1 2018-04-04 Richard Biener <rguenther@suse.de>
4 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5 from contexts for DINFO_LEVEL_TERSE and below.
7 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10 resiter_operand condition.
11 * config/nds32/nds32.md (*move<mode>): Ditto.
13 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14 Monk Chiang <sh.chiang04@gmail.com>
16 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
23 Kito Cheng <kito.cheng@gmail.com>
25 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
26 nds32_cond_code_str, output_cond_branch,
27 output_cond_branch_compare_zero, nds32_expand_cbranch,
28 nds32_expand_cstore, nds32_expand_movcc,
29 nds32_output_cbranchsi4_equality_zero,
30 nds32_output_cbranchsi4_equality_reg,
31 nds32_output_cbranchsi4_equality_reg_or_const_int,
32 nds32_output_cbranchsi4_greater_less_zero: New functions.
33 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
34 nds32_expand_cstore, nds32_expand_movcc,
35 nds32_output_cbranchsi4_equality_zero,
36 nds32_output_cbranchsi4_equality_reg,
37 nds32_output_cbranchsi4_equality_reg_or_const_int,
38 nds32_output_cbranchsi4_greater_less_zero): Declare.
39 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
40 nds32_rimm11s_operand): New predicates.
41 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
42 * config/nds32/nds32.md: Rewrite all the branch and conditional move
45 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
47 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
48 * config/nds32/nds32.md: Ditto.
49 * config/nds32/pipelines.md: Ditto.
51 2018-04-04 Richard Biener <rguenther@suse.de>
53 PR tree-optimization/85168
54 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
55 propagating abnormals.
57 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
59 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
61 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
62 Kito Cheng <kito.cheng@gmail.com>
64 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
65 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
66 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
67 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
68 * config/nds32/nds32.md (sibcall_internal): New.
69 (sibcall_register): Remove.
70 (sibcall_immediate): Remove.
71 (sibcall_value_internal): New.
72 (sibcall_value_register): Remove.
73 (sibcall_value_immediate): Remove.
74 * config/nds32/predicates.md (nds32_general_register_operand): New.
75 (nds32_call_address_operand): New.
77 2018-04-03 Jakub Jelinek <jakub@redhat.com>
79 PR rtl-optimization/85167
80 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
81 bb_defs if *split_p, instead preinitialize it to NULL.
83 PR tree-optimization/85156
84 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
85 evaluating the argument multiple times.
87 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
89 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
91 (_mm_cvtpd_ps): Likewise.
92 (_mm_cvttpd_epi32): Likewise.
93 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
94 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
95 vector, pixel, and bool following altivec.h include.
97 2018-04-03 Martin Sebor <msebor@redhat.com>
99 * doc/extend.texi (Common Function Attributes): Clarify.
100 (const attribute): Likewise.
101 (pure attribute): Likewise.
103 2018-04-03 Jakub Jelinek <jakub@redhat.com>
106 * config/i386/i386.c (ix86_expand_vector_set): Use
107 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
109 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
111 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
112 instructions when changing rounding bits to preserve precision bits
113 in the x87 control word.
115 2018-04-03 Martin Liska <mliska@suse.cz>
117 PR tree-optimization/82491
118 * rtl.h (strip_offset_and_add): Replace += suboffset with
119 poly_uint64 () + suboffset.
121 2018-03-29 Martin Liska <mliska@suse.cz>
122 Martin Jambor <mjambor@suse.cz>
125 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
126 param_type is not an integral or pointer type.
128 2018-04-03 Richard Biener <rguenther@suse.de>
130 * sese.h (recompute_all_dominators): Remove.
132 2018-04-02 Martin Sebor <msebor@redhat.com>
134 * doc/invoke.texi (-Wrestrict): Fix typos.
136 2018-04-02 Jim Wilson <jimw@sifive.com>
138 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
139 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
140 (<optab>di3, <optab>si3_extend): Likewise.
141 (<optab>si3_mask, <optab>si3_mask_1): New.
142 (<optab>di3_mask, <optab>di3_mask_1): New.
143 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
144 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
145 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
147 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
149 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
152 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
154 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
155 (nds32_canonicalize_comparison): New function.
157 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
158 Kito Cheng <kito.cheng@gmail.com>
159 Kuan-Lin Chen <kuanlinchentw@gmail.com>
161 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
162 * config/nds32/constants.md (unspec_volatile_element): Add
163 UNSPEC_VOLATILE_RELAX_GROUP.
164 * config/nds32/nds32-relax-opt.c: New file.
165 * config/nds32/nds32-predicates.c
166 (nds32_symbol_load_store_p): New function.
167 * config/nds32/nds32-protos.h
168 (nds32_symbol_load_store_p): Declare function.
169 (make_pass_nds32_relax_opt): Declare new rtl pass function.
170 * config/nds32/nds32.c
171 (nds32_register_pass): New function to register pass.
172 (nds32_register_passes): New function to register passes.
173 * config/nds32/nds32.md (relax_group): New pattern.
174 * config/nds32/nds32.opt (mrelax-hint): New option.
175 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
177 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
179 * config/nds32/t-nds32: Modify files dependency.
181 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
183 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
184 (PROFILE_HOOK): Define its implementation.
186 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
188 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
189 type and 32-bit size.
191 2018-04-01 Jakub Jelinek <jakub@redhat.com>
194 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
195 (V_128_256): New mode iterator.
196 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
197 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
198 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
200 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
201 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
203 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
206 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
207 NaN inputs correctly.
209 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
212 * config/rs6000/vsx.md (??r): New mode attribute.
213 (*vsx_mov<mode>_64bit): Use it.
214 (*vsx_mov<mode>_32bit): Likewise.
216 2018-03-30 Martin Sebor <msebor@redhat.com>
218 PR tree-optimization/84818
219 * builtins.c (check_access): Use warning_n.
221 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
224 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
226 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
229 2018-03-30 Julia Koval <julia.koval@intel.com>
232 * x86-tune.def (movx, partial_reg_dependency): Enable for
235 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
238 * lra-constraints.c (process_alt_operands): Move setting
239 this_alternative_matches below.
241 2018-03-29 Martin Liska <mliska@suse.cz>
244 * doc/invoke.texi: Document how LTO works with debug info.
245 Describe auto-load support of binutils. Mention 'x86-64'
246 as valid option value of -march option.
248 2018-03-29 Jakub Jelinek <jakub@redhat.com>
250 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
253 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
254 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
255 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
258 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
261 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
262 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
263 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
264 for RS6000_BTM_POWERPC64.
265 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
266 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
267 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
272 2018-03-28 Carl Love <cel@us.ibm.com>
275 2017-09-27 Carl Love <cel@us.ibm.com>
277 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
278 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
279 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
282 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
284 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
285 instead of __vector bool.
286 (_mm_max_pu8): Likewise.
287 (_mm_min_pi16): Likewise.
289 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
292 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
296 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
298 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
299 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
300 (div_extend): Likewise.
301 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
303 (__builtin_divweuo): Likewise.
304 (__builtin_divdeo): Likewise.
305 (__builtin_divdeuo): Likewise.
307 2018-03-28 Jakub Jelinek <jakub@redhat.com>
310 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
311 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
313 PR tree-optimization/82004
314 * gimple-match-head.c (optimize_pow_to_exp): New function.
315 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
316 Don't fold to exp if optimize_pow_to_exp is false.
318 2018-03-28 Martin Liska <mliska@suse.cz>
321 * calls.c (initialize_argument_information): Fix trailing space.
322 * common.opt: Fix typo and provide better explanation for
323 -fsanitize-coverage option.
324 * config/i386/i386.opt: Fix typo.
326 2018-03-28 Jakub Jelinek <jakub@redhat.com>
327 Martin Liska <mliska@suse.cz>
330 * gimplify.c (asan_poison_variable): Don't do the check for
331 gimplify_omp_ctxp here.
332 (gimplify_decl_expr): Do it here.
333 (gimplify_target_expr): Likewise.
335 2018-03-28 Martin Liska <mliska@suse.cz>
338 * config/i386/i386.c (ix86_function_arg_advance): Do not call
339 chkp_type_bounds_count if MPX is not enabled.
341 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
343 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
345 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
348 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
349 function to create the function decl for complex long double
350 multiply and divide for -mabi=ieeelongdouble.
351 (init_float128_ieee): Call it.
353 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
356 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
357 -fcf-protection=branch -mibt.
358 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
360 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
363 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
365 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
368 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
369 extern array declarations.
371 2018-03-27 Richard Biener <rguenther@suse.de>
374 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
375 explicit single_use checks.
377 2018-03-27 Richard Biener <rguenther@suse.de>
379 PR tree-optimization/85082
380 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
383 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
385 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
386 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
387 Turn on fasynchronous-unwind-tables and funwind-tables.
389 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
392 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
393 (*bmi_blsr_<mode>_ccz): Ditto.
395 2018-03-26 Tom de Vries <tom@codesourcery.com>
397 PR tree-optimization/85063
398 * omp-general.c (offloading_function_p): New function. Factor out
400 * omp-offload.c (pass_omp_target_link::gate): ... here.
401 * omp-general.h (offloading_function_p): Declare.
402 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
403 with attribute omp declare target for offloading functions.
405 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
407 PR tree-optimization/84005
408 * tree-data-ref.h (get_base_for_alignment): Declare.
409 * tree-data-ref.c (get_base_for_alignment_1): New function.
410 (get_base_for_alignment): Likewise.
411 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
412 get_base_for_alignment to find a suitable base object, instead
413 of always using drb->base_address.
415 2018-03-23 Jakub Jelinek <jakub@redhat.com>
418 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
419 known size by default.
421 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
424 * lra-constraints.c (process_alt_operands): Don't match BLKmode
425 and non BLKmode operands.
427 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
430 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
433 2018-03-23 Richard Biener <rguenther@suse.de>
436 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
437 we are going to emit early debug for LTO.
439 2018-03-23 Jakub Jelinek <jakub@redhat.com>
442 * function.c (match_asm_constraints_1): Don't optimize if input
443 doesn't satisfy general_operand predicate for output's mode.
446 * alias.c (write_dependence_p): Don't require for x_canonicalized
447 non-VOIDmode if x has VOIDmode.
450 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
451 just don't try to optimize it rather than assert it never happens.
453 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
455 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
456 macro expansions for definition of ST_INTERNAL_<mode> and
457 LD_INTERNAL_<mode> builtins.
458 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
460 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
462 (altivec_expand_st_builtin): Likewise.
463 (altivec_expand_builtin): Remove calls to deleted functions.
464 (rs6000_address_for_altivec): Delete this function.
465 * config/rs6000/vector.md: Remove expands for
466 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
468 2018-03-22 Sudakshina Das <sudi.das@arm.com>
471 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
472 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
473 re-computing once computed.
474 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
475 (arm_init_machine_status): Initialize
476 machine->static_chain_stack_bytes.
478 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
481 * doc/extend.texi: Add four new prototypes for vec_ld.
482 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
483 definitions for more logical presentation.
484 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
485 entries for V1TI variants of __builtin_altivec_ld builtin.
486 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
487 handling of V1TI variant of LVX icode pattern.
488 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
489 (rs6000_gimple_fold_builtin): Likewise.
490 (altivec_init_builtins): Add code to define
491 __builtin_altivec_lvx_v1ti function.
493 2018-03-22 Jakub Jelinek <jakub@redhat.com>
496 * function.c (match_asm_constraints_1): Don't do the optimization
497 if input isn't a REG, SUBREG, MEM or constant.
499 2018-03-22 Tom de Vries <tom@codesourcery.com>
501 PR tree-optimization/84956
502 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
503 bb_has_abnormal_pred.
505 2018-03-22 Jakub Jelinek <jakub@redhat.com>
508 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
509 DECL_INITIAL (decl) to decl at the end.
510 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
513 2018-03-21 Joseph Myers <joseph@codesourcery.com>
515 * doc/extend.texi (__builtin_tgmath): Document when complex
516 integer types are treated as _Complex _Float64.
518 2018-03-21 Tom de Vries <tom@codesourcery.com>
520 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
522 2018-03-21 Jakub Jelinek <jakub@redhat.com>
524 PR tree-optimization/84960
525 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
526 if it is ENTRY block, move them into single succ of ENTRY in that case.
528 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
530 PR tree-optimization/84811
531 * poly-int.h (poly_span_traits): Remove the T3 parameter and
532 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
533 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
534 (known_subrange_p): Update accordingly. Cast each value involved
535 in the size comparison, rather than casting the result of the
538 2018-03-21 Jakub Jelinek <jakub@redhat.com>
540 PR tree-optimization/84982
541 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
542 by flipping the least significant bit rather than all bits from
543 bitpos to bitpos + bitsize - 1.
545 2018-03-21 Nathan Sidwell <nathan@acm.org>
547 * doc/extend.texi (Deprecated Features): Remove mention of
548 long-deleted deprecations.
550 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
553 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
554 * configure: Regenerate.
556 2018-03-21 Tom de Vries <tom@codesourcery.com>
558 PR tree-optimization/83126
559 * tree-parloops.c (num_phis): New function.
560 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
562 2018-03-21 Nathan Sidwell <nathan@acm.org>
564 * doc/extend.texi (Deprecated Features): Update deprecated flags,
565 mention anon-struct/union members and trailing attributes.
567 2018-03-21 Bin Cheng <bin.cheng@arm.com>
569 PR tree-optimization/84969
570 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
571 builtin memset partitions if they set different rhs values.
573 2018-03-21 Jakub Jelinek <jakub@redhat.com>
575 PR rtl-optimization/84989
576 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
577 VEC_DUPLICATE with scalar result mode.
579 2018-03-21 Martin Liska <mliska@suse.cz>
582 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
583 not intended return statement.
585 2018-03-21 Martin Liska <mliska@suse.cz>
588 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
589 (chkp_find_bound_slots_1): Limit number of iterations.
591 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
594 * Minor grammar fixes for x86 options.
596 2018-03-20 Jakub Jelinek <jakub@redhat.com>
599 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
600 holding REG_CFA_RESTORE notes, instead turn them into a USE.
602 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
605 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
606 (altivec_lvx_<mode>_1op): Likewise.
607 (altivec_stvx_<mode>_2op): Likewise.
608 (altivec_stvx_<mode>_1op): Likewise.
609 (altivec_lvx_<VM2:mode>): New define_expand.
610 (altivec_stvx_<VM2:mode>): Likewise.
611 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
612 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
613 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
614 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
615 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
616 (rs6000_gen_lvx): Likewise.
617 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
618 (altivec_expand_stv_builtin): Likewise.
619 (altivec_expand_builtin): Likewise.
620 * config/rs6000/vector.md: Likewise.
622 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
625 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
628 2018-03-20 Richard Biener <rguenther@suse.de>
631 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
632 sign-conversions as zero, fall back to standard scalar_stmt
635 2018-03-20 Martin Liska <mliska@suse.cz>
638 * predict.c (rebuild_frequencies): Handle case when we have
639 PROFILE_ABSENT, but flag_guess_branch_prob is false.
641 2018-03-20 Jakub Jelinek <jakub@redhat.com>
644 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
645 flag_section_anchors.
646 * varasm.c (use_blocks_for_decl_p): Remove hack for
650 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
652 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
653 be created, use lowpart_subreg of operands[0] rather than operands[0]
655 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
656 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
657 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
658 and n constraint instead of aarch64_shift_imm_di and Usd.
659 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
660 (*aarch64_<optab>_reg_minus<mode>3): ... this.
662 2018-03-20 Sudakshina Das <sudi.das@arm.com>
665 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
666 to favor GPR over NEON registers.
667 (<shift>di3_neon): Likewise.
669 2018-03-20 Tom de Vries <tom@codesourcery.com>
672 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
673 (nvptx_process_pars): Emit bar.sync asap and alap.
675 2018-03-20 Tom de Vries <tom@codesourcery.com>
678 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
679 seen_label if seen_label is already set.
681 2018-03-20 Jakub Jelinek <jakub@redhat.com>
684 * config/i386/i386.c (fold_builtin_cpu): For features above 31
685 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
686 Use 1U instead of 1. Formatting fixes.
689 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
690 instead of TREE_TYPE (s1) for the return value.
692 2018-03-19 Jakub Jelinek <jakub@redhat.com>
694 PR tree-optimization/84946
695 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
696 bitsize + bitsize in poly_uint64 rather than poly_int64.
699 * dwarf2asm.c: Include fold-const.c.
700 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
701 of decl rather than decl itself.
703 PR rtl-optimization/84643
704 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
706 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
709 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
710 calling assemble_variable.
712 2018-03-19 Sudakshina Das <sudi.das@arm.com>
715 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
716 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
718 2018-03-19 Jim Wilson <jimw@sifive.com>
721 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
722 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
723 (riscv_first_stack_step): Likewise.
724 (riscv_option_override): Use STACK_BOUNDARY instead of
726 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
728 (BIGGEST_ALIGNMENT): Set to 128.
729 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
730 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
733 2018-03-19 Richard Biener <rguenther@suse.de>
735 PR tree-optimization/84933
736 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
737 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
739 2018-03-19 Richard Biener <rguenther@suse.de>
741 PR tree-optimization/84859
742 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
743 (cond_if_else_store_replacement): Perform sinking operation on
744 single-store BBs regardless of MAX_STORES_TO_SINK setting.
745 Generalize what a BB with a single eligible store is.
747 2018-03-19 Richard Biener <rguenther@suse.de>
749 PR tree-optimization/84929
750 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
751 chrec_is_positive against non-chrec arg.
753 2018-03-19 Tamar Christina <tamar.christina@arm.com>
756 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
758 2018-03-18 Martin Liska <mliska@suse.cz>
760 PR rtl-optimization/84635
761 * regrename.c (build_def_use): Use matches_mode only when
764 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
766 PR tree-optimization/84913
767 * tree-vect-loop.c (vectorizable_reduction): Don't try to
768 vectorize chains of COND_EXPRs.
770 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
772 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
774 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
776 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
778 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
780 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
782 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
783 Kito Cheng <kito.cheng@gmail.com>
785 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
786 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
787 (nds32_adjust_reg_alloc_order): New function.
788 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
790 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
792 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
793 nds32_print_operand, nds32_print_operand_address): Use
794 HOST_WIDE_INT_PRINT_DEC instead.
796 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
798 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
800 2018-03-17 Jakub Jelinek <jakub@redhat.com>
803 * config/i386/i386.c (initial_ix86_tune_features,
804 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
806 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
807 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
808 rather than 1u << ix86_tune. Formatting fix.
809 (ix86_option_override_internal): Change ix86_arch_mask from
810 unsigned int to unsigned HOST_WIDE_INT, initialize to
811 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
812 (ix86_function_specific_restore): Likewise.
814 2018-03-16 Jakub Jelinek <jakub@redhat.com>
817 * postreload.c (reload_combine_recognize_pattern): Perform
818 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
819 truncate_int_for_mode the result for the destination's mode.
822 * hsa-gen.c (mem_type_for_type): Fix comment typo.
823 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
825 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
828 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
831 * lra-assigns.c (lra_split_hard_reg_for): Don't use
832 regno_allocno_class_array and sorted_pseudos.
833 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
834 insns where regno is used.
