942 * diagnostic-show-locus.c (class layout_point): Convert m_line
943 from int to linenum_type.
944 (line_span::comparator): Use linenum "compare" function when
945 comparing line numbers.
946 (test_line_span): New function.
947 (layout_range::contains_point): Convert param "row" from int to
949 (layout_range::intersects_line_p): Likewise.
950 (layout::will_show_line_p): Likewise.
951 (layout::print_source_line): Likewise.
952 (layout::should_print_annotation_line_p): Likewise.
953 (layout::print_annotation_line): Likewise.
954 (layout::print_leading_fixits): Likewise.
955 (layout::annotation_line_showed_range_p): Likewise.
956 (struct line_corrections): Likewise for field m_row.
957 (line_corrections::line_corrections): Likewise for param "row".
958 (layout::print_trailing_fixits): Likewise.
959 (layout::get_state_at_point): Likewise.
960 (layout::get_x_bound_for_row): Likewise.
961 (layout::print_line): Likewise.
962 (diagnostic_show_locus): Likewise for locals "last_line" and
964 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
965 * input.c (selftest::test_linenum_comparisons): New function.
966 (selftest::input_c_tests): Call it.
967 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
968 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
969 * selftest.h (ASSERT_GT): New macro.
970 (ASSERT_GT_AT): New macro.
971 (ASSERT_LT): New macro.
972 (ASSERT_LT_AT): New macro.
974 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
976 PR rtl-optimization/84780
977 * combine.c (distribute_links): Don't make a link based on pc_rtx.
979 2018-03-14 Martin Liska <mliska@suse.cz>
981 * tree.c (record_node_allocation_statistics): Use
983 (get_stats_node_kind): New function extracted from
984 record_node_allocation_statistics.
985 (free_node): Use get_stats_node_kind.
987 2018-03-14 Richard Biener <rguenther@suse.de>
989 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
990 that the value-set of ANTIC_IN doesn't grow.
993 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
995 (BB_VISITED_WITH_VISITED_SUCCS): New define.
996 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
998 2018-03-14 Julia Koval <julia.koval@intel.com>
1000 * config.gcc (icelake-client, icelake-server): New.
1002 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1003 (initial_ix86_arch_features): Ditto.
1004 (PTA_SKYLAKE): Add SGX.
1005 (PTA_ICELAKE): Remove.
1006 (PTA_ICELAKE_CLIENT): New.
1007 (PTA_ICELAKE_SERVER): New.
1008 (ix86_option_override_internal): Split up icelake on icelake client and
1010 (get_builtin_code_for_version): Ditto.
1011 (fold_builtin_cpu): Ditto.
1012 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1013 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1014 * config/i386/i386.h (processor_type): Ditto.
1015 * doc/invoke.texi: Ditto.
1017 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1020 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1021 INTEGER_CST offset, add it together with bitpos / 8 and
1022 sign extend based on POINTER_SIZE.
1026 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1029 * config/i386/constraints.md (Yc): New register constraint.
1030 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1031 Use Yc constraint for alternative 2 of operand 0. Remove
1032 preferred_for_speed attribute.
1034 2018-03-14 Richard Biener <rguenther@suse.de>
1036 PR tree-optimization/84830
1037 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1038 with the old one to avoid oscillations.
1040 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1043 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1045 (assign_by_spills): Return a flag of reload assignment failure.
1046 Do not process the reload assignment failures. Do not spill other
1047 reload pseudos if they has the same reg class. Update n if
1049 (lra_assign): Add a return arg. Set up from the result of
1050 assign_by_spills call.
1051 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1052 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1053 usage_insns if it is not NULL.
1054 (spill_hard_reg_in_range): New function.
1055 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1056 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1057 function prototypes.
1058 (lra_assign): Change prototype.
1059 * lra.c (lra): Add code to deal with fails by splitting hard reg
1062 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1064 * config/riscv/riscv.opt (mrelax): New option.
1065 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1066 "norelax" when riscv_mrelax is disabled.
1067 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1069 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1072 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1073 reassociation for int modes.
1075 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1077 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1078 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1080 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1081 * config/aarch64/aarch64-sve.md
1082 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1083 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1084 (*extend<mode><Vwide>2): Rename to...
1085 (aarch64_sve_extend<mode><Vwide>2): ...this.
1086 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1087 renaming the old pattern to...
1088 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1090 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1091 define_expand, renaming the old pattern to...
1092 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1093 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1094 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1095 account when deciding which SVE instruction the optab should use.
1096 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1098 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1100 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1101 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1102 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1103 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1104 (tlsdesc_small_<mode>): Turn a define_expand and use
1105 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1106 (tlsdesc_small_advsimd_<mode>): ...this.
1107 (tlsdesc_small_sve_<mode>): New pattern.
1109 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1111 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1112 (UNSPEC_UMUL_HIGHPART): New constants.
1113 (MUL_HIGHPART): New int iteraor.
1114 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1115 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1117 (*<su>mul<mode>3_highpart): New define_insn.
1119 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1122 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1125 2018-03-13 Martin Liska <mliska@suse.cz>
1128 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1130 (sem_item_optimizer::~sem_item_optimizer): Release it.
1131 (sem_item_optimizer::merge_classes): Register variable aliases.
1132 (sem_item_optimizer::fixup_pt_set): New function.
1133 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1134 * ipa-icf.h: Declare new variables and functions.
1136 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1139 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1140 integer_pow2p@2 and test integer_pow2p in condition.
1141 (A < 0 ? C : 0): Similarly for @1.
1144 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1145 characters starting at p contain '\0' character, don't look beyond
1149 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1150 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1153 * reg-stack.c (change_stack): Change update_end var from int to
1154 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1155 also call set_block_for_insn on the newly added insns and rescan.
1158 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1159 on the last operand.
1162 * tree.c (stabilize_reference_1): Return save_expr (e) for
1163 STATEMENT_LIST even if it doesn't have side-effects.
1165 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1167 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1169 2018-03-12 Renlin Li <renlin.li@arm.com>
1171 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1172 aarch64_output_scalar_simd_mov_immediate.
1174 2018-03-12 Martin Sebor <msebor@redhat.com>
1176 PR tree-optimization/83456
1177 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1178 for perfectly overlapping calls to memcpy.
1179 (gimple_fold_builtin_memory_chk): Same.
1180 (gimple_fold_builtin_strcpy): Handle no-warning.
1181 (gimple_fold_builtin_stxcpy_chk): Same.
1182 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1184 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1186 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1187 parameter. Use it for SFmode.
1188 (rs6000_function_arg_advance_1): Adjust.
1189 (rs6000_function_arg): Adjust.
1190 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1192 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1194 PR rtl-optimization/84169
1195 PR rtl-optimization/84780
1196 * combine.c (can_combine_p): Check for a 2-insn combination whether
1197 the destination register is used between the two insns, too.
1199 2018-03-12 Richard Biener <rguenther@suse.de>
1201 PR tree-optimization/84803
1202 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1203 for refs DR analysis didn't process.
1205 2018-03-12 Richard Biener <rguenther@suse.de>
1207 PR tree-optimization/84777
1208 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1209 force-vectorize loops ignore whether we are optimizing for size.
1211 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1213 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1214 (TARGET_MD_ASM_ADJUST): Define.
1216 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1217 Kito Cheng <kito.cheng@gmail.com>
1218 Chung-Ju Wu <jasonwucj@gmail.com>
1220 * config/nds32/nds32.c (nds32_compute_stack_frame,
1221 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1222 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1223 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1224 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1225 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1226 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1227 * config/nds32/nds32.md (prologue, epilogue): Use macro
1228 NDS32_V3PUSH_AVAILABLE_P to do checking.
1230 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1233 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1234 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1235 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1236 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1237 addition of most attributes on !orig_type_die or the attribute not
1238 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1240 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1241 Chung-Ju Wu <jasonwucj@gmail.com>
1243 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1245 * config/nds32/nds32.opt (mvh): New option.
1247 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1248 Chung-Ju Wu <jasonwucj@gmail.com>
1250 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1252 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1253 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1256 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1257 Chung-Ju Wu <jasonwucj@gmail.com>
1259 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1261 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1262 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1264 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1265 Kito Cheng <kito.cheng@gmail.com>
1266 Chung-Ju Wu <jasonwucj@gmail.com>
1268 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1269 UNSPEC_FFMISM and UNSPEC_FLMISM.
1270 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1271 for ffb, ffmism and flmism.
1272 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1273 (unspec_ffmism): Ditto.
1274 (unspec_flmism): Ditto.
1275 (nds32_expand_builtin_impl): Check if string extension is available.
1276 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1277 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1279 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1282 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1285 * lra-assigns.c (assign_by_spills): Return a flag of reload
1286 assignment failure. Do not process the reload assignment
1287 failures. Do not spill other reload pseudos if they has the same
1289 (lra_assign): Add a return arg. Set up from the result of
1290 assign_by_spills call.
1291 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1292 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1293 usage_insns if it is not NULL.
1294 (spill_hard_reg_in_range): New function.
1295 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1296 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1297 function prototypes.
1298 (lra_assign): Change prototype.
1299 * lra.c (lra): Add code to deal with fails by splitting hard reg
1302 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1305 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1307 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1310 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1311 (dw_val_node): Add val_symbolic_view.
1312 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1313 (symview_upper_bound): New.
1314 (new_line_info_table): Initialize symviews_since_reset.
1315 (dwarf2out_source_line): Count symviews_since_reset and set
1316 symview_upper_bound.
1317 (dw_val_equal_p): Handle symview.
1318 (add_AT_symview): New.
1319 (print_dw_val): Handle symview.
1320 (attr_checksum, attr_checksum_ordered): Likewise.
1321 (same_dw_val_p, size_of_die): Likewise.
1322 (value_format, output_die): Likewise.
1323 (add_high_low_attributes): Use add_AT_symview for entry_view.
1324 (dwarf2out_finish): Reset symview_upper_bound, clear
1327 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1330 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1331 Add strict argument and use it.
1332 (rs6000_split_multireg_move): Update for new strict argument.
1333 (mem_operand_gpr): Disallow all non-offsettable addresses.
1334 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1336 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1339 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1340 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1341 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1344 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1345 decl, use remap_type if we want to use the type.
1347 2018-03-09 Martin Sebor <msebor@redhat.com>
1349 PR tree-optimization/84526
1350 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1352 (builtin_access::generic_overlap): Be prepared to handle non-array
1355 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1357 PR rtl-optimization/84682
1358 * lra-constraints.c (process_address_1): Check is_address flag
1359 for address constraints.
1360 (process_alt_operands): Likewise.
1361 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1362 preprocess_constraints.
1363 * recog.h (preprocess_constraints): Add oploc parameter.
1365 * recog.c (preprocess_constraints): Test address_operand for
1366 CT_ADDRESS constraints.
1368 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1371 * lra-assigns.c (assign_by_spills): Return a flag of reload
1372 assignment failure. Do not process the reload assignment
1373 failures. Do not spill other reload pseudos if they has the same
1375 (lra_assign): Add a return arg. Set up from the result of
1376 assign_by_spills call.
1377 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1378 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1379 usage_insns if it is not NULL.
1380 (spill_hard_reg_in_range): New function.
1381 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1382 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1383 function prototypes.
1384 (lra_assign): Change prototype.
1385 * lra.c (lra): Add code to deal with fails by splitting hard reg
1388 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1391 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1392 Accept complain bool parameter. Only emit errors if it is true.
1393 (arm_parse_cpu_option_name): Likewise.
1394 (arm_target_thumb_only): Adjust callers of the above.
1395 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1396 prototype to take a default true bool parameter.
1397 (arm_parse_arch_option_name): Likewise.
1399 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1400 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1404 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1405 * configure: Regenerate.
1406 * configure.ac ("linker --version-script option"): New.
1407 ("linker soname option"): New.
1409 2018-03-09 Richard Biener <rguenther@suse.de>
1411 PR tree-optimization/84775
1412 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1413 immediate uses of predicate stmts and mark them modified.
1416 PR tree-optimization/84178
1417 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1419 (version_loop_for_if_conversion): Delay update_ssa call.
1420 (tree_if_conversion): Delay update_ssa until after predicate
1423 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1426 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1427 when the function accesses prior frames.
1429 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1432 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1433 gen_llsym, otherwise call maybe_gen_llsym.
1436 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1437 has ',' character inside of it.
1439 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1442 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1443 as clobbering CC_REGNUM.
1445 2018-03-08 Richard Biener <rguenther@suse.de>
1448 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1449 (follow_copies_to_constant): Do not follow SSA names registered
1452 2018-03-08 Richard Biener <rguenther@suse.de>
1454 PR tree-optimization/84178
1455 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1457 (version_loop_for_if_conversion): Delay update_ssa call.
1458 (tree_if_conversion): Delay update_ssa until after predicate
1461 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1463 PR tree-optimization/84178
1464 * tree-if-conv.c (release_bb_predicate): Remove the
1465 the assertion that the stmts have NULL use_ops.
1466 Discard the statements, asserting that they haven't
1467 yet been added to a BB.
1469 2018-03-08 Richard Biener <rguenther@suse.de>
1471 PR tree-optimization/84746
1472 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1473 (phi_translate): Pass in destination ANTIC_OUT set.
1474 (phi_translate_1): Likewise. For a simplified result lookup
1475 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1476 (phi_translate_set): Adjust.
1477 (do_pre_regular_insertion): Likewise.
1478 (do_pre_partial_partial_insertion): Likewise.
1480 2018-03-08 Martin Liska <mliska@suse.cz>
1482 PR gcov-profile/84735
1483 * doc/gcov.texi: Document usage of profile files.
1484 * gcov-io.h: Document changes in the format.
1486 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1490 * dwarf2out.c (struct dw_line_info_table): Update comments for
1492 (FORCE_RESET_NEXT_VIEW): New.
1493 (FORCE_RESETTING_VIEW_P): New.
1494 (RESETTING_VIEW_P): Check for -1 too.
1495 (ZERO_VIEW_P): Likewise.
1496 (new_line_info_table): Force-reset next view.
1497 (dwarf2out_begin_function): Likewise.
1498 (dwarf2out_source_line): Simplify zero_view_p initialization.
1499 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1500 view directly. Omit view when omitting .loc at line 0.
1502 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1504 PR tree-optimization/84740
1505 * tree-switch-conversion.c (process_switch): Call build_constructors
1506 only if info.phi_count is non-zero.
1508 PR tree-optimization/84739
1509 * tree-tailcall.c (find_tail_calls): Check call arguments against
1510 DECL_ARGUMENTS (current_function_decl) rather than
1511 DECL_ARGUMENTS (func) when checking for tail recursion.
1513 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1515 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1516 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1517 Volker Reichelt's entry and add entries for people that perform
1518 GCC fuzzy testing and report numerous bugs.
1520 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1523 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1524 readonly data in sdata, if that is disabled.
1525 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1526 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1527 -mreadonly-in-sdata option.
1529 2018-03-07 Martin Sebor <msebor@redhat.com>
1531 PR tree-optimization/84468
1532 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1533 basic block when looking for nul assignment.
1535 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1538 * except.h (output_function_exception_table): Adjust prototype.
1539 * except.c (output_function_exception_table): Remove FNNAME parameter
1540 and add SECTION parameter. Ouput one part of the table at a time.
1541 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1542 the first part of the exception table and emit unwind directives.
1543 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1544 (i386_pe_seh_cold_init): Likewise.
1545 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1546 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1547 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1548 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1549 * config/i386/winnt.c: Include except.h.
1550 (struct seh_frame_state): Add reg_offset, after_prologue and
1551 in_cold_section fields.
1552 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1553 (i386_pe_seh_cold_init): New function.
1554 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1555 to seh->in_cold_section.
1556 (seh_emit_push): Record the offset of the push.
1557 (seh_emit_save): Record the offet of the save.
1558 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1559 Test seh->after_prologue to disregard the epilogue.
1560 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1561 (i386_pe_end_cold_function): New function.
1563 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1566 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1567 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1570 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1572 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1573 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1576 * multiple_target.c: Include tree-inline.h and intl.h.
1577 (expand_target_clones): Diagnose and fail if node->definition and
1578 !tree_versionable_function_p (node->decl).
1580 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1582 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1584 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1585 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1586 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1588 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1591 * combine.c (try_combine): Use reg_or_subregno instead of handling
1592 just paradoxical SUBREGs and REGs.
1594 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1596 * config/arc/arc.c (arc_finalize_pic): Remove function.
1597 (arc_must_save_register): We use single base PIC register, remove
1598 checks to save/restore the PIC register.
1599 (arc_expand_prologue): Likewise.
1600 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1602 (arc_verify_short): Likewise.
1603 (arc_attr_type): Likewise.
1604 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1605 (walk_stores): Likewise.
1606 (arc_address_cost): Make it static.
1607 (arc_verify_short): Likewise.
1608 (branch_dest): Likewise.
1609 (arc_attr_type): Likewise.
1610 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1611 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1612 (arc_final_prescan_insn): Remove inserting the nops due to
1613 hardware hazards. It is done in reorg step.
1614 (insn_length_variant_t): Remove.
1615 (insn_length_parameters_t): Likewise.
1616 (arc_insn_length_parameters): Likewise.
1617 (arc_get_insn_variants): Likewise.
1618 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1620 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1623 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1626 PR tree-optimization/84687
1627 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1629 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1631 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1633 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1634 Rename to ppc_speculation_barrier.
1635 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1636 __builtin_ppc_speculation_barrier.
1638 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1641 * combine.c (combine_simplify_rtx): Don't try to simplify if
1642 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1645 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1647 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1648 to 32 bytes when compiling for POWER9.
1650 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1653 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1654 regparm >= 3 with no arg reg available also for calls with
1655 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1658 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1660 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1662 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1665 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1667 2018-03-05 Richard Biener <rguenther@suse.de>
1669 PR tree-optimization/84486
1670 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1671 When inserting a __builtin_assume_aligned call set the LHS
1672 SSA name alignment info accordingly.
1674 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1676 PR tree-optimization/84114
1677 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1678 Avoid reassociation of FLOAT_MODE addition.
1680 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1682 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1683 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1684 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1685 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1686 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1687 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1688 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1690 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1691 __builtin_ia32_wbinvd): New builtins.
1692 (SPECIAL_ARGS2): New.
1693 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1694 (SPECIAL_ARGS2): New.
1695 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1696 (ix86_valid_target_attribute_inner_p): Ditto.
1697 (ix86_init_mmx_sse_builtins): Add special_args2.
1698 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1699 TARGET_WBNOINVD_P): New.
1700 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1701 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1702 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1703 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1704 * config/i386/pconfigintrin.h: New file.
1705 * config/i386/wbnoinvdintrin.h: Ditto.
1706 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1707 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1709 2018-03-05 Richard Biener <rguenther@suse.de>
1711 PR tree-optimization/84670
1712 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1714 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1715 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1716 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1717 doesn't grow if all successors (recursively) were visited at least
1720 2018-03-05 Richard Biener <rguenther@suse.de>
1722 PR tree-optimization/84650
1723 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1724 if executed in the loop pipeline.
1726 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1728 * doc/configfiles.texi (Configuration Files): Move info about
1729 conditionalizing $target-protos.h to...
1730 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1731 differs from $target-protos.h.
1733 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1734 Chung-Ju Wu <jasonwucj@gmail.com>
1736 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1737 * config/nds32/nds32-multiple.md (setmemsi): Define.
1738 * config/nds32/nds32-memory-manipulation.c
1739 (nds32_gen_dup_4_byte_to_word_value): New.
1740 (emit_setmem_word_loop): New.
1741 (emit_setmem_byte_loop): New.
1742 (nds32_expand_setmem_loop): New.
1743 (nds32_expand_setmem_loop_v3m): New.
1744 (nds32_expand_setmem_unroll): New.
1745 (nds32_expand_setmem): New.
1747 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1748 Chung-Ju Wu <jasonwucj@gmail.com>
1750 * config/nds32/nds32-memory-manipulation.c
1751 (nds32_emit_load_store): New.
1752 (nds32_emit_post_inc_load_store): New.
1753 (nds32_emit_mem_move): New.
1754 (nds32_emit_mem_move_block): New.
1755 (nds32_expand_movmemsi_loop_unknown_size): New.
1756 (nds32_expand_movmemsi_loop_known_size): New.
1757 (nds32_expand_movmemsi_loop): New.
1758 (nds32_expand_movmemsi_unroll): New.
1759 (nds32_expand_movmemqi): Rename ...
1760 (nds32_expand_movmemsi): ... to this.
1761 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1762 (movmemsi): ... to this.
1763 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1764 (nds32_expand_movmemsi): ... to this.
1766 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1767 Monk Chiang <sh.chiang04@gmail.com>
1768 Chung-Ju Wu <jasonwucj@gmail.com>
1770 * config/nds32/nds32-protos.h
1771 (nds32_expand_load_multiple): New arguments.
1772 (nds32_expand_store_multiple): Ditto.
1773 (nds32_valid_multiple_load_store): Rename ...
1774 (nds32_valid_multiple_load_store_p): ... to this.
1775 * config/nds32/nds32-memory-manipulation.c
1776 (nds32_expand_load_multiple): Refine implementation.
1777 (nds32_expand_store_multiple): Ditto.
1778 * config/nds32/nds32-multiple.md
1779 (load_multiple): Update nds32_expand_load_multiple interface.
1780 (store_multiple): Update nds32_expand_store_multiple interface.
1781 * config/nds32/nds32-predicates.c
1782 (nds32_valid_multiple_load_store): Rename ...
1783 (nds32_valid_multiple_load_store_p): ... to this and refine
1785 * config/nds32/predicates.md
1786 (nds32_load_multiple_and_update_address_operation): New predicate.
1787 (nds32_store_multiple_and_update_address_operation): New predicate.
1789 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1790 Chung-Ju Wu <jasonwucj@gmail.com>
1792 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1793 (combo): New attribute.
1794 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1796 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1798 * config/nds32/nds32.opt: Change -mcmodel= default value.
1800 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1801 Monk Chiang <sh.chiang04@gmail.com>
1802 Chung-Ju Wu <jasonwucj@gmail.com>
1804 * config/nds32/constants.md (unspec_element): New enum.
1805 * config/nds32/constraints.md (Umw): New constraint.
1806 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1807 * config/nds32/nds32-intrinsic.md: Likewise.
1808 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1809 (nds32_valid_smw_lwm_base_p): New.
1810 (nds32_output_smw_single_word): New.
1811 (nds32_output_lmw_single_word): New.
1812 (nds32_expand_unaligned_load): New.
1813 (nds32_expand_unaligned_store): New.
1814 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1815 (nds32_output_smw_single_word): Declare.
1816 (nds32_output_lmw_single_word): Declare.
1817 (nds32_expand_unaligned_load): Declare.
1818 (nds32_expand_unaligned_store): Declare.
1819 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1820 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1821 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1822 NDS32_BUILTIN_UASTORE_DW.
1823 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1826 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1827 Kito Cheng <kito.cheng@gmail.com>
1828 Chung-Ju Wu <jasonwucj@gmail.com>
1830 * config/nds32/nds32-intrinsic.c
1831 (nds32_expand_builtin_null_ftype_reg): Delete.
1832 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1833 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1834 (nds32_read_argument): New.
1835 (nds32_legitimize_target): Ditto.
1836 (nds32_legitimize_argument): Ditto.
1837 (nds32_check_constant_argument): Ditto.
1838 (nds32_expand_unop_builtin): Ditto.
1839 (nds32_expand_unopimm_builtin): Ditto.
1840 (nds32_expand_binop_builtin): Ditto.
1841 (nds32_builtin_decl_impl): Ditto.
1842 (builtin_description): Ditto.
1843 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1844 (nds32_init_builtins_impl): Ditto.
1845 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1846 (nds32_builtin_decl): New.
1847 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1848 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1850 2018-03-02 Jeff Law <law@redhat.com>
1852 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1853 (redundant_insn, fill_simple_delay_slots): Likewise.
1854 (fill_slots_from_thread): Likewise.
1855 * resource.c (mark_referenced_resources): Likewise.
1856 (mark_set_resources, find_dead_or_set_registers): Likewise.
1858 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1860 * substring-locations.h (format_warning_va): Formatting fix for
1862 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1864 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1865 * substring-locations.c: Include intl.h.
1866 (format_warning_va): Turned into small wrapper around
1867 format_warning_n_va, renamed to ...
1868 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1869 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1871 (format_warning_at_substring_n): New function.
1872 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1873 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1874 format_warning_at_substring with just a shorter name instead of
1875 const function pointer.
1876 (fmtwarn_n): New function.
1877 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1878 appropriate, get rid of all the fmtstr temporaries, move conditionals
1879 with G_() wrapped string literals directly into fmtwarn arguments,
1880 cast dir.len to (int), formatting fixes.
1882 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1884 * doc/invoke.texi: Remove "Cilk Plus" references.
1886 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1887 Richard Biener <rguenther@suse.de>
1890 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1891 for error or warning attributes if CALL_FROM_THUNK_P is set.
1894 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1897 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1898 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1901 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1902 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1904 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1905 instead of -1U in last predictors element's probability member.
1907 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1910 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1911 arguments if they are comparable.
1913 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1915 PR tree-optimization/84634
1916 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1917 masks and masked_loop_p with a single loop_masks, making sure it's
1918 null for bb vectorization.
1920 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1922 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1923 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1924 loop->force_vectorize to check whether there is no alias.
1926 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1929 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1931 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1933 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1934 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1935 instead of a loop around prev_real_insn.
1936 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1940 * config/i386/i386.c (ix86_print_operand): Use conditional
1941 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1944 2018-03-02 Richard Biener <rguenther@suse.de>
1946 PR tree-optimization/84427
1947 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1948 (bitmap_set_subtract_values): Rewrite to handle multiple
1951 (prune_clobbered_mems): Likewise.
1952 (phi_translate): Take edge instead of pred/phiblock.
1953 (phi_translate_1): Likewise.
1954 (phi_translate_set): Likewise. Insert all translated
1955 exprs for a value into the set, keeping possibly multiple
1956 expressions per value.
1957 (compute_antic_aux): Adjust for phi_translate changes.
1958 When intersecting union the expressions and prune those
1959 not in the final value set, keeping possibly multiple
1960 expressions per value. Do not use value-insertion
1961 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1962 all expressions. Add verification that the value-sets
1963 only shrink during iteration.
1964 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1965 (do_pre_regular_insertion): Likewise.
1966 (do_pre_partial_partial_insertion): Likewise.
1968 2018-03-02 Richard Biener <rguenther@suse.de>
1971 * config/darwin.c (saved_debug_info_level): New static global.
1972 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1973 (darwin_asm_lto_end): Restore debug info generation settings.
1975 2018-03-01 Martin Liska <mliska@suse.cz>
1978 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1981 2018-03-01 Richard Biener <rguenther@suse.de>
1984 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1985 type annotation with LTO when debug was disabled at compile-time.
1987 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1989 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1991 (mips_final_postscan_insn): Likewise.
1993 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1995 PR rtl-optimization/84528
1996 * alias.c (init_alias_target): Add commentary.
1997 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1998 a unique base value if the frame pointer is not eliminated
1999 to the stack pointer.
2001 2018-03-01 Tom de Vries <tom@codesourcery.com>
2003 PR rtl-optimization/83327
2004 * lra-int.h (hard_regs_spilled_into): Declare.
2005 * lra.c (hard_regs_spilled_into): Define.
2006 (init_reg_info): Init hard_regs_spilled_into.
2007 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2008 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2009 (process_bb_lives): Handle hard_regs_spilled_into.
2010 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2011 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2013 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2015 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2016 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2017 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2018 * config/rs6000/aix72.h: New file.
2020 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2022 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2023 instead of warning_at with conditional singular and plural messages
2027 * stor-layout.c (update_alignment_for_field): For
2028 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2029 && !DECL_PACKED (field), do the alignment update, just use
2030 only desired_align instead of MAX (type_align, desired_align)
2032 (place_field): Don't do known_align < desired_align handling
2033 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2034 is non-NULL, instead do it after rli->prev_field handling and
2035 only if not within a bitfield word. For DECL_PACKED (field)
2036 use type_align of BITS_PER_UNIT.
2038 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2040 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2041 superfluous parentheses and trailing spaces.
2043 2018-02-28 Richard Biener <rguenther@suse.de>
2045 PR tree-optimization/84584
2046 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2047 SCoPs with fake exit edge.
2049 2018-02-28 Martin Liska <mliska@suse.cz>
2052 * timevar.c (timer::print): Fix format to properly print 100%
2055 2018-02-28 Richard Biener <rguenther@suse.de>
2058 * genmatch.c (capture_info::walk_match): Do not mark
2059 captured expressions without operands as expr_p given
2060 they act more like predicates and should be subject to
2061 "lost tail" side-effect preserving.
2063 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2065 PR rtl-optimization/81611
2066 * auto-inc-dec.c (attempt_change): Move dead note from
2067 mem_insn if it's the next use of regno
2068 (find_address): Take address use of reg holding
2069 non-incremented value. Add parm to limit search to the named
2071 (merge_in_block): Attempt to use a mem insn that is the next
2072 use of the original regno.
2074 2018-02-27 Martin Sebor <msebor@redhat.com>
2077 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2078 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2080 2018-02-27 Martin Sebor <msebor@redhat.com>
2082 PR translation/84207
2083 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2084 n argument to unsigned HOST_WIDE_INT.
2085 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2086 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2087 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2088 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2090 2018-02-27 Richard Biener <rguenther@suse.de>
2092 PR tree-optimization/84512
2093 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2094 Do not use the estimate returned from record_stmt_cost for
2095 the scalar iteration cost but sum properly using add_stmt_cost.
2097 2018-02-27 Richard Biener <rguenther@suse.de>
2099 PR tree-optimization/84466
2100 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2101 Adjust last change to less strictly validate use operands.
2103 2018-02-27 Martin Liska <mliska@suse.cz>
2105 PR gcov-profile/84548
2106 * gcov.c (process_file): Allow partial overlap and consider it
2107 also as group functions.
2108 (output_lines): Properly calculate range of lines for a group.
2110 2018-02-27 Martin Liska <mliska@suse.cz>
2112 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2113 'ggc' suffixes. Change first column width.
2114 (timer::print): Fix formatting of the column.
2116 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2118 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2119 preserve inline entry blocks for the sake of debug inline
2120 entry point markers alone.
2121 (remove_unused_locals): Suggest in comments a better place to
2122 force the preservation of inline entry blocks that are
2123 otherwise unused, but do not preserve them.