836 2018-03-16 Martin Liska <mliska@suse.cz>
839 * multiple_target.c (create_dispatcher_calls): Redirect
840 reference in the symbol table.
842 2018-03-16 Martin Liska <mliska@suse.cz>
845 * multiple_target.c (create_dispatcher_calls): Redirect also
848 2018-03-16 Jakub Jelinek <jakub@redhat.com>
852 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
853 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
854 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
856 2018-03-16 Julia Koval <julia.koval@intel.com>
858 * doc/invoke.texi (Skylake Server): Add CLWB.
859 Cannonlake): Remove CLWB.
861 2018-03-16 Jakub Jelinek <jakub@redhat.com>
863 PR tree-optimization/84841
864 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
866 (FLOAT_ONE_CONST_TYPE): Define.
867 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
868 (sort_by_operand_rank): Put entries with higher constant_type last
869 rather than first to match comments.
871 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
873 * config/nios2/nios2.md (movsi_internal): Fix thinko in
876 2018-03-15 Jakub Jelinek <jakub@redhat.com>
879 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
880 check and use address of target always.
882 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
885 * config/i386/i386.c (indirect_thunk_needed): Update comments.
886 (indirect_thunk_bnd_needed): Likewise.
887 (indirect_thunks_used): Likewise.
888 (indirect_thunks_bnd_used): Likewise.
889 (indirect_return_needed): New.
890 (indirect_return_bnd_needed): Likewise.
891 (output_indirect_thunk_function): Add a bool argument for
893 (output_indirect_thunk_function): Don't generate alias for
894 function return thunk.
895 (ix86_code_end): Call output_indirect_thunk_function to generate
896 function return thunks.
897 (ix86_output_function_return): Set indirect_return_bnd_needed
898 and indirect_return_needed instead of indirect_thunk_bnd_needed
899 and indirect_thunk_needed.
901 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
903 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
904 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
905 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
907 2018-03-15 David Malcolm <dmalcolm@redhat.com>
908 Paul Hua <paul.hua.gm@gmail.com>
911 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
913 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
915 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
916 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
919 2018-03-15 Tamar Christina <tamar.christina@arm.com>
922 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
923 instead of GET_MODE_SIZE when comparing Units.
925 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
928 * varasm.c (hash_section): Return an unchangeble hash value
929 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
930 Return !aarch64_can_use_per_function_literal_pools_p ().
932 2018-03-15 Jakub Jelinek <jakub@redhat.com>
935 * optabs.c (emit_conditional_move): Pass address of cmode's copy
936 rather than address of cmode as last argument to prepare_cmp_insn.
938 2018-03-15 Julia Koval <julia.koval@intel.com>
940 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
941 F_AVX512VNNI, F_AVX512BITALG): New.
943 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
946 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
947 insn for floating-point loads and stores.
949 2018-03-14 Carl Love <cel@us.ibm.com>
951 * config/rs6000/rs6000-c.c: Add macro definitions for
952 ALTIVEC_BUILTIN_VEC_PERMXOR.
953 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
954 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
955 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
956 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
958 * config/doc/extend.texi: Add prototypes for vec_permxor.
960 2018-03-14 David Malcolm <dmalcolm@redhat.com>
963 * diagnostic-show-locus.c (class layout_point): Convert m_line
964 from int to linenum_type.
965 (line_span::comparator): Use linenum "compare" function when
966 comparing line numbers.
967 (test_line_span): New function.
968 (layout_range::contains_point): Convert param "row" from int to
970 (layout_range::intersects_line_p): Likewise.
971 (layout::will_show_line_p): Likewise.
972 (layout::print_source_line): Likewise.
973 (layout::should_print_annotation_line_p): Likewise.
974 (layout::print_annotation_line): Likewise.
975 (layout::print_leading_fixits): Likewise.
976 (layout::annotation_line_showed_range_p): Likewise.
977 (struct line_corrections): Likewise for field m_row.
978 (line_corrections::line_corrections): Likewise for param "row".
979 (layout::print_trailing_fixits): Likewise.
980 (layout::get_state_at_point): Likewise.
981 (layout::get_x_bound_for_row): Likewise.
982 (layout::print_line): Likewise.
983 (diagnostic_show_locus): Likewise for locals "last_line" and
985 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
986 * input.c (selftest::test_linenum_comparisons): New function.
987 (selftest::input_c_tests): Call it.
988 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
989 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
990 * selftest.h (ASSERT_GT): New macro.
991 (ASSERT_GT_AT): New macro.
992 (ASSERT_LT): New macro.
993 (ASSERT_LT_AT): New macro.
995 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
997 PR rtl-optimization/84780
998 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1000 2018-03-14 Martin Liska <mliska@suse.cz>
1002 * tree.c (record_node_allocation_statistics): Use
1003 get_stats_node_kind.
1004 (get_stats_node_kind): New function extracted from
1005 record_node_allocation_statistics.
1006 (free_node): Use get_stats_node_kind.
1008 2018-03-14 Richard Biener <rguenther@suse.de>
1010 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1011 that the value-set of ANTIC_IN doesn't grow.
1014 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1016 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1017 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1019 2018-03-14 Julia Koval <julia.koval@intel.com>
1021 * config.gcc (icelake-client, icelake-server): New.
1023 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1024 (initial_ix86_arch_features): Ditto.
1025 (PTA_SKYLAKE): Add SGX.
1026 (PTA_ICELAKE): Remove.
1027 (PTA_ICELAKE_CLIENT): New.
1028 (PTA_ICELAKE_SERVER): New.
1029 (ix86_option_override_internal): Split up icelake on icelake client and
1031 (get_builtin_code_for_version): Ditto.
1032 (fold_builtin_cpu): Ditto.
1033 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1034 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1035 * config/i386/i386.h (processor_type): Ditto.
1036 * doc/invoke.texi: Ditto.
1038 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1041 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1042 INTEGER_CST offset, add it together with bitpos / 8 and
1043 sign extend based on POINTER_SIZE.
1047 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1050 * config/i386/constraints.md (Yc): New register constraint.
1051 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1052 Use Yc constraint for alternative 2 of operand 0. Remove
1053 preferred_for_speed attribute.
1055 2018-03-14 Richard Biener <rguenther@suse.de>
1057 PR tree-optimization/84830
1058 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1059 with the old one to avoid oscillations.
1061 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1064 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1066 (assign_by_spills): Return a flag of reload assignment failure.
1067 Do not process the reload assignment failures. Do not spill other
1068 reload pseudos if they has the same reg class. Update n if
1070 (lra_assign): Add a return arg. Set up from the result of
1071 assign_by_spills call.
1072 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1073 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1074 usage_insns if it is not NULL.
1075 (spill_hard_reg_in_range): New function.
1076 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1077 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1078 function prototypes.
1079 (lra_assign): Change prototype.
1080 * lra.c (lra): Add code to deal with fails by splitting hard reg
1083 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1085 * config/riscv/riscv.opt (mrelax): New option.
1086 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1087 "norelax" when riscv_mrelax is disabled.
1088 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1090 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1093 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1094 reassociation for int modes.
1096 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1098 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1099 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1101 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1102 * config/aarch64/aarch64-sve.md
1103 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1104 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1105 (*extend<mode><Vwide>2): Rename to...
1106 (aarch64_sve_extend<mode><Vwide>2): ...this.
1107 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1108 renaming the old pattern to...
1109 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1111 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1112 define_expand, renaming the old pattern to...
1113 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1114 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1115 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1116 account when deciding which SVE instruction the optab should use.
1117 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1119 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1121 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1122 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1123 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1124 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1125 (tlsdesc_small_<mode>): Turn a define_expand and use
1126 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1127 (tlsdesc_small_advsimd_<mode>): ...this.
1128 (tlsdesc_small_sve_<mode>): New pattern.
1130 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1132 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1133 (UNSPEC_UMUL_HIGHPART): New constants.
1134 (MUL_HIGHPART): New int iteraor.
1135 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1136 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1138 (*<su>mul<mode>3_highpart): New define_insn.
1140 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1143 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1146 2018-03-13 Martin Liska <mliska@suse.cz>
1149 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1151 (sem_item_optimizer::~sem_item_optimizer): Release it.
1152 (sem_item_optimizer::merge_classes): Register variable aliases.
1153 (sem_item_optimizer::fixup_pt_set): New function.
1154 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1155 * ipa-icf.h: Declare new variables and functions.
1157 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1160 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1161 integer_pow2p@2 and test integer_pow2p in condition.
1162 (A < 0 ? C : 0): Similarly for @1.
1165 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1166 characters starting at p contain '\0' character, don't look beyond
1170 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1171 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1174 * reg-stack.c (change_stack): Change update_end var from int to
1175 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1176 also call set_block_for_insn on the newly added insns and rescan.
1179 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1180 on the last operand.
1183 * tree.c (stabilize_reference_1): Return save_expr (e) for
1184 STATEMENT_LIST even if it doesn't have side-effects.
1186 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1188 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1190 2018-03-12 Renlin Li <renlin.li@arm.com>
1192 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1193 aarch64_output_scalar_simd_mov_immediate.
1195 2018-03-12 Martin Sebor <msebor@redhat.com>
1197 PR tree-optimization/83456
1198 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1199 for perfectly overlapping calls to memcpy.
1200 (gimple_fold_builtin_memory_chk): Same.
1201 (gimple_fold_builtin_strcpy): Handle no-warning.
1202 (gimple_fold_builtin_stxcpy_chk): Same.
1203 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1205 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1207 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1208 parameter. Use it for SFmode.
1209 (rs6000_function_arg_advance_1): Adjust.
1210 (rs6000_function_arg): Adjust.
1211 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1213 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1215 PR rtl-optimization/84169
1216 PR rtl-optimization/84780
1217 * combine.c (can_combine_p): Check for a 2-insn combination whether
1218 the destination register is used between the two insns, too.
1220 2018-03-12 Richard Biener <rguenther@suse.de>
1222 PR tree-optimization/84803
1223 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1224 for refs DR analysis didn't process.
1226 2018-03-12 Richard Biener <rguenther@suse.de>
1228 PR tree-optimization/84777
1229 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1230 force-vectorize loops ignore whether we are optimizing for size.
1232 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1234 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1235 (TARGET_MD_ASM_ADJUST): Define.
1237 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1238 Kito Cheng <kito.cheng@gmail.com>
1239 Chung-Ju Wu <jasonwucj@gmail.com>
1241 * config/nds32/nds32.c (nds32_compute_stack_frame,
1242 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1243 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1244 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1245 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1246 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1247 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1248 * config/nds32/nds32.md (prologue, epilogue): Use macro
1249 NDS32_V3PUSH_AVAILABLE_P to do checking.
1251 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1254 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1255 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1256 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1257 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1258 addition of most attributes on !orig_type_die or the attribute not
1259 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1261 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1262 Chung-Ju Wu <jasonwucj@gmail.com>
1264 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1266 * config/nds32/nds32.opt (mvh): New option.
1268 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1269 Chung-Ju Wu <jasonwucj@gmail.com>
1271 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1273 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1274 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1277 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1278 Chung-Ju Wu <jasonwucj@gmail.com>
1280 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1282 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1283 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1285 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1286 Kito Cheng <kito.cheng@gmail.com>
1287 Chung-Ju Wu <jasonwucj@gmail.com>
1289 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1290 UNSPEC_FFMISM and UNSPEC_FLMISM.
1291 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1292 for ffb, ffmism and flmism.
1293 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1294 (unspec_ffmism): Ditto.
1295 (unspec_flmism): Ditto.
1296 (nds32_expand_builtin_impl): Check if string extension is available.
1297 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1298 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1300 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1303 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1306 * lra-assigns.c (assign_by_spills): Return a flag of reload
1307 assignment failure. Do not process the reload assignment
1308 failures. Do not spill other reload pseudos if they has the same
1310 (lra_assign): Add a return arg. Set up from the result of
1311 assign_by_spills call.
1312 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1313 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1314 usage_insns if it is not NULL.
1315 (spill_hard_reg_in_range): New function.
1316 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1317 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1318 function prototypes.
1319 (lra_assign): Change prototype.
1320 * lra.c (lra): Add code to deal with fails by splitting hard reg
1323 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1326 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1328 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1331 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1332 (dw_val_node): Add val_symbolic_view.
1333 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1334 (symview_upper_bound): New.
1335 (new_line_info_table): Initialize symviews_since_reset.
1336 (dwarf2out_source_line): Count symviews_since_reset and set
1337 symview_upper_bound.
1338 (dw_val_equal_p): Handle symview.
1339 (add_AT_symview): New.
1340 (print_dw_val): Handle symview.
1341 (attr_checksum, attr_checksum_ordered): Likewise.
1342 (same_dw_val_p, size_of_die): Likewise.
1343 (value_format, output_die): Likewise.
1344 (add_high_low_attributes): Use add_AT_symview for entry_view.
1345 (dwarf2out_finish): Reset symview_upper_bound, clear
1348 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1351 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1352 Add strict argument and use it.
1353 (rs6000_split_multireg_move): Update for new strict argument.
1354 (mem_operand_gpr): Disallow all non-offsettable addresses.
1355 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1357 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1360 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1361 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1362 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1365 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1366 decl, use remap_type if we want to use the type.
1368 2018-03-09 Martin Sebor <msebor@redhat.com>
1370 PR tree-optimization/84526
1371 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1373 (builtin_access::generic_overlap): Be prepared to handle non-array
1376 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1378 PR rtl-optimization/84682
1379 * lra-constraints.c (process_address_1): Check is_address flag
1380 for address constraints.
1381 (process_alt_operands): Likewise.
1382 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1383 preprocess_constraints.
1384 * recog.h (preprocess_constraints): Add oploc parameter.
1386 * recog.c (preprocess_constraints): Test address_operand for
1387 CT_ADDRESS constraints.
1389 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1392 * lra-assigns.c (assign_by_spills): Return a flag of reload
1393 assignment failure. Do not process the reload assignment
1394 failures. Do not spill other reload pseudos if they has the same
1396 (lra_assign): Add a return arg. Set up from the result of
1397 assign_by_spills call.
1398 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1399 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1400 usage_insns if it is not NULL.
1401 (spill_hard_reg_in_range): New function.
1402 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1403 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1404 function prototypes.
1405 (lra_assign): Change prototype.
1406 * lra.c (lra): Add code to deal with fails by splitting hard reg
1409 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1412 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1413 Accept complain bool parameter. Only emit errors if it is true.
1414 (arm_parse_cpu_option_name): Likewise.
1415 (arm_target_thumb_only): Adjust callers of the above.
1416 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1417 prototype to take a default true bool parameter.
1418 (arm_parse_arch_option_name): Likewise.
1420 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1421 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1425 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1426 * configure: Regenerate.
1427 * configure.ac ("linker --version-script option"): New.
1428 ("linker soname option"): New.
1430 2018-03-09 Richard Biener <rguenther@suse.de>
1432 PR tree-optimization/84775
1433 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1434 immediate uses of predicate stmts and mark them modified.
1437 PR tree-optimization/84178
1438 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1440 (version_loop_for_if_conversion): Delay update_ssa call.
1441 (tree_if_conversion): Delay update_ssa until after predicate
1444 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1447 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1448 when the function accesses prior frames.
1450 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1453 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1454 gen_llsym, otherwise call maybe_gen_llsym.
1457 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1458 has ',' character inside of it.
1460 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1463 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1464 as clobbering CC_REGNUM.
1466 2018-03-08 Richard Biener <rguenther@suse.de>
1469 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1470 (follow_copies_to_constant): Do not follow SSA names registered
1473 2018-03-08 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/84178
1476 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1478 (version_loop_for_if_conversion): Delay update_ssa call.
1479 (tree_if_conversion): Delay update_ssa until after predicate
1482 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1484 PR tree-optimization/84178
1485 * tree-if-conv.c (release_bb_predicate): Remove the
1486 the assertion that the stmts have NULL use_ops.
1487 Discard the statements, asserting that they haven't
1488 yet been added to a BB.
1490 2018-03-08 Richard Biener <rguenther@suse.de>
1492 PR tree-optimization/84746
1493 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1494 (phi_translate): Pass in destination ANTIC_OUT set.
1495 (phi_translate_1): Likewise. For a simplified result lookup
1496 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1497 (phi_translate_set): Adjust.
1498 (do_pre_regular_insertion): Likewise.
1499 (do_pre_partial_partial_insertion): Likewise.
1501 2018-03-08 Martin Liska <mliska@suse.cz>
1503 PR gcov-profile/84735
1504 * doc/gcov.texi: Document usage of profile files.
1505 * gcov-io.h: Document changes in the format.
1507 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1511 * dwarf2out.c (struct dw_line_info_table): Update comments for
1513 (FORCE_RESET_NEXT_VIEW): New.
1514 (FORCE_RESETTING_VIEW_P): New.
1515 (RESETTING_VIEW_P): Check for -1 too.
1516 (ZERO_VIEW_P): Likewise.
1517 (new_line_info_table): Force-reset next view.
1518 (dwarf2out_begin_function): Likewise.
1519 (dwarf2out_source_line): Simplify zero_view_p initialization.
1520 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1521 view directly. Omit view when omitting .loc at line 0.
1523 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1525 PR tree-optimization/84740
1526 * tree-switch-conversion.c (process_switch): Call build_constructors
1527 only if info.phi_count is non-zero.
1529 PR tree-optimization/84739
1530 * tree-tailcall.c (find_tail_calls): Check call arguments against
1531 DECL_ARGUMENTS (current_function_decl) rather than
1532 DECL_ARGUMENTS (func) when checking for tail recursion.
1534 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1536 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1537 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1538 Volker Reichelt's entry and add entries for people that perform
1539 GCC fuzzy testing and report numerous bugs.
1541 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1544 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1545 readonly data in sdata, if that is disabled.
1546 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1547 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1548 -mreadonly-in-sdata option.
1550 2018-03-07 Martin Sebor <msebor@redhat.com>
1552 PR tree-optimization/84468
1553 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1554 basic block when looking for nul assignment.
1556 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1559 * except.h (output_function_exception_table): Adjust prototype.
1560 * except.c (output_function_exception_table): Remove FNNAME parameter
1561 and add SECTION parameter. Ouput one part of the table at a time.
1562 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1563 the first part of the exception table and emit unwind directives.
1564 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1565 (i386_pe_seh_cold_init): Likewise.
1566 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1567 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1568 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1569 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1570 * config/i386/winnt.c: Include except.h.
1571 (struct seh_frame_state): Add reg_offset, after_prologue and
1572 in_cold_section fields.
1573 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1574 (i386_pe_seh_cold_init): New function.
1575 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1576 to seh->in_cold_section.
1577 (seh_emit_push): Record the offset of the push.
1578 (seh_emit_save): Record the offet of the save.
1579 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1580 Test seh->after_prologue to disregard the epilogue.
1581 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1582 (i386_pe_end_cold_function): New function.
1584 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1587 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1588 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1591 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1593 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1594 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1597 * multiple_target.c: Include tree-inline.h and intl.h.
1598 (expand_target_clones): Diagnose and fail if node->definition and
1599 !tree_versionable_function_p (node->decl).
1601 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1603 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1605 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1606 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1607 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1609 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1612 * combine.c (try_combine): Use reg_or_subregno instead of handling
1613 just paradoxical SUBREGs and REGs.