2125 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2127 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2129 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2132 * config/i386/constraints.md (Bs): Replace
2133 ix86_indirect_branch_register with
2134 TARGET_INDIRECT_BRANCH_REGISTER.
2136 * config/i386/i386.md (indirect_jump): Likewise.
2137 (tablejump): Likewise.
2138 (*sibcall_memory): Likewise.
2139 (*sibcall_value_memory): Likewise.
2140 Peepholes of indirect call and jump via memory: Likewise.
2141 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2142 (*sibcall_value_GOT_32): Likewise.
2143 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2144 (GOT_memory_operand): Likewise.
2145 (call_insn_operand): Likewise.
2146 (sibcall_insn_operand): Likewise.
2147 (GOT32_symbol_operand): Likewise.
2148 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2150 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2152 PR rtl-optimization/83496
2153 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2154 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2155 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2156 redundant insn, if any.
2157 (relax_delay_slots): Likewise.
2158 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2160 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2162 PR tree-optimization/83965
2163 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2164 that grouped statements are part of a reduction chain. Return
2165 true if the statement is not marked as a reduction itself but
2167 (vect_recog_dot_prod_pattern): Don't check whether the statement
2168 is part of a group here.
2169 (vect_recog_sad_pattern): Likewise.
2170 (vect_recog_widen_sum_pattern): Likewise.
2172 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2175 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2177 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2180 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2182 (ix86_output_indirect_function_return): New prototype.
2183 (ix86_split_simple_return_pop_internal): Likewise.
2184 * config/i386/i386.c (indirect_return_via_cx): New.
2185 (indirect_return_via_cx_bnd): Likewise.
2186 (indirect_thunk_name): Handle return va CX_REG.
2187 (output_indirect_thunk_function): Create alias for
2188 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2189 (ix86_output_indirect_jmp): Remove the bool argument.
2190 (ix86_output_indirect_function_return): New function.
2191 (ix86_split_simple_return_pop_internal): Likewise.
2192 * config/i386/i386.md (*indirect_jump): Don't pass false
2193 to ix86_output_indirect_jmp.
2194 (*tablejump_1): Likewise.
2195 (simple_return_pop_internal): Change it to define_insn_and_split.
2196 Call ix86_split_simple_return_pop_internal to split it for
2198 (simple_return_indirect_internal): Call
2199 ix86_output_indirect_function_return instead of
2200 ix86_output_indirect_jmp.
2202 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2205 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2206 memset and value initialization afterwards.
2208 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2210 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2212 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2215 * common/config/aarch64/aarch64-common.c
2216 (aarch_option_optimization_table[]): Switch
2217 off fomit-frame-pointer
2219 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2220 Chung-Ju Wu <jasonwucj@gmail.com>
2222 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2224 (store_multiple): Ditto.
2226 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2228 * config.gcc: Add --with-cpu support for nds32 target.
2229 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2230 * config/nds32/nds32.opt: Add -mcpu= option.
2232 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2234 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2235 isel=yes): Warn for these deprecated options.
2237 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2239 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2240 ISA_2_5_MASKS_EMBEDDED.
2242 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2244 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2245 p->max as pointers rather than using iterative_hash_expr.
2247 2018-02-23 Carl Love <cel@us.ibm.com>
2249 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2250 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2252 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2253 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2254 P8V_BUILTIN_VEC_VUNSIGNED2.
2256 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2259 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2260 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2262 * lra-constraints.c (curr_insn_transform): Set up
2263 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2265 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2266 * lra-eliminations.c (spill_pseudos): Ditto.
2267 (process_insn_for_elimination): Ditto.
2268 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2269 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2270 LRA_NON_CLOBBERED_ALT.
2272 2018-02-22 Martin Sebor <msebor@redhat.com>
2274 PR tree-optimization/84480
2275 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2276 to maybe_diag_stxncpy_trunc. Call it.
2277 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2278 from gimple_fold_builtin_strcpy. Print inlining stack.
2279 (handle_builtin_stxncpy): Print inlining stack.
2280 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2282 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2285 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2286 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2287 and -fcheck-pointer-bounds are used together.
2288 (indirect_thunk_prefix): New enum.
2289 (indirect_thunk_need_prefix): New function.
2290 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2291 "_nt" instead of "_bnd" for NOTRACK prefix.
2292 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2293 (output_indirect_thunk_function): Likewise.
2295 (ix86_code_end): Update output_indirect_thunk_function calls.
2296 (ix86_output_indirect_branch_via_reg): Replace
2297 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2298 (ix86_output_indirect_branch_via_push): Likewise.
2299 (ix86_output_function_return): Likewise.
2300 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2301 incompatible with -fcf-protection=branch and
2302 -fcheck-pointer-bounds.
2304 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2307 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2308 Change gcc_assert call to output_operand_lossage.
2310 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2312 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2314 2018-02-22 DJ Delorie <dj@redhat.com>
2315 Sebastian Perta <sebastian.perta@renesas.com>
2316 Oleg Endo <olegendo@gcc.gnu.org>
2318 * config/rx/rx.c (rx_rtx_costs): New function.
2319 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2321 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2323 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2325 2018-02-22 Martin Liska <mliska@suse.cz>
2328 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2329 Add "native" as a possible value.
2331 2018-02-22 Martin Liska <mliska@suse.cz>
2334 * config/i386/i386.c (ix86_option_override_internal):
2335 Add "native" as a possible value for -march and -mtune.
2337 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2340 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2341 to all type variants.
2343 PR tree-optimization/84503
2344 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2345 width as info->bitpos + info->bitsize - start.
2346 (merged_store_group::merge_overlapping): Simplify width computation.
2347 (check_no_overlap): New function.
2348 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2349 start + width and last_order of the group, fail if check_no_overlap
2351 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2352 to group if check_no_overlap fails.
2354 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2356 * config/rs6000/altivec.md: Delete contraint arguments to
2357 define_expand, define_split, and define_peephole2, and in
2358 define_insn_and_split if always unused.
2359 * config/rs6000/darwin.md: Ditto.
2360 * config/rs6000/dfp.md: Ditto.
2361 * config/rs6000/rs6000.md: Ditto.
2362 * config/rs6000/sync.md: Ditto.
2363 * config/rs6000/vector.md: Ditto.
2364 * config/rs6000/vsx.md: Ditto.
2366 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2368 * config/rs6000/altivec.md: Write output control strings as braced
2369 blocks instead of double-quoted strings.
2370 * config/rs6000/darwin.md: Ditto.
2371 * config/rs6000/rs6000.md: Ditto.
2372 * config/rs6000/vector.md: Ditto.
2373 * config/rs6000/vsx.md: Ditto.
2375 2018-02-21 Jason Merrill <jason@redhat.com>
2377 PR c++/84314 - ICE with templates and fastcall attribute.
2378 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2380 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2382 * ipa-cp.c (determine_versionability): Fix comment typos.
2384 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2387 * ipa-cp.c (determine_versionability): Do not version functions caling
2390 2018-02-21 Martin Liska <mliska@suse.cz>
2393 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2394 Add "native" as a possible value.
2395 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2396 the macro when native cpu detection is available.
2398 2018-02-21 Martin Liska <mliska@suse.cz>
2401 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2402 Add "native" as a possible value.
2403 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2404 when native cpu detection is available.
2406 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2407 Martin Sebor <msebor@redhat.com>
2409 PR tree-optimization/84478
2410 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2412 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2413 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2414 support which is conservatively correct, for 2 only stay conservative
2415 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2416 argument to the 2 argument get_range_strlen, adjust 6 arg
2417 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2419 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2420 (gimple_fold_builtin_strlen): Pass true as last argument to
2423 2018-02-20 Martin Sebor <msebor@redhat.com>
2426 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2427 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2428 (builtin_memref::builtin_memref): Factor out parts into
2429 set_base_and_offset and call it.
2431 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2434 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2435 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2436 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2437 search at the associated MODE_INT.
2439 2018-02-20 Jeff Law <law@redhat.com>
2442 PR tree-optimization/81592
2444 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2445 for range data rather than using global data.
2446 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2447 range data rather than using global data.
2448 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2449 pass it to children as needed.
2450 (struct directive::fmtresult): Similarly.
2451 (struct directive::set_width): Similarly.
2452 (struct directive::set_precision): Similarly.
2453 (format_integer, format_directive, parse_directive): Similarly.
2454 (format_none): Accept unnamed vr_values parameter.
2455 (format_percent, format_floating, format_character): Similarly.
2456 (format_string, format_plain): Similarly.
2457 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2458 the EVRP range analyzer for range data rather than using global data.
2459 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2460 gimple-ssa-evrp-analyze.h
2461 (class sprintf_dom_walker): Add after_dom_children member function.
2462 Add evrp_range_analyzer member.
2463 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2464 range analyzer as needed.
2465 (sprintf_dom_walker::after_dom_children): New member function.
2466 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2468 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2469 (evrp_range_analyzer::pop_to_marker): Likewise.
2471 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2473 PR tree-optimization/84419
2474 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2475 with the required type if its current type is compatible but
2478 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2481 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2482 after vectorization.
2484 2018-02-20 Martin Liska <mliska@suse.cz>
2487 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2488 possible values if we don't have a hint.
2490 2018-02-20 Martin Liska <mliska@suse.cz>
2494 * final.c (shorten_branches): Build align_tab array with one
2496 * opts.c (finish_options): Add alignment option limit check.
2497 (MAX_CODE_ALIGN): Likewise.
2498 (MAX_CODE_ALIGN_VALUE): Likewise.
2499 * doc/invoke.texi: Document maximum allowed option value for
2500 all -falign-* options.
2502 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2505 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2506 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2507 * var-tracking.c (emit_note_insn_var_location): Remove all references
2508 to NOTE_INSN_CALL_ARG_LOCATION.
2509 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2510 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2511 Use copy_rtx_if_shared.
2512 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2513 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2514 (dwarf2out_var_location): Remove handling of
2515 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2517 * final.c (final_scan_insn): Remove all references to
2518 NOTE_INSN_CALL_ARG_LOCATION.
2519 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2520 before dumping final insns.
2521 * except.c (emit_note_eh_region_end): Remove all references to
2522 NOTE_INSN_CALL_ARG_LOCATION.
2523 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2524 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2525 * config/arc/arc.c (hwloop_optimize): Likewise.
2526 * config/arm/arm.c (create_fix_barrier): Likewise.
2527 * config/s390/s390.c (s390_chunkify_start): Likewise.
2528 * config/sh/sh.c (find_barrier): Likewise.
2529 * config/i386/i386.c (rest_of_insert_endbranch,
2530 ix86_seh_fixup_eh_fallthru): Likewise.
2531 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2532 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2533 * config/frv/frv.c (frv_function_prologue): Likewise.
2534 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2536 (note_outside_basic_block_p): Remove all references to
2537 NOTE_INSN_CALL_ARG_LOCATION.
2538 * gengtype.c (adjust_field_rtx_def): Likewise.
2539 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2541 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2542 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2545 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2548 PR tree-optimization/84452
2549 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2550 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2553 2018-02-19 Martin Liska <mliska@suse.cz>
2556 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2558 2018-02-19 Martin Liska <mliska@suse.cz>
2559 Richard Sandiford <richard.sandiford@linaro.org>
2561 PR tree-optimization/82491
2562 * gimple-fold.c (get_base_constructor): Make earlier bail out
2565 2018-02-19 Carl Love <cel@us.ibm.com>
2567 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2568 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2570 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2571 P8V_BUILTIN_VEC_NEG.
2573 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2575 * config/rl78/rl78.md (movdf): New define expand.
2577 2018-02-19 Martin Liska <mliska@suse.cz>
2580 * doc/invoke.texi: Fix typo.
2581 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2583 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2585 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2586 handle rs6000_single_float and rs6000_double_float specially for
2589 2018-02-16 Jeff Law <law@redhat.com>
2591 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2594 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2597 * ipa-inline.c (inline_small_functions): Fix a typo.
2599 2018-02-16 Nathan Sidwell <nathan@acm.org>
2601 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2603 2018-02-16 Carl Love <cel@us.ibm.com>
2605 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2606 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2607 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2608 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2609 expansion to P8V_BUILTIN_VEC_FLOAT2.
2611 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2613 PR rtl-optimization/70023
2614 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2615 src_regno into account.
2617 2018-02-16 Carl Love <cel@us.ibm.com>
2619 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2620 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2621 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2622 * config/rs6000/rs6000.c: Remove case statements for
2623 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2624 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2625 and P9V_BUILTIN_VEC_VINSERT4B.
2626 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2627 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2628 * config/rs6000/vsx.md:
2629 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2632 2018-02-16 Carl Love <cel@us.ibm.com>
2634 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2636 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2638 * config/rs6000/rs6000-c.c: Add the definitions for
2639 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2640 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2641 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2642 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2643 definition for insert4b and define insn *insert3b_internal.
2644 * doc/extend.texi: Add documentation for vec_extract4b.
2646 2018-02-16 Nathan Sidwell <nathan@acm.org>
2648 * doc/extend.texi (Backwards Compatibility): Mention friend
2649 injection. Note for-scope is deprecated.
2650 * doc/invoke.texi (-ffriend-injection): Deprecate.
2652 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2654 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2655 that moved to I2, also allow destinations that are a paradoxical
2656 subreg (instead of a normal reg).
2658 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2661 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2664 2018-02-16 Richard Biener <rguenther@suse.de>
2666 PR tree-optimization/84037
2667 PR tree-optimization/84016
2669 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2670 Adjust vec_construct for the fact we need additional higher latency
2671 128bit inserts for AVX256 and AVX512 vector builds.
2672 (ix86_add_stmt_cost): Scale vector construction cost for
2675 2018-02-16 Richard Biener <rguenther@suse.de>
2677 PR tree-optimization/84417
2678 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2679 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2680 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2682 2018-02-16 Martin Liska <mliska@suse.cz>
2685 * internal-fn.def (ASAN_CHECK): Set proper flags.
2686 (ASAN_MARK): Likewise.
2688 2018-02-16 Julia Koval <julia.koval@intel.com>
2690 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2691 from PTA_CANNONLAKE.
2693 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2696 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2697 Use ++iter rather than iter++ for std::list iterators.
2698 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2699 defer deleting them until all nodes in the forest are processed. Do
2700 free even leaf nodes. Change to_process into auto_vec.
2703 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2704 * vec.h (vec_default_construct): Use memset instead of placement new
2705 if BROKEN_VALUE_INITIALIZATION is defined.
2706 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2707 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2710 PR rtl-optimization/83723
2711 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2712 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2713 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2715 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2717 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2719 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2721 PR rtl-optimization/81443
2722 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2723 from inner REGs to paradoxical SUBREGs.
2725 2018-02-16 Richard Biener <rguenther@suse.de>
2727 PR tree-optimization/84399
2728 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2729 For operands we can analyze at their definition make sure we can
2730 analyze them at each use as well.
2732 2018-02-16 Richard Biener <rguenther@suse.de>
2734 PR tree-optimization/84190
2735 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2736 volatile accesses if the decl isn't volatile.
2738 2018-02-15 Jason Merrill <jason@redhat.com>
2740 PR c++/84314 - ICE with templates and fastcall attribute.
2741 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2742 TYPE_CANONICAL on an existing type.
2744 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2746 PR tree-optimization/84383
2747 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2748 dstoff nor call operand_equal_p if dstbase is NULL.
2750 PR tree-optimization/84334
2751 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2752 also a CONSTANT_CLASS_P, punt.
2754 2018-02-14 Jim Wilson <jimw@sifive.com>
2756 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2757 first SMALL_OPERAND check. New local min_second_step. Move assert
2758 to where locals are set. Add TARGET_RVC support.
2759 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2761 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2763 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2765 2018-02-14 Martin Sebor <msebor@redhat.com>
2767 PR tree-optimization/83698
2768 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2769 arrays constrain the offset range to their bounds.
2770 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2771 (builtin_access::overlap): Avoid setting the size of overlap if it's
2773 (maybe_diag_overlap): Also consider arrays when deciding what values
2774 of offsets to include in diagnostics.
2776 2018-02-14 Martin Sebor <msebor@redhat.com>
2779 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2780 that correspond to the kind of a declaration.
2782 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2785 * config/pa/pa.md: Load address of PIC label using the linkage table
2786 if the label is nonlocal.
2788 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2790 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2791 warning message if user requests -maltivec=be.
2792 * doc/invoke.texi: Document deprecation of -maltivec=be.
2794 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2797 * config/rs6000/rs6000-c.c: Update definitions for
2798 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2799 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2801 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2804 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2805 add _get_ssp intrinsics. Remove argument from
2806 __builtin_ia32_rdssp[d|q].
2807 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2808 * config/i386/i386-builtin.def: Remove argument from
2809 __builtin_ia32_rdssp[d|q].
2810 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2811 ix86_expand_special_args_builtin for _rdssp[d|q].
2812 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2813 Clear register before usage.
2814 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2815 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2817 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2819 PR tree-optimization/84357
2820 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2821 operand 1 of an ARRAY_REF too.
2823 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2826 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2827 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2829 (set_of_reg): New struct.
2830 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2831 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2832 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2834 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2835 Split into bitclr, bitset, bitinvert patterns if appropriate.
2836 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2837 use rx_fuse_in_memory_bitop.
2838 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2839 to named insn, correct maximum insn length.
2841 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2844 * machmode.def: Define a complex mode for PARTIAL_INT.
2845 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2847 * doc/rtl.texi: Document CSPImode.
2848 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2850 (msp430_hard_regno_nregs_with_padding): Likewise.
2852 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2855 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2857 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2859 PR rtl-optimization/84169
2860 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2861 we generated a parallel as new i3 and we split that to new i2 and i3
2862 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2863 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2864 those to i2, not i1. Partially rewrite this scan code.
2866 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2869 * stor-layout.c (place_field): For variable length fields, adjust
2870 offset_align afterwards not just based on the field's alignment,
2871 but also on the size.
2874 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2875 of exps and logs in the use_exp2 case.
2877 2018-02-13 Jeff Law <law@redhat.com>
2879 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2882 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2885 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2889 * common.opt (gas-loc-support, gas-locview-support): New.
2890 (ginline-points, ginternal-reset-location-views): New.
2891 * doc/invoke.texi: Document them. Use @itemx where intended.
2892 (gvariable-location-views): Adjust.
2893 * target.def (reset_location_view): New.
2894 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2895 (TARGET_RESET_LOCATION_VIEW): New.
2896 * doc/tm.texi: Rebuilt.
2897 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2898 (dwarf2out_default_as_locview_support): New.
2899 (output_asm_line_debug_info): Use option variables.
2900 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2901 (output_loc_list): Likewise.
2902 (add_high_low_attributes): Check option variables.
2903 Don't output entry view attribute in strict mode.
2904 (gen_inlined_subroutine_die): Check option variables.
2905 (dwarf2out_inline_entry): Likewise.
2906 (init_sections_and_labels): Likewise.
2907 (dwarf2out_early_finish): Likewise.
2908 (maybe_reset_location_view): New, from...
2909 (dwarf2out_var_location): ... here. Call it.
2910 * debug.h (dwarf2out_default_as_loc_support): Declare.
2911 (dwarf2out_default_as_locview_support): Declare.
2912 * hooks.c (hook_int_rtx_insn_0): New.
2913 * hooks.h (hook_int_rtx_insn_0): Declare.
2914 * toplev.c (process_options): Take -gas-loc-support and
2915 -gas-locview-support from dwarf2out. Enable
2916 -gvariable-location-views by default only with locview
2917 assembler support. Enable -ginternal-reset-location-views by
2918 default only if the target defines the corresponding hook.
2919 Enable -ginline-points by default if location views are
2920 enabled; force it disabled if statement frontiers are
2922 * tree-inline.c (expand_call_inline): Check option variables.
2923 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2925 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2927 PR tree-optimization/84321
2928 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2929 handling. Also check whether the anti-range contains any values
2930 that satisfy the mask; switch to a VR_RANGE if not.
2932 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2935 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2937 2018-02-13 Martin Jambor <mjambor@suse.cz>
2940 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2941 of call statements, also set location of a load to a temporary.
2943 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2945 * config/rl78/rl78.c (add_vector_labels): New function.
2946 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2947 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2948 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2949 which checks that no arguments are passed.
2950 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2951 * doc/extend.texi: Documentation for the new attribute.
2953 2018-02-13 Andreas Schwab <schwab@suse.de>
2955 * config/riscv/linux.h (CPP_SPEC): Define.
2957 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2960 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2961 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2962 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2963 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2964 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2965 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2966 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2967 temporarily for AES and PCLMUL builtins.
2969 PR tree-optimization/84339
2970 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2971 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2975 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2976 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2977 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2979 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2981 * omp-simd-clone.h: New file.
2982 * omp-simd-clone.c: Include omp-simd-clone.h.
2983 (expand_simd_clones): No longer static.
2984 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2985 cgraph.h and omp-simd-clone.h.
2986 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2987 (vect_recog_widen_shift_pattern): Formatting fix.
2988 (vect_pattern_recog_1): Don't check optab for calls.
2991 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2992 operands[2] into a REG before using gen_lowpart on it.
2994 2018-02-12 Jeff Law <law@redhat.com>
2997 * config/sh/sh.c (find_barrier): Consider a sibling call
3000 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3001 successfully back substituting a reg.
3003 2018-02-12 Richard Biener <rguenther@suse.de>
3005 PR tree-optimization/84037
3006 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3007 parameter, move visited init to caller.
3008 (vect_slp_analyze_operations): Separate cost from validity
3009 check, initialize visited once for all instances.
3010 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3012 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3014 (vect_model_promotion_demotion_cost): Likewise.
3015 (vectorizable_bswap): Guard cost modeling with !slp_node
3016 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3018 (vectorizable_call): Likewise.
3019 (vectorizable_conversion): Likewise.
3020 (vectorizable_assignment): Likewise.
3021 (vectorizable_shift): Likewise.
3022 (vectorizable_operation): Likewise.
3023 (vectorizable_store): Likewise.
3024 (vectorizable_load): Likewise.
3025 (vectorizable_condition): Likewise.
3026 (vectorizable_comparison): Likewise.
3028 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3031 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3032 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3033 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3035 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3038 * params.def (inline-min-speedup): Increase from 8 to 15.
3039 (max-inline-insns-auto): Decrease from 40 to 30.
3040 * ipa-split.c (consider_split): Add some buffer for function to
3041 be considered inlining candidate.
3042 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3045 2018-02-12 Richard Biener <rguenther@suse.de>
3047 PR tree-optimization/84037
3048 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3049 matched stmts if we cannot swap the non-matched ones.
3051 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3053 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3054 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3055 _mm_maskz_scalef_round_ss): New intrinsics.
3056 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3057 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3058 __builtin_ia32_scalefss_round): Remove.
3059 (__builtin_ia32_scalefsd_mask_round,
3060 __builtin_ia32_scalefss_mask_round): New intrinsics.
3061 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3062 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3063 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3064 "<round_constraint>")): Changed to ...
3065 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3066 "<round_scalar_constraint>")): ... this.
3067 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3068 %0, %1, %2<round_op3>}"): Changed to ...
3069 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3070 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3071 %2<round_scalar_mask_op3>}"): ... this.
3072 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3074 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3076 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3077 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3078 (_mm_maskz_sqrt_round_ss): New intrinsics.
3079 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3080 (__builtin_ia32_sqrtsd_mask_round)
3081 (__builtin_ia32_sqrtss_mask_round): New builtins.
3082 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3083 (__builtin_ia32_sqrtss_round): Remove.
3084 (__builtin_ia32_sqrtsd_mask_round)
3085 (__builtin_ia32_sqrtss_mask_round): New builtins.
3086 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3087 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3088 ((match_operand:VF_128 1 "vector_operand"
3089 "xBm,<round_constraint>")): Changed to ...
3090 ((match_operand:VF_128 1 "vector_operand"
3091 "xBm,<round_scalar_constraint>")): ... this.
3092 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3093 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3094 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3095 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3096 %<iptr>1<round_scalar_mask_op3>}): ... this.
3097 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3098 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3100 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3103 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3104 Cast vec_cmpeq result to correct type.
3105 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3106 Cast vec_cmpgt result to correct type.
3108 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3110 * final.c (final_scan_insn_1): Renamed from...
3111 (final_scan_insn): ... this. New wrapper, to recover
3112 seen from the outermost call in recursive ones.
3113 * config/sparc/sparc.c (output_return): Drop seen from call.
3114 (output_sibcall): Likewise.
3115 * config/visium/visium.c (output_branch): Likewise.
3117 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3119 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3122 2018-02-10 Alan Modra <amodra@gmail.com>
3125 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3126 Specify LR as an input.
3128 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3131 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3132 remove_member_access_dummy_vars): New functions.
3133 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3134 lower_omp_1, execute_lower_omp): Use them.
3136 PR rtl-optimization/84308
3137 * shrink-wrap.c (spread_components): Release todo vector.
3139 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3141 PR rtl-optimization/57193
3142 * ira-color.c (struct allocno_color_data): Add member
3143 conflict_allocno_hard_prefs.
3144 (update_conflict_allocno_hard_prefs): New.
3145 (bucket_allocno_compare_func): Add a preference based on
3146 conflict_allocno_hard_prefs.
3147 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3148 (color_allocnos): Remove a dead code. Initiate
3149 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3151 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3154 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3155 constraint from =wa to wa. Avoid a subreg on the output operand,
3156 instead use a pseudo and subreg it in a move.
3157 (p9_xxbrd_<mode>): Changed to ...
3158 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3159 (p9_xxbrd_v2df): New expander.
3160 (p9_xxbrw_<mode>): Changed to ...
3161 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3162 (p9_xxbrw_v4sf): New expander.
3164 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3166 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3168 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3171 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3172 multiply in 32-bit mode.
3173 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3174 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3177 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3179 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3180 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3181 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3182 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3184 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3187 * dwarf2out.c (is_trivial_indirect_ref): New function.
3188 (dwarf2out_late_global_decl): Do not generate a location
3189 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3190 and that are not defined in the current unit.
3192 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3194 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3195 instead of a libcall for UNORDERED.
3197 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3200 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3201 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3203 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3206 * config/s390/s390.c (s390_set_current_function): Invoke
3207 s390_indirect_branch_settings also if fndecl didn't change.
3209 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3211 * config/rs6000/rs6000.md (blockage): Set length to zero.
3213 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3215 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3217 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3220 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3221 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3225 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3226 PARALLEL incoming that failed vt_get_decl_and_offset check.
3229 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3230 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3232 (get_variable_section): For decls in named .bss* sections pass true as
3233 second argument to bss_initializer_p.
3235 2018-02-09 Marek Polacek <polacek@redhat.com>
3236 Jakub Jelinek <jakub@redhat.com>
3239 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3240 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3241 Sync some changes from cxx_fold_indirect_ref.
3243 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3245 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3247 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3248 (BLOCK_INLINE_ENTRY_LABEL): New.
3249 (dwarf2out_var_location): Disregard inline entry markers.
3250 (inline_entry_data): New struct.
3251 (inline_entry_data_hasher): New hashtable type.
3252 (inline_entry_data_hasher::hash): New.
3253 (inline_entry_data_hasher::equal): New.
3254 (inline_entry_data_table): New variable.
3255 (add_high_low_attributes): Add DW_AT_entry_pc and
3256 DW_AT_GNU_entry_view attributes if a pending entry is found
3257 in inline_entry_data_table. Add old entry_pc attribute only
3258 if debug nonbinding markers are disabled.
3259 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3260 markers are enabled.
3261 (block_within_block_p, dwarf2out_inline_entry): New.
3262 (dwarf2out_finish): Check that no entries remained in
3263 inline_entry_data_table.
3264 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3265 (final_scan_insn, notice_source_line): Likewise.
3266 (rest_of_clean_state): Skip inline entry markers.
3267 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3269 * gimple.c (gimple_build_debug_inline_entry): New.
3270 * gimple.h (enum gimple_debug_subcode): Add
3271 GIMPLE_DEBUG_INLINE_ENTRY.
3272 (gimple_build_debug_inline_entry): Declare.
3273 (gimple_debug_inline_entry_p): New.
3274 (gimple_debug_nonbind_marker_p): Adjust.
3275 * insn-notes.def (INLINE_ENTRY): New.
3276 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3277 inline entry marker notes.
3278 (print_insn): Likewise.
3279 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3280 (INSN_DEBUG_MARKER_KIND): Likewise.
3281 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3282 * tree-inline.c (expand_call_inline): Build and insert
3283 debug_inline_entry stmt.
3284 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3285 inline entry blocks early, if nonbind markers are enabled.
3286 (dump_scope_block): Dump fragment info.
3287 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3288 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3289 (gimple_build_debug_inline_entry): New.
3290 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3291 Enable/disable inline entry points too.
3292 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3293 (DEBUG_INSN): Describe inline entry markers.
3295 * common.opt (gvariable-location-views): New.
3296 (gvariable-location-views=incompat5): New.
3297 * config.in: Rebuilt.
3298 * configure: Rebuilt.
3299 * configure.ac: Test assembler for view support.
3300 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3301 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3302 * dwarf2out.c (var_loc_view): New typedef.
3303 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3304 (dwarf2out_locviews_in_attribute): New.
3305 (dwarf2out_locviews_in_loclist): New.
3306 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3307 (enum dw_line_info_opcode): Add LI_adv_address.
3308 (struct dw_line_info_table): Add view.
3309 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3310 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3311 (zero_view_p): New variable.
3312 (ZERO_VIEW_P): New macro.
3313 (output_asm_line_debug_info): New.
3314 (struct var_loc_node): Add view.
3315 (add_AT_view_list, AT_loc_list): New.
3316 (add_var_loc_to_decl): Add view param. Test it against last.
3317 (new_loc_list): Add view params. Record them.
3318 (AT_loc_list_ptr): Handle loc and view lists.
3319 (view_list_to_loc_list_val_node): New.
3320 (print_dw_val): Handle dw_val_class_view_list.
3321 (size_of_die): Likewise.
3322 (value_format): Likewise.
3323 (loc_list_has_views): New.
3324 (gen_llsym): Set vl_symbol too.
3325 (maybe_gen_llsym, skip_loc_list_entry): New.
3326 (dwarf2out_maybe_output_loclist_view_pair): New.
3327 (output_loc_list): Output view list or entries too.
3328 (output_view_list_offset): New.
3329 (output_die): Handle dw_val_class_view_list.
3330 (output_dwarf_version): New.
3331 (output_compilation_unit_header): Use it.