1615 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1617 * config/arc/arc.c (arc_finalize_pic): Remove function.
1618 (arc_must_save_register): We use single base PIC register, remove
1619 checks to save/restore the PIC register.
1620 (arc_expand_prologue): Likewise.
1621 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1623 (arc_verify_short): Likewise.
1624 (arc_attr_type): Likewise.
1625 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1626 (walk_stores): Likewise.
1627 (arc_address_cost): Make it static.
1628 (arc_verify_short): Likewise.
1629 (branch_dest): Likewise.
1630 (arc_attr_type): Likewise.
1631 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1632 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1633 (arc_final_prescan_insn): Remove inserting the nops due to
1634 hardware hazards. It is done in reorg step.
1635 (insn_length_variant_t): Remove.
1636 (insn_length_parameters_t): Likewise.
1637 (arc_insn_length_parameters): Likewise.
1638 (arc_get_insn_variants): Likewise.
1639 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1641 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1644 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1647 PR tree-optimization/84687
1648 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1650 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1652 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1654 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1655 Rename to ppc_speculation_barrier.
1656 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1657 __builtin_ppc_speculation_barrier.
1659 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1662 * combine.c (combine_simplify_rtx): Don't try to simplify if
1663 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1666 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1668 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1669 to 32 bytes when compiling for POWER9.
1671 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1674 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1675 regparm >= 3 with no arg reg available also for calls with
1676 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1679 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1681 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1683 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1686 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1688 2018-03-05 Richard Biener <rguenther@suse.de>
1690 PR tree-optimization/84486
1691 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1692 When inserting a __builtin_assume_aligned call set the LHS
1693 SSA name alignment info accordingly.
1695 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1697 PR tree-optimization/84114
1698 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1699 Avoid reassociation of FLOAT_MODE addition.
1701 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1703 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1704 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1705 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1706 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1707 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1708 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1709 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1711 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1712 __builtin_ia32_wbinvd): New builtins.
1713 (SPECIAL_ARGS2): New.
1714 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1715 (SPECIAL_ARGS2): New.
1716 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1717 (ix86_valid_target_attribute_inner_p): Ditto.
1718 (ix86_init_mmx_sse_builtins): Add special_args2.
1719 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1720 TARGET_WBNOINVD_P): New.
1721 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1722 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1723 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1724 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1725 * config/i386/pconfigintrin.h: New file.
1726 * config/i386/wbnoinvdintrin.h: Ditto.
1727 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1728 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1730 2018-03-05 Richard Biener <rguenther@suse.de>
1732 PR tree-optimization/84670
1733 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1735 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1736 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1737 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1738 doesn't grow if all successors (recursively) were visited at least
1741 2018-03-05 Richard Biener <rguenther@suse.de>
1743 PR tree-optimization/84650
1744 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1745 if executed in the loop pipeline.
1747 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1749 * doc/configfiles.texi (Configuration Files): Move info about
1750 conditionalizing $target-protos.h to...
1751 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1752 differs from $target-protos.h.
1754 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1755 Chung-Ju Wu <jasonwucj@gmail.com>
1757 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1758 * config/nds32/nds32-multiple.md (setmemsi): Define.
1759 * config/nds32/nds32-memory-manipulation.c
1760 (nds32_gen_dup_4_byte_to_word_value): New.
1761 (emit_setmem_word_loop): New.
1762 (emit_setmem_byte_loop): New.
1763 (nds32_expand_setmem_loop): New.
1764 (nds32_expand_setmem_loop_v3m): New.
1765 (nds32_expand_setmem_unroll): New.
1766 (nds32_expand_setmem): New.
1768 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1769 Chung-Ju Wu <jasonwucj@gmail.com>
1771 * config/nds32/nds32-memory-manipulation.c
1772 (nds32_emit_load_store): New.
1773 (nds32_emit_post_inc_load_store): New.
1774 (nds32_emit_mem_move): New.
1775 (nds32_emit_mem_move_block): New.
1776 (nds32_expand_movmemsi_loop_unknown_size): New.
1777 (nds32_expand_movmemsi_loop_known_size): New.
1778 (nds32_expand_movmemsi_loop): New.
1779 (nds32_expand_movmemsi_unroll): New.
1780 (nds32_expand_movmemqi): Rename ...
1781 (nds32_expand_movmemsi): ... to this.
1782 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1783 (movmemsi): ... to this.
1784 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1785 (nds32_expand_movmemsi): ... to this.
1787 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1788 Monk Chiang <sh.chiang04@gmail.com>
1789 Chung-Ju Wu <jasonwucj@gmail.com>
1791 * config/nds32/nds32-protos.h
1792 (nds32_expand_load_multiple): New arguments.
1793 (nds32_expand_store_multiple): Ditto.
1794 (nds32_valid_multiple_load_store): Rename ...
1795 (nds32_valid_multiple_load_store_p): ... to this.
1796 * config/nds32/nds32-memory-manipulation.c
1797 (nds32_expand_load_multiple): Refine implementation.
1798 (nds32_expand_store_multiple): Ditto.
1799 * config/nds32/nds32-multiple.md
1800 (load_multiple): Update nds32_expand_load_multiple interface.
1801 (store_multiple): Update nds32_expand_store_multiple interface.
1802 * config/nds32/nds32-predicates.c
1803 (nds32_valid_multiple_load_store): Rename ...
1804 (nds32_valid_multiple_load_store_p): ... to this and refine
1806 * config/nds32/predicates.md
1807 (nds32_load_multiple_and_update_address_operation): New predicate.
1808 (nds32_store_multiple_and_update_address_operation): New predicate.
1810 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1811 Chung-Ju Wu <jasonwucj@gmail.com>
1813 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1814 (combo): New attribute.
1815 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1817 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1819 * config/nds32/nds32.opt: Change -mcmodel= default value.
1821 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1822 Monk Chiang <sh.chiang04@gmail.com>
1823 Chung-Ju Wu <jasonwucj@gmail.com>
1825 * config/nds32/constants.md (unspec_element): New enum.
1826 * config/nds32/constraints.md (Umw): New constraint.
1827 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1828 * config/nds32/nds32-intrinsic.md: Likewise.
1829 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1830 (nds32_valid_smw_lwm_base_p): New.
1831 (nds32_output_smw_single_word): New.
1832 (nds32_output_lmw_single_word): New.
1833 (nds32_expand_unaligned_load): New.
1834 (nds32_expand_unaligned_store): New.
1835 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1836 (nds32_output_smw_single_word): Declare.
1837 (nds32_output_lmw_single_word): Declare.
1838 (nds32_expand_unaligned_load): Declare.
1839 (nds32_expand_unaligned_store): Declare.
1840 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1841 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1842 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1843 NDS32_BUILTIN_UASTORE_DW.
1844 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1847 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1848 Kito Cheng <kito.cheng@gmail.com>
1849 Chung-Ju Wu <jasonwucj@gmail.com>
1851 * config/nds32/nds32-intrinsic.c
1852 (nds32_expand_builtin_null_ftype_reg): Delete.
1853 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1854 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1855 (nds32_read_argument): New.
1856 (nds32_legitimize_target): Ditto.
1857 (nds32_legitimize_argument): Ditto.
1858 (nds32_check_constant_argument): Ditto.
1859 (nds32_expand_unop_builtin): Ditto.
1860 (nds32_expand_unopimm_builtin): Ditto.
1861 (nds32_expand_binop_builtin): Ditto.
1862 (nds32_builtin_decl_impl): Ditto.
1863 (builtin_description): Ditto.
1864 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1865 (nds32_init_builtins_impl): Ditto.
1866 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1867 (nds32_builtin_decl): New.
1868 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1869 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1871 2018-03-02 Jeff Law <law@redhat.com>
1873 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1874 (redundant_insn, fill_simple_delay_slots): Likewise.
1875 (fill_slots_from_thread): Likewise.
1876 * resource.c (mark_referenced_resources): Likewise.
1877 (mark_set_resources, find_dead_or_set_registers): Likewise.
1879 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1881 * substring-locations.h (format_warning_va): Formatting fix for
1883 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1885 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1886 * substring-locations.c: Include intl.h.
1887 (format_warning_va): Turned into small wrapper around
1888 format_warning_n_va, renamed to ...
1889 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1890 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1892 (format_warning_at_substring_n): New function.
1893 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1894 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1895 format_warning_at_substring with just a shorter name instead of
1896 const function pointer.
1897 (fmtwarn_n): New function.
1898 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1899 appropriate, get rid of all the fmtstr temporaries, move conditionals
1900 with G_() wrapped string literals directly into fmtwarn arguments,
1901 cast dir.len to (int), formatting fixes.
1903 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1905 * doc/invoke.texi: Remove "Cilk Plus" references.
1907 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1908 Richard Biener <rguenther@suse.de>
1911 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1912 for error or warning attributes if CALL_FROM_THUNK_P is set.
1915 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1918 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1919 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1922 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1923 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1925 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1926 instead of -1U in last predictors element's probability member.
1928 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1931 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1932 arguments if they are comparable.
1934 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1936 PR tree-optimization/84634
1937 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1938 masks and masked_loop_p with a single loop_masks, making sure it's
1939 null for bb vectorization.
1941 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1943 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1944 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1945 loop->force_vectorize to check whether there is no alias.
1947 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1950 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1952 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1954 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1955 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1956 instead of a loop around prev_real_insn.
1957 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1961 * config/i386/i386.c (ix86_print_operand): Use conditional
1962 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1965 2018-03-02 Richard Biener <rguenther@suse.de>
1967 PR tree-optimization/84427
1968 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1969 (bitmap_set_subtract_values): Rewrite to handle multiple
1972 (prune_clobbered_mems): Likewise.
1973 (phi_translate): Take edge instead of pred/phiblock.
1974 (phi_translate_1): Likewise.
1975 (phi_translate_set): Likewise. Insert all translated
1976 exprs for a value into the set, keeping possibly multiple
1977 expressions per value.
1978 (compute_antic_aux): Adjust for phi_translate changes.
1979 When intersecting union the expressions and prune those
1980 not in the final value set, keeping possibly multiple
1981 expressions per value. Do not use value-insertion
1982 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1983 all expressions. Add verification that the value-sets
1984 only shrink during iteration.
1985 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1986 (do_pre_regular_insertion): Likewise.
1987 (do_pre_partial_partial_insertion): Likewise.
1989 2018-03-02 Richard Biener <rguenther@suse.de>
1992 * config/darwin.c (saved_debug_info_level): New static global.
1993 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1994 (darwin_asm_lto_end): Restore debug info generation settings.
1996 2018-03-01 Martin Liska <mliska@suse.cz>
1999 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2002 2018-03-01 Richard Biener <rguenther@suse.de>
2005 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2006 type annotation with LTO when debug was disabled at compile-time.
2008 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2010 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2012 (mips_final_postscan_insn): Likewise.
2014 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2016 PR rtl-optimization/84528
2017 * alias.c (init_alias_target): Add commentary.
2018 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2019 a unique base value if the frame pointer is not eliminated
2020 to the stack pointer.
2022 2018-03-01 Tom de Vries <tom@codesourcery.com>
2024 PR rtl-optimization/83327
2025 * lra-int.h (hard_regs_spilled_into): Declare.
2026 * lra.c (hard_regs_spilled_into): Define.
2027 (init_reg_info): Init hard_regs_spilled_into.
2028 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2029 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2030 (process_bb_lives): Handle hard_regs_spilled_into.
2031 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2032 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2034 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2036 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2037 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2038 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2039 * config/rs6000/aix72.h: New file.
2041 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2043 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2044 instead of warning_at with conditional singular and plural messages
2048 * stor-layout.c (update_alignment_for_field): For
2049 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2050 && !DECL_PACKED (field), do the alignment update, just use
2051 only desired_align instead of MAX (type_align, desired_align)
2053 (place_field): Don't do known_align < desired_align handling
2054 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2055 is non-NULL, instead do it after rli->prev_field handling and
2056 only if not within a bitfield word. For DECL_PACKED (field)
2057 use type_align of BITS_PER_UNIT.
2059 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2061 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2062 superfluous parentheses and trailing spaces.
2064 2018-02-28 Richard Biener <rguenther@suse.de>
2066 PR tree-optimization/84584
2067 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2068 SCoPs with fake exit edge.
2070 2018-02-28 Martin Liska <mliska@suse.cz>
2073 * timevar.c (timer::print): Fix format to properly print 100%
2076 2018-02-28 Richard Biener <rguenther@suse.de>
2079 * genmatch.c (capture_info::walk_match): Do not mark
2080 captured expressions without operands as expr_p given
2081 they act more like predicates and should be subject to
2082 "lost tail" side-effect preserving.
2084 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2086 PR rtl-optimization/81611
2087 * auto-inc-dec.c (attempt_change): Move dead note from
2088 mem_insn if it's the next use of regno
2089 (find_address): Take address use of reg holding
2090 non-incremented value. Add parm to limit search to the named
2092 (merge_in_block): Attempt to use a mem insn that is the next
2093 use of the original regno.
2095 2018-02-27 Martin Sebor <msebor@redhat.com>
2098 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2099 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2101 2018-02-27 Martin Sebor <msebor@redhat.com>
2103 PR translation/84207
2104 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2105 n argument to unsigned HOST_WIDE_INT.
2106 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2107 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2108 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2109 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2111 2018-02-27 Richard Biener <rguenther@suse.de>
2113 PR tree-optimization/84512
2114 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2115 Do not use the estimate returned from record_stmt_cost for
2116 the scalar iteration cost but sum properly using add_stmt_cost.
2118 2018-02-27 Richard Biener <rguenther@suse.de>
2120 PR tree-optimization/84466
2121 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2122 Adjust last change to less strictly validate use operands.
2124 2018-02-27 Martin Liska <mliska@suse.cz>
2126 PR gcov-profile/84548
2127 * gcov.c (process_file): Allow partial overlap and consider it
2128 also as group functions.
2129 (output_lines): Properly calculate range of lines for a group.
2131 2018-02-27 Martin Liska <mliska@suse.cz>
2133 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2134 'ggc' suffixes. Change first column width.
2135 (timer::print): Fix formatting of the column.
2137 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2139 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2140 preserve inline entry blocks for the sake of debug inline
2141 entry point markers alone.
2142 (remove_unused_locals): Suggest in comments a better place to
2143 force the preservation of inline entry blocks that are
2144 otherwise unused, but do not preserve them.
2146 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2148 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2150 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2153 * config/i386/constraints.md (Bs): Replace
2154 ix86_indirect_branch_register with
2155 TARGET_INDIRECT_BRANCH_REGISTER.
2157 * config/i386/i386.md (indirect_jump): Likewise.
2158 (tablejump): Likewise.
2159 (*sibcall_memory): Likewise.
2160 (*sibcall_value_memory): Likewise.
2161 Peepholes of indirect call and jump via memory: Likewise.
2162 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2163 (*sibcall_value_GOT_32): Likewise.
2164 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2165 (GOT_memory_operand): Likewise.
2166 (call_insn_operand): Likewise.
2167 (sibcall_insn_operand): Likewise.
2168 (GOT32_symbol_operand): Likewise.
2169 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2171 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2173 PR rtl-optimization/83496
2174 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2175 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2176 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2177 redundant insn, if any.
2178 (relax_delay_slots): Likewise.
2179 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2181 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2183 PR tree-optimization/83965
2184 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2185 that grouped statements are part of a reduction chain. Return
2186 true if the statement is not marked as a reduction itself but
2188 (vect_recog_dot_prod_pattern): Don't check whether the statement
2189 is part of a group here.
2190 (vect_recog_sad_pattern): Likewise.
2191 (vect_recog_widen_sum_pattern): Likewise.
2193 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2196 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2198 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2201 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2203 (ix86_output_indirect_function_return): New prototype.
2204 (ix86_split_simple_return_pop_internal): Likewise.
2205 * config/i386/i386.c (indirect_return_via_cx): New.
2206 (indirect_return_via_cx_bnd): Likewise.
2207 (indirect_thunk_name): Handle return va CX_REG.
2208 (output_indirect_thunk_function): Create alias for
2209 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2210 (ix86_output_indirect_jmp): Remove the bool argument.
2211 (ix86_output_indirect_function_return): New function.
2212 (ix86_split_simple_return_pop_internal): Likewise.
2213 * config/i386/i386.md (*indirect_jump): Don't pass false
2214 to ix86_output_indirect_jmp.
2215 (*tablejump_1): Likewise.
2216 (simple_return_pop_internal): Change it to define_insn_and_split.
2217 Call ix86_split_simple_return_pop_internal to split it for
2219 (simple_return_indirect_internal): Call
2220 ix86_output_indirect_function_return instead of
2221 ix86_output_indirect_jmp.
2223 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2226 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2227 memset and value initialization afterwards.
2229 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2231 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2233 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2236 * common/config/aarch64/aarch64-common.c
2237 (aarch_option_optimization_table[]): Switch
2238 off fomit-frame-pointer
2240 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2241 Chung-Ju Wu <jasonwucj@gmail.com>
2243 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2245 (store_multiple): Ditto.
2247 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2249 * config.gcc: Add --with-cpu support for nds32 target.
2250 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2251 * config/nds32/nds32.opt: Add -mcpu= option.
2253 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2255 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2256 isel=yes): Warn for these deprecated options.
2258 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2260 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2261 ISA_2_5_MASKS_EMBEDDED.
2263 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2265 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2266 p->max as pointers rather than using iterative_hash_expr.
2268 2018-02-23 Carl Love <cel@us.ibm.com>
2270 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2271 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2273 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2274 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2275 P8V_BUILTIN_VEC_VUNSIGNED2.
2277 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2280 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2281 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2283 * lra-constraints.c (curr_insn_transform): Set up
2284 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2286 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2287 * lra-eliminations.c (spill_pseudos): Ditto.
2288 (process_insn_for_elimination): Ditto.
2289 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2290 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2291 LRA_NON_CLOBBERED_ALT.
2293 2018-02-22 Martin Sebor <msebor@redhat.com>
2295 PR tree-optimization/84480
2296 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2297 to maybe_diag_stxncpy_trunc. Call it.
2298 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2299 from gimple_fold_builtin_strcpy. Print inlining stack.
2300 (handle_builtin_stxncpy): Print inlining stack.
2301 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2303 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2306 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2307 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2308 and -fcheck-pointer-bounds are used together.
2309 (indirect_thunk_prefix): New enum.
2310 (indirect_thunk_need_prefix): New function.
2311 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2312 "_nt" instead of "_bnd" for NOTRACK prefix.
2313 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2314 (output_indirect_thunk_function): Likewise.
2316 (ix86_code_end): Update output_indirect_thunk_function calls.
2317 (ix86_output_indirect_branch_via_reg): Replace
2318 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2319 (ix86_output_indirect_branch_via_push): Likewise.
2320 (ix86_output_function_return): Likewise.
2321 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2322 incompatible with -fcf-protection=branch and
2323 -fcheck-pointer-bounds.
2325 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2328 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2329 Change gcc_assert call to output_operand_lossage.
2331 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2333 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2335 2018-02-22 DJ Delorie <dj@redhat.com>
2336 Sebastian Perta <sebastian.perta@renesas.com>
2337 Oleg Endo <olegendo@gcc.gnu.org>
2339 * config/rx/rx.c (rx_rtx_costs): New function.
2340 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2342 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2344 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2346 2018-02-22 Martin Liska <mliska@suse.cz>
2349 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2350 Add "native" as a possible value.
2352 2018-02-22 Martin Liska <mliska@suse.cz>
2355 * config/i386/i386.c (ix86_option_override_internal):
2356 Add "native" as a possible value for -march and -mtune.
2358 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2361 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2362 to all type variants.
2364 PR tree-optimization/84503
2365 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2366 width as info->bitpos + info->bitsize - start.
2367 (merged_store_group::merge_overlapping): Simplify width computation.
2368 (check_no_overlap): New function.
2369 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2370 start + width and last_order of the group, fail if check_no_overlap
2372 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2373 to group if check_no_overlap fails.
2375 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2377 * config/rs6000/altivec.md: Delete contraint arguments to
2378 define_expand, define_split, and define_peephole2, and in
2379 define_insn_and_split if always unused.
2380 * config/rs6000/darwin.md: Ditto.
2381 * config/rs6000/dfp.md: Ditto.
2382 * config/rs6000/rs6000.md: Ditto.
2383 * config/rs6000/sync.md: Ditto.
2384 * config/rs6000/vector.md: Ditto.
2385 * config/rs6000/vsx.md: Ditto.
2387 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2389 * config/rs6000/altivec.md: Write output control strings as braced
2390 blocks instead of double-quoted strings.
2391 * config/rs6000/darwin.md: Ditto.
2392 * config/rs6000/rs6000.md: Ditto.
2393 * config/rs6000/vector.md: Ditto.
2394 * config/rs6000/vsx.md: Ditto.
2396 2018-02-21 Jason Merrill <jason@redhat.com>
2398 PR c++/84314 - ICE with templates and fastcall attribute.
2399 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2401 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2403 * ipa-cp.c (determine_versionability): Fix comment typos.
2405 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2408 * ipa-cp.c (determine_versionability): Do not version functions caling
2411 2018-02-21 Martin Liska <mliska@suse.cz>
2414 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2415 Add "native" as a possible value.
2416 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2417 the macro when native cpu detection is available.
2419 2018-02-21 Martin Liska <mliska@suse.cz>
2422 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2423 Add "native" as a possible value.
2424 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2425 when native cpu detection is available.
2427 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2428 Martin Sebor <msebor@redhat.com>
2430 PR tree-optimization/84478
2431 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2433 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2434 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2435 support which is conservatively correct, for 2 only stay conservative
2436 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2437 argument to the 2 argument get_range_strlen, adjust 6 arg
2438 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2440 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2441 (gimple_fold_builtin_strlen): Pass true as last argument to
2444 2018-02-20 Martin Sebor <msebor@redhat.com>
2447 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2448 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2449 (builtin_memref::builtin_memref): Factor out parts into
2450 set_base_and_offset and call it.
2452 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2455 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2456 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2457 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2458 search at the associated MODE_INT.
2460 2018-02-20 Jeff Law <law@redhat.com>
2463 PR tree-optimization/81592
2465 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2466 for range data rather than using global data.
2467 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2468 range data rather than using global data.
2469 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2470 pass it to children as needed.
2471 (struct directive::fmtresult): Similarly.
2472 (struct directive::set_width): Similarly.
2473 (struct directive::set_precision): Similarly.
2474 (format_integer, format_directive, parse_directive): Similarly.
2475 (format_none): Accept unnamed vr_values parameter.
2476 (format_percent, format_floating, format_character): Similarly.
2477 (format_string, format_plain): Similarly.
2478 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2479 the EVRP range analyzer for range data rather than using global data.
2480 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2481 gimple-ssa-evrp-analyze.h
2482 (class sprintf_dom_walker): Add after_dom_children member function.
2483 Add evrp_range_analyzer member.
2484 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2485 range analyzer as needed.
2486 (sprintf_dom_walker::after_dom_children): New member function.
2487 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2489 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2490 (evrp_range_analyzer::pop_to_marker): Likewise.
2492 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2494 PR tree-optimization/84419
2495 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2496 with the required type if its current type is compatible but
2499 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2502 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2503 after vectorization.
2505 2018-02-20 Martin Liska <mliska@suse.cz>
2508 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2509 possible values if we don't have a hint.
2511 2018-02-20 Martin Liska <mliska@suse.cz>
2515 * final.c (shorten_branches): Build align_tab array with one
2517 * opts.c (finish_options): Add alignment option limit check.
2518 (MAX_CODE_ALIGN): Likewise.
2519 (MAX_CODE_ALIGN_VALUE): Likewise.
2520 * doc/invoke.texi: Document maximum allowed option value for
2521 all -falign-* options.
2523 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2526 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2527 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2528 * var-tracking.c (emit_note_insn_var_location): Remove all references
2529 to NOTE_INSN_CALL_ARG_LOCATION.
2530 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2531 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2532 Use copy_rtx_if_shared.
2533 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2534 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2535 (dwarf2out_var_location): Remove handling of
2536 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2538 * final.c (final_scan_insn): Remove all references to
2539 NOTE_INSN_CALL_ARG_LOCATION.
2540 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2541 before dumping final insns.
2542 * except.c (emit_note_eh_region_end): Remove all references to
2543 NOTE_INSN_CALL_ARG_LOCATION.
2544 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2545 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2546 * config/arc/arc.c (hwloop_optimize): Likewise.
2547 * config/arm/arm.c (create_fix_barrier): Likewise.
2548 * config/s390/s390.c (s390_chunkify_start): Likewise.
2549 * config/sh/sh.c (find_barrier): Likewise.
2550 * config/i386/i386.c (rest_of_insert_endbranch,
2551 ix86_seh_fixup_eh_fallthru): Likewise.
2552 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2553 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2554 * config/frv/frv.c (frv_function_prologue): Likewise.
2555 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2557 (note_outside_basic_block_p): Remove all references to
2558 NOTE_INSN_CALL_ARG_LOCATION.
2559 * gengtype.c (adjust_field_rtx_def): Likewise.
2560 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2562 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2563 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2566 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2569 PR tree-optimization/84452
2570 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2571 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2574 2018-02-19 Martin Liska <mliska@suse.cz>
2577 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2579 2018-02-19 Martin Liska <mliska@suse.cz>
2580 Richard Sandiford <richard.sandiford@linaro.org>
2582 PR tree-optimization/82491
2583 * gimple-fold.c (get_base_constructor): Make earlier bail out
2586 2018-02-19 Carl Love <cel@us.ibm.com>
2588 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2589 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2591 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2592 P8V_BUILTIN_VEC_NEG.
2594 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2596 * config/rl78/rl78.md (movdf): New define expand.
2598 2018-02-19 Martin Liska <mliska@suse.cz>
2601 * doc/invoke.texi: Fix typo.
2602 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2604 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2606 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2607 handle rs6000_single_float and rs6000_double_float specially for
2610 2018-02-16 Jeff Law <law@redhat.com>
2612 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2615 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2618 * ipa-inline.c (inline_small_functions): Fix a typo.
2620 2018-02-16 Nathan Sidwell <nathan@acm.org>
2622 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2624 2018-02-16 Carl Love <cel@us.ibm.com>
2626 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2627 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2628 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2629 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2630 expansion to P8V_BUILTIN_VEC_FLOAT2.
2632 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2634 PR rtl-optimization/70023
2635 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2636 src_regno into account.
2638 2018-02-16 Carl Love <cel@us.ibm.com>
2640 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2641 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2642 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2643 * config/rs6000/rs6000.c: Remove case statements for
2644 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2645 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2646 and P9V_BUILTIN_VEC_VINSERT4B.
2647 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2648 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2649 * config/rs6000/vsx.md:
2650 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2653 2018-02-16 Carl Love <cel@us.ibm.com>
2655 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2657 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2659 * config/rs6000/rs6000-c.c: Add the definitions for
2660 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2661 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2662 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2663 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2664 definition for insert4b and define insn *insert3b_internal.
2665 * doc/extend.texi: Add documentation for vec_extract4b.
2667 2018-02-16 Nathan Sidwell <nathan@acm.org>
2669 * doc/extend.texi (Backwards Compatibility): Mention friend
2670 injection. Note for-scope is deprecated.
2671 * doc/invoke.texi (-ffriend-injection): Deprecate.
2673 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2675 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2676 that moved to I2, also allow destinations that are a paradoxical
2677 subreg (instead of a normal reg).
2679 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2682 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2685 2018-02-16 Richard Biener <rguenther@suse.de>
2687 PR tree-optimization/84037
2688 PR tree-optimization/84016
2690 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2691 Adjust vec_construct for the fact we need additional higher latency
2692 128bit inserts for AVX256 and AVX512 vector builds.
2693 (ix86_add_stmt_cost): Scale vector construction cost for
2696 2018-02-16 Richard Biener <rguenther@suse.de>
2698 PR tree-optimization/84417
2699 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2700 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2701 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2703 2018-02-16 Martin Liska <mliska@suse.cz>
2706 * internal-fn.def (ASAN_CHECK): Set proper flags.
2707 (ASAN_MARK): Likewise.
2709 2018-02-16 Julia Koval <julia.koval@intel.com>
2711 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2712 from PTA_CANNONLAKE.
2714 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2717 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2718 Use ++iter rather than iter++ for std::list iterators.
2719 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2720 defer deleting them until all nodes in the forest are processed. Do
2721 free even leaf nodes. Change to_process into auto_vec.
2724 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2725 * vec.h (vec_default_construct): Use memset instead of placement new
2726 if BROKEN_VALUE_INITIALIZATION is defined.
2727 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2728 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2731 PR rtl-optimization/83723
2732 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2733 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2734 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2736 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2738 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2740 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2742 PR rtl-optimization/81443
2743 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2744 from inner REGs to paradoxical SUBREGs.
2746 2018-02-16 Richard Biener <rguenther@suse.de>
2748 PR tree-optimization/84399
2749 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2750 For operands we can analyze at their definition make sure we can
2751 analyze them at each use as well.
2753 2018-02-16 Richard Biener <rguenther@suse.de>
2755 PR tree-optimization/84190
2756 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2757 volatile accesses if the decl isn't volatile.
2759 2018-02-15 Jason Merrill <jason@redhat.com>
2761 PR c++/84314 - ICE with templates and fastcall attribute.
2762 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2763 TYPE_CANONICAL on an existing type.
2765 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2767 PR tree-optimization/84383
2768 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2769 dstoff nor call operand_equal_p if dstbase is NULL.
2771 PR tree-optimization/84334
2772 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2773 also a CONSTANT_CLASS_P, punt.
2775 2018-02-14 Jim Wilson <jimw@sifive.com>
2777 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2778 first SMALL_OPERAND check. New local min_second_step. Move assert
2779 to where locals are set. Add TARGET_RVC support.
2780 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2782 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2784 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2786 2018-02-14 Martin Sebor <msebor@redhat.com>
2788 PR tree-optimization/83698
2789 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2790 arrays constrain the offset range to their bounds.
2791 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2792 (builtin_access::overlap): Avoid setting the size of overlap if it's
2794 (maybe_diag_overlap): Also consider arrays when deciding what values
2795 of offsets to include in diagnostics.
2797 2018-02-14 Martin Sebor <msebor@redhat.com>
2800 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2801 that correspond to the kind of a declaration.
2803 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2806 * config/pa/pa.md: Load address of PIC label using the linkage table
2807 if the label is nonlocal.
2809 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2811 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2812 warning message if user requests -maltivec=be.
2813 * doc/invoke.texi: Document deprecation of -maltivec=be.
2815 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2818 * config/rs6000/rs6000-c.c: Update definitions for
2819 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2820 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2822 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2825 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2826 add _get_ssp intrinsics. Remove argument from
2827 __builtin_ia32_rdssp[d|q].
2828 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2829 * config/i386/i386-builtin.def: Remove argument from
2830 __builtin_ia32_rdssp[d|q].
2831 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2832 ix86_expand_special_args_builtin for _rdssp[d|q].
2833 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2834 Clear register before usage.
2835 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2836 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2838 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2840 PR tree-optimization/84357
2841 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2842 operand 1 of an ARRAY_REF too.
2844 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2847 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2848 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2850 (set_of_reg): New struct.
2851 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2852 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2853 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2855 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2856 Split into bitclr, bitset, bitinvert patterns if appropriate.
2857 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2858 use rx_fuse_in_memory_bitop.
2859 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2860 to named insn, correct maximum insn length.
2862 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2865 * machmode.def: Define a complex mode for PARTIAL_INT.
2866 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2868 * doc/rtl.texi: Document CSPImode.
2869 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2871 (msp430_hard_regno_nregs_with_padding): Likewise.
2873 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2876 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2878 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2880 PR rtl-optimization/84169
2881 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2882 we generated a parallel as new i3 and we split that to new i2 and i3
2883 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2884 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2885 those to i2, not i1. Partially rewrite this scan code.
2887 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2890 * stor-layout.c (place_field): For variable length fields, adjust
2891 offset_align afterwards not just based on the field's alignment,
2892 but also on the size.
2895 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2896 of exps and logs in the use_exp2 case.
2898 2018-02-13 Jeff Law <law@redhat.com>
2900 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2903 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2906 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2910 * common.opt (gas-loc-support, gas-locview-support): New.
2911 (ginline-points, ginternal-reset-location-views): New.
2912 * doc/invoke.texi: Document them. Use @itemx where intended.
2913 (gvariable-location-views): Adjust.
2914 * target.def (reset_location_view): New.
2915 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2916 (TARGET_RESET_LOCATION_VIEW): New.
2917 * doc/tm.texi: Rebuilt.
2918 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2919 (dwarf2out_default_as_locview_support): New.
2920 (output_asm_line_debug_info): Use option variables.
2921 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2922 (output_loc_list): Likewise.
2923 (add_high_low_attributes): Check option variables.
2924 Don't output entry view attribute in strict mode.
2925 (gen_inlined_subroutine_die): Check option variables.
2926 (dwarf2out_inline_entry): Likewise.
2927 (init_sections_and_labels): Likewise.
2928 (dwarf2out_early_finish): Likewise.
2929 (maybe_reset_location_view): New, from...
2930 (dwarf2out_var_location): ... here. Call it.
2931 * debug.h (dwarf2out_default_as_loc_support): Declare.
2932 (dwarf2out_default_as_locview_support): Declare.
2933 * hooks.c (hook_int_rtx_insn_0): New.
2934 * hooks.h (hook_int_rtx_insn_0): Declare.
2935 * toplev.c (process_options): Take -gas-loc-support and
2936 -gas-locview-support from dwarf2out. Enable
2937 -gvariable-location-views by default only with locview
2938 assembler support. Enable -ginternal-reset-location-views by
2939 default only if the target defines the corresponding hook.
2940 Enable -ginline-points by default if location views are
2941 enabled; force it disabled if statement frontiers are
2943 * tree-inline.c (expand_call_inline): Check option variables.
2944 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2946 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2948 PR tree-optimization/84321
2949 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2950 handling. Also check whether the anti-range contains any values
2951 that satisfy the mask; switch to a VR_RANGE if not.
2953 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2956 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2958 2018-02-13 Martin Jambor <mjambor@suse.cz>
2961 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2962 of call statements, also set location of a load to a temporary.
2964 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2966 * config/rl78/rl78.c (add_vector_labels): New function.
2967 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2968 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2969 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2970 which checks that no arguments are passed.
2971 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2972 * doc/extend.texi: Documentation for the new attribute.
2974 2018-02-13 Andreas Schwab <schwab@suse.de>
2976 * config/riscv/linux.h (CPP_SPEC): Define.
2978 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2981 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2982 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2983 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2984 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2985 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2986 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2987 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2988 temporarily for AES and PCLMUL builtins.
2990 PR tree-optimization/84339
2991 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2992 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2996 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2997 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2998 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3000 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3002 * omp-simd-clone.h: New file.
3003 * omp-simd-clone.c: Include omp-simd-clone.h.
3004 (expand_simd_clones): No longer static.
3005 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3006 cgraph.h and omp-simd-clone.h.
3007 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3008 (vect_recog_widen_shift_pattern): Formatting fix.
3009 (vect_pattern_recog_1): Don't check optab for calls.
3012 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3013 operands[2] into a REG before using gen_lowpart on it.
3015 2018-02-12 Jeff Law <law@redhat.com>
3018 * config/sh/sh.c (find_barrier): Consider a sibling call
3021 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3022 successfully back substituting a reg.
3024 2018-02-12 Richard Biener <rguenther@suse.de>
3026 PR tree-optimization/84037
3027 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3028 parameter, move visited init to caller.
3029 (vect_slp_analyze_operations): Separate cost from validity
3030 check, initialize visited once for all instances.
3031 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3033 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3035 (vect_model_promotion_demotion_cost): Likewise.
3036 (vectorizable_bswap): Guard cost modeling with !slp_node
3037 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3039 (vectorizable_call): Likewise.
3040 (vectorizable_conversion): Likewise.
3041 (vectorizable_assignment): Likewise.
3042 (vectorizable_shift): Likewise.
3043 (vectorizable_operation): Likewise.
3044 (vectorizable_store): Likewise.
3045 (vectorizable_load): Likewise.
3046 (vectorizable_condition): Likewise.
3047 (vectorizable_comparison): Likewise.
3049 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3052 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3053 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3054 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3056 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3059 * params.def (inline-min-speedup): Increase from 8 to 15.
3060 (max-inline-insns-auto): Decrease from 40 to 30.
3061 * ipa-split.c (consider_split): Add some buffer for function to
3062 be considered inlining candidate.
3063 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3066 2018-02-12 Richard Biener <rguenther@suse.de>
3068 PR tree-optimization/84037
3069 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3070 matched stmts if we cannot swap the non-matched ones.
3072 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3074 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3075 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3076 _mm_maskz_scalef_round_ss): New intrinsics.
3077 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3078 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3079 __builtin_ia32_scalefss_round): Remove.
3080 (__builtin_ia32_scalefsd_mask_round,
3081 __builtin_ia32_scalefss_mask_round): New intrinsics.
3082 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3083 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3084 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3085 "<round_constraint>")): Changed to ...
3086 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3087 "<round_scalar_constraint>")): ... this.
3088 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3089 %0, %1, %2<round_op3>}"): Changed to ...
3090 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3091 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3092 %2<round_scalar_mask_op3>}"): ... this.
3093 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3095 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3097 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3098 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3099 (_mm_maskz_sqrt_round_ss): New intrinsics.
3100 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3101 (__builtin_ia32_sqrtsd_mask_round)
3102 (__builtin_ia32_sqrtss_mask_round): New builtins.
3103 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3104 (__builtin_ia32_sqrtss_round): Remove.
3105 (__builtin_ia32_sqrtsd_mask_round)
3106 (__builtin_ia32_sqrtss_mask_round): New builtins.
3107 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3108 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3109 ((match_operand:VF_128 1 "vector_operand"
3110 "xBm,<round_constraint>")): Changed to ...
3111 ((match_operand:VF_128 1 "vector_operand"
3112 "xBm,<round_scalar_constraint>")): ... this.
3113 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3114 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3115 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3116 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3117 %<iptr>1<round_scalar_mask_op3>}): ... this.
3118 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3119 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3121 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3124 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3125 Cast vec_cmpeq result to correct type.