3332 (output_skeleton_debug_sections): Likewise.
3333 (output_rnglists, output_line_info): Likewise.
3334 (output_pubnames, output_aranges): Update version comments.
3335 (output_one_line_info_table): Output view numbers in asm comments.
3336 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3337 Call maybe_gen_llsym.
3338 (loc_list_from_tree_1): Adjust.
3339 (add_AT_location_description): Create view list attribute if
3340 needed, check it's absent otherwise.
3341 (convert_cfa_to_fb_loc_list): Adjust.
3342 (maybe_emit_file): Call output_asm_line_debug_info for test.
3343 (dwarf2out_var_location): Reset views as needed. Precompute
3344 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3345 attribute. Set view.
3346 (new_line_info_table): Reset next view.
3347 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3348 (dwarf2out_source_line): Likewise. Output view resets and labels to
3349 the assembler, or select appropriate line info opcodes.
3350 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3351 (optimize_string_length): Catch it. Adjust.
3352 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3353 dw_val_class_view_list, and remove it if no longer needed.
3354 (hash_loc_list): Hash view numbers.
3355 (loc_list_hasher::equal): Compare them.
3356 (optimize_location_lists): Check whether a view list symbol is
3357 needed, and whether the locview attribute is present, and
3358 whether they match. Remove the locview attribute if no longer
3360 (index_location_lists): Call skip_loc_list_entry for test.
3361 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3362 Use output_dwarf_version.
3363 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3364 (struct dw_val_node): Add val_view_list.
3365 * final.c (SEEN_NEXT_VIEW): New.
3366 (set_next_view_needed): New.
3367 (clear_next_view_needed): New.
3368 (maybe_output_next_view): New.
3369 (final_start_function): Rename to...
3370 (final_start_function_1): ... this. Take pointer to FIRST,
3371 add SEEN parameter. Emit param bindings in the initial view.
3372 (final_start_function): Reintroduce SEEN-less interface.
3373 (final): Rename to...
3374 (final_1): ... this. Take SEEN parameter. Output final pending
3375 next view at the end.
3376 (final): Reintroduce seen-less interface.
3377 (final_scan_insn): Output pending next view before switching
3378 sections or ending a block. Mark the next view as needed when
3379 outputting variable locations. Notify debug backend of section
3380 changes, and of location view changes.
3381 (rest_of_handle_final): Adjust.
3382 * toplev.c (process_options): Autodetect value for debug variable
3383 location views option. Warn on incompat5 without -gdwarf-5.
3384 * doc/invoke.texi (gvariable-location-views): New.
3385 (gvariable-location-views=incompat5): New.
3386 (gno-variable-location-views): New.
3388 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3390 PR tree-optimization/84136
3391 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3392 that the result of find_edge is non-NULL.
3394 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3397 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3398 storing integer register in SImode. Fix cost of 256 and 512
3399 byte aligned SSE register store.
3401 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3403 * config/i386/i386.c (ix86_multiplication_cost): Fix
3404 multiplication cost for TARGET_AVX512DQ.
3406 2018-02-08 Marek Polacek <polacek@redhat.com>
3408 PR tree-optimization/84238
3409 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3412 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3414 PR tree-optimization/84265
3415 * tree-vect-stmts.c (vectorizable_store): Don't treat
3416 VMAT_CONTIGUOUS accesses as grouped.
3417 (vectorizable_load): Likewise.
3419 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3421 PR tree-optimization/81635
3422 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3423 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3424 (test_round_for_mask): New functions.
3425 (wide_int_cc_tests): Call test_round_for_mask.
3426 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3427 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3428 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3429 range returned by get_range_info.
3431 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3434 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3435 * symtab.c: Include builtins.h
3436 (symtab_node::output_to_lto_symbol_table_p): Move here
3437 from lto-streamer-out.c:output_symbol_p.
3438 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3439 (output_symbol_p): Move all logic to symtab.c
3440 (produce_symtab): Update.
3442 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3444 * config/s390/s390-opts.h (enum indirect_branch): Define.
3445 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3446 (s390_indirect_branch_via_thunk)
3447 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3448 (enum s390_indirect_branch_type): Define.
3449 * config/s390/s390.c (struct s390_frame_layout, struct
3450 machine_function): Remove.
3451 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3452 (indirect_branch_table_label_no, indirect_branch_table_name):
3454 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3455 (enum s390_indirect_branch_option): Define.
3456 (s390_return_addr_from_memory): New function.
3457 (s390_handle_string_attribute): New function.
3458 (s390_attribute_table): Add new attribute handler.
3459 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3460 (s390_indirect_branch_via_thunk): New function.
3461 (s390_indirect_branch_via_inline_thunk): New function.
3462 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3463 sibling call optimization for non z10 compiles.
3464 (s390_emit_call): Force indirect branch target to be a single
3465 register. Add r1 clobber for non-z10 compiles.
3466 (s390_emit_epilogue): Emit return jump via return_use expander.
3467 (s390_reorg): Handle JUMP_INSNs as execute targets.
3468 (s390_option_override_internal): Perform validity checks for the
3469 new command line options.
3470 (s390_indirect_branch_attrvalue): New function.
3471 (s390_indirect_branch_settings): New function.
3472 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3473 (s390_output_indirect_thunk_function): New function.
3474 (s390_code_end): Implement target hook.
3475 (s390_case_values_threshold): Implement target hook.
3476 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3478 * config/s390/s390.h (struct s390_frame_layout)
3479 (struct machine_function): Move here from s390.c.
3480 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3481 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3482 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3483 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3484 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3485 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3486 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3487 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3488 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3489 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3490 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3491 (mnemonic attribute): Add values which aren't recognized
3493 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3494 pattern for branch conversion. Fix mnemonic attribute.
3495 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3496 indirect branch via thunk if requested.
3497 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3498 ("*indirect_jump"): Disable for branch conversion using out of
3500 ("indirect_jump_via_thunk<mode>_z10")
3501 ("indirect_jump_via_thunk<mode>")
3502 ("indirect_jump_via_inlinethunk<mode>_z10")
3503 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3504 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3505 ("casesi_jump_via_inlinethunk<mode>_z10")
3506 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3507 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3508 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3509 ("*indirect2_jump"): Disable for branch conversion.
3510 ("casesi_jump"): Turn into expander and expand patterns for branch
3512 ("return_use"): New expander.
3513 ("*return"): Emit return via thunk and rename it to ...
3514 ("*return<mode>"): ... this one.
3515 * config/s390/s390.opt: Add new options and and enum for the
3518 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3520 * lra-constraints.c (match_reload): Unconditionally use
3521 gen_lowpart_SUBREG, rather than selecting between that
3522 and equivalent gen_rtx_SUBREG code.
3524 2018-02-08 Richard Biener <rguenther@suse.de>
3526 PR tree-optimization/84233
3527 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3528 changed flag instead of boguously re-using phi_inserted.
3530 2018-02-08 Martin Jambor <mjambor@suse.cz>
3532 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3533 static local variables.
3535 2018-02-08 Richard Biener <rguenther@suse.de>
3537 PR tree-optimization/84278
3538 * tree-vect-stmts.c (vectorizable_store): When looking for
3539 smaller vector types to perform grouped strided loads/stores
3540 make sure the mode is supported by the target.
3541 (vectorizable_load): Likewise.
3543 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3545 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3546 Increase LDP/STP opportunities by adding adjacent callee-saves.
3548 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3550 PR rtl-optimization/84068
3551 PR rtl-optimization/83459
3552 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3554 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3556 PR tree-optimization/84224
3557 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3558 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3561 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3564 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3565 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3566 count, remove one USE.
3568 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3570 * doc/install.texi (Configuration): Document the
3571 --with-long-double-format={ibm,ieee} PowerPC configuration
3575 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3576 Convert from define_expand to be define_insn_and_split. Rework
3577 float/double/_Float128 conversions to QI/HI/SImode to work with
3578 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3579 conversions to QI/HImode types did a store and then a load to
3580 truncate the value. For conversions to VSX registers, don't split
3581 the insn, instead emit the code directly. Use the code iterator
3582 any_fix to combine signed and unsigned conversions.
3583 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3584 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3585 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3586 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3587 (fix_<mode>di2_hw): Likewise.
3588 (fixuns_<mode>di2_hw): Likewise.
3589 (fix_<mode>si2_hw): Likewise.
3590 (fixuns_<mode>si2_hw): Likewise.
3591 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3592 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3593 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3594 fix<uns>_trunc<SFDF:mode>si2_p8.
3595 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3597 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3598 (fix<uns>_<mode>_mem): Likewise.
3599 (fctiw<u>z_<mode>_mem): Likewise.
3600 (fix<uns>_<mode>_mem): Likewise.
3601 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3602 the register allocator from doing a direct move to the GPRs to do
3603 a store, and instead use the ISA 3.0 store byte/half-word from
3604 vector register instruction. For IEEE 128-bit floating point,
3605 also optimize stores of 32-bit ints.
3606 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3608 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3610 * genextract.c (push_pathstr_operand): New function to support
3612 (walk_rtx): Call push_pathstr_operand.
3613 (print_path): Support [a-zA-Z].
3615 2018-02-07 Richard Biener <rguenther@suse.de>
3617 PR tree-optimization/84037
3618 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3619 (cse_and_gimplify_to_preheader): Declare.
3620 (vect_get_place_in_interleaving_chain): Likewise.
3621 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3623 (_loop_vec_info::~_loop_vec_info): Delete it.
3624 (cse_and_gimplify_to_preheader): New function.
3625 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3626 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3627 (vectorizable_load): Likewise. For grouped stores always base
3628 the IV on the first element.
3629 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3630 condition before gimplifying.
3632 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3634 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3635 *DIV_EXPR and *MOD_EXPR.
3637 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3640 * config/i386/i386.c (ix86_option_override_internal): Mask out
3641 the CF_SET bit when checking -fcf-protection.
3643 2018-02-07 Tom de Vries <tom@codesourcery.com>
3646 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3649 2018-02-07 Richard Biener <rguenther@suse.de>
3651 PR tree-optimization/84204
3652 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3655 PR tree-optimization/84205
3656 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3657 special-case isl_ast_op_zdiv_r.
3659 PR tree-optimization/84223
3660 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3661 Only add conditions from within the region.
3662 (gather_bbs::after_dom_children): Adjust.
3664 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3667 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3668 * config/avr/avr.md: Only post-reload split REG-REG moves if
3669 either register is GENERAL_REG_P.
3671 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3673 PR tree-optimization/84235
3674 * tree-ssa-scopedtables.c
3675 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3676 if the subtraction is performed in floating point type where NaNs are
3677 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3678 build 1. Formatting fix.
3680 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3683 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3684 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3685 and skip it regardless of bb boundaries. Use CALL_P macro,
3686 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3687 unnecessarily, formatting fix.
3689 2018-02-06 Michael Collison <michael.collison@arm.com>
3691 * config/arm/thumb2.md:
3692 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3693 (*thumb_mov_notscc): Ditto.
3695 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3698 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3699 unsigned_fix, not "s".
3701 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3703 * configure.ac (gcc_fn_eh_frame_ro): New function.
3704 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3705 correct .eh_frame permissions.
3706 * configure: Regenerate.
3708 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3710 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3713 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3715 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3716 Display warning message for -mno-speculate-indirect-jumps.
3718 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3720 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3722 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3724 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3726 PR tree-optimization/84225
3727 * tree-eh.c (find_trapping_overflow): Only call
3728 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3730 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3733 * config/i386/i386.c: Reimplement the check of possible options
3734 -mibt/-mshstk conbination. Change error messages.
3735 * doc/invoke.texi: Fix a typo: remove extra '='.
3737 2018-02-06 Marek Polacek <polacek@redhat.com>
3739 PR tree-optimization/84228
3740 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3742 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3745 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3746 emitted arch directives.
3747 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3748 __ARM_FEATURE_COPROC before changing architectures.
3750 2018-02-06 Richard Biener <rguenther@suse.de>
3752 * config/i386/i386.c (print_reg): Fix typo.
3753 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3755 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3757 * configure: Regenerate.
3759 2018-02-05 Martin Sebor <msebor@redhat.com>
3761 PR tree-optimization/83369
3762 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3765 2018-02-05 Martin Liska <mliska@suse.cz>
3767 * doc/invoke.texi: Cherry-pick upstream r323995.
3769 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3771 * ira.c (ira_init_register_move_cost): Adjust comment.
3773 2018-02-05 Martin Liska <mliska@suse.cz>
3775 PR gcov-profile/84137
3776 * doc/gcov.texi: Fix typo in documentation.
3778 2018-02-05 Martin Liska <mliska@suse.cz>
3780 PR gcov-profile/83879
3781 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3782 using dlopen functionality.
3784 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3786 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3787 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3788 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3789 _mm_maskz_range_round_ss): New intrinsics.
3790 (__builtin_ia32_rangesd128_round)
3791 (__builtin_ia32_rangess128_round): Remove.
3792 (__builtin_ia32_rangesd128_mask_round,
3793 __builtin_ia32_rangess128_mask_round): New builtins.
3794 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3795 __builtin_ia32_rangess128_round): Remove.
3796 (__builtin_ia32_rangesd128_mask_round,
3797 __builtin_ia32_rangess128_mask_round): New builtins.
3798 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3799 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3800 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3801 "<round_saeonly_constraint>")): Changed to ...
3802 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3803 "<round_saeonly_scalar_constraint>")): ... this.
3804 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3805 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3806 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3807 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3808 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3810 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3812 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3814 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3815 Remove all values except native, 8540 and 8548.
3817 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3819 * config/i386/i386.c (ix86_output_function_return): Pass
3820 INVALID_REGNUM, instead of -1, as invalid register number to
3821 indirect_thunk_name and output_indirect_thunk.
3823 2018-02-02 Julia Koval <julia.koval@intel.com>
3825 * config.gcc: Add -march=icelake.
3826 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3827 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3828 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3829 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3830 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3831 (processor_target_table): Add icelake.
3832 (ix86_option_override_internal): Handle new PTAs.
3833 (get_builtin_code_for_version): Handle icelake.
3834 (M_INTEL_COREI7_ICELAKE): New.
3835 (fold_builtin_cpu): Handle icelake.
3836 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3837 * doc/invoke.texi: Add -march=icelake.
3839 2018-02-02 Julia Koval <julia.koval@intel.com>
3841 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3842 to wide_int_bitmask.
3843 * wide-int-bitmask.h: New.
3845 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3848 * config/i386/i386.md: Replace Pmode with word_mode in
3849 builtin_setjmp_setup and builtin_longjmp to support x32.
3851 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3855 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3857 (rs6000_supported_cpu_names): New static variable.
3858 (linux_cpu_translation_table): Likewise.
3859 (elf_platform) <cpu>: Define new static variable and use it.
3860 Translate kernel AT_PLATFORM name to canonical name if needed.
3861 Error if platform name is unknown.
3863 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3866 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3868 2018-02-01 Jeff Law <law@redhat.com>
3871 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3872 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3873 the scratch if RELEASE_VIA_POP is false.
3874 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3875 If we have to save a temporary register, decrement SIZE appropriately.
3876 Pass new arguments to release_scratch_register_on_entry.
3877 (ix86_adjust_stack_and_probe): Likewise.
3878 (ix86_emit_probe_stack_range): Pass new arguments to
3879 release_scratch_register_on_entry.
3881 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3883 PR rtl-optimization/84157
3884 * combine.c (change_zero_ext): Use REG_P predicate in
3885 front of HARD_REGISTER_P predicate.
3887 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3889 * config/avr/avr.c (avr_option_override): Move disabling of
3890 -fdelete-null-pointer-checks to...
3891 * common/config/avr/avr-common.c (avr_option_optimization_table):
3894 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3896 PR tree-optimization/81635
3897 * tree-data-ref.c (split_constant_offset_1): For types that
3898 wrap on overflow, try to use range info to prove that wrapping
3901 2018-02-01 Renlin Li <renlin.li@arm.com>
3904 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3906 (aarch64_register_move_cost): Likewise.
3907 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3909 (REG_CLASS_NAMES): Likewise.
3910 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3911 TAILCALL_ADDR_REGS. Remove IP registers.
3912 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3914 2018-02-01 Richard Biener <rguenther@suse.de>
3916 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3917 for specifying RPO order and allow NULL for that.
3918 * domwalk.c (dom_walker::dom_walker): Likewise.
3919 (dom_walker::walk): Handle NULL RPO order.
3920 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3922 (rewrite_update_dom_walker): Likewise.
3923 (mark_def_dom_walker): Likewise.
3925 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3927 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3928 (aarch64_maybe_expand_sve_subreg_move): Declare.
3929 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3930 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3932 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3933 that are semantically a reverse operation.
3934 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3935 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3936 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3938 (aarch64_can_change_mode_class): For big-endian, forbid changes
3939 between two SVE modes if they have different element sizes.
3941 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3943 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3944 the TImode handling for big-endian targets.
3946 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3948 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3949 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3951 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3952 Remove BSWAP handing for big-endian targets and use the form of
3953 LD1RQ appropariate for the mode.
3955 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3957 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3958 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3961 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3964 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3965 check for operands that need to go through aarch64_sve_reload_be.
3967 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3969 PR tree-optimization/81661
3970 PR tree-optimization/84117
3971 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3972 * tree-eh.c: Include gimplify.h.
3973 (find_trapping_overflow, replace_trapping_overflow,
3974 rewrite_to_non_trapping_overflow): New functions.
3975 * tree-vect-loop.c: Include tree-eh.h.
3976 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3977 * tree-data-ref.c: Include tree-eh.h.
3978 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3980 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3982 PR rtl-optimization/84123
3983 * combine.c (change_zero_ext): Check if hard register satisfies
3984 can_change_dest_mode before calling gen_lowpart_SUBREG.
3986 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3989 * ira.c (ira_init_register_move_cost): Remove assert.
3991 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3993 PR rtl-optimization/84071
3994 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3995 * doc/tm.texi: Regenerate.
3997 2018-01-31 Richard Biener <rguenther@suse.de>
3999 PR tree-optimization/84132
4000 * tree-data-ref.c (analyze_miv_subscript): Properly
4001 check whether evolution_function_is_affine_multivariate_p
4002 before calling gcd_of_steps_may_divide_p.
4004 2018-01-31 Julia Koval <julia.koval@intel.com>
4007 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4008 * config/i386/i386.md (rdpid_rex64) New.
4009 (rdpid): Make 32bit only.
4011 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4014 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4015 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4017 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4020 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4022 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4024 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4026 PR rtl-optimization/84071
4027 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4028 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4030 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4032 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4033 (arc_attribute_table): Add 'aux' attribute.
4034 (arc_in_small_data_p): Consider aux like variables.
4035 (arc_is_aux_reg_p): New function.
4036 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4037 (arc_get_aux_arg): New function.
4038 (prepare_move_operands): Handle aux-register access.
4039 (arc_handle_aux_attribute): New function.
4040 * doc/extend.texi (ARC Variable attributes): Add subsection.
4042 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4044 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4045 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4046 (arc_attribute_table): Add 'uncached' attribute.
4047 (arc_print_operand): Print '.di' flag for uncached memory
4049 (arc_in_small_data_p): Do not consider for small data the uncached
4051 (arc_is_uncached_mem_p): New function.
4052 * config/arc/predicates.md (compact_store_memory_operand): Check
4053 for uncached memory accesses.
4054 (nonvol_nonimm_operand): Likewise.
4055 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4057 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4060 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4061 falign-loops=): Add Optimization flag.
4063 2018-01-30 Jeff Law <law@redhat.com>
4066 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4067 INT_REGISTERS_SAVED. Check it prior to calling
4068 get_scratch_register_on_entry.
4069 (ix86_adjust_stack_and_probe): Similarly.
4070 (ix86_emit_probe_stack_range): Similarly.
4071 (ix86_expand_prologue): Corresponding changes.
4073 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4076 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4077 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4079 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4082 * lra-constraints.c (curr_insn_transform): Process AND in the
4085 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4087 PR rtl-optimization/83986
4088 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4089 dependence against last_pending_memory_flush in addition to
4092 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4094 PR tree-optimization/81611
4095 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4098 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4101 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4104 2018-01-30 Richard Biener <rguenther@suse.de>
4105 Jakub Jelinek <jakub@redhat.com>
4107 PR tree-optimization/84111
4108 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4109 inner loops added during recursion, as they don't have up-to-date
4112 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4115 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4116 (can_inline_edge_by_limits_p): ... here.
4117 (can_early_inline_edge_p, check_callers,
4118 update_caller_keys, update_callee_keys, recursive_inlining,
4119 add_new_edges_to_heap, speculation_useful_p,
4120 inline_small_functions,
4121 inline_small_functions, flatten_function,
4122 inline_to_all_callers_1): Update.
4124 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4126 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4129 2018-01-30 Richard Biener <rguenther@suse.de>
4131 PR tree-optimization/83008
4132 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4133 invariant and constant vector uses in stmts when they need
4136 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4139 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4140 * configure: Regenerate.
4142 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4144 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4146 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4147 Use gen_rtx_REG rather than gen_lowpart.
4149 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4151 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4152 rather than 0 when creating partial subregs.
4154 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4156 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4159 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4162 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4163 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4164 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4165 flags. This restores the settings used before the 2017-07-24.
4166 Turning off pre increment/decrement/modify allows IVOPTS to
4167 optimize DF/SF loops where the index is an int.
4169 2018-01-29 Richard Biener <rguenther@suse.de>
4170 Kelvin Nilsen <kelvin@gcc.gnu.org>
4173 * tree-vect-stmts.c (vectorizable_call): Don't call
4174 targetm.vectorize_builtin_md_vectorized_function if callee is
4177 2018-01-22 Carl Love <cel@us.ibm.com>
4179 * doc/extend.tex: Fix typo in second arg in
4180 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4182 2018-01-29 Richard Biener <rguenther@suse.de>
4184 PR tree-optimization/84086
4185 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4186 (flush_ssaname_freelist): When SSA names were released reset
4187 the SCEV hash table.
4189 2018-01-29 Richard Biener <rguenther@suse.de>
4191 PR tree-optimization/84057
4192 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4193 removed paths when removing edges.
4195 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4197 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4198 -mfunction-return=@var{choice}.
4200 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4203 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4205 (layout::print_source_line): Likewise.
4206 (test_get_line_width_without_trailing_whitespace): Add test cases.
4208 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4211 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4214 2018-01-26 Jim Wilson <jimw@sifive.com>
4216 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4218 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4221 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4223 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4224 and CMP + SUB-immediate -> SUBS.
4226 2018-01-26 Martin Sebor <msebor@redhat.com>
4228 PR tree-optimization/83896
4229 * tree-ssa-strlen.c (get_string_len): Rename...
4230 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4231 Avoid assuming length is constant.
4232 (handle_char_store): Use HOST_WIDE_INT for string length.
4234 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4237 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4238 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4240 2018-01-26 Richard Biener <rguenther@suse.de>
4242 PR rtl-optimization/84003
4243 * dse.c (record_store): Only record redundant stores when
4244 the earlier store aliases at least all accesses the later one does.
4246 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4248 PR rtl-optimization/83985
4249 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4250 REG_CFA_RESTORE insns.
4251 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4252 REG_CFA_RESTORE insns here.
4255 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4256 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4258 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4260 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4261 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4262 (arc_init): Likewise.
4263 (arc_override_options): Likewise.
4264 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4266 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4268 * config/arc/arc.h (TARGET_DBNZ): Define.
4269 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4270 properly set the tune attribute.
4271 (dbnz): Use TARGET_DBNZ guard.
4272 * config/arc/arc.opt (mtune): Add core3 option.
4274 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4276 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4277 recognize new pic like addresses.
4278 (arc_delegitimize_address): Clean up.
4280 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4282 * config/arc/arc-arches.def: Option mrf16 valid for all
4284 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4285 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4286 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4287 * config/arc/arc-tables.opt: Regenerate.
4288 * config/arc/arc.c (arc_conditional_register_usage): Handle
4289 reduced register file case.
4290 (arc_file_start): Set must have build attributes.
4291 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4293 * config/arc/arc.opt (mrf16): Add new option.
4294 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4295 * config/arc/genmultilib.awk: Handle new mrf16 option.
4296 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4297 * config/arc/t-multilib: Regenerate.
4298 * doc/invoke.texi (ARC Options): Document mrf16 option.
4300 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4302 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4303 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4304 (arc_attribute_table): Add 'secure_call' attribute.
4305 (arc_print_operand): Print secure call operand.
4306 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4308 (arc_is_secure_call_p): New function. * config/arc/arc.md
4309 (call_i): Add support for sjli instruction.
4310 (call_value_i): Likewise.
4311 * config/arc/constraints.md (Csc): New constraint.
4313 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4314 John Eric Martin <John.Martin@emmicro-us.com>
4316 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4317 * config/arc/arc.c (_arc_jli_section): New struct.
4318 (arc_jli_section): New type.
4319 (rc_jli_sections): New static variable.
4320 (arc_handle_jli_attribute): New function.
4321 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4322 (arc_file_end): New function.
4323 (TARGET_ASM_FILE_END): Define.
4324 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4325 (arc_add_jli_section): New function.
4326 (jli_call_scan): Likewise.
4327 (arc_reorg): Call jli_call_scan.
4328 (arc_output_addsi): Remove 'S' from printing asm operand.
4329 (arc_is_jli_call_p): New function.
4330 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4332 (movhi_insn): Likewise.
4333 (movsi_insn): Likewise.
4334 (movsi_set_cc_insn): Likewise.
4335 (loadqi_update): Likewise.
4336 (load_zeroextendqisi_update): Likewise.
4337 (load_signextendqisi_update): Likewise.
4338 (loadhi_update): Likewise.
4339 (load_zeroextendhisi_update): Likewise.
4340 (load_signextendhisi_update): Likewise.
4341 (loadsi_update): Likewise.
4342 (loadsf_update): Likewise.
4343 (movsicc_insn): Likewise.
4344 (bset_insn): Likewise.
4345 (bxor_insn): Likewise.
4346 (bclr_insn): Likewise.
4347 (bmsk_insn): Likewise.
4348 (bicsi3_insn): Likewise.
4349 (cmpsi_cc_c_insn): Likewise.
4350 (movsi_ne): Likewise.
4351 (movsi_cond_exec): Likewise.
4352 (clrsbsi2): Likewise.
4363 (call_i): Remove 'S' asm letter, add jli instruction.
4364 (call_value_i): Likewise.
4365 * config/arc/arc.op (mjli-always): New option.
4366 * config/arc/constraints.md (Cji): New constraint.
4367 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4369 (subsf3_fpx): Likewise.
4370 (mulsf3_fpx): Likewise.
4371 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4373 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4375 * doc/invoke.texi (ARC): Document mjli-always option.
4377 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4379 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4380 avoid addition with 0 and use incw and decw where possible.
4382 2018-01-26 Richard Biener <rguenther@suse.de>
4384 PR tree-optimization/81082
4385 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4386 association if it requires casting to unsigned.
4387 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4388 from fold_plusminus_mult_expr to catch important cases late when
4389 range info is available.
4391 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4393 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4394 * configure.ac (hidden_linkonce): New test.
4395 * configure: Regenerate.
4396 * config.in: Regenerate.
4398 2018-01-26 Julia Koval <julia.koval@intel.com>
4400 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4401 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4402 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4403 _mm_mask_bitshuffle_epi64_mask): Fix type.
4404 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4405 USI_FTYPE_V4DI_V4DI_USI): Remove.
4406 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4407 __builtin_ia32_vpshufbitqmb256_mask,
4408 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4409 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4410 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4412 2018-01-26 Alan Modra <amodra@gmail.com>
4415 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4416 UNSPEC_VBPERMQ. Sort other unspecs.
4418 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4420 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4422 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4425 * predict.c (drop_profile): Do not push/pop cfun; update also
4427 (handle_missing_profiles): Fix logic looking for zero profiles.
4429 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4432 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4433 on functions with #pragma omp declare simd or functions with simd
4435 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4436 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4437 Remove trailing \n from warning_at calls.
4439 2018-01-25 Tom de Vries <tom@codesourcery.com>
4442 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4443 for neutered workers.
4445 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4448 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4449 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4451 2018-01-24 Jeff Law <law@redhat.com>
4454 * i386.c (get_probe_interval): Move to earlier point.
4455 (ix86_compute_frame_layout): If -fstack-clash-protection and
4456 the frame is larger than the probe interval, then use pushes
4457 to save registers rather than reg->mem moves.
4458 (ix86_expand_prologue): Remove conditional for int_registers_saved
4461 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4464 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4465 min/max for never referenced object.
4467 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4470 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4472 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4473 attributes from DECL_ATTRIBUTES (decl) without affecting
4474 DECL_ATTRIBUTES (current_function_decl).
4475 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4476 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4478 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4480 PR tree-optimization/83979
4481 * fold-const.c (fold_comparison): Use constant_boolean_node
4482 instead of boolean_{true,false}_node.
4484 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4486 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4489 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4491 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4492 Simplify the clause that sets the length attribute.
4493 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4494 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4495 clause that sets the length attribute.
4496 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4498 2018-01-24 Tom de Vries <tom@codesourcery.com>
4501 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4502 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4503 Add strict parameter.
4504 (prevent_branch_around_nothing): Insert dummy insn between branch to
4505 label and label with no ptx insn inbetween.
4506 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4508 2018-01-24 Tom de Vries <tom@codesourcery.com>
4511 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4512 for neutered threads in warp.
4513 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4515 2018-01-24 Richard Biener <rguenther@suse.de>
4517 PR tree-optimization/83176
4518 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4521 2018-01-24 Richard Biener <rguenther@suse.de>
4523 PR tree-optimization/82819
4524 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4525 code generating pluses that are no-ops in the target precision.
4527 2018-01-24 Richard Biener <rguenther@suse.de>
4530 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4532 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4534 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4535 to merge probabilities.
4536 * predict.c (probably_never_executed): Also mark as cold functions
4537 with global 0 profile and guessed local profile.
4538 * profile-count.c (profile_probability::combine_with_count): New
4540 * profile-count.h (profile_probability::operator*,
4541 profile_probability::operator*=, profile_probability::operator/,
4542 profile_probability::operator/=): Reduce precision to adjusted
4543 and set value to guessed on contradictory divisions.
4544 (profile_probability::combine_with_freq): Remove.
4545 (profile_probability::combine_wiht_count): Declare.
4546 (profile_count::force_nonzero):: Set to adjusted.
4547 (profile_count::probability_in):: Set quality to adjusted.