3126 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3127 Cast vec_cmpgt result to correct type.
3129 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3131 * final.c (final_scan_insn_1): Renamed from...
3132 (final_scan_insn): ... this. New wrapper, to recover
3133 seen from the outermost call in recursive ones.
3134 * config/sparc/sparc.c (output_return): Drop seen from call.
3135 (output_sibcall): Likewise.
3136 * config/visium/visium.c (output_branch): Likewise.
3138 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3140 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3143 2018-02-10 Alan Modra <amodra@gmail.com>
3146 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3147 Specify LR as an input.
3149 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3152 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3153 remove_member_access_dummy_vars): New functions.
3154 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3155 lower_omp_1, execute_lower_omp): Use them.
3157 PR rtl-optimization/84308
3158 * shrink-wrap.c (spread_components): Release todo vector.
3160 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3162 PR rtl-optimization/57193
3163 * ira-color.c (struct allocno_color_data): Add member
3164 conflict_allocno_hard_prefs.
3165 (update_conflict_allocno_hard_prefs): New.
3166 (bucket_allocno_compare_func): Add a preference based on
3167 conflict_allocno_hard_prefs.
3168 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3169 (color_allocnos): Remove a dead code. Initiate
3170 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3172 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3175 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3176 constraint from =wa to wa. Avoid a subreg on the output operand,
3177 instead use a pseudo and subreg it in a move.
3178 (p9_xxbrd_<mode>): Changed to ...
3179 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3180 (p9_xxbrd_v2df): New expander.
3181 (p9_xxbrw_<mode>): Changed to ...
3182 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3183 (p9_xxbrw_v4sf): New expander.
3185 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3187 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3189 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3192 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3193 multiply in 32-bit mode.
3194 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3195 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3198 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3200 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3201 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3202 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3203 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3205 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3208 * dwarf2out.c (is_trivial_indirect_ref): New function.
3209 (dwarf2out_late_global_decl): Do not generate a location
3210 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3211 and that are not defined in the current unit.
3213 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3215 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3216 instead of a libcall for UNORDERED.
3218 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3221 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3222 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3224 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3227 * config/s390/s390.c (s390_set_current_function): Invoke
3228 s390_indirect_branch_settings also if fndecl didn't change.
3230 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3232 * config/rs6000/rs6000.md (blockage): Set length to zero.
3234 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3236 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3238 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3241 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3242 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3246 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3247 PARALLEL incoming that failed vt_get_decl_and_offset check.
3250 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3251 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3253 (get_variable_section): For decls in named .bss* sections pass true as
3254 second argument to bss_initializer_p.
3256 2018-02-09 Marek Polacek <polacek@redhat.com>
3257 Jakub Jelinek <jakub@redhat.com>
3260 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3261 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3262 Sync some changes from cxx_fold_indirect_ref.
3264 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3266 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3268 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3269 (BLOCK_INLINE_ENTRY_LABEL): New.
3270 (dwarf2out_var_location): Disregard inline entry markers.
3271 (inline_entry_data): New struct.
3272 (inline_entry_data_hasher): New hashtable type.
3273 (inline_entry_data_hasher::hash): New.
3274 (inline_entry_data_hasher::equal): New.
3275 (inline_entry_data_table): New variable.
3276 (add_high_low_attributes): Add DW_AT_entry_pc and
3277 DW_AT_GNU_entry_view attributes if a pending entry is found
3278 in inline_entry_data_table. Add old entry_pc attribute only
3279 if debug nonbinding markers are disabled.
3280 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3281 markers are enabled.
3282 (block_within_block_p, dwarf2out_inline_entry): New.
3283 (dwarf2out_finish): Check that no entries remained in
3284 inline_entry_data_table.
3285 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3286 (final_scan_insn, notice_source_line): Likewise.
3287 (rest_of_clean_state): Skip inline entry markers.
3288 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3290 * gimple.c (gimple_build_debug_inline_entry): New.
3291 * gimple.h (enum gimple_debug_subcode): Add
3292 GIMPLE_DEBUG_INLINE_ENTRY.
3293 (gimple_build_debug_inline_entry): Declare.
3294 (gimple_debug_inline_entry_p): New.
3295 (gimple_debug_nonbind_marker_p): Adjust.
3296 * insn-notes.def (INLINE_ENTRY): New.
3297 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3298 inline entry marker notes.
3299 (print_insn): Likewise.
3300 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3301 (INSN_DEBUG_MARKER_KIND): Likewise.
3302 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3303 * tree-inline.c (expand_call_inline): Build and insert
3304 debug_inline_entry stmt.
3305 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3306 inline entry blocks early, if nonbind markers are enabled.
3307 (dump_scope_block): Dump fragment info.
3308 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3309 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3310 (gimple_build_debug_inline_entry): New.
3311 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3312 Enable/disable inline entry points too.
3313 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3314 (DEBUG_INSN): Describe inline entry markers.
3316 * common.opt (gvariable-location-views): New.
3317 (gvariable-location-views=incompat5): New.
3318 * config.in: Rebuilt.
3319 * configure: Rebuilt.
3320 * configure.ac: Test assembler for view support.
3321 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3322 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3323 * dwarf2out.c (var_loc_view): New typedef.
3324 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3325 (dwarf2out_locviews_in_attribute): New.
3326 (dwarf2out_locviews_in_loclist): New.
3327 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3328 (enum dw_line_info_opcode): Add LI_adv_address.
3329 (struct dw_line_info_table): Add view.
3330 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3331 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3332 (zero_view_p): New variable.
3333 (ZERO_VIEW_P): New macro.
3334 (output_asm_line_debug_info): New.
3335 (struct var_loc_node): Add view.
3336 (add_AT_view_list, AT_loc_list): New.
3337 (add_var_loc_to_decl): Add view param. Test it against last.
3338 (new_loc_list): Add view params. Record them.
3339 (AT_loc_list_ptr): Handle loc and view lists.
3340 (view_list_to_loc_list_val_node): New.
3341 (print_dw_val): Handle dw_val_class_view_list.
3342 (size_of_die): Likewise.
3343 (value_format): Likewise.
3344 (loc_list_has_views): New.
3345 (gen_llsym): Set vl_symbol too.
3346 (maybe_gen_llsym, skip_loc_list_entry): New.
3347 (dwarf2out_maybe_output_loclist_view_pair): New.
3348 (output_loc_list): Output view list or entries too.
3349 (output_view_list_offset): New.
3350 (output_die): Handle dw_val_class_view_list.
3351 (output_dwarf_version): New.
3352 (output_compilation_unit_header): Use it.
3353 (output_skeleton_debug_sections): Likewise.
3354 (output_rnglists, output_line_info): Likewise.
3355 (output_pubnames, output_aranges): Update version comments.
3356 (output_one_line_info_table): Output view numbers in asm comments.
3357 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3358 Call maybe_gen_llsym.
3359 (loc_list_from_tree_1): Adjust.
3360 (add_AT_location_description): Create view list attribute if
3361 needed, check it's absent otherwise.
3362 (convert_cfa_to_fb_loc_list): Adjust.
3363 (maybe_emit_file): Call output_asm_line_debug_info for test.
3364 (dwarf2out_var_location): Reset views as needed. Precompute
3365 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3366 attribute. Set view.
3367 (new_line_info_table): Reset next view.
3368 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3369 (dwarf2out_source_line): Likewise. Output view resets and labels to
3370 the assembler, or select appropriate line info opcodes.
3371 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3372 (optimize_string_length): Catch it. Adjust.
3373 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3374 dw_val_class_view_list, and remove it if no longer needed.
3375 (hash_loc_list): Hash view numbers.
3376 (loc_list_hasher::equal): Compare them.
3377 (optimize_location_lists): Check whether a view list symbol is
3378 needed, and whether the locview attribute is present, and
3379 whether they match. Remove the locview attribute if no longer
3381 (index_location_lists): Call skip_loc_list_entry for test.
3382 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3383 Use output_dwarf_version.
3384 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3385 (struct dw_val_node): Add val_view_list.
3386 * final.c (SEEN_NEXT_VIEW): New.
3387 (set_next_view_needed): New.
3388 (clear_next_view_needed): New.
3389 (maybe_output_next_view): New.
3390 (final_start_function): Rename to...
3391 (final_start_function_1): ... this. Take pointer to FIRST,
3392 add SEEN parameter. Emit param bindings in the initial view.
3393 (final_start_function): Reintroduce SEEN-less interface.
3394 (final): Rename to...
3395 (final_1): ... this. Take SEEN parameter. Output final pending
3396 next view at the end.
3397 (final): Reintroduce seen-less interface.
3398 (final_scan_insn): Output pending next view before switching
3399 sections or ending a block. Mark the next view as needed when
3400 outputting variable locations. Notify debug backend of section
3401 changes, and of location view changes.
3402 (rest_of_handle_final): Adjust.
3403 * toplev.c (process_options): Autodetect value for debug variable
3404 location views option. Warn on incompat5 without -gdwarf-5.
3405 * doc/invoke.texi (gvariable-location-views): New.
3406 (gvariable-location-views=incompat5): New.
3407 (gno-variable-location-views): New.
3409 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3411 PR tree-optimization/84136
3412 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3413 that the result of find_edge is non-NULL.
3415 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3418 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3419 storing integer register in SImode. Fix cost of 256 and 512
3420 byte aligned SSE register store.
3422 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3424 * config/i386/i386.c (ix86_multiplication_cost): Fix
3425 multiplication cost for TARGET_AVX512DQ.
3427 2018-02-08 Marek Polacek <polacek@redhat.com>
3429 PR tree-optimization/84238
3430 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3433 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3435 PR tree-optimization/84265
3436 * tree-vect-stmts.c (vectorizable_store): Don't treat
3437 VMAT_CONTIGUOUS accesses as grouped.
3438 (vectorizable_load): Likewise.
3440 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3442 PR tree-optimization/81635
3443 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3444 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3445 (test_round_for_mask): New functions.
3446 (wide_int_cc_tests): Call test_round_for_mask.
3447 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3448 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3449 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3450 range returned by get_range_info.
3452 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3455 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3456 * symtab.c: Include builtins.h
3457 (symtab_node::output_to_lto_symbol_table_p): Move here
3458 from lto-streamer-out.c:output_symbol_p.
3459 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3460 (output_symbol_p): Move all logic to symtab.c
3461 (produce_symtab): Update.
3463 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3465 * config/s390/s390-opts.h (enum indirect_branch): Define.
3466 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3467 (s390_indirect_branch_via_thunk)
3468 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3469 (enum s390_indirect_branch_type): Define.
3470 * config/s390/s390.c (struct s390_frame_layout, struct
3471 machine_function): Remove.
3472 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3473 (indirect_branch_table_label_no, indirect_branch_table_name):
3475 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3476 (enum s390_indirect_branch_option): Define.
3477 (s390_return_addr_from_memory): New function.
3478 (s390_handle_string_attribute): New function.
3479 (s390_attribute_table): Add new attribute handler.
3480 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3481 (s390_indirect_branch_via_thunk): New function.
3482 (s390_indirect_branch_via_inline_thunk): New function.
3483 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3484 sibling call optimization for non z10 compiles.
3485 (s390_emit_call): Force indirect branch target to be a single
3486 register. Add r1 clobber for non-z10 compiles.
3487 (s390_emit_epilogue): Emit return jump via return_use expander.
3488 (s390_reorg): Handle JUMP_INSNs as execute targets.
3489 (s390_option_override_internal): Perform validity checks for the
3490 new command line options.
3491 (s390_indirect_branch_attrvalue): New function.
3492 (s390_indirect_branch_settings): New function.
3493 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3494 (s390_output_indirect_thunk_function): New function.
3495 (s390_code_end): Implement target hook.
3496 (s390_case_values_threshold): Implement target hook.
3497 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3499 * config/s390/s390.h (struct s390_frame_layout)
3500 (struct machine_function): Move here from s390.c.
3501 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3502 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3503 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3504 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3505 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3506 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3507 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3508 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3509 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3510 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3511 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3512 (mnemonic attribute): Add values which aren't recognized
3514 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3515 pattern for branch conversion. Fix mnemonic attribute.
3516 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3517 indirect branch via thunk if requested.
3518 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3519 ("*indirect_jump"): Disable for branch conversion using out of
3521 ("indirect_jump_via_thunk<mode>_z10")
3522 ("indirect_jump_via_thunk<mode>")
3523 ("indirect_jump_via_inlinethunk<mode>_z10")
3524 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3525 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3526 ("casesi_jump_via_inlinethunk<mode>_z10")
3527 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3528 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3529 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3530 ("*indirect2_jump"): Disable for branch conversion.
3531 ("casesi_jump"): Turn into expander and expand patterns for branch
3533 ("return_use"): New expander.
3534 ("*return"): Emit return via thunk and rename it to ...
3535 ("*return<mode>"): ... this one.
3536 * config/s390/s390.opt: Add new options and and enum for the
3539 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3541 * lra-constraints.c (match_reload): Unconditionally use
3542 gen_lowpart_SUBREG, rather than selecting between that
3543 and equivalent gen_rtx_SUBREG code.
3545 2018-02-08 Richard Biener <rguenther@suse.de>
3547 PR tree-optimization/84233
3548 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3549 changed flag instead of boguously re-using phi_inserted.
3551 2018-02-08 Martin Jambor <mjambor@suse.cz>
3553 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3554 static local variables.
3556 2018-02-08 Richard Biener <rguenther@suse.de>
3558 PR tree-optimization/84278
3559 * tree-vect-stmts.c (vectorizable_store): When looking for
3560 smaller vector types to perform grouped strided loads/stores
3561 make sure the mode is supported by the target.
3562 (vectorizable_load): Likewise.
3564 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3566 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3567 Increase LDP/STP opportunities by adding adjacent callee-saves.
3569 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3571 PR rtl-optimization/84068
3572 PR rtl-optimization/83459
3573 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3575 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3577 PR tree-optimization/84224
3578 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3579 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3582 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3585 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3586 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3587 count, remove one USE.
3589 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3591 * doc/install.texi (Configuration): Document the
3592 --with-long-double-format={ibm,ieee} PowerPC configuration
3596 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3597 Convert from define_expand to be define_insn_and_split. Rework
3598 float/double/_Float128 conversions to QI/HI/SImode to work with
3599 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3600 conversions to QI/HImode types did a store and then a load to
3601 truncate the value. For conversions to VSX registers, don't split
3602 the insn, instead emit the code directly. Use the code iterator
3603 any_fix to combine signed and unsigned conversions.
3604 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3605 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3606 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3607 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3608 (fix_<mode>di2_hw): Likewise.
3609 (fixuns_<mode>di2_hw): Likewise.
3610 (fix_<mode>si2_hw): Likewise.
3611 (fixuns_<mode>si2_hw): Likewise.
3612 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3613 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3614 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3615 fix<uns>_trunc<SFDF:mode>si2_p8.
3616 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3618 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3619 (fix<uns>_<mode>_mem): Likewise.
3620 (fctiw<u>z_<mode>_mem): Likewise.
3621 (fix<uns>_<mode>_mem): Likewise.
3622 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3623 the register allocator from doing a direct move to the GPRs to do
3624 a store, and instead use the ISA 3.0 store byte/half-word from
3625 vector register instruction. For IEEE 128-bit floating point,
3626 also optimize stores of 32-bit ints.
3627 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3629 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3631 * genextract.c (push_pathstr_operand): New function to support
3633 (walk_rtx): Call push_pathstr_operand.
3634 (print_path): Support [a-zA-Z].
3636 2018-02-07 Richard Biener <rguenther@suse.de>
3638 PR tree-optimization/84037
3639 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3640 (cse_and_gimplify_to_preheader): Declare.
3641 (vect_get_place_in_interleaving_chain): Likewise.
3642 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3644 (_loop_vec_info::~_loop_vec_info): Delete it.
3645 (cse_and_gimplify_to_preheader): New function.
3646 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3647 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3648 (vectorizable_load): Likewise. For grouped stores always base
3649 the IV on the first element.
3650 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3651 condition before gimplifying.
3653 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3655 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3656 *DIV_EXPR and *MOD_EXPR.
3658 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3661 * config/i386/i386.c (ix86_option_override_internal): Mask out
3662 the CF_SET bit when checking -fcf-protection.
3664 2018-02-07 Tom de Vries <tom@codesourcery.com>
3667 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3670 2018-02-07 Richard Biener <rguenther@suse.de>
3672 PR tree-optimization/84204
3673 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3676 PR tree-optimization/84205
3677 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3678 special-case isl_ast_op_zdiv_r.
3680 PR tree-optimization/84223
3681 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3682 Only add conditions from within the region.
3683 (gather_bbs::after_dom_children): Adjust.
3685 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3688 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3689 * config/avr/avr.md: Only post-reload split REG-REG moves if
3690 either register is GENERAL_REG_P.
3692 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3694 PR tree-optimization/84235
3695 * tree-ssa-scopedtables.c
3696 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3697 if the subtraction is performed in floating point type where NaNs are
3698 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3699 build 1. Formatting fix.
3701 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3704 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3705 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3706 and skip it regardless of bb boundaries. Use CALL_P macro,
3707 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3708 unnecessarily, formatting fix.
3710 2018-02-06 Michael Collison <michael.collison@arm.com>
3712 * config/arm/thumb2.md:
3713 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3714 (*thumb_mov_notscc): Ditto.
3716 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3719 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3720 unsigned_fix, not "s".
3722 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3724 * configure.ac (gcc_fn_eh_frame_ro): New function.
3725 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3726 correct .eh_frame permissions.
3727 * configure: Regenerate.
3729 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3731 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3734 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3736 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3737 Display warning message for -mno-speculate-indirect-jumps.
3739 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3741 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3743 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3745 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3747 PR tree-optimization/84225
3748 * tree-eh.c (find_trapping_overflow): Only call
3749 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3751 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3754 * config/i386/i386.c: Reimplement the check of possible options
3755 -mibt/-mshstk conbination. Change error messages.
3756 * doc/invoke.texi: Fix a typo: remove extra '='.
3758 2018-02-06 Marek Polacek <polacek@redhat.com>
3760 PR tree-optimization/84228
3761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3763 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3766 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3767 emitted arch directives.
3768 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3769 __ARM_FEATURE_COPROC before changing architectures.
3771 2018-02-06 Richard Biener <rguenther@suse.de>
3773 * config/i386/i386.c (print_reg): Fix typo.
3774 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3776 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3778 * configure: Regenerate.
3780 2018-02-05 Martin Sebor <msebor@redhat.com>
3782 PR tree-optimization/83369
3783 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3786 2018-02-05 Martin Liska <mliska@suse.cz>
3788 * doc/invoke.texi: Cherry-pick upstream r323995.
3790 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3792 * ira.c (ira_init_register_move_cost): Adjust comment.
3794 2018-02-05 Martin Liska <mliska@suse.cz>
3796 PR gcov-profile/84137
3797 * doc/gcov.texi: Fix typo in documentation.
3799 2018-02-05 Martin Liska <mliska@suse.cz>
3801 PR gcov-profile/83879
3802 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3803 using dlopen functionality.
3805 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3807 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3808 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3809 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3810 _mm_maskz_range_round_ss): New intrinsics.