4548 * tree-ssa-tail-merge.c (replace_block_by): Use
4551 2018-01-23 Andrew Waterman <andrew@sifive.com>
4552 Jim Wilson <jimw@sifive.com>
4554 * config/riscv/riscv.c (riscv_stack_boundary): New.
4555 (riscv_option_override): Set riscv_stack_boundary. Handle
4556 riscv_preferred_stack_boundary_arg.
4557 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4558 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4559 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4560 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4561 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4562 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4564 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4567 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4568 of struct ix86_frame.
4569 (ix86_expand_epilogue): Likewise. Add a local variable for
4570 the reg_save_offset field in struct ix86_frame.
4572 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4574 PR tree-optimization/82604
4575 * tree-loop-distribution.c (enum partition_kind): New enum item
4576 PKIND_PARTIAL_MEMSET.
4577 (partition_builtin_p): Support above new enum item.
4578 (generate_code_for_partition): Ditto.
4579 (compute_access_range): Differentiate cases that equality can be
4580 proven at all loops, the innermost loops or no loops.
4581 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4582 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4583 (finalize_partitions, distribute_loop): Don't fuse partition of
4584 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4585 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4588 2018-01-23 Martin Liska <mliska@suse.cz>
4590 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4591 order to ignore the predictor.
4592 (PRED_POLYMORPHIC_CALL): Likewise.
4593 (PRED_RECURSIVE_CALL): Likewise.
4595 2018-01-23 Martin Liska <mliska@suse.cz>
4597 * tree-profile.c (tree_profiling): Print function header to
4598 aware reader which function we are working on.
4599 * value-prof.c (gimple_find_values_to_profile): Do not print
4600 not interesting value histograms.
4602 2018-01-23 Martin Liska <mliska@suse.cz>
4604 * profile-count.h (enum profile_quality): Add
4605 profile_uninitialized as the first value. Do not number values
4606 as they are zero based.
4607 (profile_count::verify): Update sanity check.
4608 (profile_probability::verify): Likewise.
4610 2018-01-23 Nathan Sidwell <nathan@acm.org>
4612 * doc/invoke.texi (ffor-scope): Deprecate.
4614 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4616 PR tree-optimization/83510
4617 * domwalk.c (set_all_edges_as_executable): New function.
4618 (dom_walker::dom_walker): Convert bool param
4619 "skip_unreachable_blocks" to enum reachability. Move setup of
4620 edge flags to set_all_edges_as_executable and only do it when
4621 reachability is REACHABLE_BLOCKS.
4622 * domwalk.h (enum dom_walker::reachability): New enum.
4623 (dom_walker::dom_walker): Convert bool param
4624 "skip_unreachable_blocks" to enum reachability.
4625 (set_all_edges_as_executable): New decl.
4626 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4627 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4629 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4630 but converting true to REACHABLE_BLOCKS.
4631 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4633 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4634 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4635 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4637 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4638 if check_all_array_refs will be called.
4640 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4642 * tree.c (selftest::test_location_wrappers): Add more test
4645 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4647 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4648 (selftest::test_bit_in_range): Likewise.
4650 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4653 * doc/sourcebuild.texi (vect_float): Say that the selector
4654 only describes the situation when -funsafe-math-optimizations is on.
4655 (vect_float_strict): Document.
4657 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4659 PR tree-optimization/83965
4660 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4661 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4662 instead of checking only for a reduction.
4663 (vect_recog_widen_sum_pattern): Likewise.
4665 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4667 * predict.c (probably_never_executed): Only use precise profile info.
4668 (compute_function_frequency): Skip after inlining hack since we now
4669 have quality checking.
4671 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4673 * profile-count.h (profile_probability::very_unlikely,
4674 profile_probability::unlikely, profile_probability::even): Set
4675 precision to guessed.
4677 2018-01-23 Richard Biener <rguenther@suse.de>
4679 PR tree-optimization/83963
4680 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4681 Properly terminate dominator walk when crossing the exit edge not
4682 when visiting its source block.
4684 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4687 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4688 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4690 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4692 PR tree-optimization/83957
4693 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4694 semicolon after for body surrounded by braces.
4696 PR tree-optimization/83081
4697 * profile-count.h (profile_probability::split): New method.
4698 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4699 Use profile_probability::split.
4700 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4701 when splitting a single conditional jump into 2.
4703 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4705 PR tree-optimization/69452
4706 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4709 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4711 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4712 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4713 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4715 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4717 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4718 * config/rl78/rl78.md (movdi): New define_expand.
4719 * config/rl78/rl78.c (rl78_split_movdi): New function.
4721 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4724 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4726 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4727 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4728 128-bit to produce an UNSPEC move to get the double word with the
4729 signbit and then a shift directly to do signbit.
4730 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4731 implementation with a new version that just does either a direct
4732 move or a regular move. Move memory interface to separate insns.
4733 Move insns so they are next to the expander.
4734 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4735 with signbit move. Split big and little endian case.
4736 (signbit<mode>2_dm_mem_le): Likewise.
4737 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4738 (signbit<mode>2_dm2): Likewise.
4740 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4742 * config/rl78/rl78.md (anddi3): New define_expand.
4744 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4746 * config/rl78/rl78.md (umindi3): New define_expand.
4748 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4750 * config/rl78/rl78.md (smindi3): New define_expand.
4752 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4754 * config/rl78/rl78.md (smaxdi3): New define_expand.
4756 2018-01-22 Carl Love <cel@us.ibm.com>
4758 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4759 LVX_V1TI): Add macro expansion.
4760 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4761 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4762 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4763 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4764 Change check to determine if the instruction is a byte reversing
4765 entry. Fix typo in comment.
4766 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4767 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4768 Add def_builtin calls for new builtins.
4769 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4770 Add define_insn expansion.
4772 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4774 * config/rl78/rl78.md (umaxdi3): New define_expand.
4776 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4778 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4779 for non-QImode registers.
4781 2018-01-22 Richard Biener <rguenther@suse.de>
4783 PR tree-optimization/83963
4784 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4785 including the loop exit block.
4786 (scop_detection::merge_sese): Likewise.
4787 (scop_detection::add_scop): Do it here instead.
4789 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4791 * doc/sourcebuild.texi (arm_softfloat): Document.
4793 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4796 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4797 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4798 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4800 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4801 David Edelsohn <dje.gcc@gmail.com>
4804 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4805 Change "crset eq" to "crset 2".
4806 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4807 (*call_indirect_aix<mode>_nospec): Likewise.
4808 (*call_value_indirect_aix<mode>_nospec): Likewise.
4809 (*call_indirect_elfv2<mode>_nospec): Likewise.
4810 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4811 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4812 change assembly output from . to $.
4813 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4814 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4815 (*tablejump<mode>_internal1_nospec): Likewise.
4817 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4820 * config/sh/sh_optimize_sett_clrt.cc:
4821 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4823 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4825 PR tree-optimization/83940
4826 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4827 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4828 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4829 use it to pass back the definition type.
4830 (vect_check_store_rhs): Likewise rhs_dt_out.
4831 (vect_build_gather_load_calls): Add a mask_dt argument and use
4832 it instead of a call to vect_is_simple_use.
4833 (vectorizable_store): Update calls to vect_check_load_store_mask
4834 and vect_check_store_rhs. Use the dt returned by the latter instead
4835 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4836 instead of calls to vect_is_simple_use. Pass the scalar rather
4837 than the vector operand to vect_is_simple_use when handling
4838 second and subsequent copies of an rhs value.
4839 (vectorizable_load): Update calls to vect_check_load_store_mask
4840 and vect_build_gather_load_calls. Use the cached mask_dt and
4841 gs_info.offset_dt instead of calls to vect_is_simple_use.
4843 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4846 * tree-emutls.c: Include gimplify.h.
4847 (lower_emutls_2): New function.
4848 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4849 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4850 it before further processing.
4853 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4854 UINTVAL (trueop1) instead of INTVAL (op1).
4856 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4860 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4861 INCOMING_FRAME_SP_OFFSET if not defined.
4862 (scan_trace): Add ENTRY argument. If true and
4863 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4864 emit a note to adjust the CFA offset.
4865 (create_cfi_notes): Adjust scan_trace callers.
4866 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4867 INCOMING_FRAME_SP_OFFSET in the CIE.
4868 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4869 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4871 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4872 * doc/tm.texi: Regenerated.
4874 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4876 PR rtl-optimization/83147
4877 * lra-constraints.c (remove_inheritance_pseudos): Use
4878 lra_substitute_pseudo_within_insn.
4880 2018-01-19 Tom de Vries <tom@codesourcery.com>
4881 Cesar Philippidis <cesar@codesourcery.com>
4884 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4886 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4889 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4890 spaces for function labels.
4892 2018-01-19 Martin Liska <mliska@suse.cz>
4894 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4895 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4896 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4897 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4898 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4899 (PRED_CONST_RETURN): Change from 69 to 65.
4900 (PRED_NULL_RETURN): Change from 91 to 71.
4901 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4902 (PRED_LOOP_GUARD): Change from 66 to 73.
4904 2018-01-19 Martin Liska <mliska@suse.cz>
4906 * predict.c (predict_insn_def): Add new assert.
4907 (struct branch_predictor): Change type to signed integer.
4908 (test_prediction_value_range): Amend test to cover
4910 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4911 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4912 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4913 (PRED_LOOP_IV_COMPARE): Likewise.
4914 * predict.h (PROB_UNINITIALIZED): Define new constant.
4916 2018-01-19 Martin Liska <mliska@suse.cz>
4918 * predict.c (dump_prediction): Add new format for
4919 analyze_brprob.py script which is enabled with -details
4921 * profile-count.h (precise_p): New function.
4923 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4925 PR tree-optimization/83922
4926 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4927 there are no statements that need masking.
4928 (vect_active_double_reduction_p): New function.
4929 (vect_analyze_loop_operations): Use it when handling phis that
4930 are not in the loop header.
4932 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4934 PR tree-optimization/83914
4935 * tree-vect-loop.c (vectorizable_induction): Don't convert
4936 init_expr or apply the peeling adjustment for inductions
4937 that are nested within the vectorized loop.
4939 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4941 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4944 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4948 * function.h (gimplify_parameters): Add gimple_seq * argument.
4949 * function.c: Include gimple.h and options.h.
4950 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4951 for the added local temporaries if needed.
4952 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4953 if there are any parameter cleanups, wrap whole body into a
4954 try/finally with the cleanups.
4956 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4959 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4960 Use GET_MODE_CLASS for scalar floating point.
4962 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4966 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4967 Fix call of call_cgraph_insertion_hooks.
4969 2018-01-18 Martin Sebor <msebor@redhat.com>
4971 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4973 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4976 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4979 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4982 * common.opt: (-ffile-prefix-map): New option.
4983 * opts.c (common_handle_option): Defer it.
4984 * opts-global.c (handle_common_deferred_options): Handle it.
4985 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4986 * file-prefix-map.h: New file.
4987 (remap_debug_filename, add_debug_prefix_map): ...here.
4988 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4989 * final.c (debug_prefix_map, add_debug_prefix_map
4990 remap_debug_filename): Move to...
4991 * file-prefix-map.c: New file.
4992 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4993 generalize, get rid of alloca(), use strrchr() instead of strchr().
4994 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4995 Implement in terms of add_prefix_map().
4996 (remap_macro_filename, remap_debug_filename): Implement in term of
4998 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4999 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5000 * dbxout.c: Include file-prefix-map.h.
5001 * varasm.c: Likewise.
5002 * vmsdbgout.c: Likewise.
5003 * xcoffout.c: Likewise.
5004 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5005 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5006 * doc/invoke.texi (-ffile-prefix-map): Document.
5007 (-fdebug-prefix-map): Update description.
5009 2018-01-18 Martin Liska <mliska@suse.cz>
5011 * config/i386/i386.c (indirect_thunk_name): Document that also
5013 (output_indirect_thunk): Document why both instructions
5014 (pause and lfence) are generated.
5016 2018-01-18 Richard Biener <rguenther@suse.de>
5018 PR tree-optimization/83887
5019 * graphite-scop-detection.c
5020 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5021 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5022 (scop_detection::merge_sese): Re-implement with a flood-fill
5023 algorithm that properly finds a SESE region if it exists.
5025 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5028 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5029 pointer_diff optimizations use view_convert instead of convert.
5031 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5033 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5034 Generate different code for -mno-speculate-indirect-jumps.
5035 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5036 (*call_indirect_aix<mode>): Disable for
5037 -mno-speculate-indirect-jumps.
5038 (*call_indirect_aix<mode>_nospec): New define_insn.
5039 (*call_value_indirect_aix<mode>): Disable for
5040 -mno-speculate-indirect-jumps.
5041 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5042 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5043 -mno-speculate-indirect-jumps.
5044 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5046 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5048 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5049 long double type, set the flags for noting the default long double
5050 type, even if we don't pass or return a long double type.
5052 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5055 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5058 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5060 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5061 support for merge[hl].
5062 (fold_mergehl_helper): New helper function.
5063 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5064 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5065 (altivec_vmrglw_direct): Add xxmrglw insn.
5067 2018-01-17 Andrew Waterman <andrew@sifive.com>
5069 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5070 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5072 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5075 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5076 call the lto_location_cache before reading the
5077 DECL_SOURCE_LOCATION of the types.
5079 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5080 Richard Sandiford <richard.sandiford@linaro.org>
5082 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5083 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5084 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5085 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5086 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5088 * config/aarch64/constraints.md (aarch64_movti_operand):
5090 * config/aarch64/predicates.md (Uti): Add new constraint.
5092 2018-01-17 Carl Love <cel@us.ibm.com>
5094 * config/rs6000/vsx.md (define_expand xl_len_r,
5095 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5096 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5098 (define_expand, define_insn): Move the shift left from the
5099 define_insn to the define_expand for lxvl and stxvl instructions.
5100 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5101 and XL_LEN_R definitions to PURE.
5103 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5105 * config/i386/i386.c (indirect_thunk_name): Declare regno
5106 as unsigned int. Compare regno with INVALID_REGNUM.
5107 (output_indirect_thunk): Ditto.
5108 (output_indirect_thunk_function): Ditto.
5109 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5110 in the call to output_indirect_thunk_function.
5112 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5115 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5116 rather than the size of inner_type to determine the stack slot size
5117 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5119 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5122 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5125 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5127 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5128 endian Linux systems to optionally enable multilibs for selecting
5129 the long double type if the user configured an explicit type.
5130 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5131 have no long double multilibs if not defined.
5132 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5133 warn if the user used -mabi={ieee,ibm}longdouble and we built
5134 multilibs for long double.
5135 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5136 appropriate multilib option.
5137 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5139 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5140 for building long double multilibs.
5141 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5143 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5145 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5148 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5150 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5153 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5156 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5159 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5161 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5162 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5164 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5166 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5167 different rtl trees depending on TARGET_64BIT.
5168 (rs6000_gen_lvx): Likewise.
5170 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5172 * config/visium/visium.md (nop): Tweak comment.
5173 (hazard_nop): Likewise.
5175 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5177 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5178 -mspeculate-indirect-jumps.
5179 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5180 for -mno-speculate-indirect-jumps.
5181 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5182 (*call_value_indirect_elfv2<mode>): Disable for
5183 -mno-speculate-indirect-jumps.
5184 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5185 (indirect_jump): Emit different RTL for
5186 -mno-speculate-indirect-jumps.
5187 (*indirect_jump<mode>): Disable for
5188 -mno-speculate-indirect-jumps.
5189 (*indirect_jump<mode>_nospec): New define_insn.
5190 (tablejump): Emit different RTL for
5191 -mno-speculate-indirect-jumps.
5192 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5193 (tablejumpsi_nospec): New define_expand.
5194 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5195 (tablejumpdi_nospec): New define_expand.
5196 (*tablejump<mode>_internal1): Disable for
5197 -mno-speculate-indirect-jumps.
5198 (*tablejump<mode>_internal1_nospec): New define_insn.
5199 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5202 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5204 * caller-save.c (insert_save): Drop unnecessary parameter. All
5207 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5208 Richard Biener <rguenth@suse.de>
5211 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5212 return early, inline manually is_gimple_sizepos. Make sure if we
5213 call gimplify_expr we don't end up with a gimple constant.
5214 * tree.c (variably_modified_type_p): Don't return true for
5215 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5216 * gimplify.h (is_gimple_sizepos): Remove.
5218 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5220 PR tree-optimization/83857
5221 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5222 vectorizable_live_operation for pure SLP statements.
5223 (vectorizable_live_operation): Handle PHIs.
5225 2018-01-16 Richard Biener <rguenther@suse.de>
5227 PR tree-optimization/83867
5228 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5229 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5231 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5234 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5235 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5236 If off is not INTEGER_CST, issue a may not be aligned warning
5237 rather than isn't aligned. Use isn%'t rather than isn't.
5238 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5240 <case MULT_EXPR>: Improve the case when bottom and one of the
5241 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5242 operand, in that case check if the other operand is multiple of
5243 bottom divided by the INTEGER_CST operand.
5245 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5248 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5249 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5250 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5251 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5252 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5253 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5254 (pa_function_arg_size): New function.
5256 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5258 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5259 in a separate statement.
5261 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5263 PR tree-optimization/83847
5264 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5265 group gathers and scatters.
5267 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5269 PR rtl-optimization/86620
5270 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5272 PR rtl-optimization/83213
5273 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5274 to last if both are JUMP_INSNs.
5276 PR tree-optimization/83843
5277 * gimple-ssa-store-merging.c
5278 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5279 store_immediate_info for bswap/nop orig_stores.
5281 2018-01-15 Andrew Waterman <andrew@sifive.com>
5283 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5285 <UDIV>: Increase cost if !TARGET_DIV.
5287 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5289 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5290 (define_attr "cr_logical_3op"): New.
5291 (cceq_ior_compare): Adjust.
5292 (cceq_ior_compare_complement): Adjust.
5293 (*cceq_rev_compare): Adjust.
5294 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5295 (is_cracked_insn): Adjust.
5296 (insn_must_be_first_in_group): Adjust.
5297 * config/rs6000/40x.md: Adjust.
5298 * config/rs6000/440.md: Adjust.
5299 * config/rs6000/476.md: Adjust.
5300 * config/rs6000/601.md: Adjust.
5301 * config/rs6000/603.md: Adjust.
5302 * config/rs6000/6xx.md: Adjust.
5303 * config/rs6000/7450.md: Adjust.
5304 * config/rs6000/7xx.md: Adjust.
5305 * config/rs6000/8540.md: Adjust.
5306 * config/rs6000/cell.md: Adjust.
5307 * config/rs6000/e300c2c3.md: Adjust.
5308 * config/rs6000/e500mc.md: Adjust.
5309 * config/rs6000/e500mc64.md: Adjust.
5310 * config/rs6000/e5500.md: Adjust.
5311 * config/rs6000/e6500.md: Adjust.
5312 * config/rs6000/mpc.md: Adjust.
5313 * config/rs6000/power4.md: Adjust.
5314 * config/rs6000/power5.md: Adjust.
5315 * config/rs6000/power6.md: Adjust.
5316 * config/rs6000/power7.md: Adjust.
5317 * config/rs6000/power8.md: Adjust.
5318 * config/rs6000/power9.md: Adjust.
5319 * config/rs6000/rs64.md: Adjust.
5320 * config/rs6000/titan.md: Adjust.
5322 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5324 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5325 ix86_indirect_branch_register logic.
5327 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5329 * config/i386/constraints.md (Bs): Update
5330 ix86_indirect_branch_register check. Don't check
5331 ix86_indirect_branch_register with GOT_memory_operand.
5333 * config/i386/predicates.md (GOT_memory_operand): Don't check
5334 ix86_indirect_branch_register here.
5335 (GOT32_symbol_operand): Likewise.
5337 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5339 * config/i386/predicates.md (constant_call_address_operand):
5340 Rewrite ix86_indirect_branch_register logic.
5341 (sibcall_insn_operand): Likewise.
5343 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5345 * config/i386/constraints.md (Bs): Replace
5346 ix86_indirect_branch_thunk_register with
5347 ix86_indirect_branch_register.
5349 * config/i386/i386.md (indirect_jump): Likewise.
5350 (tablejump): Likewise.
5351 (*sibcall_memory): Likewise.
5352 (*sibcall_value_memory): Likewise.
5353 Peepholes of indirect call and jump via memory: Likewise.
5354 * config/i386/i386.opt: Likewise.
5355 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5356 (GOT_memory_operand): Likewise.
5357 (call_insn_operand): Likewise.
5358 (sibcall_insn_operand): Likewise.
5359 (GOT32_symbol_operand): Likewise.
5361 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5364 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5365 type rather than type addr's type points to.
5366 (expand_omp_atomic_mutex): Likewise.
5367 (expand_omp_atomic): Likewise.
5369 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5372 * config/i386/i386.c (output_indirect_thunk_function): Use
5373 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5374 for __x86_return_thunk.
5376 2018-01-15 Richard Biener <rguenther@suse.de>
5379 * expmed.c (extract_bit_field_1): Fix typo.
5381 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5384 * config/arm/iterators.md (VF): New mode iterator.
5385 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5386 Remove integer-related logic from pattern.
5387 (neon_vabd<mode>_3): Likewise.
5389 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5392 * common.opt (fstrict-overflow): No longer an alias.
5393 (fwrapv-pointer): New option.
5394 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5395 also for pointer types based on flag_wrapv_pointer.
5396 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5397 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5398 opts->x_flag_wrapv got set.
5399 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5400 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5401 POINTER_TYPE_OVERFLOW_UNDEFINED.
5402 * match.pd: Likewise in address comparison pattern.
5403 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5405 2018-01-15 Richard Biener <rguenther@suse.de>
5408 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5409 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5410 Reset type names to their identifier if their TYPE_DECL doesn't
5411 have linkage (and thus is used for ODR and devirt).
5412 (save_debug_info_for_decl): Remove.
5413 (save_debug_info_for_type): Likewise.
5414 (add_tree_to_fld_list): Adjust.
5415 * tree-pretty-print.c (dump_generic_node): Make dumping of
5416 type names more robust.
5418 2018-01-15 Richard Biener <rguenther@suse.de>
5420 * BASE-VER: Bump to 8.0.1.
5422 2018-01-14 Martin Sebor <msebor@redhat.com>
5425 * builtins.c (check_access): Avoid warning when the no-warning bit
5428 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5430 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5431 * ira-color (allocno_hard_regs_compare): Likewise.
5433 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5436 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5437 Use .pushsection/.popsection.
5439 2018-01-14 Martin Sebor <msebor@redhat.com>
5442 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5444 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5446 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5447 entry from extra_headers.
5448 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5449 extra_headers, make the list bitwise identical to the i?86-*-* one.
5451 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5453 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5454 -mcmodel=large with -mindirect-branch=thunk,
5455 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5456 -mfunction-return=thunk-extern.
5457 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5458 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5459 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5461 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5463 * config/i386/i386.c (print_reg): Print the name of the full
5464 integer register without '%'.
5465 (ix86_print_operand): Handle 'V'.
5466 * doc/extend.texi: Document 'V' modifier.
5468 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5470 * config/i386/constraints.md (Bs): Disallow memory operand for
5471 -mindirect-branch-register.
5473 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5474 (GOT_memory_operand): Likewise.
5475 (call_insn_operand): Likewise.
5476 (sibcall_insn_operand): Likewise.
5477 (GOT32_symbol_operand): Likewise.
5478 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5479 for -mindirect-branch-register.
5480 (tablejump): Likewise.
5481 (*sibcall_memory): Likewise.
5482 (*sibcall_value_memory): Likewise.
5483 Disallow peepholes of indirect call and jump via memory for
5484 -mindirect-branch-register.
5485 (*call_pop): Replace m with Bw.
5486 (*call_value_pop): Likewise.
5487 (*sibcall_pop_memory): Replace m with Bs.
5488 * config/i386/i386.opt (mindirect-branch-register): New option.
5489 * doc/invoke.texi: Document -mindirect-branch-register option.
5491 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5493 * config/i386/i386-protos.h (ix86_output_function_return): New.
5494 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5495 set function_return_type.
5496 (indirect_thunk_name): Add ret_p to indicate thunk for function
5498 (output_indirect_thunk_function): Pass false to
5499 indirect_thunk_name.
5500 (ix86_output_indirect_branch_via_reg): Likewise.
5501 (ix86_output_indirect_branch_via_push): Likewise.
5502 (output_indirect_thunk_function): Create alias for function
5503 return thunk if regno < 0.
5504 (ix86_output_function_return): New function.
5505 (ix86_handle_fndecl_attribute): Handle function_return.
5506 (ix86_attribute_table): Add function_return.
5507 * config/i386/i386.h (machine_function): Add
5508 function_return_type.
5509 * config/i386/i386.md (simple_return_internal): Use
5510 ix86_output_function_return.
5511 (simple_return_internal_long): Likewise.
5512 * config/i386/i386.opt (mfunction-return=): New option.
5513 (indirect_branch): Mention -mfunction-return=.
5514 * doc/extend.texi: Document function_return function attribute.
5515 * doc/invoke.texi: Document -mfunction-return= option.
5517 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5519 * config/i386/i386-opts.h (indirect_branch): New.
5520 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5521 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5522 with local indirect jump when converting indirect call and jump.
5523 (ix86_set_indirect_branch_type): New.
5524 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5525 (indirectlabelno): New.
5526 (indirect_thunk_needed): Likewise.
5527 (indirect_thunk_bnd_needed): Likewise.
5528 (indirect_thunks_used): Likewise.
5529 (indirect_thunks_bnd_used): Likewise.
5530 (INDIRECT_LABEL): Likewise.
5531 (indirect_thunk_name): Likewise.
5532 (output_indirect_thunk): Likewise.
5533 (output_indirect_thunk_function): Likewise.
5534 (ix86_output_indirect_branch_via_reg): Likewise.
5535 (ix86_output_indirect_branch_via_push): Likewise.
5536 (ix86_output_indirect_branch): Likewise.
5537 (ix86_output_indirect_jmp): Likewise.
5538 (ix86_code_end): Call output_indirect_thunk_function if needed.
5539 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5541 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5542 (ix86_attribute_table): Add indirect_branch.
5543 * config/i386/i386.h (machine_function): Add indirect_branch_type
5544 and has_local_indirect_jump.
5545 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5547 (tablejump): Likewise.
5548 (*indirect_jump): Use ix86_output_indirect_jmp.
5549 (*tablejump_1): Likewise.
5550 (simple_return_indirect_internal): Likewise.
5551 * config/i386/i386.opt (mindirect-branch=): New option.
5552 (indirect_branch): New.
5555 (thunk-inline): Likewise.
5556 (thunk-extern): Likewise.
5557 * doc/extend.texi: Document indirect_branch function attribute.
5558 * doc/invoke.texi: Document -mindirect-branch= option.
5560 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5563 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5565 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5567 * ipa-inline.c (want_inline_small_function_p): Return false if
5568 inlining has already failed with CIF_FINAL_ERROR.
5569 (update_caller_keys): Call want_inline_small_function_p before
5571 (update_callee_keys): Likewise.
5573 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5575 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5577 (rs6000_quadword_masked_address_p): Likewise.
5578 (quad_aligned_load_p): Likewise.
5579 (quad_aligned_store_p): Likewise.
5580 (const_load_sequence_p): Add comment to describe the outer-most loop.
5581 (mimic_memory_attributes_and_flags): New function.
5582 (rs6000_gen_stvx): Likewise.
5583 (replace_swapped_aligned_store): Likewise.
5584 (rs6000_gen_lvx): Likewise.
5585 (replace_swapped_aligned_load): Likewise.
5586 (replace_swapped_load_constant): Capitalize argument name in
5587 comment describing this function.
5588 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5589 and stores that access quad-word aligned addresses and replace
5590 with stvx or lvx instructions when appropriate.
5591 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5592 New function prototype.
5593 (rs6000_quadword_masked_address_p): Likewise.
5594 (rs6000_gen_lvx): Likewise.
5595 (rs6000_gen_stvx): Likewise.
5596 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5597 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5598 when memory address is aligned.
5599 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5600 this split to select lvx instruction when memory address is aligned.
5601 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5602 instruction when memory address is aligned.
5603 (*vsx_le_perm_load_v16qi): Likewise.
5604 (four unnamed splitters): Modify to select the stvx instruction
5605 when memory is aligned.
5607 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5609 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5610 which appears in the queue multiple times.
5612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5613 Alan Hayward <alan.hayward@arm.com>
5614 David Sherwood <david.sherwood@arm.com>
5616 * tree-vectorizer.h (vec_lower_bound): New structure.
5617 (_loop_vec_info): Add check_nonzero and lower_bounds.
5618 (LOOP_VINFO_CHECK_NONZERO): New macro.
5619 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5620 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5621 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5622 fields. Make seg_len the distance travelled, not including the
5624 (dr_direction_indicator): Declare.
5625 (dr_zero_step_indicator): Likewise.
5626 (dr_known_forward_stride_p): Likewise.
5627 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5629 (runtime_alias_check_p): Allow runtime alias checks with
5631 (operator ==): Compare access_size and align.
5632 (prune_runtime_alias_test_list): Rework for new distinction between
5633 the access_size and seg_len.
5634 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5636 (get_segment_min_max): New function.
5637 (create_intersect_range_checks): Use it.
5638 (dr_step_indicator): New function.
5639 (dr_direction_indicator): Likewise.
5640 (dr_zero_step_indicator): Likewise.
5641 (dr_known_forward_stride_p): Likewise.
5642 * tree-loop-distribution.c (data_ref_segment_size): Return
5643 DR_STEP * (niters - 1).
5644 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5646 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5647 (vect_preserves_scalar_order_p): New function, split out from...
5648 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5649 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5650 (vect_vfa_access_size): New function.
5651 (vect_vfa_align): Likewise.
5652 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5653 (dump_lower_bound): New function.
5654 (vect_check_lower_bound): Likewise.
5655 (vect_small_gap_p): Likewise.
5656 (vectorizable_with_step_bound_p): Likewise.
5657 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5658 depencies if the vectorization factor is 1. Convert the checks
5659 for nonzero steps into checks on the bounds of DR_STEP. Try using
5660 a bunds check for variable steps if the minimum required step is
5661 relatively small. Update calls to the dr_with_seg_len
5662 constructor and to vect_compile_time_alias.
5663 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5665 (vect_loop_versioning): Call it.
5666 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5668 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5671 Alan Hayward <alan.hayward@arm.com>
5672 David Sherwood <david.sherwood@arm.com>
5674 * doc/sourcebuild.texi (vect_scatter_store): Document.
5675 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5677 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5679 * genopinit.c (main): Add supports_vec_scatter_store and
5680 supports_vec_scatter_store_cached to target_optabs.
5681 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5682 IFN_MASK_SCATTER_STORE.
5683 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5685 * internal-fn.h (internal_store_fn_p): Declare.
5686 (internal_fn_stored_value_index): Likewise.
5687 * internal-fn.c (scatter_store_direct): New macro.
5688 (expand_scatter_store_optab_fn): New function.
5689 (direct_scatter_store_optab_supported_p): New macro.
5690 (internal_store_fn_p): New function.
5691 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5692 IFN_MASK_SCATTER_STORE.
5693 (internal_fn_mask_index): Likewise.
5694 (internal_fn_stored_value_index): New function.
5695 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5697 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5698 * optabs-query.c (supports_vec_scatter_store_p): New function.
5699 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5700 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5701 true for scatter stores.
5702 (vect_gather_scatter_fn_p): Handle scatter stores too.
5703 (vect_check_gather_scatter): Consider using scatter stores if
5704 supports_vec_scatter_store_p.
5705 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5707 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5708 internal_fn_stored_value_index.
5709 (check_load_store_masking): Handle scatter stores too.
5710 (vect_get_store_rhs): Make public.
5711 (vectorizable_call): Use internal_store_fn_p.
5712 (vectorizable_store): Handle scatter store internal functions.
5713 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5714 when deciding whether the end of the group has been reached.
5715 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5716 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5717 (mask_scatter_store<mode>): New insns.
5719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5720 Alan Hayward <alan.hayward@arm.com>
5721 David Sherwood <david.sherwood@arm.com>
5723 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5724 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5725 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5727 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5728 Use vect_truncate_gather_scatter_offset if we can't treat the
5729 operation as a normal gather load or scatter store.
5730 (get_group_load_store_type): Take the gather_scatter_info
5731 as argument. Try using a gather load or scatter store for
5732 single-element groups.
5733 (get_load_store_type): Update calls to get_group_load_store_type
5734 and vect_use_strided_gather_scatters_p.
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 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5741 optional tree argument.
5742 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5744 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5745 but continue to use the current value as a fallback.
5746 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5747 to compare the updates.
5748 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5749 (get_load_store_type): Use it when handling a strided access.
5750 (vect_get_strided_load_store_ops): New function.
5751 (vect_get_data_ptr_increment): Likewise.
5752 (vectorizable_load): Handle strided gather loads. Always pass
5753 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5756 Alan Hayward <alan.hayward@arm.com>
5757 David Sherwood <david.sherwood@arm.com>
5759 * doc/md.texi (gather_load@var{m}): Document.
5760 (mask_gather_load@var{m}): Likewise.
5761 * genopinit.c (main): Add supports_vec_gather_load and
5762 supports_vec_gather_load_cached to target_optabs.
5763 * optabs-tree.c (init_tree_optimization_optabs): Use
5764 ggc_cleared_alloc to allocate target_optabs.
5765 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5766 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5768 * internal-fn.h (internal_load_fn_p): Declare.
5769 (internal_gather_scatter_fn_p): Likewise.
5770 (internal_fn_mask_index): Likewise.
5771 (internal_gather_scatter_fn_supported_p): Likewise.
5772 * internal-fn.c (gather_load_direct): New macro.
5773 (expand_gather_load_optab_fn): New function.
5774 (direct_gather_load_optab_supported_p): New macro.
5775 (direct_internal_fn_optab): New function.
5776 (internal_load_fn_p): Likewise.
5777 (internal_gather_scatter_fn_p): Likewise.
5778 (internal_fn_mask_index): Likewise.
5779 (internal_gather_scatter_fn_supported_p): Likewise.
5780 * optabs-query.c (supports_at_least_one_mode_p): New function.
5781 (supports_vec_gather_load_p): Likewise.
5782 * optabs-query.h (supports_vec_gather_load_p): Declare.
5783 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5784 and memory_type field.
5785 (NUM_PATTERNS): Bump to 15.
5786 * tree-vect-data-refs.c: Include internal-fn.h.
5787 (vect_gather_scatter_fn_p): New function.
5788 (vect_describe_gather_scatter_call): Likewise.
5789 (vect_check_gather_scatter): Try using internal functions for
5790 gather loads. Recognize existing calls to a gather load function.
5791 (vect_analyze_data_refs): Consider using gather loads if
5792 supports_vec_gather_load_p.
5793 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5794 (vect_get_gather_scatter_offset_type): Likewise.
5795 (vect_convert_mask_for_vectype): Likewise.
5796 (vect_add_conversion_to_patterm): Likewise.
5797 (vect_try_gather_scatter_pattern): Likewise.
5798 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5799 (vect_vect_recog_func_ptrs): Add it.
5800 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5801 internal_fn_mask_index and internal_gather_scatter_fn_p.
5802 (check_load_store_masking): Take the gather_scatter_info as an
5803 argument and handle gather loads.
5804 (vect_get_gather_scatter_ops): New function.
5805 (vectorizable_call): Check internal_load_fn_p.
5806 (vectorizable_load): Likewise. Handle gather load internal
5808 (vectorizable_store): Update call to check_load_store_masking.
5809 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5810 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5811 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5812 (aarch64_gather_scale_operand_d): New predicates.
5813 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5814 (mask_gather_load<mode>): New insns.
5816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5817 Alan Hayward <alan.hayward@arm.com>
5818 David Sherwood <david.sherwood@arm.com>
5820 * optabs.def (fold_left_plus_optab): New optab.
5821 * doc/md.texi (fold_left_plus_@var{m}): Document.
5822 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5823 * internal-fn.c (fold_left_direct): Define.
5824 (expand_fold_left_optab_fn): Likewise.
5825 (direct_fold_left_optab_supported_p): Likewise.
5826 * fold-const-call.c (fold_const_fold_left): New function.
5827 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5828 * tree-parloops.c (valid_reduction_p): New function.
5829 (gather_scalar_reductions): Use it.
5830 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5831 (vect_finish_replace_stmt): Declare.
5832 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5833 (needs_fold_left_reduction_p): New function, split out from...
5834 (vect_is_simple_reduction): ...here. Accept reductions that
5835 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5836 (vect_force_simple_reduction): Also store the reduction type in
5837 the assignment's STMT_VINFO_REDUC_TYPE.
5838 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5839 (merge_with_identity): New function.
5840 (vect_expand_fold_left): Likewise.
5841 (vectorize_fold_left_reduction): Likewise.
5842 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5843 scalar phi in place for it. Check for target support and reject
5844 cases that would reassociate the operation. Defer the transform
5845 phase to vectorize_fold_left_reduction.
5846 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5847 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5848 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5852 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5853 call to ifc_temp_var.
5855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5856 Alan Hayward <alan.hayward@arm.com>
5857 David Sherwood <david.sherwood@arm.com>
5859 * target.def (legitimize_address_displacement): Take the original
5860 offset as a poly_int.
5861 * targhooks.h (default_legitimize_address_displacement): Update
5863 * targhooks.c (default_legitimize_address_displacement): Likewise.
5864 * doc/tm.texi: Regenerate.
5865 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5866 as an argument, moving assert of ad->disp == ad->disp_term to...
5867 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5868 Try calling targetm.legitimize_address_displacement before expanding
5869 the address rather than afterwards, and adjust for the new interface.
5870 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5871 Match the new hook interface. Handle SVE addresses.
5872 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5877 * Makefile.in (OBJS): Add early-remat.o.
5878 * target.def (select_early_remat_modes): New hook.
5879 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5880 * doc/tm.texi: Regenerate.
5881 * targhooks.h (default_select_early_remat_modes): Declare.
5882 * targhooks.c (default_select_early_remat_modes): New function.
5883 * timevar.def (TV_EARLY_REMAT): New timevar.
5884 * passes.def (pass_early_remat): New pass.
5885 * tree-pass.h (make_pass_early_remat): Declare.
5886 * early-remat.c: New file.
5887 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5889 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5892 Alan Hayward <alan.hayward@arm.com>
5893 David Sherwood <david.sherwood@arm.com>
5895 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5896 vfm1 with a bound_epilog parameter.
5897 (vect_do_peeling): Update calls accordingly, and move the prologue
5898 call earlier in the function. Treat the base bound_epilog as 0 for
5899 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5900 this base when peeling for gaps.
5901 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5902 with fully-masked loops.
5903 (vect_estimate_min_profitable_iters): Handle the single peeled
5904 iteration in that case.
5906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5907 Alan Hayward <alan.hayward@arm.com>
5908 David Sherwood <david.sherwood@arm.com>
5910 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5911 single-element interleaving even if the size is not a power of 2.
5912 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5913 accesses for single-element interleaving if the group size is
5916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5917 Alan Hayward <alan.hayward@arm.com>
5918 David Sherwood <david.sherwood@arm.com>
5920 * doc/md.texi (fold_extract_last_@var{m}): Document.
5921 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5922 * optabs.def (fold_extract_last_optab): New optab.
5923 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5924 * internal-fn.c (fold_extract_direct): New macro.
5925 (expand_fold_extract_optab_fn): Likewise.
5926 (direct_fold_extract_optab_supported_p): Likewise.
5927 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5928 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5929 EXTRACT_LAST_REDUCTION.
5930 (get_initial_def_for_reduction): Do not create an initial vector
5931 for EXTRACT_LAST_REDUCTION reductions.
5932 (vectorizable_reduction): Leave the scalar phi in place for
5933 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5934 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5935 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5936 transform phase to vectorizable_condition.
5937 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5939 (vect_finish_stmt_generation): ...here.
5940 (vect_finish_replace_stmt): New function.
5941 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5942 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5944 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5947 Alan Hayward <alan.hayward@arm.com>
5948 David Sherwood <david.sherwood@arm.com>
5950 * doc/md.texi (extract_last_@var{m}): Document.
5951 * optabs.def (extract_last_optab): New optab.
5952 * internal-fn.def (EXTRACT_LAST): New internal function.
5953 * internal-fn.c (cond_unary_direct): New macro.
5954 (expand_cond_unary_optab_fn): Likewise.
5955 (direct_cond_unary_optab_supported_p): Likewise.
5956 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5957 loops using EXTRACT_LAST.
5958 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5959 (extract_last_<mode>): ...this optab.
5960 (vec_extract<mode><Vel>): Update accordingly.
5962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5963 Alan Hayward <alan.hayward@arm.com>
5964 David Sherwood <david.sherwood@arm.com>
5966 * target.def (empty_mask_is_expensive): New hook.
5967 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5968 * doc/tm.texi: Regenerate.
5969 * targhooks.h (default_empty_mask_is_expensive): Declare.
5970 * targhooks.c (default_empty_mask_is_expensive): New function.
5971 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5972 if the target says that empty masks are expensive.
5973 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5975 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5978 Alan Hayward <alan.hayward@arm.com>
5979 David Sherwood <david.sherwood@arm.com>
5981 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5982 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5983 (vect_use_loop_mask_for_alignment_p): New function.
5984 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5985 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5986 niters_skip argument. Make sure that the first niters_skip elements
5987 of the first iteration are inactive.
5988 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5989 Update call to vect_set_loop_masks_directly.
5990 (get_misalign_in_elems): New function, split out from...
5991 (vect_gen_prolog_loop_niters): ...here.
5992 (vect_update_init_of_dr): Take a code argument that specifies whether
5993 the adjustment should be added or subtracted.
5994 (vect_update_init_of_drs): Likewise.
5995 (vect_prepare_for_masked_peels): New function.
5996 (vect_do_peeling): Skip prologue peeling if we're using a mask
5997 instead. Update call to vect_update_inits_of_drs.
5998 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6000 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6001 alignment. Do not include the number of peeled iterations in
6002 the minimum threshold in that case.
6003 (vectorizable_induction): Adjust the start value down by
6004 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6005 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6006 Take the number of skipped iterations into account when calculating
6008 * tree-vect-stmts.c (vect_gen_while_not): New function.
6010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6011 Alan Hayward <alan.hayward@arm.com>
6012 David Sherwood <david.sherwood@arm.com>
6014 * doc/sourcebuild.texi (vect_fully_masked): Document.
6015 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6017 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6019 (vect_analyze_loop_2): ...here. Don't check the vectorization
6020 factor against the number of loop iterations if the loop is
6023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6024 Alan Hayward <alan.hayward@arm.com>
6025 David Sherwood <david.sherwood@arm.com>
6027 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6028 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6029 (dump_groups): Update accordingly.
6030 (iv_use::mem_type): New member variable.
6031 (address_p): New function.
6032 (record_use): Add a mem_type argument and initialize the new
6034 (record_group_use): Add a mem_type argument. Use address_p.
6035 Remove obsolete null checks of base_object. Update call to record_use.
6036 (find_interesting_uses_op): Update call to record_group_use.
6037 (find_interesting_uses_cond): Likewise.
6038 (find_interesting_uses_address): Likewise.
6039 (get_mem_type_for_internal_fn): New function.
6040 (find_address_like_use): Likewise.
6041 (find_interesting_uses_stmt): Try find_address_like_use before
6042 calling find_interesting_uses_op.
6043 (addr_offset_valid_p): Use the iv mem_type field as the type
6044 of the addressed memory.
6045 (add_autoinc_candidates): Likewise.
6046 (get_address_cost): Likewise.
6047 (split_small_address_groups_p): Use address_p.
6048 (split_address_groups): Likewise.
6049 (add_iv_candidate_for_use): Likewise.
6050 (autoinc_possible_for_pair): Likewise.
6051 (rewrite_groups): Likewise.
6052 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6053 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6054 (get_alias_ptr_type_for_ptr_address): New function.
6055 (rewrite_use_address): Rewrite address uses in calls that were
6056 identified by find_address_like_use.
6058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6059 Alan Hayward <alan.hayward@arm.com>
6060 David Sherwood <david.sherwood@arm.com>
6062 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6064 * gimple-expr.h (is_gimple_addressable: Likewise.
6065 * gimple-expr.c (is_gimple_address): Likewise.
6066 * internal-fn.c (expand_call_mem_ref): New function.
6067 (expand_mask_load_optab_fn): Use it.
6068 (expand_mask_store_optab_fn): Likewise.
6070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6071 Alan Hayward <alan.hayward@arm.com>
6072 David Sherwood <david.sherwood@arm.com>
6074 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6075 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6076 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6077 (cond_umax@var{mode}): Document.
6078 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6079 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6080 (cond_umin_optab, cond_umax_optab): New optabs.
6081 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6082 (COND_IOR, COND_XOR): New internal functions.
6083 * internal-fn.h (get_conditional_internal_fn): Declare.
6084 * internal-fn.c (cond_binary_direct): New macro.
6085 (expand_cond_binary_optab_fn): Likewise.
6086 (direct_cond_binary_optab_supported_p): Likewise.
6087 (get_conditional_internal_fn): New function.
6088 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6089 Cope with reduction statements that are vectorized as calls rather
6091 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6092 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6093 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6094 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6095 (UNSPEC_COND_EOR): New unspecs.
6096 (optab): Add mappings for them.
6097 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6098 (sve_int_op, sve_fp_op): New int attributes.
6100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6101 Alan Hayward <alan.hayward@arm.com>
6102 David Sherwood <david.sherwood@arm.com>
6104 * optabs.def (while_ult_optab): New optab.
6105 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6106 * internal-fn.def (WHILE_ULT): New internal function.
6107 * internal-fn.h (direct_internal_fn_supported_p): New override
6108 that takes two types as argument.
6109 * internal-fn.c (while_direct): New macro.
6110 (expand_while_optab_fn): New function.
6111 (convert_optab_supported_p): Likewise.
6112 (direct_while_optab_supported_p): New macro.
6113 * wide-int.h (wi::udiv_ceil): New function.
6114 * tree-vectorizer.h (rgroup_masks): New structure.
6115 (vec_loop_masks): New typedef.
6116 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6118 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6119 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6120 (vect_max_vf): New function.
6121 (slpeel_make_loop_iterate_ntimes): Delete.
6122 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6123 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6124 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6125 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6126 internal-fn.h, stor-layout.h and optabs-query.h.
6127 (vect_set_loop_mask): New function.
6128 (add_preheader_seq): Likewise.
6129 (add_header_seq): Likewise.
6130 (interleave_supported_p): Likewise.
6131 (vect_maybe_permute_loop_masks): Likewise.
6132 (vect_set_loop_masks_directly): Likewise.
6133 (vect_set_loop_condition_masked): Likewise.
6134 (vect_set_loop_condition_unmasked): New function, split out from
6135 slpeel_make_loop_iterate_ntimes.
6136 (slpeel_make_loop_iterate_ntimes): Rename to..
6137 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6138 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6139 (vect_do_peeling): Update call accordingly.
6140 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6142 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6143 mask_compare_type, can_fully_mask_p and fully_masked_p.
6144 (release_vec_loop_masks): New function.
6145 (_loop_vec_info): Use it to free the loop masks.
6146 (can_produce_all_loop_masks_p): New function.
6147 (vect_get_max_nscalars_per_iter): Likewise.
6148 (vect_verify_full_masking): Likewise.
6149 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6150 retries, and free the mask rgroups before retrying. Check loop-wide
6151 reasons for disallowing fully-masked loops. Make the final decision
6152 about whether use a fully-masked loop or not.
6153 (vect_estimate_min_profitable_iters): Do not assume that peeling
6154 for the number of iterations will be needed for fully-masked loops.
6155 (vectorizable_reduction): Disable fully-masked loops.
6156 (vectorizable_live_operation): Likewise.
6157 (vect_halve_mask_nunits): New function.
6158 (vect_double_mask_nunits): Likewise.
6159 (vect_record_loop_mask): Likewise.
6160 (vect_get_loop_mask): Likewise.
6161 (vect_transform_loop): Handle the case in which the final loop
6162 iteration might handle a partial vector. Call vect_set_loop_condition
6163 instead of slpeel_make_loop_iterate_ntimes.
6164 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6165 (check_load_store_masking): New function.
6166 (prepare_load_store_mask): Likewise.
6167 (vectorizable_store): Handle fully-masked loops.
6168 (vectorizable_load): Likewise.
6169 (supportable_widening_operation): Use vect_halve_mask_nunits for
6171 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6172 (vect_gen_while): New function.
6173 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6174 (aarch64_uqdec<mode>): New insn.
6176 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6177 Alan Hayward <alan.hayward@arm.com>
6178 David Sherwood <david.sherwood@arm.com>
6180 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6181 (reduc_xor_scal_optab): New optabs.
6182 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6183 (reduc_xor_scal_@var{m}): Document.
6184 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6185 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6187 * fold-const-call.c (fold_const_call): Handle them.
6188 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6189 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6190 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6191 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6192 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6193 (UNSPEC_XORV): New unspecs.
6194 (optab): Add entries for them.
6195 (BITWISEV): New int iterator.
6196 (bit_reduc_op): New int attributes.
6198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6199 Alan Hayward <alan.hayward@arm.com>
6200 David Sherwood <david.sherwood@arm.com>
6202 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6203 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6204 * optabs.def (vec_shl_insert_optab): New optab.
6205 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6206 (duplicate_and_interleave): Likewise.
6207 * tree-vect-loop.c: Include internal-fn.h.
6208 (neutral_op_for_slp_reduction): New function, split out from
6209 get_initial_defs_for_reduction.
6210 (get_initial_def_for_reduction): Handle option 2 for variable-length
6211 vectors by loading the neutral value into a vector and then shifting
6212 the initial value into element 0.
6213 (get_initial_defs_for_reduction): Replace the code argument with
6214 the neutral value calculated by neutral_op_for_slp_reduction.
6215 Use gimple_build_vector for constant-length vectors.
6216 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6217 but the first group_size elements have a neutral value.
6218 Use duplicate_and_interleave otherwise.
6219 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6220 Update call to get_initial_defs_for_reduction. Handle SLP
6221 reductions for variable-length vectors by creating one vector
6222 result for each scalar result, with the elements associated
6223 with other scalar results stubbed out with the neutral value.
6224 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6225 Require IFN_VEC_SHL_INSERT for double reductions on
6226 variable-length vectors, or SLP reductions that have
6227 a neutral value. Require can_duplicate_and_interleave_p
6228 support for variable-length unchained SLP reductions if there
6229 is no neutral value, such as for MIN/MAX reductions. Also require
6230 the number of vector elements to be a multiple of the number of
6231 SLP statements when doing variable-length unchained SLP reductions.
6232 Update call to vect_create_epilog_for_reduction.
6233 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6234 and remove initial values.
6235 (duplicate_and_interleave): Make public.
6236 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6237 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6240 Alan Hayward <alan.hayward@arm.com>
6241 David Sherwood <david.sherwood@arm.com>
6243 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6244 (can_duplicate_and_interleave_p): New function.
6245 (vect_get_and_check_slp_defs): Take the vector of statements
6246 rather than just the current one. Remove excess parentheses.
6247 Restriction rejectinon of vect_constant_def and vect_external_def
6248 for variable-length vectors to boolean types, or types for which
6249 can_duplicate_and_interleave_p is false.
6250 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6251 (duplicate_and_interleave): New function.
6252 (vect_get_constant_vectors): Use gimple_build_vector for
6253 constant-length vectors and suitable variable-length constant
6254 vectors. Use duplicate_and_interleave for other variable-length
6255 vectors. Don't defer the update when inserting new statements.
6257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6258 Alan Hayward <alan.hayward@arm.com>
6259 David Sherwood <david.sherwood@arm.com>
6261 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6262 min_profitable_iters doesn't go negative.
6264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6265 Alan Hayward <alan.hayward@arm.com>
6266 David Sherwood <david.sherwood@arm.com>
6268 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6269 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6270 * optabs.def (vec_mask_load_lanes_optab): New optab.
6271 (vec_mask_store_lanes_optab): Likewise.
6272 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6273 (MASK_STORE_LANES): Likewise.
6274 * internal-fn.c (mask_load_lanes_direct): New macro.
6275 (mask_store_lanes_direct): Likewise.
6276 (expand_mask_load_optab_fn): Handle masked operations.
6277 (expand_mask_load_lanes_optab_fn): New macro.
6278 (expand_mask_store_optab_fn): Handle masked operations.
6279 (expand_mask_store_lanes_optab_fn): New macro.
6280 (direct_mask_load_lanes_optab_supported_p): Likewise.
6281 (direct_mask_store_lanes_optab_supported_p): Likewise.
6282 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6284 (vect_load_lanes_supported): Likewise.
6285 * tree-vect-data-refs.c (strip_conversion): New function.
6286 (can_group_stmts_p): Likewise.
6287 (vect_analyze_data_ref_accesses): Use it instead of checking
6288 for a pair of assignments.
6289 (vect_store_lanes_supported): Take a masked_p parameter.
6290 (vect_load_lanes_supported): Likewise.
6291 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6292 vect_store_lanes_supported and vect_load_lanes_supported.
6293 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6294 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6295 parameter. Don't allow gaps for masked accesses.
6296 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6297 and vect_load_lanes_supported.
6298 (get_load_store_type): Take a masked_p parameter and update
6299 call to get_group_load_store_type.
6300 (vectorizable_store): Update call to get_load_store_type.
6301 Handle IFN_MASK_STORE_LANES.
6302 (vectorizable_load): Update call to get_load_store_type.
6303 Handle IFN_MASK_LOAD_LANES.
6305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6306 Alan Hayward <alan.hayward@arm.com>
6307 David Sherwood <david.sherwood@arm.com>
6309 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6311 * config/aarch64/aarch64-protos.h
6312 (aarch64_sve_struct_memory_operand_p): Declare.
6313 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6314 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6315 (VPRED, vpred): Handle SVE structure modes.
6316 * config/aarch64/constraints.md (Utx): New constraint.
6317 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6318 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6319 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6320 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6321 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6322 structure modes. Split into pieces after RA.
6323 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6324 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6326 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6327 SVE structure modes.
6328 (aarch64_classify_address): Likewise.
6329 (sizetochar): Move earlier in file.
6330 (aarch64_print_operand): Handle SVE register lists.
6331 (aarch64_array_mode): New function.
6332 (aarch64_sve_struct_memory_operand_p): Likewise.
6333 (TARGET_ARRAY_MODE): Redefine.
6335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6336 Alan Hayward <alan.hayward@arm.com>
6337 David Sherwood <david.sherwood@arm.com>
6339 * target.def (array_mode): New target hook.
6340 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6341 * doc/tm.texi: Regenerate.
6342 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6343 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6344 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6346 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6350 Alan Hayward <alan.hayward@arm.com>
6351 David Sherwood <david.sherwood@arm.com>
6353 * fold-const.c (fold_binary_loc): Check the argument types
6354 rather than the result type when testing for a vector operation.
6356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6358 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6359 * doc/tm.texi: Regenerate.
6361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6362 Alan Hayward <alan.hayward@arm.com>
6363 David Sherwood <david.sherwood@arm.com>
6365 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6366 (sve): Document new AArch64 extension.
6367 * doc/md.texi (w): Extend the description of the AArch64
6368 constraint to include SVE vectors.
6369 (Upl, Upa): Document new AArch64 predicate constraints.
6370 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6372 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6373 (msve-vector-bits=): New option.
6374 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6375 SVE when these are disabled.
6376 (sve): New extension.
6377 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6378 modes. Adjust their number of units based on aarch64_sve_vg.
6379 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6380 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6381 aarch64_addr_query_type.
6382 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6383 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6384 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6385 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6386 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6387 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6388 (aarch64_simd_imm_zero_p): Delete.
6389 (aarch64_check_zero_based_sve_index_immediate): Declare.
6390 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6391 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6392 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6393 (aarch64_sve_float_mul_immediate_p): Likewise.
6394 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6396 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6397 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6398 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6399 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6400 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6401 (aarch64_regmode_natural_size): Likewise.
6402 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6403 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6405 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6406 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6407 for VG and the SVE predicate registers.
6408 (V_ALIASES): Add a "z"-prefixed alias.
6409 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6410 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6411 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6412 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6413 (REG_CLASS_NAMES): Add entries for them.
6414 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6415 and the predicate registers.
6416 (aarch64_sve_vg): Declare.
6417 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6418 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6419 (REGMODE_NATURAL_SIZE): Define.
6420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6422 * config/aarch64/aarch64.c: Include cfgrtl.h.
6423 (simd_immediate_info): Add a constructor for series vectors,
6424 and an associated step field.
6425 (aarch64_sve_vg): New variable.
6426 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6427 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6428 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6429 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6430 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6431 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6432 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6433 (aarch64_get_mask_mode): New functions.
6434 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6435 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6436 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6437 predicate modes and predicate registers. Explicitly restrict
6438 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6439 to store a vector mode if it is recognized by
6440 aarch64_classify_vector_mode.
6441 (aarch64_regmode_natural_size): New function.
6442 (aarch64_hard_regno_caller_save_mode): Return the original mode
6444 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6445 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6446 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6447 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6449 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6450 does not overlap dest if the function is frame-related. Handle
6452 (aarch64_split_add_offset): New function.
6453 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6454 them aarch64_add_offset.
6455 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6456 and update call to aarch64_sub_sp.
6457 (aarch64_add_cfa_expression): New function.
6458 (aarch64_expand_prologue): Pass extra temporary registers to the
6459 functions above. Handle the case in which we need to emit new
6460 DW_CFA_expressions for registers that were originally saved
6461 relative to the stack pointer, but now have to be expressed
6462 relative to the frame pointer.
6463 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6465 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6466 IP0 and IP1 values for SVE frames.
6467 (aarch64_expand_vec_series): New function.
6468 (aarch64_expand_sve_widened_duplicate): Likewise.
6469 (aarch64_expand_sve_const_vector): Likewise.
6470 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6471 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6472 into the register, rather than emitting a SET directly.
6473 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6474 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6475 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6476 (offset_9bit_signed_scaled_p): New functions.
6477 (aarch64_replicate_bitmask_imm): New function.
6478 (aarch64_bitmask_imm): Use it.
6479 (aarch64_cannot_force_const_mem): Reject expressions involving
6480 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6481 (aarch64_classify_index): Handle SVE indices, by requiring
6482 a plain register index with a scale that matches the element size.
6483 (aarch64_classify_address): Handle SVE addresses. Assert that
6484 the mode of the address is VOIDmode or an integer mode.
6485 Update call to aarch64_classify_symbol.
6486 (aarch64_classify_symbolic_expression): Update call to
6487 aarch64_classify_symbol.
6488 (aarch64_const_vec_all_in_range_p): New function.
6489 (aarch64_print_vector_float_operand): Likewise.
6490 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6491 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6492 and the FP immediates 1.0 and 0.5.
6493 (aarch64_print_address_internal): Handle SVE addresses.
6494 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6495 (aarch64_regno_regclass): Handle predicate registers.
6496 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6498 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6499 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6500 (aarch64_convert_sve_vector_bits): New function.
6501 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6502 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6504 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6505 Handle SVE vector and predicate modes. Accept VL-based constants
6506 that need only one temporary register, and VL offsets that require
6507 no temporary registers.
6508 (aarch64_conditional_register_usage): Mark the predicate registers
6509 as fixed if SVE isn't available.
6510 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6511 Return true for SVE vector and predicate modes.
6512 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6513 rather than an unsigned int. Handle SVE modes.
6514 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6516 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6518 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6519 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6520 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6521 (aarch64_sve_float_mul_immediate_p): New functions.
6522 (aarch64_sve_valid_immediate): New function.
6523 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6524 Explicitly reject structure modes. Check for INDEX constants.
6525 Handle PTRUE and PFALSE constants.
6526 (aarch64_check_zero_based_sve_index_immediate): New function.
6527 (aarch64_simd_imm_zero_p): Delete.
6528 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6529 vector modes. Accept constants in the range of CNT[BHWD].
6530 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6531 ask for an Advanced SIMD mode.
6532 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6533 (aarch64_simd_vector_alignment): Handle SVE predicates.
6534 (aarch64_vectorize_preferred_vector_alignment): New function.
6535 (aarch64_simd_vector_alignment_reachable): Use it instead of
6537 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6538 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6540 (MAX_VECT_LEN): Delete.
6541 (expand_vec_perm_d): Add a vec_flags field.
6542 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6543 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6544 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6546 (aarch64_evpc_rev): Rename to...
6547 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6548 (aarch64_evpc_rev_global): New function.
6549 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6550 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6552 (aarch64_evpc_sve_tbl): New function.