3811 (__builtin_ia32_rangesd128_round)
3812 (__builtin_ia32_rangess128_round): Remove.
3813 (__builtin_ia32_rangesd128_mask_round,
3814 __builtin_ia32_rangess128_mask_round): New builtins.
3815 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3816 __builtin_ia32_rangess128_round): Remove.
3817 (__builtin_ia32_rangesd128_mask_round,
3818 __builtin_ia32_rangess128_mask_round): New builtins.
3819 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3820 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3821 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3822 "<round_saeonly_constraint>")): Changed to ...
3823 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3824 "<round_saeonly_scalar_constraint>")): ... this.
3825 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3826 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3827 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3828 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3829 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3831 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3833 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3835 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3836 Remove all values except native, 8540 and 8548.
3838 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3840 * config/i386/i386.c (ix86_output_function_return): Pass
3841 INVALID_REGNUM, instead of -1, as invalid register number to
3842 indirect_thunk_name and output_indirect_thunk.
3844 2018-02-02 Julia Koval <julia.koval@intel.com>
3846 * config.gcc: Add -march=icelake.
3847 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3848 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3849 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3850 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3851 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3852 (processor_target_table): Add icelake.
3853 (ix86_option_override_internal): Handle new PTAs.
3854 (get_builtin_code_for_version): Handle icelake.
3855 (M_INTEL_COREI7_ICELAKE): New.
3856 (fold_builtin_cpu): Handle icelake.
3857 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3858 * doc/invoke.texi: Add -march=icelake.
3860 2018-02-02 Julia Koval <julia.koval@intel.com>
3862 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3863 to wide_int_bitmask.
3864 * wide-int-bitmask.h: New.
3866 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3869 * config/i386/i386.md: Replace Pmode with word_mode in
3870 builtin_setjmp_setup and builtin_longjmp to support x32.
3872 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3876 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3878 (rs6000_supported_cpu_names): New static variable.
3879 (linux_cpu_translation_table): Likewise.
3880 (elf_platform) <cpu>: Define new static variable and use it.
3881 Translate kernel AT_PLATFORM name to canonical name if needed.
3882 Error if platform name is unknown.
3884 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3887 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3889 2018-02-01 Jeff Law <law@redhat.com>
3892 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3893 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3894 the scratch if RELEASE_VIA_POP is false.
3895 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3896 If we have to save a temporary register, decrement SIZE appropriately.
3897 Pass new arguments to release_scratch_register_on_entry.
3898 (ix86_adjust_stack_and_probe): Likewise.
3899 (ix86_emit_probe_stack_range): Pass new arguments to
3900 release_scratch_register_on_entry.
3902 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3904 PR rtl-optimization/84157
3905 * combine.c (change_zero_ext): Use REG_P predicate in
3906 front of HARD_REGISTER_P predicate.
3908 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3910 * config/avr/avr.c (avr_option_override): Move disabling of
3911 -fdelete-null-pointer-checks to...
3912 * common/config/avr/avr-common.c (avr_option_optimization_table):
3915 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3917 PR tree-optimization/81635
3918 * tree-data-ref.c (split_constant_offset_1): For types that
3919 wrap on overflow, try to use range info to prove that wrapping
3922 2018-02-01 Renlin Li <renlin.li@arm.com>
3925 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3927 (aarch64_register_move_cost): Likewise.
3928 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3930 (REG_CLASS_NAMES): Likewise.
3931 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3932 TAILCALL_ADDR_REGS. Remove IP registers.
3933 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3935 2018-02-01 Richard Biener <rguenther@suse.de>
3937 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3938 for specifying RPO order and allow NULL for that.
3939 * domwalk.c (dom_walker::dom_walker): Likewise.
3940 (dom_walker::walk): Handle NULL RPO order.
3941 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3943 (rewrite_update_dom_walker): Likewise.
3944 (mark_def_dom_walker): Likewise.
3946 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3948 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3949 (aarch64_maybe_expand_sve_subreg_move): Declare.
3950 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3951 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3953 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3954 that are semantically a reverse operation.
3955 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3956 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3957 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3959 (aarch64_can_change_mode_class): For big-endian, forbid changes
3960 between two SVE modes if they have different element sizes.
3962 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3964 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3965 the TImode handling for big-endian targets.
3967 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3969 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3970 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3972 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3973 Remove BSWAP handing for big-endian targets and use the form of
3974 LD1RQ appropariate for the mode.
3976 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3978 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3979 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3982 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3985 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3986 check for operands that need to go through aarch64_sve_reload_be.
3988 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3990 PR tree-optimization/81661
3991 PR tree-optimization/84117
3992 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3993 * tree-eh.c: Include gimplify.h.
3994 (find_trapping_overflow, replace_trapping_overflow,
3995 rewrite_to_non_trapping_overflow): New functions.
3996 * tree-vect-loop.c: Include tree-eh.h.
3997 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3998 * tree-data-ref.c: Include tree-eh.h.
3999 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4001 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4003 PR rtl-optimization/84123
4004 * combine.c (change_zero_ext): Check if hard register satisfies
4005 can_change_dest_mode before calling gen_lowpart_SUBREG.
4007 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4010 * ira.c (ira_init_register_move_cost): Remove assert.
4012 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4014 PR rtl-optimization/84071
4015 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4016 * doc/tm.texi: Regenerate.
4018 2018-01-31 Richard Biener <rguenther@suse.de>
4020 PR tree-optimization/84132
4021 * tree-data-ref.c (analyze_miv_subscript): Properly
4022 check whether evolution_function_is_affine_multivariate_p
4023 before calling gcd_of_steps_may_divide_p.
4025 2018-01-31 Julia Koval <julia.koval@intel.com>
4028 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4029 * config/i386/i386.md (rdpid_rex64) New.
4030 (rdpid): Make 32bit only.
4032 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4035 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4036 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4038 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4041 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4043 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4045 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4047 PR rtl-optimization/84071
4048 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4049 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4051 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4053 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4054 (arc_attribute_table): Add 'aux' attribute.
4055 (arc_in_small_data_p): Consider aux like variables.
4056 (arc_is_aux_reg_p): New function.
4057 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4058 (arc_get_aux_arg): New function.
4059 (prepare_move_operands): Handle aux-register access.
4060 (arc_handle_aux_attribute): New function.
4061 * doc/extend.texi (ARC Variable attributes): Add subsection.
4063 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4065 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4066 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4067 (arc_attribute_table): Add 'uncached' attribute.
4068 (arc_print_operand): Print '.di' flag for uncached memory
4070 (arc_in_small_data_p): Do not consider for small data the uncached
4072 (arc_is_uncached_mem_p): New function.
4073 * config/arc/predicates.md (compact_store_memory_operand): Check
4074 for uncached memory accesses.
4075 (nonvol_nonimm_operand): Likewise.
4076 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4078 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4081 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4082 falign-loops=): Add Optimization flag.
4084 2018-01-30 Jeff Law <law@redhat.com>
4087 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4088 INT_REGISTERS_SAVED. Check it prior to calling
4089 get_scratch_register_on_entry.
4090 (ix86_adjust_stack_and_probe): Similarly.
4091 (ix86_emit_probe_stack_range): Similarly.
4092 (ix86_expand_prologue): Corresponding changes.
4094 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4097 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4098 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4100 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4103 * lra-constraints.c (curr_insn_transform): Process AND in the
4106 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4108 PR rtl-optimization/83986
4109 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4110 dependence against last_pending_memory_flush in addition to
4113 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4115 PR tree-optimization/81611
4116 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4119 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4122 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4125 2018-01-30 Richard Biener <rguenther@suse.de>
4126 Jakub Jelinek <jakub@redhat.com>
4128 PR tree-optimization/84111
4129 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4130 inner loops added during recursion, as they don't have up-to-date
4133 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4136 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4137 (can_inline_edge_by_limits_p): ... here.
4138 (can_early_inline_edge_p, check_callers,
4139 update_caller_keys, update_callee_keys, recursive_inlining,
4140 add_new_edges_to_heap, speculation_useful_p,
4141 inline_small_functions,
4142 inline_small_functions, flatten_function,
4143 inline_to_all_callers_1): Update.
4145 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4147 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4150 2018-01-30 Richard Biener <rguenther@suse.de>
4152 PR tree-optimization/83008
4153 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4154 invariant and constant vector uses in stmts when they need
4157 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4160 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4161 * configure: Regenerate.
4163 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4165 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4167 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4168 Use gen_rtx_REG rather than gen_lowpart.
4170 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4172 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4173 rather than 0 when creating partial subregs.
4175 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4177 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4180 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4183 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4184 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4185 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4186 flags. This restores the settings used before the 2017-07-24.
4187 Turning off pre increment/decrement/modify allows IVOPTS to
4188 optimize DF/SF loops where the index is an int.
4190 2018-01-29 Richard Biener <rguenther@suse.de>
4191 Kelvin Nilsen <kelvin@gcc.gnu.org>
4194 * tree-vect-stmts.c (vectorizable_call): Don't call
4195 targetm.vectorize_builtin_md_vectorized_function if callee is
4198 2018-01-22 Carl Love <cel@us.ibm.com>
4200 * doc/extend.tex: Fix typo in second arg in
4201 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4203 2018-01-29 Richard Biener <rguenther@suse.de>
4205 PR tree-optimization/84086
4206 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4207 (flush_ssaname_freelist): When SSA names were released reset
4208 the SCEV hash table.
4210 2018-01-29 Richard Biener <rguenther@suse.de>
4212 PR tree-optimization/84057
4213 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4214 removed paths when removing edges.
4216 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4218 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4219 -mfunction-return=@var{choice}.
4221 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4224 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4226 (layout::print_source_line): Likewise.
4227 (test_get_line_width_without_trailing_whitespace): Add test cases.
4229 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4232 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4235 2018-01-26 Jim Wilson <jimw@sifive.com>
4237 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4239 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4242 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4244 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4245 and CMP + SUB-immediate -> SUBS.
4247 2018-01-26 Martin Sebor <msebor@redhat.com>
4249 PR tree-optimization/83896
4250 * tree-ssa-strlen.c (get_string_len): Rename...
4251 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4252 Avoid assuming length is constant.
4253 (handle_char_store): Use HOST_WIDE_INT for string length.
4255 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4258 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4259 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4261 2018-01-26 Richard Biener <rguenther@suse.de>
4263 PR rtl-optimization/84003
4264 * dse.c (record_store): Only record redundant stores when
4265 the earlier store aliases at least all accesses the later one does.
4267 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4269 PR rtl-optimization/83985
4270 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4271 REG_CFA_RESTORE insns.
4272 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4273 REG_CFA_RESTORE insns here.
4276 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4277 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4279 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4281 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4282 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4283 (arc_init): Likewise.
4284 (arc_override_options): Likewise.
4285 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4287 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4289 * config/arc/arc.h (TARGET_DBNZ): Define.
4290 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4291 properly set the tune attribute.
4292 (dbnz): Use TARGET_DBNZ guard.
4293 * config/arc/arc.opt (mtune): Add core3 option.
4295 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4297 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4298 recognize new pic like addresses.
4299 (arc_delegitimize_address): Clean up.
4301 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4303 * config/arc/arc-arches.def: Option mrf16 valid for all
4305 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4306 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4307 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4308 * config/arc/arc-tables.opt: Regenerate.
4309 * config/arc/arc.c (arc_conditional_register_usage): Handle
4310 reduced register file case.
4311 (arc_file_start): Set must have build attributes.
4312 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4314 * config/arc/arc.opt (mrf16): Add new option.
4315 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4316 * config/arc/genmultilib.awk: Handle new mrf16 option.
4317 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4318 * config/arc/t-multilib: Regenerate.
4319 * doc/invoke.texi (ARC Options): Document mrf16 option.
4321 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4323 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4324 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4325 (arc_attribute_table): Add 'secure_call' attribute.
4326 (arc_print_operand): Print secure call operand.
4327 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4329 (arc_is_secure_call_p): New function. * config/arc/arc.md
4330 (call_i): Add support for sjli instruction.
4331 (call_value_i): Likewise.
4332 * config/arc/constraints.md (Csc): New constraint.
4334 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4335 John Eric Martin <John.Martin@emmicro-us.com>
4337 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4338 * config/arc/arc.c (_arc_jli_section): New struct.
4339 (arc_jli_section): New type.
4340 (rc_jli_sections): New static variable.
4341 (arc_handle_jli_attribute): New function.
4342 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4343 (arc_file_end): New function.
4344 (TARGET_ASM_FILE_END): Define.
4345 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4346 (arc_add_jli_section): New function.
4347 (jli_call_scan): Likewise.
4348 (arc_reorg): Call jli_call_scan.
4349 (arc_output_addsi): Remove 'S' from printing asm operand.
4350 (arc_is_jli_call_p): New function.
4351 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4353 (movhi_insn): Likewise.
4354 (movsi_insn): Likewise.
4355 (movsi_set_cc_insn): Likewise.
4356 (loadqi_update): Likewise.
4357 (load_zeroextendqisi_update): Likewise.
4358 (load_signextendqisi_update): Likewise.
4359 (loadhi_update): Likewise.
4360 (load_zeroextendhisi_update): Likewise.
4361 (load_signextendhisi_update): Likewise.
4362 (loadsi_update): Likewise.
4363 (loadsf_update): Likewise.
4364 (movsicc_insn): Likewise.
4365 (bset_insn): Likewise.
4366 (bxor_insn): Likewise.
4367 (bclr_insn): Likewise.
4368 (bmsk_insn): Likewise.
4369 (bicsi3_insn): Likewise.
4370 (cmpsi_cc_c_insn): Likewise.
4371 (movsi_ne): Likewise.
4372 (movsi_cond_exec): Likewise.
4373 (clrsbsi2): Likewise.
4384 (call_i): Remove 'S' asm letter, add jli instruction.
4385 (call_value_i): Likewise.
4386 * config/arc/arc.op (mjli-always): New option.
4387 * config/arc/constraints.md (Cji): New constraint.
4388 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4390 (subsf3_fpx): Likewise.
4391 (mulsf3_fpx): Likewise.
4392 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4394 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4396 * doc/invoke.texi (ARC): Document mjli-always option.
4398 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4400 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4401 avoid addition with 0 and use incw and decw where possible.
4403 2018-01-26 Richard Biener <rguenther@suse.de>
4405 PR tree-optimization/81082
4406 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4407 association if it requires casting to unsigned.
4408 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4409 from fold_plusminus_mult_expr to catch important cases late when
4410 range info is available.
4412 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4414 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4415 * configure.ac (hidden_linkonce): New test.
4416 * configure: Regenerate.
4417 * config.in: Regenerate.
4419 2018-01-26 Julia Koval <julia.koval@intel.com>
4421 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4422 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4423 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4424 _mm_mask_bitshuffle_epi64_mask): Fix type.
4425 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4426 USI_FTYPE_V4DI_V4DI_USI): Remove.
4427 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4428 __builtin_ia32_vpshufbitqmb256_mask,
4429 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4430 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4431 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4433 2018-01-26 Alan Modra <amodra@gmail.com>
4436 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4437 UNSPEC_VBPERMQ. Sort other unspecs.
4439 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4441 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4443 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4446 * predict.c (drop_profile): Do not push/pop cfun; update also
4448 (handle_missing_profiles): Fix logic looking for zero profiles.
4450 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4453 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4454 on functions with #pragma omp declare simd or functions with simd
4456 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4457 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4458 Remove trailing \n from warning_at calls.
4460 2018-01-25 Tom de Vries <tom@codesourcery.com>
4463 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4464 for neutered workers.
4466 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4469 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4470 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4472 2018-01-24 Jeff Law <law@redhat.com>
4475 * i386.c (get_probe_interval): Move to earlier point.
4476 (ix86_compute_frame_layout): If -fstack-clash-protection and
4477 the frame is larger than the probe interval, then use pushes
4478 to save registers rather than reg->mem moves.
4479 (ix86_expand_prologue): Remove conditional for int_registers_saved
4482 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4485 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4486 min/max for never referenced object.
4488 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4491 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4493 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4494 attributes from DECL_ATTRIBUTES (decl) without affecting
4495 DECL_ATTRIBUTES (current_function_decl).
4496 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4497 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4499 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4501 PR tree-optimization/83979
4502 * fold-const.c (fold_comparison): Use constant_boolean_node
4503 instead of boolean_{true,false}_node.
4505 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4507 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4510 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4512 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4513 Simplify the clause that sets the length attribute.
4514 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4515 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4516 clause that sets the length attribute.
4517 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4519 2018-01-24 Tom de Vries <tom@codesourcery.com>
4522 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4523 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4524 Add strict parameter.
4525 (prevent_branch_around_nothing): Insert dummy insn between branch to
4526 label and label with no ptx insn inbetween.
4527 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4529 2018-01-24 Tom de Vries <tom@codesourcery.com>
4532 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4533 for neutered threads in warp.
4534 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4536 2018-01-24 Richard Biener <rguenther@suse.de>
4538 PR tree-optimization/83176
4539 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4542 2018-01-24 Richard Biener <rguenther@suse.de>
4544 PR tree-optimization/82819
4545 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4546 code generating pluses that are no-ops in the target precision.
4548 2018-01-24 Richard Biener <rguenther@suse.de>
4551 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4553 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4555 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4556 to merge probabilities.
4557 * predict.c (probably_never_executed): Also mark as cold functions
4558 with global 0 profile and guessed local profile.
4559 * profile-count.c (profile_probability::combine_with_count): New
4561 * profile-count.h (profile_probability::operator*,
4562 profile_probability::operator*=, profile_probability::operator/,
4563 profile_probability::operator/=): Reduce precision to adjusted
4564 and set value to guessed on contradictory divisions.
4565 (profile_probability::combine_with_freq): Remove.
4566 (profile_probability::combine_wiht_count): Declare.
4567 (profile_count::force_nonzero):: Set to adjusted.
4568 (profile_count::probability_in):: Set quality to adjusted.
4569 * tree-ssa-tail-merge.c (replace_block_by): Use
4572 2018-01-23 Andrew Waterman <andrew@sifive.com>
4573 Jim Wilson <jimw@sifive.com>
4575 * config/riscv/riscv.c (riscv_stack_boundary): New.
4576 (riscv_option_override): Set riscv_stack_boundary. Handle
4577 riscv_preferred_stack_boundary_arg.
4578 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4579 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4580 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4581 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4582 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4583 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4585 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4588 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4589 of struct ix86_frame.
4590 (ix86_expand_epilogue): Likewise. Add a local variable for
4591 the reg_save_offset field in struct ix86_frame.
4593 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4595 PR tree-optimization/82604
4596 * tree-loop-distribution.c (enum partition_kind): New enum item
4597 PKIND_PARTIAL_MEMSET.
4598 (partition_builtin_p): Support above new enum item.
4599 (generate_code_for_partition): Ditto.
4600 (compute_access_range): Differentiate cases that equality can be
4601 proven at all loops, the innermost loops or no loops.
4602 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4603 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4604 (finalize_partitions, distribute_loop): Don't fuse partition of
4605 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4606 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4609 2018-01-23 Martin Liska <mliska@suse.cz>
4611 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4612 order to ignore the predictor.
4613 (PRED_POLYMORPHIC_CALL): Likewise.
4614 (PRED_RECURSIVE_CALL): Likewise.