6553 (aarch64_expand_vec_perm_const_1): Update after rename of
6554 aarch64_evpc_rev. Handle SVE permutes too, trying
6555 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6556 than aarch64_evpc_tbl.
6557 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6558 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6559 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6560 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6561 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6562 (aarch64_expand_sve_vcond): New functions.
6563 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6564 of aarch64_vector_mode_p.
6565 (aarch64_dwarf_poly_indeterminate_value): New function.
6566 (aarch64_compute_pressure_classes): Likewise.
6567 (aarch64_can_change_mode_class): Likewise.
6568 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6569 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6570 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6571 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6572 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6573 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6574 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6575 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6577 (Dn, Dl, Dr): Accept const as well as const_vector.
6578 (Dz): Likewise. Compare against CONST0_RTX.
6579 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6580 of "vector" where appropriate.
6581 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6582 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6583 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6584 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6585 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6586 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6587 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6588 (v_int_equiv): Extend to SVE modes.
6589 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6591 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6592 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6593 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6594 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6595 (SVE_COND_FP_CMP): New int iterators.
6596 (perm_hilo): Handle the new unpack unspecs.
6597 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6599 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6600 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6601 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6602 (aarch64_equality_operator, aarch64_constant_vector_operand)
6603 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6604 (aarch64_sve_nonimmediate_operand): Likewise.
6605 (aarch64_sve_general_operand): Likewise.
6606 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6607 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6608 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6609 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6610 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6611 (aarch64_sve_float_arith_immediate): Likewise.
6612 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6613 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6614 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6615 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6616 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6617 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6618 (aarch64_sve_float_arith_operand): Likewise.
6619 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6620 (aarch64_sve_float_mul_operand): Likewise.
6621 (aarch64_sve_vec_perm_operand): Likewise.
6622 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6623 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6624 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6625 as well as const_vector.
6626 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6627 in file. Use CONST0_RTX and CONSTM1_RTX.
6628 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6629 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6630 Use aarch64_simd_imm_zero.
6631 * config/aarch64/aarch64-sve.md: New file.
6632 * config/aarch64/aarch64.md: Include it.
6633 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6634 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6635 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6636 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6637 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6638 (sve): New attribute.
6639 (enabled): Disable instructions with the sve attribute unless
6641 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6642 aarch64_expand_mov_immediate.
6643 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6644 CNT[BHSD] immediates.
6645 (movti): Split CONST_POLY_INT moves into two halves.
6646 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6647 Split additions that need a temporary here if the destination
6648 is the stack pointer.
6649 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6650 (*add<mode>3_poly_1): New instruction.
6651 (set_clobber_cc): New expander.
6653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6655 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6656 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6657 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6658 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6659 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6660 Change innermode from fixed_mode_size to machine_mode.
6661 (simplify_subreg): Update call accordingly. Handle a constant-sized
6662 subreg of a variable-length CONST_VECTOR.
6664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6665 Alan Hayward <alan.hayward@arm.com>
6666 David Sherwood <david.sherwood@arm.com>
6668 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6669 (add_offset_to_base): New function, split out from...
6670 (create_mem_ref): ...here. When handling a scale other than 1,
6671 check first whether the address is valid without the offset.
6672 Add it into the base if so, leaving the index and scale as-is.
6674 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6677 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6678 fold_for_warn before checking if arg2 is INTEGER_CST.
6680 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6682 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6683 (store_multiple_operation): Delete.
6684 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6685 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6686 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6687 guarded by TARGET_STRING.
6688 (rs6000_output_load_multiple): Delete.
6689 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6690 OPTION_MASK_STRING / TARGET_STRING handling.
6691 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6692 (const rs6000_opt_masks) <"string">: Change mask to 0.
6693 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6694 (MASK_STRING): Delete.
6695 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6697 (load_multiple): Delete.
6704 (store_multiple): Delete.
6711 (movmemsi_8reg): Delete.
6712 (corresponding unnamed define_insn): Delete.
6713 (movmemsi_6reg): Delete.
6714 (corresponding unnamed define_insn): Delete.
6715 (movmemsi_4reg): Delete.
6716 (corresponding unnamed define_insn): Delete.
6717 (movmemsi_2reg): Delete.
6718 (corresponding unnamed define_insn): Delete.
6719 (movmemsi_1reg): Delete.
6720 (corresponding unnamed define_insn): Delete.
6721 * config/rs6000/rs6000.opt (mno-string): New.
6722 (mstring): Replace by deprecation warning stub.
6723 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6725 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6727 * regrename.c (regrename_do_replace): If replacing the same
6728 reg multiple times, try to reuse last created gen_raw_REG.
6731 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6732 main to workaround a bug in GDB.
6734 2018-01-12 Tom de Vries <tom@codesourcery.com>
6737 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6739 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6741 PR rtl-optimization/80481
6742 * ira-color.c (get_cap_member): New function.
6743 (allocnos_conflict_by_live_ranges_p): Use it.
6744 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6745 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6747 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6750 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6751 (*saddl_se_1): Ditto.
6753 (*ssubl_se_1): Ditto.
6755 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6757 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6758 rather than wi::to_widest for DR_INITs.
6759 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6760 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6761 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6763 (vect_analyze_group_access_1): Note that here.
6765 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6767 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6768 polynomial type sizes.
6770 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6772 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6773 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6774 (gimple_add_tmp_var): Likewise.
6776 2018-01-12 Martin Liska <mliska@suse.cz>
6778 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6779 (gimple_alloc_sizes): Likewise.
6780 (dump_gimple_statistics): Use PRIu64 in printf format.
6781 * gimple.h: Change uint64_t to int.
6783 2018-01-12 Martin Liska <mliska@suse.cz>
6785 * tree-core.h: Use uint64_t instead of int.
6786 * tree.c (tree_node_counts): Likewise.
6787 (tree_node_sizes): Likewise.
6788 (dump_tree_statistics): Use PRIu64 in printf format.
6790 2018-01-12 Martin Liska <mliska@suse.cz>
6792 * Makefile.in: As qsort_chk is implemented in vec.c, add
6793 vec.o to linkage of gencfn-macros.
6794 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6795 passing the info to record_node_allocation_statistics.
6796 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6798 * ggc-common.c (struct ggc_usage): Add operator== and use
6799 it in operator< and compare function.
6800 * mem-stats.h (struct mem_usage): Likewise.
6801 * vec.c (struct vec_usage): Remove operator< and compare
6802 function. Can be simply inherited.
6804 2018-01-12 Martin Jambor <mjambor@suse.cz>
6807 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6808 * tree-ssa-math-opts.c: Include domwalk.h.
6809 (convert_mult_to_fma_1): New function.
6810 (fma_transformation_info): New type.
6811 (fma_deferring_state): Likewise.
6812 (cancel_fma_deferring): New function.
6813 (result_of_phi): Likewise.
6814 (last_fma_candidate_feeds_initial_phi): Likewise.
6815 (convert_mult_to_fma): Added deferring logic, split actual
6816 transformation to convert_mult_to_fma_1.
6817 (math_opts_dom_walker): New type.
6818 (math_opts_dom_walker::after_dom_children): New method, body moved
6819 here from pass_optimize_widening_mul::execute, added deferring logic
6821 (pass_optimize_widening_mul::execute): Moved most of code to
6822 math_opts_dom_walker::after_dom_children.
6823 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6824 * config/i386/i386.c (ix86_option_override_internal): Added
6825 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6827 2018-01-12 Richard Biener <rguenther@suse.de>
6830 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6831 inline instance vars.
6833 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6836 * config/rx/rx.c (rx_is_restricted_memory_address):
6839 2018-01-12 Richard Biener <rguenther@suse.de>
6841 PR tree-optimization/80846
6842 * target.def (split_reduction): New target hook.
6843 * targhooks.c (default_split_reduction): New function.
6844 * targhooks.h (default_split_reduction): Declare.
6845 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6846 target requests first reduce vectors by combining low and high
6848 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6849 (get_vectype_for_scalar_type_and_size): Export.
6850 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6851 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6852 * doc/tm.texi: Regenerate.
6853 * config/i386/i386.c (ix86_split_reduction): Implement
6854 TARGET_VECTORIZE_SPLIT_REDUCTION.
6856 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6859 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6860 in PIC mode except for TARGET_VXWORKS_RTP.
6861 * config/sparc/sparc.c: Include cfgrtl.h.
6862 (TARGET_INIT_PIC_REG): Define.
6863 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6864 (sparc_pic_register_p): New predicate.
6865 (sparc_legitimate_address_p): Use it.
6866 (sparc_legitimize_pic_address): Likewise.
6867 (sparc_delegitimize_address): Likewise.
6868 (sparc_mode_dependent_address_p): Likewise.
6869 (gen_load_pcrel_sym): Remove 4th parameter.
6870 (load_got_register): Adjust call to above. Remove obsolete stuff.
6871 (sparc_expand_prologue): Do not call load_got_register here.
6872 (sparc_flat_expand_prologue): Likewise.
6873 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6874 (sparc_use_pseudo_pic_reg): New function.
6875 (sparc_init_pic_reg): Likewise.
6876 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6877 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6879 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6881 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6882 Add item for branch_cost.
6884 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6886 PR rtl-optimization/83565
6887 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6888 not extend the result to a larger mode for rotate operations.
6889 (num_sign_bit_copies1): Likewise.
6891 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6894 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6896 Use values-Xc.o for -pedantic.
6897 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6899 2018-01-12 Martin Liska <mliska@suse.cz>
6902 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6904 (possible_polymorphic_call_targets): Use it.
6905 (ipa_devirt): Likewise.
6907 2018-01-12 Martin Liska <mliska@suse.cz>
6909 * profile-count.h (enum profile_quality): Use 0 as invalid
6910 enum value of profile_quality.
6912 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6914 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6915 -mext-string options.
6917 2018-01-12 Richard Biener <rguenther@suse.de>
6919 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6920 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6921 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6923 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6925 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6927 * configure.ac (--with-long-double-format): Add support for the
6928 configuration option to change the default long double format on
6930 * config.gcc (powerpc*-linux*-*): Likewise.
6931 * configure: Regenerate.
6932 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6933 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6934 used without modification.
6936 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6938 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6939 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6940 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6941 MISC_BUILTIN_SPEC_BARRIER.
6942 (rs6000_init_builtins): Likewise.
6943 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6945 (speculation_barrier): New define_insn.
6946 * doc/extend.texi: Document __builtin_speculation_barrier.
6948 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6951 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6952 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6953 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6955 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6956 integral modes instead of "ss" and "sd".
6957 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6958 vectors with 32-bit and 64-bit elements.
6959 (vecdupssescalarmodesuffix): New mode attribute.
6960 (vec_dup<mode>): Use it.
6962 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6965 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6966 frame if argument is passed on stack.
6968 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6971 * ree.c (combine_reaching_defs): Optimize also
6972 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6973 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6975 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6978 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6980 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6983 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6984 after they are computed.
6986 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6988 PR tree-optimization/83695
6989 * gimple-loop-linterchange.cc
6990 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6991 reset cached scev information after interchange.
6992 (pass_linterchange::execute): Remove call to scev_reset_htab.
6994 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6996 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6997 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6998 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6999 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7000 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7001 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7002 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7003 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7004 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7005 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7006 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7007 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7008 (V_lane_reg): Likewise.
7009 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7011 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7012 (vfmal_lane_low<mode>_intrinsic,
7013 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7014 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7015 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7016 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7017 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7018 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7020 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7022 * config/arm/arm-cpus.in (fp16fml): New feature.
7023 (ALL_SIMD): Add fp16fml.
7024 (armv8.2-a): Add fp16fml as an option.
7025 (armv8.3-a): Likewise.
7026 (armv8.4-a): Add fp16fml as part of fp16.
7027 * config/arm/arm.h (TARGET_FP16FML): Define.
7028 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7030 * config/arm/arm-modes.def (V2HF): Define.
7031 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7032 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7033 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7034 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7035 vfmsl_low, vfmsl_high): New set of builtins.
7036 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7037 (vfml_op): New code attribute.
7038 (VFMLHALVES): New int iterator.
7039 (VFML, VFMLSEL): New mode attributes.
7040 (V_reg): Define mapping for V2HF.
7041 (V_hi, V_lo): New mode attributes.
7042 (VF_constraint): Likewise.
7043 (vfml_half, vfml_half_selector): New int attributes.
7044 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7046 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7047 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7049 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7050 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7051 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7052 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7054 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7055 Document new effective target and option set.
7057 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7059 * config/arm/arm-cpus.in (armv8_4): New feature.
7060 (ARMv8_4a): New fgroup.
7061 (armv8.4-a): New arch.
7062 * config/arm/arm-tables.opt: Regenerate.
7063 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7064 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7065 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7066 Add matching rules for -march=armv8.4-a and extensions.
7067 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7069 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7072 * config/rx/rx.md (BW): New mode attribute.
7073 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7075 2018-01-11 Richard Biener <rguenther@suse.de>
7077 PR tree-optimization/83435
7078 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7079 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7080 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7082 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7083 Alan Hayward <alan.hayward@arm.com>
7084 David Sherwood <david.sherwood@arm.com>
7086 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7088 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7089 (aarch64_print_address_internal): Use it to check for a zero offset.
7091 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7092 Alan Hayward <alan.hayward@arm.com>
7093 David Sherwood <david.sherwood@arm.com>
7095 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7096 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7097 Return a poly_int64 rather than a HOST_WIDE_INT.
7098 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7099 rather than a HOST_WIDE_INT.
7100 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7101 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7102 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7103 final_offset from HOST_WIDE_INT to poly_int64.
7104 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7105 to_constant when getting the number of units in an Advanced SIMD
7107 (aarch64_builtin_vectorized_function): Check for a constant number
7109 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7111 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7112 attribute instead of GET_MODE_NUNITS.
7113 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7114 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7115 GET_MODE_SIZE for fixed-size registers.
7116 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7117 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7118 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7119 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7120 (aarch64_print_operand, aarch64_print_address_internal)
7121 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7122 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7123 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7124 Handle polynomial GET_MODE_SIZE.
7125 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7126 wider than SImode without modification.
7127 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7128 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7129 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7130 passing and returning SVE modes.
7131 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7132 rather than GEN_INT.
7133 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7134 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7135 (aarch64_allocate_and_probe_stack_space): Likewise.
7136 (aarch64_layout_frame): Cope with polynomial offsets.
7137 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7138 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7140 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7141 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7142 poly_int64 rather than a HOST_WIDE_INT.
7143 (aarch64_get_separate_components, aarch64_process_components)
7144 (aarch64_expand_prologue, aarch64_expand_epilogue)
7145 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7146 (aarch64_anchor_offset): New function, split out from...
7147 (aarch64_legitimize_address): ...here.
7148 (aarch64_builtin_vectorization_cost): Handle polynomial
7149 TYPE_VECTOR_SUBPARTS.
7150 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7152 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7153 number of elements from the PARALLEL rather than the mode.
7154 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7155 rather than GET_MODE_BITSIZE.
7156 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7157 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7158 (aarch64_expand_vec_perm_const_1): Handle polynomial
7159 d->perm.length () and d->perm elements.
7160 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7161 Apply to_constant to d->perm elements.
7162 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7163 polynomial CONST_VECTOR_NUNITS.
7164 (aarch64_move_pointer): Take amount as a poly_int64 rather
7166 (aarch64_progress_pointer): Avoid temporary variable.
7167 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7168 the mode attribute instead of GET_MODE.
7170 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7171 Alan Hayward <alan.hayward@arm.com>
7172 David Sherwood <david.sherwood@arm.com>
7174 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7175 x exists before using it.
7176 (aarch64_add_constant_internal): Rename to...
7177 (aarch64_add_offset_1): ...this. Replace regnum with separate
7178 src and dest rtxes. Handle the case in which they're different,
7179 including when the offset is zero. Replace scratchreg with an rtx.
7180 Use 2 additions if there is no spare register into which we can
7181 move a 16-bit constant.
7182 (aarch64_add_constant): Delete.
7183 (aarch64_add_offset): Replace reg with separate src and dest
7184 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7185 Use aarch64_add_offset_1.
7186 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7187 an rtx rather than an int. Take the delta as a poly_int64
7188 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7189 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7190 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7191 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7192 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7194 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7195 aarch64_add_constant.
7197 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7199 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7200 Use scalar_float_mode.
7202 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7204 * config/aarch64/aarch64-simd.md
7205 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7206 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7207 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7208 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7209 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7210 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7211 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7212 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7213 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7214 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7216 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7219 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7220 targ_options->x_arm_arch_string is non NULL.
7222 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7224 * config/aarch64/aarch64.h
7225 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7227 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7230 * expmed.c (emit_store_flag_force): Swap if const op0
7231 and change VOIDmode to mode of op0.
7233 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7235 PR rtl-optimization/83761
7236 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7237 than bytes to mode_for_size.
7239 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7242 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7243 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7246 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7249 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7250 when in layout mode.
7251 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7252 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7255 2018-01-10 Michael Collison <michael.collison@arm.com>
7257 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7258 * config/aarch64/aarch64-option-extension.def: Add
7259 AARCH64_OPT_EXTENSION of 'fp16fml'.
7260 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7261 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7262 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7263 * config/aarch64/constraints.md (Ui7): New constraint.
7264 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7265 (VFMLA_SEL_W): Ditto.
7268 (VFMLA16_LOW): New int iterator.
7269 (VFMLA16_HIGH): Ditto.
7270 (UNSPEC_FMLAL): New unspec.
7271 (UNSPEC_FMLSL): Ditto.
7272 (UNSPEC_FMLAL2): Ditto.
7273 (UNSPEC_FMLSL2): Ditto.
7274 (f16mac): New code attribute.
7275 * config/aarch64/aarch64-simd-builtins.def
7276 (aarch64_fmlal_lowv2sf): Ditto.
7277 (aarch64_fmlsl_lowv2sf): Ditto.
7278 (aarch64_fmlalq_lowv4sf): Ditto.
7279 (aarch64_fmlslq_lowv4sf): Ditto.
7280 (aarch64_fmlal_highv2sf): Ditto.
7281 (aarch64_fmlsl_highv2sf): Ditto.
7282 (aarch64_fmlalq_highv4sf): Ditto.
7283 (aarch64_fmlslq_highv4sf): Ditto.
7284 (aarch64_fmlal_lane_lowv2sf): Ditto.
7285 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7286 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7287 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7288 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7289 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7290 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7291 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7292 (aarch64_fmlal_lane_highv2sf): Ditto.
7293 (aarch64_fmlsl_lane_highv2sf): Ditto.
7294 (aarch64_fmlal_laneq_highv2sf): Ditto.
7295 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7296 (aarch64_fmlalq_lane_highv4sf): Ditto.
7297 (aarch64_fmlsl_lane_highv4sf): Ditto.
7298 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7299 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7300 * config/aarch64/aarch64-simd.md:
7301 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7302 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7303 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7304 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7305 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7306 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7307 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7308 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7309 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7310 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7311 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7312 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7313 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7314 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7315 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7316 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7317 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7318 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7319 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7320 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7321 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7322 (vfmlsl_low_u32): Ditto.
7323 (vfmlalq_low_u32): Ditto.
7324 (vfmlslq_low_u32): Ditto.
7325 (vfmlal_high_u32): Ditto.
7326 (vfmlsl_high_u32): Ditto.
7327 (vfmlalq_high_u32): Ditto.
7328 (vfmlslq_high_u32): Ditto.
7329 (vfmlal_lane_low_u32): Ditto.
7330 (vfmlsl_lane_low_u32): Ditto.
7331 (vfmlal_laneq_low_u32): Ditto.
7332 (vfmlsl_laneq_low_u32): Ditto.
7333 (vfmlalq_lane_low_u32): Ditto.
7334 (vfmlslq_lane_low_u32): Ditto.
7335 (vfmlalq_laneq_low_u32): Ditto.
7336 (vfmlslq_laneq_low_u32): Ditto.
7337 (vfmlal_lane_high_u32): Ditto.
7338 (vfmlsl_lane_high_u32): Ditto.
7339 (vfmlal_laneq_high_u32): Ditto.
7340 (vfmlsl_laneq_high_u32): Ditto.
7341 (vfmlalq_lane_high_u32): Ditto.
7342 (vfmlslq_lane_high_u32): Ditto.
7343 (vfmlalq_laneq_high_u32): Ditto.
7344 (vfmlslq_laneq_high_u32): Ditto.
7345 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7346 (AARCH64_FL_FOR_ARCH8_4): New.
7347 (AARCH64_ISA_F16FML): New ISA flag.
7348 (TARGET_F16FML): New feature flag for fp16fml.
7349 (doc/invoke.texi): Document new fp16fml option.
7351 2018-01-10 Michael Collison <michael.collison@arm.com>
7353 * config/aarch64/aarch64-builtins.c:
7354 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7355 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7356 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7357 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7358 (AARCH64_ISA_SHA3): New ISA flag.
7359 (TARGET_SHA3): New feature flag for sha3.
7360 * config/aarch64/iterators.md (sha512_op): New int attribute.
7361 (CRYPTO_SHA512): New int iterator.
7362 (UNSPEC_SHA512H): New unspec.
7363 (UNSPEC_SHA512H2): Ditto.
7364 (UNSPEC_SHA512SU0): Ditto.
7365 (UNSPEC_SHA512SU1): Ditto.
7366 * config/aarch64/aarch64-simd-builtins.def
7367 (aarch64_crypto_sha512hqv2di): New builtin.
7368 (aarch64_crypto_sha512h2qv2di): Ditto.
7369 (aarch64_crypto_sha512su0qv2di): Ditto.
7370 (aarch64_crypto_sha512su1qv2di): Ditto.
7371 (aarch64_eor3qv8hi): Ditto.
7372 (aarch64_rax1qv2di): Ditto.
7373 (aarch64_xarqv2di): Ditto.
7374 (aarch64_bcaxqv8hi): Ditto.
7375 * config/aarch64/aarch64-simd.md:
7376 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7377 (aarch64_crypto_sha512su0qv2di): Ditto.
7378 (aarch64_crypto_sha512su1qv2di): Ditto.
7379 (aarch64_eor3qv8hi): Ditto.
7380 (aarch64_rax1qv2di): Ditto.
7381 (aarch64_xarqv2di): Ditto.
7382 (aarch64_bcaxqv8hi): Ditto.
7383 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7384 (vsha512h2q_u64): Ditto.
7385 (vsha512su0q_u64): Ditto.
7386 (vsha512su1q_u64): Ditto.
7387 (veor3q_u16): Ditto.
7388 (vrax1q_u64): Ditto.
7390 (vbcaxq_u16): Ditto.
7391 * config/arm/types.md (crypto_sha512): New type attribute.
7392 (crypto_sha3): Ditto.
7393 (doc/invoke.texi): Document new sha3 option.
7395 2018-01-10 Michael Collison <michael.collison@arm.com>
7397 * config/aarch64/aarch64-builtins.c:
7398 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7399 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7400 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7401 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7402 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7403 (AARCH64_ISA_SM4): New ISA flag.
7404 (TARGET_SM4): New feature flag for sm4.
7405 * config/aarch64/aarch64-simd-builtins.def
7406 (aarch64_sm3ss1qv4si): Ditto.
7407 (aarch64_sm3tt1aq4si): Ditto.
7408 (aarch64_sm3tt1bq4si): Ditto.
7409 (aarch64_sm3tt2aq4si): Ditto.
7410 (aarch64_sm3tt2bq4si): Ditto.
7411 (aarch64_sm3partw1qv4si): Ditto.
7412 (aarch64_sm3partw2qv4si): Ditto.
7413 (aarch64_sm4eqv4si): Ditto.
7414 (aarch64_sm4ekeyqv4si): Ditto.
7415 * config/aarch64/aarch64-simd.md:
7416 (aarch64_sm3ss1qv4si): Ditto.
7417 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7418 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7419 (aarch64_sm4eqv4si): Ditto.
7420 (aarch64_sm4ekeyqv4si): Ditto.
7421 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7422 (sm3part_op): Ditto.
7423 (CRYPTO_SM3TT): Ditto.
7424 (CRYPTO_SM3PART): Ditto.
7425 (UNSPEC_SM3SS1): New unspec.
7426 (UNSPEC_SM3TT1A): Ditto.
7427 (UNSPEC_SM3TT1B): Ditto.
7428 (UNSPEC_SM3TT2A): Ditto.
7429 (UNSPEC_SM3TT2B): Ditto.
7430 (UNSPEC_SM3PARTW1): Ditto.
7431 (UNSPEC_SM3PARTW2): Ditto.
7432 (UNSPEC_SM4E): Ditto.
7433 (UNSPEC_SM4EKEY): Ditto.
7434 * config/aarch64/constraints.md (Ui2): New constraint.
7435 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7436 * config/arm/types.md (crypto_sm3): New type attribute.
7437 (crypto_sm4): Ditto.
7438 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7439 (vsm3tt1aq_u32): Ditto.
7440 (vsm3tt1bq_u32): Ditto.
7441 (vsm3tt2aq_u32): Ditto.
7442 (vsm3tt2bq_u32): Ditto.
7443 (vsm3partw1q_u32): Ditto.
7444 (vsm3partw2q_u32): Ditto.
7445 (vsm4eq_u32): Ditto.
7446 (vsm4ekeyq_u32): Ditto.
7447 (doc/invoke.texi): Document new sm4 option.
7449 2018-01-10 Michael Collison <michael.collison@arm.com>
7451 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7452 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7453 (AARCH64_FL_FOR_ARCH8_4): New.
7454 (AARCH64_FL_V8_4): New flag.
7455 (doc/invoke.texi): Document new armv8.4-a option.
7457 2018-01-10 Michael Collison <michael.collison@arm.com>
7459 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7460 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7461 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7462 * config/aarch64/aarch64-option-extension.def: Add
7463 AARCH64_OPT_EXTENSION of 'sha2'.
7464 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7465 (crypto): Disable sha2 and aes if crypto disabled.
7466 (crypto): Enable aes and sha2 if enabled.
7467 (simd): Disable sha2 and aes if simd disabled.
7468 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7470 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7471 (TARGET_SHA2): New feature flag for sha2.
7472 (TARGET_AES): New feature flag for aes.
7473 * config/aarch64/aarch64-simd.md:
7474 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7475 conditional on TARGET_AES.
7476 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7477 (aarch64_crypto_sha1hsi): Make pattern conditional
7479 (aarch64_crypto_sha1hv4si): Ditto.
7480 (aarch64_be_crypto_sha1hv4si): Ditto.
7481 (aarch64_crypto_sha1su1v4si): Ditto.
7482 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7483 (aarch64_crypto_sha1su0v4si): Ditto.
7484 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7485 (aarch64_crypto_sha256su0v4si): Ditto.
7486 (aarch64_crypto_sha256su1v4si): Ditto.
7487 (doc/invoke.texi): Document new aes and sha2 options.
7489 2018-01-10 Martin Sebor <msebor@redhat.com>
7491 PR tree-optimization/83781
7492 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7495 2018-01-11 Martin Sebor <msebor@gmail.com>
7496 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7498 PR tree-optimization/83501
7499 PR tree-optimization/81703
7501 * tree-ssa-strlen.c (get_string_cst): Rename...
7502 (get_string_len): ...to this. Handle global constants.
7503 (handle_char_store): Adjust.
7505 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7506 Jim Wilson <jimw@sifive.com>
7508 * config/riscv/riscv-protos.h (riscv_output_return): New.
7509 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7510 (riscv_attribute_table, riscv_output_return),
7511 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7512 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7513 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7515 (riscv_expand_prologue): Add early return for naked function.
7516 (riscv_expand_epilogue): Likewise.
7517 (riscv_function_ok_for_sibcall): Return false for naked function.
7518 (riscv_set_current_function): New.
7519 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7520 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7521 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7522 * doc/extend.texi (RISC-V Function Attributes): New.
7524 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7526 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7527 check for 128-bit long double before checking TCmode.
7528 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7529 128-bit long doubles before checking TFmode or TCmode.
7530 (FLOAT128_IBM_P): Likewise.
7532 2018-01-10 Martin Sebor <msebor@redhat.com>
7534 PR tree-optimization/83671
7535 * builtins.c (c_strlen): Unconditionally return zero for the empty
7537 Use -Warray-bounds for warnings.
7538 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7539 for non-constant array indices with COMPONENT_REF, arrays of
7540 arrays, and pointers to arrays.
7541 (gimple_fold_builtin_strlen): Determine and set length range for
7542 non-constant character arrays.
7544 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7547 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7550 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7552 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7554 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7557 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7558 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7559 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7560 indexed_or_indirect_operand predicate.
7561 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7562 (*vsx_le_perm_load_v8hi): Likewise.
7563 (*vsx_le_perm_load_v16qi): Likewise.
7564 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7565 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7566 (*vsx_le_perm_store_v8hi): Likewise.
7567 (*vsx_le_perm_store_v16qi): Likewise.
7568 (eight unnamed splitters): Likewise.
7570 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7572 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7573 * config/rs6000/emmintrin.h: Likewise.
7574 * config/rs6000/mmintrin.h: Likewise.
7575 * config/rs6000/xmmintrin.h: Likewise.
7577 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7580 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7582 * tree.c (tree_nop_conversion): Return true for location wrapper
7584 (maybe_wrap_with_location): New function.
7585 (selftest::check_strip_nops): New function.
7586 (selftest::test_location_wrappers): New function.
7587 (selftest::tree_c_tests): Call it.
7588 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7589 (maybe_wrap_with_location): New decl.
7590 (EXPR_LOCATION_WRAPPER_P): New macro.
7591 (location_wrapper_p): New inline function.
7592 (tree_strip_any_location_wrapper): New inline function.
7594 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7597 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7598 stack_realign_offset for the largest alignment of stack slot
7600 (ix86_find_max_used_stack_alignment): New function.
7601 (ix86_finalize_stack_frame_flags): Use it. Set
7602 max_used_stack_alignment if we don't realign stack.
7603 * config/i386/i386.h (machine_function): Add
7604 max_used_stack_alignment.
7606 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7608 * config/arm/arm.opt (-mbranch-cost): New option.
7609 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7612 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7615 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7616 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7618 2018-01-10 Richard Biener <rguenther@suse.de>
7621 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7622 early out so it also covers the case where we have a non-NULL
7625 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7627 PR tree-optimization/83753
7628 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7629 for non-strided grouped accesses if the number of elements is 1.
7631 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7634 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7635 * i386.h (TARGET_USE_GATHER): Define.
7636 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7638 2018-01-10 Martin Liska <mliska@suse.cz>
7641 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7642 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7644 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7645 CLEANUP_NO_PARTITIONING is not set.