4616 2018-01-23 Martin Liska <mliska@suse.cz>
4618 * tree-profile.c (tree_profiling): Print function header to
4619 aware reader which function we are working on.
4620 * value-prof.c (gimple_find_values_to_profile): Do not print
4621 not interesting value histograms.
4623 2018-01-23 Martin Liska <mliska@suse.cz>
4625 * profile-count.h (enum profile_quality): Add
4626 profile_uninitialized as the first value. Do not number values
4627 as they are zero based.
4628 (profile_count::verify): Update sanity check.
4629 (profile_probability::verify): Likewise.
4631 2018-01-23 Nathan Sidwell <nathan@acm.org>
4633 * doc/invoke.texi (ffor-scope): Deprecate.
4635 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4637 PR tree-optimization/83510
4638 * domwalk.c (set_all_edges_as_executable): New function.
4639 (dom_walker::dom_walker): Convert bool param
4640 "skip_unreachable_blocks" to enum reachability. Move setup of
4641 edge flags to set_all_edges_as_executable and only do it when
4642 reachability is REACHABLE_BLOCKS.
4643 * domwalk.h (enum dom_walker::reachability): New enum.
4644 (dom_walker::dom_walker): Convert bool param
4645 "skip_unreachable_blocks" to enum reachability.
4646 (set_all_edges_as_executable): New decl.
4647 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4648 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4650 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4651 but converting true to REACHABLE_BLOCKS.
4652 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4654 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4655 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4656 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4658 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4659 if check_all_array_refs will be called.
4661 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4663 * tree.c (selftest::test_location_wrappers): Add more test
4666 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4668 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4669 (selftest::test_bit_in_range): Likewise.
4671 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4674 * doc/sourcebuild.texi (vect_float): Say that the selector
4675 only describes the situation when -funsafe-math-optimizations is on.
4676 (vect_float_strict): Document.
4678 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4680 PR tree-optimization/83965
4681 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4682 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4683 instead of checking only for a reduction.
4684 (vect_recog_widen_sum_pattern): Likewise.
4686 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4688 * predict.c (probably_never_executed): Only use precise profile info.
4689 (compute_function_frequency): Skip after inlining hack since we now
4690 have quality checking.
4692 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4694 * profile-count.h (profile_probability::very_unlikely,
4695 profile_probability::unlikely, profile_probability::even): Set
4696 precision to guessed.
4698 2018-01-23 Richard Biener <rguenther@suse.de>
4700 PR tree-optimization/83963
4701 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4702 Properly terminate dominator walk when crossing the exit edge not
4703 when visiting its source block.
4705 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4708 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4709 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4711 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4713 PR tree-optimization/83957
4714 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4715 semicolon after for body surrounded by braces.
4717 PR tree-optimization/83081
4718 * profile-count.h (profile_probability::split): New method.
4719 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4720 Use profile_probability::split.
4721 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4722 when splitting a single conditional jump into 2.
4724 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4726 PR tree-optimization/69452
4727 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4730 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4732 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4733 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4734 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4736 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4738 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4739 * config/rl78/rl78.md (movdi): New define_expand.
4740 * config/rl78/rl78.c (rl78_split_movdi): New function.
4742 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4745 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4747 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4748 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4749 128-bit to produce an UNSPEC move to get the double word with the
4750 signbit and then a shift directly to do signbit.
4751 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4752 implementation with a new version that just does either a direct
4753 move or a regular move. Move memory interface to separate insns.
4754 Move insns so they are next to the expander.
4755 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4756 with signbit move. Split big and little endian case.
4757 (signbit<mode>2_dm_mem_le): Likewise.
4758 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4759 (signbit<mode>2_dm2): Likewise.
4761 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4763 * config/rl78/rl78.md (anddi3): New define_expand.
4765 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4767 * config/rl78/rl78.md (umindi3): New define_expand.
4769 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4771 * config/rl78/rl78.md (smindi3): New define_expand.
4773 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4775 * config/rl78/rl78.md (smaxdi3): New define_expand.
4777 2018-01-22 Carl Love <cel@us.ibm.com>
4779 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4780 LVX_V1TI): Add macro expansion.
4781 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4782 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4783 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4784 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4785 Change check to determine if the instruction is a byte reversing
4786 entry. Fix typo in comment.
4787 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4788 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4789 Add def_builtin calls for new builtins.
4790 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4791 Add define_insn expansion.
4793 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4795 * config/rl78/rl78.md (umaxdi3): New define_expand.
4797 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4799 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4800 for non-QImode registers.
4802 2018-01-22 Richard Biener <rguenther@suse.de>
4804 PR tree-optimization/83963
4805 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4806 including the loop exit block.
4807 (scop_detection::merge_sese): Likewise.
4808 (scop_detection::add_scop): Do it here instead.
4810 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4812 * doc/sourcebuild.texi (arm_softfloat): Document.
4814 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4817 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4818 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4819 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4821 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4822 David Edelsohn <dje.gcc@gmail.com>
4825 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4826 Change "crset eq" to "crset 2".
4827 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4828 (*call_indirect_aix<mode>_nospec): Likewise.
4829 (*call_value_indirect_aix<mode>_nospec): Likewise.
4830 (*call_indirect_elfv2<mode>_nospec): Likewise.
4831 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4832 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4833 change assembly output from . to $.
4834 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4835 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4836 (*tablejump<mode>_internal1_nospec): Likewise.
4838 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4841 * config/sh/sh_optimize_sett_clrt.cc:
4842 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4844 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4846 PR tree-optimization/83940
4847 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4848 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4849 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4850 use it to pass back the definition type.
4851 (vect_check_store_rhs): Likewise rhs_dt_out.
4852 (vect_build_gather_load_calls): Add a mask_dt argument and use
4853 it instead of a call to vect_is_simple_use.
4854 (vectorizable_store): Update calls to vect_check_load_store_mask
4855 and vect_check_store_rhs. Use the dt returned by the latter instead
4856 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4857 instead of calls to vect_is_simple_use. Pass the scalar rather
4858 than the vector operand to vect_is_simple_use when handling
4859 second and subsequent copies of an rhs value.
4860 (vectorizable_load): Update calls to vect_check_load_store_mask
4861 and vect_build_gather_load_calls. Use the cached mask_dt and
4862 gs_info.offset_dt instead of calls to vect_is_simple_use.
4864 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4867 * tree-emutls.c: Include gimplify.h.
4868 (lower_emutls_2): New function.
4869 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4870 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4871 it before further processing.
4874 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4875 UINTVAL (trueop1) instead of INTVAL (op1).
4877 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4881 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4882 INCOMING_FRAME_SP_OFFSET if not defined.
4883 (scan_trace): Add ENTRY argument. If true and
4884 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4885 emit a note to adjust the CFA offset.
4886 (create_cfi_notes): Adjust scan_trace callers.
4887 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4888 INCOMING_FRAME_SP_OFFSET in the CIE.
4889 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4890 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4892 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4893 * doc/tm.texi: Regenerated.
4895 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4897 PR rtl-optimization/83147
4898 * lra-constraints.c (remove_inheritance_pseudos): Use
4899 lra_substitute_pseudo_within_insn.
4901 2018-01-19 Tom de Vries <tom@codesourcery.com>
4902 Cesar Philippidis <cesar@codesourcery.com>
4905 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4907 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4910 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4911 spaces for function labels.
4913 2018-01-19 Martin Liska <mliska@suse.cz>
4915 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4916 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4917 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4918 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4919 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4920 (PRED_CONST_RETURN): Change from 69 to 65.
4921 (PRED_NULL_RETURN): Change from 91 to 71.
4922 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4923 (PRED_LOOP_GUARD): Change from 66 to 73.
4925 2018-01-19 Martin Liska <mliska@suse.cz>
4927 * predict.c (predict_insn_def): Add new assert.
4928 (struct branch_predictor): Change type to signed integer.
4929 (test_prediction_value_range): Amend test to cover
4931 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4932 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4933 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4934 (PRED_LOOP_IV_COMPARE): Likewise.
4935 * predict.h (PROB_UNINITIALIZED): Define new constant.
4937 2018-01-19 Martin Liska <mliska@suse.cz>
4939 * predict.c (dump_prediction): Add new format for
4940 analyze_brprob.py script which is enabled with -details
4942 * profile-count.h (precise_p): New function.
4944 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4946 PR tree-optimization/83922
4947 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4948 there are no statements that need masking.
4949 (vect_active_double_reduction_p): New function.
4950 (vect_analyze_loop_operations): Use it when handling phis that
4951 are not in the loop header.
4953 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4955 PR tree-optimization/83914
4956 * tree-vect-loop.c (vectorizable_induction): Don't convert
4957 init_expr or apply the peeling adjustment for inductions
4958 that are nested within the vectorized loop.
4960 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4962 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4965 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4969 * function.h (gimplify_parameters): Add gimple_seq * argument.
4970 * function.c: Include gimple.h and options.h.
4971 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4972 for the added local temporaries if needed.
4973 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4974 if there are any parameter cleanups, wrap whole body into a
4975 try/finally with the cleanups.
4977 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4980 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4981 Use GET_MODE_CLASS for scalar floating point.
4983 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4987 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4988 Fix call of call_cgraph_insertion_hooks.
4990 2018-01-18 Martin Sebor <msebor@redhat.com>
4992 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4994 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4997 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5000 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5003 * common.opt: (-ffile-prefix-map): New option.
5004 * opts.c (common_handle_option): Defer it.
5005 * opts-global.c (handle_common_deferred_options): Handle it.
5006 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5007 * file-prefix-map.h: New file.
5008 (remap_debug_filename, add_debug_prefix_map): ...here.
5009 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5010 * final.c (debug_prefix_map, add_debug_prefix_map
5011 remap_debug_filename): Move to...
5012 * file-prefix-map.c: New file.
5013 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5014 generalize, get rid of alloca(), use strrchr() instead of strchr().
5015 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5016 Implement in terms of add_prefix_map().
5017 (remap_macro_filename, remap_debug_filename): Implement in term of
5019 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5020 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5021 * dbxout.c: Include file-prefix-map.h.
5022 * varasm.c: Likewise.
5023 * vmsdbgout.c: Likewise.
5024 * xcoffout.c: Likewise.
5025 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5026 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5027 * doc/invoke.texi (-ffile-prefix-map): Document.
5028 (-fdebug-prefix-map): Update description.
5030 2018-01-18 Martin Liska <mliska@suse.cz>
5032 * config/i386/i386.c (indirect_thunk_name): Document that also
5034 (output_indirect_thunk): Document why both instructions
5035 (pause and lfence) are generated.
5037 2018-01-18 Richard Biener <rguenther@suse.de>
5039 PR tree-optimization/83887
5040 * graphite-scop-detection.c
5041 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5042 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5043 (scop_detection::merge_sese): Re-implement with a flood-fill
5044 algorithm that properly finds a SESE region if it exists.
5046 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5049 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5050 pointer_diff optimizations use view_convert instead of convert.
5052 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5054 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5055 Generate different code for -mno-speculate-indirect-jumps.
5056 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5057 (*call_indirect_aix<mode>): Disable for
5058 -mno-speculate-indirect-jumps.
5059 (*call_indirect_aix<mode>_nospec): New define_insn.
5060 (*call_value_indirect_aix<mode>): Disable for
5061 -mno-speculate-indirect-jumps.
5062 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5063 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5064 -mno-speculate-indirect-jumps.
5065 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5067 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5069 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5070 long double type, set the flags for noting the default long double
5071 type, even if we don't pass or return a long double type.
5073 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5076 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5079 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5081 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5082 support for merge[hl].
5083 (fold_mergehl_helper): New helper function.
5084 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5085 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5086 (altivec_vmrglw_direct): Add xxmrglw insn.
5088 2018-01-17 Andrew Waterman <andrew@sifive.com>
5090 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5091 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5093 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5096 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5097 call the lto_location_cache before reading the
5098 DECL_SOURCE_LOCATION of the types.
5100 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5101 Richard Sandiford <richard.sandiford@linaro.org>
5103 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5104 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5105 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5106 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5107 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5109 * config/aarch64/constraints.md (aarch64_movti_operand):
5111 * config/aarch64/predicates.md (Uti): Add new constraint.
5113 2018-01-17 Carl Love <cel@us.ibm.com>
5115 * config/rs6000/vsx.md (define_expand xl_len_r,
5116 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5117 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5119 (define_expand, define_insn): Move the shift left from the
5120 define_insn to the define_expand for lxvl and stxvl instructions.
5121 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5122 and XL_LEN_R definitions to PURE.
5124 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5126 * config/i386/i386.c (indirect_thunk_name): Declare regno
5127 as unsigned int. Compare regno with INVALID_REGNUM.
5128 (output_indirect_thunk): Ditto.
5129 (output_indirect_thunk_function): Ditto.
5130 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5131 in the call to output_indirect_thunk_function.
5133 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5136 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5137 rather than the size of inner_type to determine the stack slot size
5138 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5140 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5143 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5146 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5148 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5149 endian Linux systems to optionally enable multilibs for selecting
5150 the long double type if the user configured an explicit type.
5151 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5152 have no long double multilibs if not defined.
5153 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5154 warn if the user used -mabi={ieee,ibm}longdouble and we built
5155 multilibs for long double.
5156 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5157 appropriate multilib option.
5158 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5160 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5161 for building long double multilibs.
5162 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5164 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5166 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5169 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5171 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5174 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5177 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5180 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5182 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5183 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5185 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5187 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5188 different rtl trees depending on TARGET_64BIT.
5189 (rs6000_gen_lvx): Likewise.
5191 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5193 * config/visium/visium.md (nop): Tweak comment.
5194 (hazard_nop): Likewise.
5196 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5198 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5199 -mspeculate-indirect-jumps.
5200 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5201 for -mno-speculate-indirect-jumps.
5202 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5203 (*call_value_indirect_elfv2<mode>): Disable for
5204 -mno-speculate-indirect-jumps.
5205 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5206 (indirect_jump): Emit different RTL for
5207 -mno-speculate-indirect-jumps.
5208 (*indirect_jump<mode>): Disable for
5209 -mno-speculate-indirect-jumps.
5210 (*indirect_jump<mode>_nospec): New define_insn.
5211 (tablejump): Emit different RTL for
5212 -mno-speculate-indirect-jumps.
5213 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5214 (tablejumpsi_nospec): New define_expand.
5215 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5216 (tablejumpdi_nospec): New define_expand.
5217 (*tablejump<mode>_internal1): Disable for
5218 -mno-speculate-indirect-jumps.
5219 (*tablejump<mode>_internal1_nospec): New define_insn.
5220 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5223 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5225 * caller-save.c (insert_save): Drop unnecessary parameter. All
5228 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5229 Richard Biener <rguenth@suse.de>
5232 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5233 return early, inline manually is_gimple_sizepos. Make sure if we
5234 call gimplify_expr we don't end up with a gimple constant.
5235 * tree.c (variably_modified_type_p): Don't return true for
5236 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5237 * gimplify.h (is_gimple_sizepos): Remove.
5239 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5241 PR tree-optimization/83857
5242 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5243 vectorizable_live_operation for pure SLP statements.
5244 (vectorizable_live_operation): Handle PHIs.
5246 2018-01-16 Richard Biener <rguenther@suse.de>
5248 PR tree-optimization/83867
5249 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5250 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5252 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5255 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5256 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5257 If off is not INTEGER_CST, issue a may not be aligned warning
5258 rather than isn't aligned. Use isn%'t rather than isn't.
5259 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5261 <case MULT_EXPR>: Improve the case when bottom and one of the
5262 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5263 operand, in that case check if the other operand is multiple of
5264 bottom divided by the INTEGER_CST operand.
5266 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5269 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5270 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5271 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5272 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5273 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5274 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5275 (pa_function_arg_size): New function.
5277 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5279 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5280 in a separate statement.
5282 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5284 PR tree-optimization/83847
5285 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5286 group gathers and scatters.
5288 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5290 PR rtl-optimization/86620
5291 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5293 PR rtl-optimization/83213
5294 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5295 to last if both are JUMP_INSNs.
5297 PR tree-optimization/83843
5298 * gimple-ssa-store-merging.c
5299 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5300 store_immediate_info for bswap/nop orig_stores.
5302 2018-01-15 Andrew Waterman <andrew@sifive.com>
5304 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5306 <UDIV>: Increase cost if !TARGET_DIV.
5308 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5310 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5311 (define_attr "cr_logical_3op"): New.
5312 (cceq_ior_compare): Adjust.
5313 (cceq_ior_compare_complement): Adjust.
5314 (*cceq_rev_compare): Adjust.
5315 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5316 (is_cracked_insn): Adjust.
5317 (insn_must_be_first_in_group): Adjust.
5318 * config/rs6000/40x.md: Adjust.
5319 * config/rs6000/440.md: Adjust.
5320 * config/rs6000/476.md: Adjust.
5321 * config/rs6000/601.md: Adjust.
5322 * config/rs6000/603.md: Adjust.
5323 * config/rs6000/6xx.md: Adjust.
5324 * config/rs6000/7450.md: Adjust.
5325 * config/rs6000/7xx.md: Adjust.
5326 * config/rs6000/8540.md: Adjust.
5327 * config/rs6000/cell.md: Adjust.
5328 * config/rs6000/e300c2c3.md: Adjust.
5329 * config/rs6000/e500mc.md: Adjust.
5330 * config/rs6000/e500mc64.md: Adjust.
5331 * config/rs6000/e5500.md: Adjust.
5332 * config/rs6000/e6500.md: Adjust.
5333 * config/rs6000/mpc.md: Adjust.
5334 * config/rs6000/power4.md: Adjust.
5335 * config/rs6000/power5.md: Adjust.
5336 * config/rs6000/power6.md: Adjust.
5337 * config/rs6000/power7.md: Adjust.
5338 * config/rs6000/power8.md: Adjust.
5339 * config/rs6000/power9.md: Adjust.
5340 * config/rs6000/rs64.md: Adjust.
5341 * config/rs6000/titan.md: Adjust.
5343 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5345 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5346 ix86_indirect_branch_register logic.
5348 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5350 * config/i386/constraints.md (Bs): Update
5351 ix86_indirect_branch_register check. Don't check
5352 ix86_indirect_branch_register with GOT_memory_operand.
5354 * config/i386/predicates.md (GOT_memory_operand): Don't check
5355 ix86_indirect_branch_register here.
5356 (GOT32_symbol_operand): Likewise.
5358 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5360 * config/i386/predicates.md (constant_call_address_operand):
5361 Rewrite ix86_indirect_branch_register logic.
5362 (sibcall_insn_operand): Likewise.
5364 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5366 * config/i386/constraints.md (Bs): Replace
5367 ix86_indirect_branch_thunk_register with
5368 ix86_indirect_branch_register.
5370 * config/i386/i386.md (indirect_jump): Likewise.
5371 (tablejump): Likewise.
5372 (*sibcall_memory): Likewise.
5373 (*sibcall_value_memory): Likewise.
5374 Peepholes of indirect call and jump via memory: Likewise.
5375 * config/i386/i386.opt: Likewise.
5376 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5377 (GOT_memory_operand): Likewise.
5378 (call_insn_operand): Likewise.
5379 (sibcall_insn_operand): Likewise.