7647 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7649 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7650 for vectors, as a partial revert of r254296.
7651 * rtl.h (const_vec_p): Delete.
7652 (const_vec_duplicate_p): Don't test for vector CONSTs.
7653 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7654 * expmed.c (make_tree): Likewise.
7657 * common.md (E, F): Use CONSTANT_P instead of checking for
7659 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7660 checking for CONST_VECTOR.
7662 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7665 * predict.c (force_edge_cold): Handle in more sane way edges
7668 2018-01-09 Carl Love <cel@us.ibm.com>
7670 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7672 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7673 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7674 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7675 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7676 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7677 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7678 * config/rs6000/rs6000-protos.h: Add extern defition for
7679 rs6000_generate_float2_double_code.
7680 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7682 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7683 (float2_v2df): Add define_expand.
7685 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7688 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7689 op_mode in the force_to_mode call.
7691 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7693 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7694 instead of checking each element individually.
7695 (aarch64_evpc_uzp): Likewise.
7696 (aarch64_evpc_zip): Likewise.
7697 (aarch64_evpc_ext): Likewise.
7698 (aarch64_evpc_rev): Likewise.
7699 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7700 instead of checking each element individually. Return true without
7702 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7703 whether all selected elements come from the same input, instead of
7704 checking each element individually. Remove calls to gen_rtx_REG,
7705 start_sequence and end_sequence and instead assert that no rtl is
7708 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7710 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7711 order of HIGH and CONST checks.
7713 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7715 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7716 if the destination isn't an SSA_NAME.
7718 2018-01-09 Richard Biener <rguenther@suse.de>
7720 PR tree-optimization/83668
7721 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7723 (canonicalize_loop_form): ... here, renamed from ...
7724 (canonicalize_loop_closed_ssa_form): ... this and amended to
7725 swap successor edges for loop exit blocks to make us use
7726 the RPO order we need for initial schedule generation.
7728 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7730 PR tree-optimization/64811
7731 * match.pd: When optimizing comparisons with Inf, avoid
7732 introducing or losing exceptions from comparisons with NaN.
7734 2018-01-09 Martin Liska <mliska@suse.cz>
7737 * asan.c (shadow_mem_size): Add gcc_assert.
7739 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7741 Don't save registers in main().
7744 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7745 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7746 * config/avr/avr.c (avr_set_current_function): Don't error if
7747 naked, OS_task or OS_main are specified at the same time.
7748 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7750 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7752 * common/config/avr/avr-common.c (avr_option_optimization_table):
7753 Switch on -mmain-is-OS_task for optimizing compilations.
7755 2018-01-09 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/83572
7758 * graphite.c: Include cfganal.h.
7759 (graphite_transform_loops): Connect infinite loops to exit
7760 and remove fake edges at the end.
7762 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7764 * ipa-inline.c (edge_badness): Revert accidental checkin.
7766 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7769 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7770 symbols; not inline clones.
7772 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7775 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7776 hard registers. Formatting fixes.
7778 PR preprocessor/83722
7779 * gcc.c (try_generate_repro): Pass
7780 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7781 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7784 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7785 Kito Cheng <kito.cheng@gmail.com>
7787 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7788 (riscv_leaf_function_p): Delete.
7789 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7791 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7793 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7795 (do_ifelse): New function.
7796 (do_isel): New function.
7797 (do_sub3): New function.
7798 (do_add3): New function.
7799 (do_load_mask_compare): New function.
7800 (do_overlap_load_compare): New function.
7801 (expand_compare_loop): New function.
7802 (expand_block_compare): Call expand_compare_loop() when appropriate.
7803 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7805 (-mblock-compare-inline-loop-limit): New option.
7807 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7810 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7811 Reverse order of second and third operands in first alternative.
7812 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7813 of first and second elements in UNSPEC_VPERMR vector.
7814 (altivec_expand_vec_perm_le): Likewise.
7816 2018-01-08 Jeff Law <law@redhat.com>
7818 PR rtl-optimizatin/81308
7819 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7820 (process_switch): If group_case_labels makes a change, then set
7822 (pass_convert_switch::execute): If a switch is converted, then
7823 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7825 PR rtl-optimization/81308
7826 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7829 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7831 PR target/83663 - Revert r255946
7832 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7833 generation for cases where splatting a value is not useful.
7834 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7835 across a vec_duplicate and a paradoxical subreg forming a vector
7836 mode to a vec_concat.
7838 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7840 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7841 -march=armv8.3-a variants.
7842 * config/arm/t-multilib: Likewise.
7843 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7845 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7847 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7849 (cceq_ior_compare_complement): Give it a name so I can use it, and
7850 change boolean_or_operator predicate to boolean_operator so it can
7851 be used to generate a crand.
7852 (eqne): New code iterator.
7853 (bd/bd_neg): New code_attrs.
7854 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7855 a single define_insn.
7856 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7857 decrement (bdnzt/bdnzf/bdzt/bdzf).
7858 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7859 with the new names of the branch decrement patterns, and added the
7860 names of the branch decrement conditional patterns.
7862 2018-01-08 Richard Biener <rguenther@suse.de>
7864 PR tree-optimization/83563
7865 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7868 2018-01-08 Richard Biener <rguenther@suse.de>
7871 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7873 2018-01-08 Richard Biener <rguenther@suse.de>
7875 PR tree-optimization/83685
7876 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7877 references to abnormals.
7879 2018-01-08 Richard Biener <rguenther@suse.de>
7882 * dwarf2out.c (output_indirect_strings): Handle empty
7883 skeleton_debug_str_hash.
7884 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7886 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7888 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7889 (emit_store_direct): Likewise.
7890 (arc_trampoline_adjust_address): Likewise.
7891 (arc_asm_trampoline_template): New function.
7892 (arc_initialize_trampoline): Use asm_trampoline_template.
7893 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7894 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7895 * config/arc/arc.md (flush_icache): Delete pattern.
7897 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7899 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7900 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7903 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7906 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7907 by not USED_FOR_TARGET.
7908 (make_pass_resolve_sw_modes): Likewise.
7910 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7912 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7915 2018-01-08 Richard Biener <rguenther@suse.de>
7918 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7920 2018-01-08 Richard Biener <rguenther@suse.de>
7923 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7925 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7928 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7929 basic blocks with a small number of successors.
7930 (convert_control_dep_chain_into_preds): Improve handling of
7932 (dump_predicates): Split apart into...
7933 (dump_pred_chain): ...here...
7934 (dump_pred_info): ...and here.
7935 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7936 (can_chain_union_be_invalidated_p): Improve check for invalidation
7938 (uninit_uses_cannot_happen): Avoid unnecessary if
7939 convert_control_dep_chain_into_preds yielded nothing.
7941 2018-01-06 Martin Sebor <msebor@redhat.com>
7943 PR tree-optimization/83640
7944 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7945 subtracting negative offset from size.
7946 (builtin_access::overlap): Adjust offset bounds of the access to fall
7947 within the size of the object if possible.
7949 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7951 PR rtl-optimization/83699
7952 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7953 extract_bit_field_as_subreg to cases in which the extracted
7954 value is also a vector.
7956 * lra-constraints.c (process_alt_operands): Test for the equivalence
7957 substitutions when detecting a possible reload cycle.
7959 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7962 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7963 by default if flag_selective_schedling{,2}. Formatting fixes.
7965 PR rtl-optimization/83682
7966 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7967 if it has non-VECTOR_MODE element mode.
7968 (vec_duplicate_p): Likewise.
7971 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7972 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7974 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7977 * config/i386/i386-builtin.def
7978 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7979 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7980 Require also OPTION_MASK_ISA_AVX512F in addition to
7981 OPTION_MASK_ISA_GFNI.
7982 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7983 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7984 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7985 to OPTION_MASK_ISA_GFNI.
7986 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7987 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7988 OPTION_MASK_ISA_AVX512BW.
7989 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7990 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7991 addition to OPTION_MASK_ISA_GFNI.
7992 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7993 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7994 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7995 to OPTION_MASK_ISA_GFNI.
7996 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7997 a requirement for all ISAs rather than any of them with a few
7999 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8001 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8002 bitmasks to be enabled with 3 exceptions, instead of requiring any
8003 enabled ISA with lots of exceptions.
8004 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8005 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8006 Change avx512bw in isa attribute to avx512f.
8007 * config/i386/sgxintrin.h: Add license boilerplate.
8008 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8009 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8010 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8011 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8013 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8014 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8015 temporarily sse2 rather than sse if not enabled already.
8018 * config/i386/sse.md (VI248_VLBW): Rename to ...
8019 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8020 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8021 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8022 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8023 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8024 mode iterator instead of VI248_VLBW.
8026 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8028 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8029 (record_modified): Skip clobbers; add debug output.
8030 (param_change_prob): Use sreal frequencies.
8032 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8034 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8035 punt for user-aligned variables.
8037 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8039 * tree-chrec.c (chrec_contains_symbols): Return true for
8042 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8045 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8046 of (x|y) == x for BICS pattern.
8048 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8050 PR tree-optimization/83605
8051 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8052 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8055 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8057 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8058 * config/epiphany/rtems.h: New file.
8060 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8061 Uros Bizjak <ubizjak@gmail.com>
8064 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8065 QIreg_operand instead of register_operand predicate.
8066 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8067 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8068 comments instead of -fmitigate[-_]rop.
8070 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8073 * cgraphunit.c (symbol_table::compile): Switch to text_section
8074 before calling assembly_start debug hook.
8075 * run-rtl-passes.c (run_rtl_passes): Likewise.
8078 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8080 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8081 range_int_cst_p rather than !symbolic_range_p before calling
8082 extract_range_from_multiplicative_op_1.
8084 2018-01-04 Jeff Law <law@redhat.com>
8086 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8087 redundant test in assertion.
8089 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8091 * doc/rtl.texi: Document machine_mode wrapper classes.
8093 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8095 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8098 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8100 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8101 the VEC_PERM_EXPR fold to fail.
8103 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8106 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8107 to switched_sections.
8109 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8112 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8115 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8118 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8119 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8121 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8124 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8125 is BLKmode and bitpos not zero or mode change is needed.
8127 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8130 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8133 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8136 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8137 instead of MULT rtx. Update all corresponding splitters.
8139 (*ssub<modesuffix>): Ditto.
8141 (*cmp_sadd_di): Update split patterns.
8142 (*cmp_sadd_si): Ditto.
8143 (*cmp_sadd_sidi): Ditto.
8144 (*cmp_ssub_di): Ditto.
8145 (*cmp_ssub_si): Ditto.
8146 (*cmp_ssub_sidi): Ditto.
8147 * config/alpha/predicates.md (const23_operand): New predicate.
8148 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8149 Look for ASHIFT, not MULT inner operand.
8150 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8152 2018-01-04 Martin Liska <mliska@suse.cz>
8154 PR gcov-profile/83669
8155 * gcov.c (output_intermediate_file): Add version to intermediate
8157 * doc/gcov.texi: Document new field 'version' in intermediate
8158 file format. Fix location of '-k' option of gcov command.
8160 2018-01-04 Martin Liska <mliska@suse.cz>
8163 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8165 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8167 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8169 2018-01-03 Martin Sebor <msebor@redhat.com>
8171 PR tree-optimization/83655
8172 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8173 checking calls with invalid arguments.
8175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8177 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8178 (vectorizable_mask_load_store): Delete.
8179 (vectorizable_call): Return false for masked loads and stores.
8180 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8181 instead of gimple_assign_rhs1.
8182 (vectorizable_load): Handle IFN_MASK_LOAD.
8183 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8187 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8189 (vectorizable_mask_load_store): ...here.
8190 (vectorizable_load): ...and here.
8192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8194 * tree-vect-stmts.c (vect_build_all_ones_mask)
8195 (vect_build_zero_merge_argument): New functions, split out from...
8196 (vectorizable_load): ...here.
8198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8200 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8202 (vectorizable_mask_load_store): ...here.
8203 (vectorizable_store): ...and here.
8205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8207 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8209 (vectorizable_mask_load_store): ...here.
8211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8213 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8214 (vect_model_store_cost): Take a vec_load_store_type instead of a
8216 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8217 (vect_model_store_cost): Take a vec_load_store_type instead of a
8219 (vectorizable_mask_load_store): Update accordingly.
8220 (vectorizable_store): Likewise.
8221 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8225 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8226 IFN_MASK_LOAD calls here rather than...
8227 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8230 Alan Hayward <alan.hayward@arm.com>
8231 David Sherwood <david.sherwood@arm.com>
8233 * expmed.c (extract_bit_field_1): For vector extracts,
8234 fall back to extract_bit_field_as_subreg if vec_extract
8237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8238 Alan Hayward <alan.hayward@arm.com>
8239 David Sherwood <david.sherwood@arm.com>
8241 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8242 they are variable or constant sized.
8243 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8244 slots for constant-sized data.
8246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8247 Alan Hayward <alan.hayward@arm.com>
8248 David Sherwood <david.sherwood@arm.com>
8250 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8251 handling COND_EXPRs with boolean comparisons, try to find a better
8252 basis for the mask type than the boolean itself.
8254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8256 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8257 is calculated and how it can be overridden.
8258 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8259 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8261 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8265 Alan Hayward <alan.hayward@arm.com>
8266 David Sherwood <david.sherwood@arm.com>
8268 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8269 Remove the mode argument.
8270 (aarch64_simd_valid_immediate): Remove the mode and inverse
8272 * config/aarch64/iterators.md (bitsize): New iterator.
8273 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8274 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8275 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8276 aarch64_simd_valid_immediate.
8277 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8278 (aarch64_reg_or_bic_imm): Likewise.
8279 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8280 with an insn_type enum and msl with a modifier_type enum.
8281 Replace element_width with a scalar_mode. Change the shift
8282 to unsigned int. Add constructors for scalar_float_mode and
8283 scalar_int_mode elements.
8284 (aarch64_vect_float_const_representable_p): Delete.
8285 (aarch64_can_const_movi_rtx_p)
8286 (aarch64_simd_scalar_immediate_valid_for_move)
8287 (aarch64_simd_make_constant): Update call to
8288 aarch64_simd_valid_immediate.
8289 (aarch64_advsimd_valid_immediate_hs): New function.
8290 (aarch64_advsimd_valid_immediate): Likewise.
8291 (aarch64_simd_valid_immediate): Remove mode and inverse
8292 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8293 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8294 and aarch64_float_const_representable_p on the result.
8295 (aarch64_output_simd_mov_immediate): Remove mode argument.
8296 Update call to aarch64_simd_valid_immediate and use of
8297 simd_immediate_info.
8298 (aarch64_output_scalar_simd_mov_immediate): Update call
8301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8302 Alan Hayward <alan.hayward@arm.com>
8303 David Sherwood <david.sherwood@arm.com>
8305 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8306 (mode_nunits): Likewise CONST_MODE_NUNITS.
8307 * machmode.def (ADJUST_NUNITS): Document.
8308 * genmodes.c (mode_data::need_nunits_adj): New field.
8309 (blank_mode): Update accordingly.
8310 (adj_nunits): New variable.
8311 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8313 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8314 listed in adj_nunits.
8315 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8316 listed in adj_nunits. Don't emit case statements for such modes.
8317 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8318 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8319 nothing if adj_nunits is nonnull.
8320 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8321 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8322 (emit_mode_fbit): Update use of print_maybe_const_decl.
8323 (emit_move_size): Likewise. Treat the array as non-const
8325 (emit_mode_adjustments): Handle adj_nunits.
8327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8329 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8330 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8331 (VECTOR_MODES): Use it.
8332 (make_vector_modes): Take the prefix as an argument.
8334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8335 Alan Hayward <alan.hayward@arm.com>
8336 David Sherwood <david.sherwood@arm.com>
8338 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8339 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8340 for MODE_VECTOR_BOOL.
8341 * machmode.def (VECTOR_BOOL_MODE): Document.
8342 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8343 (make_vector_bool_mode): New function.
8344 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8346 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8347 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8349 * stor-layout.c (int_mode_for_mode): Likewise.
8350 * tree.c (build_vector_type_for_mode): Likewise.
8351 * varasm.c (output_constant_pool_2): Likewise.
8352 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8353 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8354 for MODE_VECTOR_BOOL.
8355 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8356 of mode class checks.
8357 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8358 instead of a list of mode class checks.
8359 (expand_vector_scalar_condition): Likewise.
8360 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8363 Alan Hayward <alan.hayward@arm.com>
8364 David Sherwood <david.sherwood@arm.com>
8366 * machmode.h (mode_size): Change from unsigned short to
8368 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8369 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8370 or if measurement_type is not polynomial.
8371 (fixed_size_mode::includes_p): Check for constant-sized modes.
8372 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8373 return a poly_uint16 rather than an unsigned short.
8374 (emit_mode_size): Change the type of mode_size from unsigned short
8375 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8376 (emit_mode_adjustments): Cope with polynomial vector sizes.
8377 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8379 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8381 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8382 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8383 * caller-save.c (setup_save_areas): Likewise.
8384 (replace_reg_with_saved_mem): Likewise.
8385 * calls.c (emit_library_call_value_1): Likewise.
8386 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8387 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8388 (gen_lowpart_for_combine): Likewise.
8389 * convert.c (convert_to_integer_1): Likewise.
8390 * cse.c (equiv_constant, cse_insn): Likewise.
8391 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8392 (cselib_subst_to_values): Likewise.
8393 * dce.c (word_dce_process_block): Likewise.
8394 * df-problems.c (df_word_lr_mark_ref): Likewise.
8395 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8396 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8397 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8398 (rtl_for_decl_location): Likewise.
8399 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8400 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8401 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8402 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8403 (expand_expr_real_1): Likewise.
8404 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8405 (pad_below): Likewise.
8406 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8407 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8408 * ira.c (get_subreg_tracking_sizes): Likewise.
8409 * ira-build.c (ira_create_allocno_objects): Likewise.
8410 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8411 (ira_sort_regnos_for_alter_reg): Likewise.
8412 * ira-costs.c (record_operand_costs): Likewise.
8413 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8414 (resolve_simple_move): Likewise.
8415 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8416 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8417 (lra_constraints): Likewise.
8418 (CONST_POOL_OK_P): Reject variable-sized modes.
8419 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8420 (add_pseudo_to_slot, lra_spill): Likewise.
8421 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8422 * optabs-query.c (get_best_extraction_insn): Likewise.
8423 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8424 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8425 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8426 * recog.c (offsettable_address_addr_space_p): Likewise.
8427 * regcprop.c (maybe_mode_change): Likewise.
8428 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8429 * regrename.c (build_def_use): Likewise.
8430 * regstat.c (dump_reg_info): Likewise.
8431 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8432 (find_reloads, find_reloads_subreg_address): Likewise.
8433 * reload1.c (eliminate_regs_1): Likewise.
8434 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8435 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8436 (simplify_binary_operation_1, simplify_subreg): Likewise.
8437 * targhooks.c (default_function_arg_padding): Likewise.
8438 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8439 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8440 (verify_gimple_assign_ternary): Likewise.
8441 * tree-inline.c (estimate_move_cost): Likewise.
8442 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8443 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8444 (get_address_cost_ainc): Likewise.
8445 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8446 (vect_supportable_dr_alignment): Likewise.
8447 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8448 (vectorizable_reduction): Likewise.
8449 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8450 (vectorizable_operation, vectorizable_load): Likewise.
8451 * tree.c (build_same_sized_truth_vector_type): Likewise.
8452 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8453 * var-tracking.c (emit_note_insn_var_location): Likewise.
8454 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8455 (ADDR_VEC_ALIGN): Likewise.
8457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8458 Alan Hayward <alan.hayward@arm.com>
8459 David Sherwood <david.sherwood@arm.com>
8461 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8463 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8464 or if measurement_type is polynomial.
8465 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8466 * combine.c (make_extraction): Likewise.
8467 * dse.c (find_shift_sequence): Likewise.
8468 * dwarf2out.c (mem_loc_descriptor): Likewise.
8469 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8470 (extract_bit_field, extract_low_bits): Likewise.
8471 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8472 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8473 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8474 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8475 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8476 * reload.c (find_reloads): Likewise.
8477 * reload1.c (alter_reg): Likewise.
8478 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8479 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8480 * tree-if-conv.c (predicate_mem_writes): Likewise.
8481 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8482 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8483 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8484 * valtrack.c (dead_debug_insert_temp): Likewise.
8485 * varasm.c (mergeable_constant_section): Likewise.
8486 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8489 Alan Hayward <alan.hayward@arm.com>
8490 David Sherwood <david.sherwood@arm.com>
8492 * expr.c (expand_assignment): Cope with polynomial mode sizes
8493 when assigning to a CONCAT.
8495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8496 Alan Hayward <alan.hayward@arm.com>
8497 David Sherwood <david.sherwood@arm.com>
8499 * machmode.h (mode_precision): Change from unsigned short to
8501 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8503 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8504 or if measurement_type is not polynomial.
8505 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8506 in which the mode is already known to be a scalar_int_mode.
8507 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8508 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8510 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8511 for GET_MODE_PRECISION.
8512 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8513 for GET_MODE_PRECISION.
8514 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8516 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8517 (expand_field_assignment, make_extraction): Likewise.
8518 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8519 (get_last_value): Likewise.
8520 * convert.c (convert_to_integer_1): Likewise.
8521 * cse.c (cse_insn): Likewise.
8522 * expr.c (expand_expr_real_1): Likewise.
8523 * lra-constraints.c (simplify_operand_subreg): Likewise.
8524 * optabs-query.c (can_atomic_load_p): Likewise.
8525 * optabs.c (expand_atomic_load): Likewise.
8526 (expand_atomic_store): Likewise.
8527 * ree.c (combine_reaching_defs): Likewise.
8528 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8529 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8530 * tree.h (type_has_mode_precision_p): Likewise.
8531 * ubsan.c (instrument_si_overflow): Likewise.
8533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8534 Alan Hayward <alan.hayward@arm.com>
8535 David Sherwood <david.sherwood@arm.com>
8537 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8538 polynomial numbers of units.
8539 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8540 (valid_vector_subparts_p): New function.
8541 (build_vector_type): Remove temporary shim and take the number
8542 of units as a poly_uint64 rather than an int.
8543 (build_opaque_vector_type): Take the number of units as a
8544 poly_uint64 rather than an int.
8545 * tree.c (build_vector_from_ctor): Handle polynomial
8546 TYPE_VECTOR_SUBPARTS.
8547 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8548 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8549 (build_vector_from_val): If the number of units is variable,
8550 use build_vec_duplicate_cst for constant operands and
8551 VEC_DUPLICATE_EXPR otherwise.
8552 (make_vector_type): Remove temporary is_constant ().
8553 (build_vector_type, build_opaque_vector_type): Take the number of
8554 units as a poly_uint64 rather than an int.
8555 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8557 * cfgexpand.c (expand_debug_expr): Likewise.
8558 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8559 (store_constructor, expand_expr_real_1): Likewise.
8560 (const_scalar_mask_from_tree): Likewise.
8561 * fold-const-call.c (fold_const_reduction): Likewise.
8562 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8563 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8564 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8565 (fold_relational_const): Likewise.
8566 (native_interpret_vector): Likewise. Change the size from an
8567 int to an unsigned int.
8568 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8569 TYPE_VECTOR_SUBPARTS.
8570 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8571 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8572 duplicating a non-constant operand into a variable-length vector.
8573 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8574 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8575 * ipa-icf.c (sem_variable::equals): Likewise.
8576 * match.pd: Likewise.
8577 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8578 * print-tree.c (print_node): Likewise.
8579 * stor-layout.c (layout_type): Likewise.
8580 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8581 * tree-cfg.c (verify_gimple_comparison): Likewise.
8582 (verify_gimple_assign_binary): Likewise.
8583 (verify_gimple_assign_ternary): Likewise.
8584 (verify_gimple_assign_single): Likewise.
8585 * tree-pretty-print.c (dump_generic_node): Likewise.
8586 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8587 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8588 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8589 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8590 (vect_shift_permute_load_chain): Likewise.
8591 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8592 (expand_vector_condition, optimize_vector_constructor): Likewise.
8593 (lower_vec_perm, get_compute_type): Likewise.
8594 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8595 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8596 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8597 (vect_recog_mask_conversion_pattern): Likewise.
8598 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8599 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8600 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8601 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8602 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8603 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8604 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8605 (supportable_widening_operation): Likewise.
8606 (supportable_narrowing_operation): Likewise.
8607 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8609 * varasm.c (output_constant): Likewise.
8611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8612 Alan Hayward <alan.hayward@arm.com>
8613 David Sherwood <david.sherwood@arm.com>
8615 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8616 so that both the length == 3 and length != 3 cases set up their
8617 own permute vectors. Add comments explaining why we know the
8618 number of elements is constant.
8619 (vect_permute_load_chain): Likewise.
8621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8622 Alan Hayward <alan.hayward@arm.com>
8623 David Sherwood <david.sherwood@arm.com>
8625 * machmode.h (mode_nunits): Change from unsigned char to
8627 (ONLY_FIXED_SIZE_MODES): New macro.
8628 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8629 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8630 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8632 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8633 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8634 or if measurement_type is not polynomial.
8635 * genmodes.c (ZERO_COEFFS): New macro.
8636 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8638 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8639 Use ZERO_COEFFS when emitting initializers.
8640 * data-streamer.h (bp_pack_poly_value): New function.
8641 (bp_unpack_poly_value): Likewise.
8642 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8643 for GET_MODE_NUNITS.
8644 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8645 for GET_MODE_NUNITS.
8646 * tree.c (make_vector_type): Remove temporary shim and make
8647 the real function take the number of units as a poly_uint64
8649 (build_vector_type_for_mode): Handle polynomial nunits.
8650 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8651 * emit-rtl.c (const_vec_series_p_1): Likewise.
8652 (gen_rtx_CONST_VECTOR): Likewise.
8653 * fold-const.c (test_vec_duplicate_folding): Likewise.
8654 * genrecog.c (validate_pattern): Likewise.
8655 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8656 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8657 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8658 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8659 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8660 * rtlanal.c (subreg_get_info): Likewise.
8661 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8662 (vect_grouped_load_supported): Likewise.
8663 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8664 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8665 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8666 (simplify_const_unary_operation, simplify_binary_operation_1)
8667 (simplify_const_binary_operation, simplify_ternary_operation)
8668 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8669 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8670 instead of CONST_VECTOR_NUNITS.
8671 * varasm.c (output_constant_pool_2): Likewise.
8672 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8673 explicit-encoded elements in the XVEC for variable-length vectors.
8675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8677 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8680 Alan Hayward <alan.hayward@arm.com>
8681 David Sherwood <david.sherwood@arm.com>
8683 * coretypes.h (fixed_size_mode): Declare.
8684 (fixed_size_mode_pod): New typedef.
8685 * builtins.h (target_builtins::x_apply_args_mode)
8686 (target_builtins::x_apply_result_mode): Change type to
8687 fixed_size_mode_pod.
8688 * builtins.c (apply_args_size, apply_result_size, result_vector)
8689 (expand_builtin_apply_args_1, expand_builtin_apply)
8690 (expand_builtin_return): Update accordingly.
8692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8694 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8695 * cselib.c (cselib_hash_rtx): Likewise.
8696 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8697 CONST_VECTOR encoding.
8699 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8700 Jeff Law <law@redhat.com>
8703 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8704 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8705 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8706 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8709 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8710 explicitly probe *sp in a noreturn function if there were any callee
8711 register saves or frame pointer is needed.
8713 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8716 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8717 BLKmode for ternary, binary or unary expressions.
8720 * var-tracking.c (delete_vta_debug_insn): New inline function.
8721 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8722 insns from get_insns () to NULL instead of each bb separately.
8723 Use delete_vta_debug_insn. No longer static.
8724 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8725 delete_vta_debug_insns callers.
8726 * rtl.h (delete_vta_debug_insns): Declare.
8727 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8728 instead of variable_tracking_main.
8730 2018-01-03 Martin Sebor <msebor@redhat.com>
8732 PR tree-optimization/83603
8733 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8734 arguments past the endof the argument list in functions declared
8735 without a prototype.
8736 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8737 Avoid checking when arguments are null.
8739 2018-01-03 Martin Sebor <msebor@redhat.com>
8742 * doc/extend.texi (attribute const): Fix a typo.
8743 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8744 issuing -Wsuggest-attribute for void functions.
8746 2018-01-03 Martin Sebor <msebor@redhat.com>
8748 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8749 offset_int::from instead of wide_int::to_shwi.
8750 (maybe_diag_overlap): Remove assertion.
8751 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8752 * gimple-ssa-sprintf.c (format_directive): Same.
8753 (parse_directive): Same.
8754 (sprintf_dom_walker::compute_format_length): Same.
8755 (try_substitute_return_value): Same.
8757 2018-01-03 Jeff Law <law@redhat.com>
8760 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8761 non-constant residual for zero at runtime and avoid probing in
8762 that case. Reorganize code for trailing problem to mirror handling
8765 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8767 PR tree-optimization/83501
8768 * tree-ssa-strlen.c (get_string_cst): New.
8769 (handle_char_store): Call get_string_cst.
8771 2018-01-03 Martin Liska <mliska@suse.cz>
8773 PR tree-optimization/83593
8774 * tree-ssa-strlen.c: Include tree-cfg.h.
8775 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8776 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8777 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8779 (strlen_dom_walker::before_dom_children): Call
8780 gimple_purge_dead_eh_edges. Dump tranformation with details
8782 (strlen_dom_walker::before_dom_children): Update call by adding
8783 new argument cleanup_eh.
8784 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8786 2018-01-03 Martin Liska <mliska@suse.cz>
8789 * cif-code.def (VARIADIC_THUNK): New enum value.
8790 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8793 2018-01-03 Jan Beulich <jbeulich@suse.com>
8795 * sse.md (mov<mode>_internal): Tighten condition for when to use
8796 vmovdqu<ssescalarsize> for TI and OI modes.
8798 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8800 Update copyright years.
8802 2018-01-03 Martin Liska <mliska@suse.cz>
8805 * ipa-visibility.c (function_and_variable_visibility): Skip
8806 functions with noipa attribure.
8808 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8810 * gcc.c (process_command): Update copyright notice dates.
8811 * gcov-dump.c (print_version): Ditto.
8812 * gcov.c (print_version): Ditto.
8813 * gcov-tool.c (print_version): Ditto.
8814 * gengtype.c (create_file): Ditto.
8815 * doc/cpp.texi: Bump @copying's copyright year.
8816 * doc/cppinternals.texi: Ditto.
8817 * doc/gcc.texi: Ditto.
8818 * doc/gccint.texi: Ditto.
8819 * doc/gcov.texi: Ditto.
8820 * doc/install.texi: Ditto.
8821 * doc/invoke.texi: Ditto.
8823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8825 * vector-builder.h (vector_builder::m_full_nelts): Change from
8826 unsigned int to poly_uint64.
8827 (vector_builder::full_nelts): Update prototype accordingly.
8828 (vector_builder::new_vector): Likewise.
8829 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8830 (vector_builder::operator ==): Likewise.
8831 (vector_builder::finalize): Likewise.
8832 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8833 Take the number of elements as a poly_uint64 rather than an
8835 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8836 from unsigned int to poly_uint64.
8837 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8838 (vec_perm_indices::new_vector): Likewise.
8839 (vec_perm_indices::length): Likewise.
8840 (vec_perm_indices::nelts_per_input): Likewise.
8841 (vec_perm_indices::input_nelts): Likewise.
8842 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8843 number of elements per input as a poly_uint64 rather than an
8844 unsigned int. Use the original encoding for variable-length
8845 vectors, rather than clamping each individual element.
8846 For the second and subsequent elements in each pattern,
8847 clamp the step and base before clamping their sum.
8848 (vec_perm_indices::series_p): Handle polynomial element counts.
8849 (vec_perm_indices::all_in_range_p): Likewise.
8850 (vec_perm_indices_to_tree): Likewise.
8851 (vec_perm_indices_to_rtx): Likewise.
8852 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8853 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8854 (tree_vector_builder::new_binary_operation): Handle polynomial
8855 element counts. Return false if we need to know the number
8856 of elements at compile time.
8857 * fold-const.c (fold_vec_perm): Punt if the number of elements
8858 isn't known at compile time.
8860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8862 * vec-perm-indices.h (vec_perm_builder): Change element type
8863 from HOST_WIDE_INT to poly_int64.
8864 (vec_perm_indices::element_type): Update accordingly.
8865 (vec_perm_indices::clamp): Handle polynomial element_types.
8866 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8867 (vec_perm_indices::all_in_range_p): Likewise.
8868 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8870 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8871 polynomial vec_perm_indices element types.
8872 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8873 * fold-const.c (fold_vec_perm): Likewise.
8874 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8875 * tree-vect-generic.c (lower_vec_perm): Likewise.
8876 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8877 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8878 element type to HOST_WIDE_INT.
8880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8881 Alan Hayward <alan.hayward@arm.com>
8882 David Sherwood <david.sherwood@arm.com>
8884 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8885 rather than an int. Use plus_constant.
8886 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8887 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8894 a HOST_WIDE_INT to a poly_int64.
8896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8897 Alan Hayward <alan.hayward@arm.com>
8898 David Sherwood <david.sherwood@arm.com>
8900 * calls.c (load_register_parameters): Cope with polynomial
8901 mode sizes. Require a constant size for BLKmode parameters
8902 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8903 forces a parameter to be padded at the lsb end in order to
8904 fill a complete number of words, require the parameter size
8905 to be ordered wrt UNITS_PER_WORD.
8907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8908 Alan Hayward <alan.hayward@arm.com>
8909 David Sherwood <david.sherwood@arm.com>
8911 * reload1.c (spill_stack_slot_width): Change element type
8912 from unsigned int to poly_uint64_pod.
8913 (alter_reg): Treat mode sizes as polynomial.
8915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8916 Alan Hayward <alan.hayward@arm.com>
8917 David Sherwood <david.sherwood@arm.com>
8919 * reload.c (complex_word_subreg_p): New function.
8920 (reload_inner_reg_of_subreg, push_reload): Use it.
8922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8923 Alan Hayward <alan.hayward@arm.com>
8924 David Sherwood <david.sherwood@arm.com>
8926 * lra-constraints.c (process_alt_operands): Reject matched
8927 operands whose sizes aren't ordered.
8928 (match_reload): Refer to this check here.
8930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8931 Alan Hayward <alan.hayward@arm.com>
8932 David Sherwood <david.sherwood@arm.com>
8934 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8935 that the mode size is in the set {1, 2, 4, 8, 16}.
8937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8938 Alan Hayward <alan.hayward@arm.com>
8939 David Sherwood <david.sherwood@arm.com>
8941 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8942 Use plus_constant instead of gen_rtx_PLUS.
8944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8945 Alan Hayward <alan.hayward@arm.com>
8946 David Sherwood <david.sherwood@arm.com>
8948 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8949 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8950 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8951 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8952 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8953 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8954 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8955 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8956 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8957 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8959 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8960 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8961 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8962 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8963 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8964 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8965 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8966 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8967 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8968 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8970 * expr.c (emit_move_resolve_push): Treat the input and result
8971 of PUSH_ROUNDING as a poly_int64.
8972 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8973 (emit_push_insn): Likewise.
8974 * lra-eliminations.c (mark_not_eliminable): Likewise.
8975 * recog.c (push_operand): Likewise.
8976 * reload1.c (elimination_effects): Likewise.
8977 * rtlanal.c (nonzero_bits1): Likewise.
8978 * calls.c (store_one_arg): Likewise. Require the padding to be
8979 known at compile time.
8981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8982 Alan Hayward <alan.hayward@arm.com>
8983 David Sherwood <david.sherwood@arm.com>
8985 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8986 Use plus_constant instead of gen_rtx_PLUS.
8988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8989 Alan Hayward <alan.hayward@arm.com>
8990 David Sherwood <david.sherwood@arm.com>
8992 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8996 Alan Hayward <alan.hayward@arm.com>
8997 David Sherwood <david.sherwood@arm.com>
8999 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9000 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9001 via stack temporaries. Treat the mode size as polynomial too.
9003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9004 Alan Hayward <alan.hayward@arm.com>
9005 David Sherwood <david.sherwood@arm.com>
9007 * expr.c (expand_expr_real_2): When handling conversions involving
9008 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9009 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9010 as a poly_uint64 too.
9012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9013 Alan Hayward <alan.hayward@arm.com>
9014 David Sherwood <david.sherwood@arm.com>
9016 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9019 Alan Hayward <alan.hayward@arm.com>
9020 David Sherwood <david.sherwood@arm.com>
9022 * combine.c (can_change_dest_mode): Handle polynomial
9023 REGMODE_NATURAL_SIZE.
9024 * expmed.c (store_bit_field_1): Likewise.
9025 * expr.c (store_constructor): Likewise.
9026 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9027 and polynomial REGMODE_NATURAL_SIZE.
9028 (gen_lowpart_common): Likewise.
9029 * reginfo.c (record_subregs_of_mode): Likewise.
9030 * rtlanal.c (read_modify_subreg_p): Likewise.
9032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9033 Alan Hayward <alan.hayward@arm.com>
9034 David Sherwood <david.sherwood@arm.com>
9036 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9037 numbers of elements.
9039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9040 Alan Hayward <alan.hayward@arm.com>
9041 David Sherwood <david.sherwood@arm.com>
9043 * match.pd: Cope with polynomial numbers of vector elements.
9045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9046 Alan Hayward <alan.hayward@arm.com>
9047 David Sherwood <david.sherwood@arm.com>
9049 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9050 in a POINTER_PLUS_EXPR.
9052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9053 Alan Hayward <alan.hayward@arm.com>
9054 David Sherwood <david.sherwood@arm.com>
9056 * omp-simd-clone.c (simd_clone_subparts): New function.
9057 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9058 (ipa_simd_modify_function_body): Likewise.
9060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9061 Alan Hayward <alan.hayward@arm.com>
9062 David Sherwood <david.sherwood@arm.com>
9064 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9065 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9066 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9067 (expand_vector_condition, vector_element): Likewise.
9068 (subparts_gt): New function.
9069 (get_compute_type): Use subparts_gt.
9070 (count_type_subparts): Delete.
9071 (expand_vector_operations_1): Use subparts_gt instead of
9072 count_type_subparts.
9074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9075 Alan Hayward <alan.hayward@arm.com>
9076 David Sherwood <david.sherwood@arm.com>
9078 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9079 (vect_compile_time_alias): ...this new function. Do the calculation
9080 on poly_ints rather than trees.
9081 (vect_prune_runtime_alias_test_list): Update call accordingly.
9083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9084 Alan Hayward <alan.hayward@arm.com>
9085 David Sherwood <david.sherwood@arm.com>
9087 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9089 (vect_schedule_slp_instance): Likewise.
9091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9092 Alan Hayward <alan.hayward@arm.com>
9093 David Sherwood <david.sherwood@arm.com>
9095 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9096 constant and extern definitions for variable-length vectors.
9097 (vect_get_constant_vectors): Note that the number of units
9098 is known to be constant.
9100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9101 Alan Hayward <alan.hayward@arm.com>
9102 David Sherwood <david.sherwood@arm.com>
9104 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9105 of units as polynomial. Choose between WIDE and NARROW based
9108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9109 Alan Hayward <alan.hayward@arm.com>
9110 David Sherwood <david.sherwood@arm.com>
9112 * tree-vect-stmts.c (simd_clone_subparts): New function.
9113 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9116 Alan Hayward <alan.hayward@arm.com>
9117 David Sherwood <david.sherwood@arm.com>
9119 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9120 vectors as polynomial. Use build_index_vector for
9123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9124 Alan Hayward <alan.hayward@arm.com>
9125 David Sherwood <david.sherwood@arm.com>
9127 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9128 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9129 for variable-length vectors.
9130 (vectorizable_mask_load_store): Treat the number of units as
9131 polynomial, asserting that it is constant if the condition has
9132 already been enforced.
9133 (vectorizable_store, vectorizable_load): Likewise.
9135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9136 Alan Hayward <alan.hayward@arm.com>
9137 David Sherwood <david.sherwood@arm.com>
9139 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9140 of units as polynomial. Punt if we can't tell at compile time
9141 which vector contains the final result.
9143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9144 Alan Hayward <alan.hayward@arm.com>
9145 David Sherwood <david.sherwood@arm.com>
9147 * tree-vect-loop.c (vectorizable_induction): Treat the number
9148 of units as polynomial. Punt on SLP inductions. Use an integer
9149 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9150 cast of such a series for variable-length floating-point
9153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9154 Alan Hayward <alan.hayward@arm.com>
9155 David Sherwood <david.sherwood@arm.com>
9157 * tree.h (build_index_vector): Declare.
9158 * tree.c (build_index_vector): New function.
9159 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9160 of units as polynomial, forcibly converting it to a constant if
9161 vectorizable_reduction has already enforced the condition.
9162 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9163 to create a {1,2,3,...} vector.
9164 (vectorizable_reduction): Treat the number of units as polynomial.
9165 Choose vectype_in based on the largest scalar element size rather
9166 than the smallest number of units. Enforce the restrictions
9169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9170 Alan Hayward <alan.hayward@arm.com>
9171 David Sherwood <david.sherwood@arm.com>
9173 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9174 number of units as polynomial.
9176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9177 Alan Hayward <alan.hayward@arm.com>
9178 David Sherwood <david.sherwood@arm.com>
9180 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9181 * target.def (autovectorize_vector_sizes): Return the vector sizes
9182 by pointer, using vector_sizes rather than a bitmask.
9183 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9184 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9185 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9187 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9188 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9189 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9190 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9191 * omp-general.c (omp_max_vf): Likewise.
9192 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9193 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9194 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9195 * tree-vect-slp.c (vect_slp_bb): Likewise.
9196 * doc/tm.texi: Regenerate.
9197 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9199 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9200 the vector size as a poly_uint64 rather than an unsigned int.
9201 (current_vector_size): Change from an unsigned int to a poly_uint64.
9202 (get_vectype_for_scalar_type): Update accordingly.
9203 * tree.h (build_truth_vector_type): Take the size and number of
9204 units as a poly_uint64 rather than an unsigned int.
9205 (build_vector_type): Add a temporary overload that takes
9206 the number of units as a poly_uint64 rather than an unsigned int.
9207 * tree.c (make_vector_type): Likewise.
9208 (build_truth_vector_type): Take the number of units as a poly_uint64
9209 rather than an unsigned int.
9211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9212 Alan Hayward <alan.hayward@arm.com>
9213 David Sherwood <david.sherwood@arm.com>
9215 * target.def (get_mask_mode): Take the number of units and length
9216 as poly_uint64s rather than unsigned ints.
9217 * targhooks.h (default_get_mask_mode): Update accordingly.
9218 * targhooks.c (default_get_mask_mode): Likewise.
9219 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9220 * doc/tm.texi: Regenerate.
9222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9223 Alan Hayward <alan.hayward@arm.com>
9224 David Sherwood <david.sherwood@arm.com>
9226 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9227 * omp-general.c (omp_max_vf): Likewise.
9228 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9229 (expand_omp_simd): Handle polynomial safelen.
9230 * omp-low.c (omplow_simd_context): Add a default constructor.
9231 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9232 (lower_rec_simd_input_clauses): Update accordingly.
9233 (lower_rec_input_clauses): Likewise.
9235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9236 Alan Hayward <alan.hayward@arm.com>
9237 David Sherwood <david.sherwood@arm.com>
9239 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9240 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9241 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9242 (vect_analyze_slp_cost): Likewise.
9243 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9244 (vect_model_load_cost): Likewise.
9246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9247 Alan Hayward <alan.hayward@arm.com>
9248 David Sherwood <david.sherwood@arm.com>
9250 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9251 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9252 from an unsigned int * to a poly_uint64_pod *.
9253 (calculate_unrolling_factor): New function.
9254 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9257 Alan Hayward <alan.hayward@arm.com>
9258 David Sherwood <david.sherwood@arm.com>
9260 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9261 from an unsigned int to a poly_uint64.
9262 (_loop_vec_info::slp_unrolling_factor): Likewise.
9263 (_loop_vec_info::vectorization_factor): Change from an int
9265 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9266 (vect_get_num_vectors): New function.
9267 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9268 (vect_get_num_copies): Use vect_get_num_vectors.
9269 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9270 to an unsigned int *.
9271 (vect_analyze_data_refs): Change min_vf from an int * to a
9273 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9274 than an unsigned HOST_WIDE_INT.
9275 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9276 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9277 to an unsigned int *.
9278 (vect_analyze_data_ref_dependences): Likewise.
9279 (vect_compute_data_ref_alignment): Handle polynomial vf.
9280 (vect_enhance_data_refs_alignment): Likewise.
9281 (vect_prune_runtime_alias_test_list): Likewise.
9282 (vect_shift_permute_load_chain): Likewise.
9283 (vect_supportable_dr_alignment): Likewise.
9284 (dependence_distance_ge_vf): Take the vectorization factor as a
9285 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9286 (vect_analyze_data_refs): Change min_vf from an int * to a
9288 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9289 vfm1 as a poly_uint64 rather than an int. Make the same change
9290 for the returned bound_scalar.
9291 (vect_gen_vector_loop_niters): Handle polynomial vf.
9292 (vect_do_peeling): Likewise. Update call to
9293 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9294 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9296 * tree-vect-loop.c (vect_determine_vectorization_factor)
9297 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9298 (vect_get_known_peeling_cost): Likewise.
9299 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9300 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9301 (vect_transform_loop): Likewise. Use the lowest possible VF when
9302 updating the upper bounds of the loop.
9303 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9305 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9306 polynomial unroll factors.
9307 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9308 (vect_make_slp_decision): Likewise.
9309 (vect_supported_load_permutation_p): Likewise, and polynomial
9311 (vect_analyze_slp_cost): Handle polynomial vf.
9312 (vect_slp_analyze_node_operations): Likewise.
9313 (vect_slp_analyze_bb_1): Likewise.
9314 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9315 than an unsigned HOST_WIDE_INT.
9316 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9317 (vectorizable_load): Handle polynomial vf.
9318 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9320 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9323 Alan Hayward <alan.hayward@arm.com>
9324 David Sherwood <david.sherwood@arm.com>
9326 * match.pd: Handle bit operations involving three constants
9327 and try to fold one pair.
9329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9331 * tree-vect-loop-manip.c: Include gimple-fold.h.
9332 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9333 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9334 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9335 Add a path that uses a step of VF instead of 1, but disable it
9337 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9338 and niters_no_overflow parameters. Update calls to
9339 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9340 Create a new SSA name if the latter choses to use a ste other
9341 than zero, and return it via niters_vector_mult_vf_var.
9342 * tree-vect-loop.c (vect_transform_loop): Update calls to
9343 vect_do_peeling, vect_gen_vector_loop_niters and
9344 slpeel_make_loop_iterate_ntimes.
9345 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9346 (vect_gen_vector_loop_niters): Update declarations after above changes.
9348 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9350 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9351 128-bit round to integer instructions.
9352 (ceil<mode>2): Likewise.
9353 (btrunc<mode>2): Likewise.
9354 (round<mode>2): Likewise.
9356 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9358 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9359 unaligned VSX load/store on P8/P9.
9360 (expand_block_clear): Allow the use of unaligned VSX
9361 load/store on P8/P9.
9363 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9365 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9367 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9368 swap associated with both a load and a store.
9370 2018-01-02 Andrew Waterman <andrew@sifive.com>
9372 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9373 * config/riscv/riscv.md (clear_cache): Use it.
9375 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9377 * web.c: Remove out-of-date comment.
9379 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9381 * expr.c (fixup_args_size_notes): Check that any existing
9382 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9383 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9384 (emit_single_push_insn): ...here.
9386 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9388 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9389 (const_vector_encoded_nelts): New function.
9390 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9391 (const_vector_int_elt, const_vector_elt): Declare.
9392 * emit-rtl.c (const_vector_int_elt_1): New function.
9393 (const_vector_elt): Likewise.
9394 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9395 of CONST_VECTOR_ELT.
9397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9399 * expr.c: Include rtx-vector-builder.h.
9400 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9401 directly on the tree encoding.
9402 (const_vector_from_tree): Likewise.
9403 * optabs.c: Include rtx-vector-builder.h.
9404 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9405 sequence of "u" values.
9406 * vec-perm-indices.c: Include rtx-vector-builder.h.
9407 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9408 directly on the vec_perm_indices encoding.
9410 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9412 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9413 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9414 * rtx-vector-builder.h: New file.
9415 * rtx-vector-builder.c: Likewise.
9416 * rtl.h (rtx_def::u2): Add a const_vector field.
9417 (CONST_VECTOR_NPATTERNS): New macro.
9418 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9419 (CONST_VECTOR_DUPLICATE_P): Likewise.
9420 (CONST_VECTOR_STEPPED_P): Likewise.
9421 (CONST_VECTOR_ENCODED_ELT): Likewise.
9422 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9423 (unwrap_const_vec_duplicate): Likewise.
9424 (const_vec_series_p): Check for a non-duplicated vector encoding.
9425 Say that the function only returns true for integer vectors.
9426 * emit-rtl.c: Include rtx-vector-builder.h.
9427 (gen_const_vec_duplicate_1): Delete.
9428 (gen_const_vector): Call gen_const_vec_duplicate instead of
9429 gen_const_vec_duplicate_1.
9430 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9431 (gen_const_vec_duplicate): Use rtx_vector_builder.
9432 (gen_const_vec_series): Likewise.
9433 (gen_rtx_CONST_VECTOR): Likewise.
9434 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9435 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9436 Build a new vector rather than modifying a CONST_VECTOR in-place.
9437 (handle_special_swappables): Update call accordingly.
9438 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9439 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9440 Build a new vector rather than modifying a CONST_VECTOR in-place.
9441 (handle_special_swappables): Update call accordingly.
9443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9445 * simplify-rtx.c (simplify_const_binary_operation): Use
9446 CONST_VECTOR_ELT instead of XVECEXP.
9448 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9450 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9451 the selector elements to be different from the data elements
9452 if the selector is a VECTOR_CST.
9453 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9454 ssizetype for the selector.
9456 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9458 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9459 before testing each element individually.
9460 * tree-vect-generic.c (lower_vec_perm): Likewise.
9462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9464 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9465 * selftest-run-tests.c (selftest::run_tests): Call it.
9466 * vector-builder.h (vector_builder::operator ==): New function.
9467 (vector_builder::operator !=): Likewise.
9468 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9469 (vec_perm_indices::all_from_input_p): New function.
9470 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9471 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9472 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9473 instead of reading the VECTOR_CST directly. Detect whether both
9474 vector inputs are the same before constructing the vec_perm_indices,
9475 and update the number of inputs argument accordingly. Use the
9476 utility functions added above. Only construct sel2 if we need to.
9478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9480 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9481 the broadcast of the low byte.
9482 (expand_mult_highpart): Use an explicit encoding for the permutes.
9483 * optabs-query.c (can_mult_highpart_p): Likewise.
9484 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9485 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9486 (vectorizable_bswap): Likewise.
9487 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9488 explicit encoding for the power-of-2 permutes.
9489 (vect_permute_store_chain): Likewise.
9490 (vect_grouped_load_supported): Likewise.
9491 (vect_permute_load_chain): Likewise.
9493 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9495 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9496 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9497 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9498 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9499 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9500 (vect_gen_perm_mask_any): Likewise.
9502 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9504 * int-vector-builder.h: New file.
9505 * vec-perm-indices.h: Include int-vector-builder.h.
9506 (vec_perm_indices): Redefine as an int_vector_builder.
9507 (auto_vec_perm_indices): Delete.
9508 (vec_perm_builder): Redefine as a stand-alone class.
9509 (vec_perm_indices::vec_perm_indices): New function.
9510 (vec_perm_indices::clamp): Likewise.
9511 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9512 (vec_perm_indices::new_vector): New function.
9513 (vec_perm_indices::new_expanded_vector): Update for new
9514 vec_perm_indices class.
9515 (vec_perm_indices::rotate_inputs): New function.
9516 (vec_perm_indices::all_in_range_p): Operate directly on the
9517 encoded form, without computing elided elements.
9518 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9519 encoding. Update for new vec_perm_indices class.
9520 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9521 the given vec_perm_builder.
9522 (expand_vec_perm_var): Update vec_perm_builder constructor.
9523 (expand_mult_highpart): Use vec_perm_builder instead of
9524 auto_vec_perm_indices.
9525 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9526 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9527 or double series encoding as appropriate.
9528 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9529 vec_perm_indices instead of auto_vec_perm_indices.
9530 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9531 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9532 (vect_permute_store_chain): Likewise.
9533 (vect_grouped_load_supported): Likewise.
9534 (vect_permute_load_chain): Likewise.
9535 (vect_shift_permute_load_chain): Likewise.
9536 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9537 (vect_transform_slp_perm_load): Likewise.
9538 (vect_schedule_slp_instance): Likewise.
9539 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9540 (vectorizable_mask_load_store): Likewise.
9541 (vectorizable_bswap): Likewise.
9542 (vectorizable_store): Likewise.
9543 (vectorizable_load): Likewise.
9544 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9545 vec_perm_indices instead of auto_vec_perm_indices. Use
9546 tree_to_vec_perm_builder to read the vector from a tree.
9547 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9548 vec_perm_builder instead of a vec_perm_indices.
9549 (have_whole_vector_shift): Use vec_perm_builder and
9550 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9551 truncation to calc_vec_perm_mask_for_shift.
9552 (vect_create_epilog_for_reduction): Likewise.
9553 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9554 from auto_vec_perm_indices to vec_perm_indices.
9555 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9556 instead of changing individual elements.
9557 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9558 the vector in d.perm.
9559 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9560 from auto_vec_perm_indices to vec_perm_indices.
9561 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9562 instead of changing individual elements.
9563 (arm_vectorize_vec_perm_const): Use new_vector to install
9564 the vector in d.perm.
9565 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9566 Update vec_perm_builder constructor.
9567 (rs6000_expand_interleave): Likewise.
9568 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9569 (rs6000_expand_interleave): Likewise.
9571 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9573 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9574 to qimode could truncate the indices.
9575 * optabs.c (expand_vec_perm_var): Likewise.
9577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9579 * Makefile.in (OBJS): Add vec-perm-indices.o.
9580 * vec-perm-indices.h: New file.
9581 * vec-perm-indices.c: Likewise.
9582 * target.h (vec_perm_indices): Replace with a forward class
9584 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9585 * optabs.h: Include vec-perm-indices.h.
9586 (expand_vec_perm): Delete.
9587 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9588 (expand_vec_perm_const): Declare.
9589 * target.def (vec_perm_const_ok): Replace with...
9590 (vec_perm_const): ...this new hook.
9591 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9592 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9593 * doc/tm.texi: Regenerate.
9594 * optabs.def (vec_perm_const): Delete.
9595 * doc/md.texi (vec_perm_const): Likewise.
9596 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9597 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9598 expand_vec_perm for constant permutation vectors. Assert that
9599 the mode of variable permutation vectors is the integer equivalent
9600 of the mode that is being permuted.
9601 * optabs-query.h (selector_fits_mode_p): Declare.
9602 * optabs-query.c: Include vec-perm-indices.h.
9603 (selector_fits_mode_p): New function.
9604 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9605 is defined, instead of checking whether the vec_perm_const_optab
9606 exists. Use targetm.vectorize.vec_perm_const instead of
9607 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9608 fit in the vector mode before using a variable permute.
9609 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9610 vec_perm_indices instead of an rtx.
9611 (expand_vec_perm): Replace with...
9612 (expand_vec_perm_const): ...this new function. Take the selector
9613 as a vec_perm_indices rather than an rtx. Also take the mode of
9614 the selector. Update call to shift_amt_for_vec_perm_mask.
9615 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9616 Use vec_perm_indices::new_expanded_vector to expand the original
9617 selector into bytes. Check whether the indices fit in the vector
9618 mode before using a variable permute.
9619 (expand_vec_perm_var): Make global.
9620 (expand_mult_highpart): Use expand_vec_perm_const.
9621 * fold-const.c: Includes vec-perm-indices.h.
9622 * tree-ssa-forwprop.c: Likewise.
9623 * tree-vect-data-refs.c: Likewise.
9624 * tree-vect-generic.c: Likewise.
9625 * tree-vect-loop.c: Likewise.
9626 * tree-vect-slp.c: Likewise.
9627 * tree-vect-stmts.c: Likewise.
9628 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9630 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9631 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9632 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9633 (aarch64_vectorize_vec_perm_const): ...this new function.
9634 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9635 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9636 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9637 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9638 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9639 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9640 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9642 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9643 check for NEON modes.
9644 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9645 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9646 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9647 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9649 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9650 the old VEC_PERM_CONST conditions.
9651 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9652 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9653 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9654 (ia64_vectorize_vec_perm_const_ok): Merge into...
9655 (ia64_vectorize_vec_perm_const): ...this new function.
9656 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9657 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9658 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9659 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9660 * config/mips/mips.c (mips_expand_vec_perm_const)
9661 (mips_vectorize_vec_perm_const_ok): Merge into...
9662 (mips_vectorize_vec_perm_const): ...this new function.
9663 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9664 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9665 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9666 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9667 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9668 (rs6000_expand_vec_perm_const): Delete.
9669 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9671 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9672 (altivec_expand_vec_perm_const_le): Take each operand individually.
9673 Operate on constant selectors rather than rtxes.
9674 (altivec_expand_vec_perm_const): Likewise. Update call to
9675 altivec_expand_vec_perm_const_le.
9676 (rs6000_expand_vec_perm_const): Delete.
9677 (rs6000_vectorize_vec_perm_const_ok): Delete.
9678 (rs6000_vectorize_vec_perm_const): New function.
9679 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9680 an element count and rtx array.
9681 (rs6000_expand_extract_even): Update call accordingly.
9682 (rs6000_expand_interleave): Likewise.
9683 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9684 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9685 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9686 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9687 (rs6000_expand_vec_perm_const): Delete.
9688 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9689 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9690 (altivec_expand_vec_perm_const_le): Take each operand individually.
9691 Operate on constant selectors rather than rtxes.
9692 (altivec_expand_vec_perm_const): Likewise. Update call to
9693 altivec_expand_vec_perm_const_le.
9694 (rs6000_expand_vec_perm_const): Delete.
9695 (rs6000_vectorize_vec_perm_const_ok): Delete.
9696 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9697 reference to the SPE evmerge intructions.
9698 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9699 an element count and rtx array.
9700 (rs6000_expand_extract_even): Update call accordingly.
9701 (rs6000_expand_interleave): Likewise.
9702 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9703 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9705 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9707 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9709 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9710 vector mode and that that mode matches the mode of the data
9712 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9713 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9714 directly using expand_vec_perm_1 when forcing selectors into
9716 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9720 * optabs-query.h (can_vec_perm_p): Delete.
9721 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9722 * optabs-query.c (can_vec_perm_p): Split into...
9723 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9724 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9725 particular selector is valid.
9726 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9727 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9728 (vect_grouped_load_supported): Likewise.
9729 (vect_shift_permute_load_chain): Likewise.
9730 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9731 (vect_transform_slp_perm_load): Likewise.
9732 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9733 (vectorizable_bswap): Likewise.
9734 (vect_gen_perm_mask_checked): Likewise.
9735 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9736 implementations of variable permutation vectors into account
9737 when deciding which selector to use.
9738 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9739 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9740 with a false third argument.
9741 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9742 to test whether the constant selector is valid and can_vec_perm_var_p
9743 to test whether a variable selector is valid.
9745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9747 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9748 * optabs-query.c (can_vec_perm_p): Likewise.
9749 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9750 instead of vec_perm_indices.
9751 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9752 (vect_gen_perm_mask_checked): Likewise,
9753 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9754 (vect_gen_perm_mask_checked): Likewise,
9756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9758 * optabs-query.h (qimode_for_vec_perm): Declare.
9759 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9760 (qimode_for_vec_perm): ...this new function.
9761 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9763 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9765 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9766 does not have a conditional at the top.
9768 2018-01-02 Richard Biener <rguenther@suse.de>
9770 * ipa-inline.c (big_speedup_p): Fix expression.
9772 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9775 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9778 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9782 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9783 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9784 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9785 cond_taken_branch_cost 3->4.
9787 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9789 PR tree-optimization/83581
9790 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9791 TODO_cleanup_cfg if any changes have been made.
9794 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9795 convert_modes if target mode has the right side, but different mode
9799 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9800 last argument when extracting from CONCAT. If either from_real or
9801 from_imag is NULL, use expansion through memory. If result is not
9802 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9803 the parts directly to inner mode, if even that fails, use expansion
9807 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9808 check for bswap in mode rather than HImode and use that in expand_unop
9811 Copyright (C) 2018 Free Software Foundation, Inc.
9813 Copying and distribu