5380 (GOT32_symbol_operand): Likewise.
5382 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5385 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5386 type rather than type addr's type points to.
5387 (expand_omp_atomic_mutex): Likewise.
5388 (expand_omp_atomic): Likewise.
5390 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5393 * config/i386/i386.c (output_indirect_thunk_function): Use
5394 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5395 for __x86_return_thunk.
5397 2018-01-15 Richard Biener <rguenther@suse.de>
5400 * expmed.c (extract_bit_field_1): Fix typo.
5402 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5405 * config/arm/iterators.md (VF): New mode iterator.
5406 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5407 Remove integer-related logic from pattern.
5408 (neon_vabd<mode>_3): Likewise.
5410 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5413 * common.opt (fstrict-overflow): No longer an alias.
5414 (fwrapv-pointer): New option.
5415 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5416 also for pointer types based on flag_wrapv_pointer.
5417 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5418 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5419 opts->x_flag_wrapv got set.
5420 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5421 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5422 POINTER_TYPE_OVERFLOW_UNDEFINED.
5423 * match.pd: Likewise in address comparison pattern.
5424 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5426 2018-01-15 Richard Biener <rguenther@suse.de>
5429 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5430 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5431 Reset type names to their identifier if their TYPE_DECL doesn't
5432 have linkage (and thus is used for ODR and devirt).
5433 (save_debug_info_for_decl): Remove.
5434 (save_debug_info_for_type): Likewise.
5435 (add_tree_to_fld_list): Adjust.
5436 * tree-pretty-print.c (dump_generic_node): Make dumping of
5437 type names more robust.
5439 2018-01-15 Richard Biener <rguenther@suse.de>
5441 * BASE-VER: Bump to 8.0.1.
5443 2018-01-14 Martin Sebor <msebor@redhat.com>
5446 * builtins.c (check_access): Avoid warning when the no-warning bit
5449 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5451 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5452 * ira-color (allocno_hard_regs_compare): Likewise.
5454 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5457 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5458 Use .pushsection/.popsection.
5460 2018-01-14 Martin Sebor <msebor@redhat.com>
5463 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5465 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5467 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5468 entry from extra_headers.
5469 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5470 extra_headers, make the list bitwise identical to the i?86-*-* one.
5472 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5474 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5475 -mcmodel=large with -mindirect-branch=thunk,
5476 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5477 -mfunction-return=thunk-extern.
5478 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5479 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5480 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5482 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5484 * config/i386/i386.c (print_reg): Print the name of the full
5485 integer register without '%'.
5486 (ix86_print_operand): Handle 'V'.
5487 * doc/extend.texi: Document 'V' modifier.
5489 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5491 * config/i386/constraints.md (Bs): Disallow memory operand for
5492 -mindirect-branch-register.
5494 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5495 (GOT_memory_operand): Likewise.
5496 (call_insn_operand): Likewise.
5497 (sibcall_insn_operand): Likewise.
5498 (GOT32_symbol_operand): Likewise.
5499 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5500 for -mindirect-branch-register.
5501 (tablejump): Likewise.
5502 (*sibcall_memory): Likewise.
5503 (*sibcall_value_memory): Likewise.
5504 Disallow peepholes of indirect call and jump via memory for
5505 -mindirect-branch-register.
5506 (*call_pop): Replace m with Bw.
5507 (*call_value_pop): Likewise.
5508 (*sibcall_pop_memory): Replace m with Bs.
5509 * config/i386/i386.opt (mindirect-branch-register): New option.
5510 * doc/invoke.texi: Document -mindirect-branch-register option.
5512 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5514 * config/i386/i386-protos.h (ix86_output_function_return): New.
5515 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5516 set function_return_type.
5517 (indirect_thunk_name): Add ret_p to indicate thunk for function
5519 (output_indirect_thunk_function): Pass false to
5520 indirect_thunk_name.
5521 (ix86_output_indirect_branch_via_reg): Likewise.
5522 (ix86_output_indirect_branch_via_push): Likewise.
5523 (output_indirect_thunk_function): Create alias for function
5524 return thunk if regno < 0.
5525 (ix86_output_function_return): New function.
5526 (ix86_handle_fndecl_attribute): Handle function_return.
5527 (ix86_attribute_table): Add function_return.
5528 * config/i386/i386.h (machine_function): Add
5529 function_return_type.
5530 * config/i386/i386.md (simple_return_internal): Use
5531 ix86_output_function_return.
5532 (simple_return_internal_long): Likewise.
5533 * config/i386/i386.opt (mfunction-return=): New option.
5534 (indirect_branch): Mention -mfunction-return=.
5535 * doc/extend.texi: Document function_return function attribute.
5536 * doc/invoke.texi: Document -mfunction-return= option.
5538 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5540 * config/i386/i386-opts.h (indirect_branch): New.
5541 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5542 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5543 with local indirect jump when converting indirect call and jump.
5544 (ix86_set_indirect_branch_type): New.
5545 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5546 (indirectlabelno): New.
5547 (indirect_thunk_needed): Likewise.
5548 (indirect_thunk_bnd_needed): Likewise.
5549 (indirect_thunks_used): Likewise.
5550 (indirect_thunks_bnd_used): Likewise.
5551 (INDIRECT_LABEL): Likewise.
5552 (indirect_thunk_name): Likewise.
5553 (output_indirect_thunk): Likewise.
5554 (output_indirect_thunk_function): Likewise.
5555 (ix86_output_indirect_branch_via_reg): Likewise.
5556 (ix86_output_indirect_branch_via_push): Likewise.
5557 (ix86_output_indirect_branch): Likewise.
5558 (ix86_output_indirect_jmp): Likewise.
5559 (ix86_code_end): Call output_indirect_thunk_function if needed.
5560 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5562 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5563 (ix86_attribute_table): Add indirect_branch.
5564 * config/i386/i386.h (machine_function): Add indirect_branch_type
5565 and has_local_indirect_jump.
5566 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5568 (tablejump): Likewise.
5569 (*indirect_jump): Use ix86_output_indirect_jmp.
5570 (*tablejump_1): Likewise.
5571 (simple_return_indirect_internal): Likewise.
5572 * config/i386/i386.opt (mindirect-branch=): New option.
5573 (indirect_branch): New.
5576 (thunk-inline): Likewise.
5577 (thunk-extern): Likewise.
5578 * doc/extend.texi: Document indirect_branch function attribute.
5579 * doc/invoke.texi: Document -mindirect-branch= option.
5581 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5584 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5586 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5588 * ipa-inline.c (want_inline_small_function_p): Return false if
5589 inlining has already failed with CIF_FINAL_ERROR.
5590 (update_caller_keys): Call want_inline_small_function_p before
5592 (update_callee_keys): Likewise.
5594 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5596 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5598 (rs6000_quadword_masked_address_p): Likewise.
5599 (quad_aligned_load_p): Likewise.
5600 (quad_aligned_store_p): Likewise.
5601 (const_load_sequence_p): Add comment to describe the outer-most loop.
5602 (mimic_memory_attributes_and_flags): New function.
5603 (rs6000_gen_stvx): Likewise.
5604 (replace_swapped_aligned_store): Likewise.
5605 (rs6000_gen_lvx): Likewise.
5606 (replace_swapped_aligned_load): Likewise.
5607 (replace_swapped_load_constant): Capitalize argument name in
5608 comment describing this function.
5609 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5610 and stores that access quad-word aligned addresses and replace
5611 with stvx or lvx instructions when appropriate.
5612 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5613 New function prototype.
5614 (rs6000_quadword_masked_address_p): Likewise.
5615 (rs6000_gen_lvx): Likewise.
5616 (rs6000_gen_stvx): Likewise.
5617 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5618 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5619 when memory address is aligned.
5620 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5621 this split to select lvx instruction when memory address is aligned.
5622 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5623 instruction when memory address is aligned.
5624 (*vsx_le_perm_load_v16qi): Likewise.
5625 (four unnamed splitters): Modify to select the stvx instruction
5626 when memory is aligned.
5628 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5630 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5631 which appears in the queue multiple times.
5633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5634 Alan Hayward <alan.hayward@arm.com>
5635 David Sherwood <david.sherwood@arm.com>
5637 * tree-vectorizer.h (vec_lower_bound): New structure.
5638 (_loop_vec_info): Add check_nonzero and lower_bounds.
5639 (LOOP_VINFO_CHECK_NONZERO): New macro.
5640 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5641 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5642 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5643 fields. Make seg_len the distance travelled, not including the
5645 (dr_direction_indicator): Declare.
5646 (dr_zero_step_indicator): Likewise.
5647 (dr_known_forward_stride_p): Likewise.
5648 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5650 (runtime_alias_check_p): Allow runtime alias checks with
5652 (operator ==): Compare access_size and align.
5653 (prune_runtime_alias_test_list): Rework for new distinction between
5654 the access_size and seg_len.
5655 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5657 (get_segment_min_max): New function.
5658 (create_intersect_range_checks): Use it.
5659 (dr_step_indicator): New function.
5660 (dr_direction_indicator): Likewise.
5661 (dr_zero_step_indicator): Likewise.
5662 (dr_known_forward_stride_p): Likewise.
5663 * tree-loop-distribution.c (data_ref_segment_size): Return
5664 DR_STEP * (niters - 1).
5665 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5667 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5668 (vect_preserves_scalar_order_p): New function, split out from...
5669 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5670 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5671 (vect_vfa_access_size): New function.
5672 (vect_vfa_align): Likewise.
5673 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5674 (dump_lower_bound): New function.
5675 (vect_check_lower_bound): Likewise.
5676 (vect_small_gap_p): Likewise.
5677 (vectorizable_with_step_bound_p): Likewise.
5678 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5679 depencies if the vectorization factor is 1. Convert the checks
5680 for nonzero steps into checks on the bounds of DR_STEP. Try using
5681 a bunds check for variable steps if the minimum required step is
5682 relatively small. Update calls to the dr_with_seg_len
5683 constructor and to vect_compile_time_alias.
5684 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5686 (vect_loop_versioning): Call it.
5687 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5689 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5692 Alan Hayward <alan.hayward@arm.com>
5693 David Sherwood <david.sherwood@arm.com>
5695 * doc/sourcebuild.texi (vect_scatter_store): Document.
5696 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5698 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5700 * genopinit.c (main): Add supports_vec_scatter_store and
5701 supports_vec_scatter_store_cached to target_optabs.
5702 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5703 IFN_MASK_SCATTER_STORE.
5704 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5706 * internal-fn.h (internal_store_fn_p): Declare.
5707 (internal_fn_stored_value_index): Likewise.
5708 * internal-fn.c (scatter_store_direct): New macro.
5709 (expand_scatter_store_optab_fn): New function.
5710 (direct_scatter_store_optab_supported_p): New macro.
5711 (internal_store_fn_p): New function.
5712 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5713 IFN_MASK_SCATTER_STORE.
5714 (internal_fn_mask_index): Likewise.
5715 (internal_fn_stored_value_index): New function.
5716 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5718 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5719 * optabs-query.c (supports_vec_scatter_store_p): New function.
5720 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5721 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5722 true for scatter stores.
5723 (vect_gather_scatter_fn_p): Handle scatter stores too.
5724 (vect_check_gather_scatter): Consider using scatter stores if
5725 supports_vec_scatter_store_p.
5726 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5728 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5729 internal_fn_stored_value_index.
5730 (check_load_store_masking): Handle scatter stores too.
5731 (vect_get_store_rhs): Make public.
5732 (vectorizable_call): Use internal_store_fn_p.
5733 (vectorizable_store): Handle scatter store internal functions.
5734 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5735 when deciding whether the end of the group has been reached.
5736 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5737 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5738 (mask_scatter_store<mode>): New insns.
5740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5741 Alan Hayward <alan.hayward@arm.com>
5742 David Sherwood <david.sherwood@arm.com>
5744 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5745 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5746 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5748 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5749 Use vect_truncate_gather_scatter_offset if we can't treat the
5750 operation as a normal gather load or scatter store.
5751 (get_group_load_store_type): Take the gather_scatter_info
5752 as argument. Try using a gather load or scatter store for
5753 single-element groups.
5754 (get_load_store_type): Update calls to get_group_load_store_type
5755 and vect_use_strided_gather_scatters_p.
5757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5758 Alan Hayward <alan.hayward@arm.com>
5759 David Sherwood <david.sherwood@arm.com>
5761 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5762 optional tree argument.
5763 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5765 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5766 but continue to use the current value as a fallback.
5767 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5768 to compare the updates.
5769 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5770 (get_load_store_type): Use it when handling a strided access.
5771 (vect_get_strided_load_store_ops): New function.
5772 (vect_get_data_ptr_increment): Likewise.
5773 (vectorizable_load): Handle strided gather loads. Always pass
5774 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5777 Alan Hayward <alan.hayward@arm.com>
5778 David Sherwood <david.sherwood@arm.com>
5780 * doc/md.texi (gather_load@var{m}): Document.
5781 (mask_gather_load@var{m}): Likewise.
5782 * genopinit.c (main): Add supports_vec_gather_load and
5783 supports_vec_gather_load_cached to target_optabs.
5784 * optabs-tree.c (init_tree_optimization_optabs): Use
5785 ggc_cleared_alloc to allocate target_optabs.
5786 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5787 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5789 * internal-fn.h (internal_load_fn_p): Declare.
5790 (internal_gather_scatter_fn_p): Likewise.
5791 (internal_fn_mask_index): Likewise.
5792 (internal_gather_scatter_fn_supported_p): Likewise.
5793 * internal-fn.c (gather_load_direct): New macro.
5794 (expand_gather_load_optab_fn): New function.
5795 (direct_gather_load_optab_supported_p): New macro.
5796 (direct_internal_fn_optab): New function.
5797 (internal_load_fn_p): Likewise.
5798 (internal_gather_scatter_fn_p): Likewise.
5799 (internal_fn_mask_index): Likewise.
5800 (internal_gather_scatter_fn_supported_p): Likewise.
5801 * optabs-query.c (supports_at_least_one_mode_p): New function.
5802 (supports_vec_gather_load_p): Likewise.
5803 * optabs-query.h (supports_vec_gather_load_p): Declare.
5804 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5805 and memory_type field.
5806 (NUM_PATTERNS): Bump to 15.
5807 * tree-vect-data-refs.c: Include internal-fn.h.
5808 (vect_gather_scatter_fn_p): New function.
5809 (vect_describe_gather_scatter_call): Likewise.
5810 (vect_check_gather_scatter): Try using internal functions for
5811 gather loads. Recognize existing calls to a gather load function.
5812 (vect_analyze_data_refs): Consider using gather loads if
5813 supports_vec_gather_load_p.
5814 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5815 (vect_get_gather_scatter_offset_type): Likewise.
5816 (vect_convert_mask_for_vectype): Likewise.
5817 (vect_add_conversion_to_patterm): Likewise.
5818 (vect_try_gather_scatter_pattern): Likewise.
5819 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5820 (vect_vect_recog_func_ptrs): Add it.
5821 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5822 internal_fn_mask_index and internal_gather_scatter_fn_p.
5823 (check_load_store_masking): Take the gather_scatter_info as an
5824 argument and handle gather loads.
5825 (vect_get_gather_scatter_ops): New function.
5826 (vectorizable_call): Check internal_load_fn_p.
5827 (vectorizable_load): Likewise. Handle gather load internal
5829 (vectorizable_store): Update call to check_load_store_masking.
5830 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5831 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5832 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5833 (aarch64_gather_scale_operand_d): New predicates.
5834 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5835 (mask_gather_load<mode>): New insns.
5837 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5838 Alan Hayward <alan.hayward@arm.com>
5839 David Sherwood <david.sherwood@arm.com>
5841 * optabs.def (fold_left_plus_optab): New optab.
5842 * doc/md.texi (fold_left_plus_@var{m}): Document.
5843 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5844 * internal-fn.c (fold_left_direct): Define.
5845 (expand_fold_left_optab_fn): Likewise.
5846 (direct_fold_left_optab_supported_p): Likewise.
5847 * fold-const-call.c (fold_const_fold_left): New function.
5848 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5849 * tree-parloops.c (valid_reduction_p): New function.
5850 (gather_scalar_reductions): Use it.
5851 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5852 (vect_finish_replace_stmt): Declare.
5853 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5854 (needs_fold_left_reduction_p): New function, split out from...
5855 (vect_is_simple_reduction): ...here. Accept reductions that
5856 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5857 (vect_force_simple_reduction): Also store the reduction type in
5858 the assignment's STMT_VINFO_REDUC_TYPE.
5859 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5860 (merge_with_identity): New function.
5861 (vect_expand_fold_left): Likewise.
5862 (vectorize_fold_left_reduction): Likewise.
5863 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5864 scalar phi in place for it. Check for target support and reject
5865 cases that would reassociate the operation. Defer the transform
5866 phase to vectorize_fold_left_reduction.
5867 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5868 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5869 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5873 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5874 call to ifc_temp_var.
5876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5877 Alan Hayward <alan.hayward@arm.com>
5878 David Sherwood <david.sherwood@arm.com>
5880 * target.def (legitimize_address_displacement): Take the original
5881 offset as a poly_int.
5882 * targhooks.h (default_legitimize_address_displacement): Update
5884 * targhooks.c (default_legitimize_address_displacement): Likewise.
5885 * doc/tm.texi: Regenerate.
5886 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5887 as an argument, moving assert of ad->disp == ad->disp_term to...
5888 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5889 Try calling targetm.legitimize_address_displacement before expanding
5890 the address rather than afterwards, and adjust for the new interface.
5891 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5892 Match the new hook interface. Handle SVE addresses.
5893 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5898 * Makefile.in (OBJS): Add early-remat.o.
5899 * target.def (select_early_remat_modes): New hook.
5900 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5901 * doc/tm.texi: Regenerate.
5902 * targhooks.h (default_select_early_remat_modes): Declare.
5903 * targhooks.c (default_select_early_remat_modes): New function.
5904 * timevar.def (TV_EARLY_REMAT): New timevar.
5905 * passes.def (pass_early_remat): New pass.
5906 * tree-pass.h (make_pass_early_remat): Declare.
5907 * early-remat.c: New file.
5908 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5910 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5913 Alan Hayward <alan.hayward@arm.com>
5914 David Sherwood <david.sherwood@arm.com>
5916 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5917 vfm1 with a bound_epilog parameter.
5918 (vect_do_peeling): Update calls accordingly, and move the prologue
5919 call earlier in the function. Treat the base bound_epilog as 0 for
5920 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5921 this base when peeling for gaps.
5922 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5923 with fully-masked loops.
5924 (vect_estimate_min_profitable_iters): Handle the single peeled
5925 iteration in that case.
5927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5928 Alan Hayward <alan.hayward@arm.com>
5929 David Sherwood <david.sherwood@arm.com>
5931 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5932 single-element interleaving even if the size is not a power of 2.
5933 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5934 accesses for single-element interleaving if the group size is
5937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5938 Alan Hayward <alan.hayward@arm.com>
5939 David Sherwood <david.sherwood@arm.com>
5941 * doc/md.texi (fold_extract_last_@var{m}): Document.
5942 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5943 * optabs.def (fold_extract_last_optab): New optab.
5944 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5945 * internal-fn.c (fold_extract_direct): New macro.
5946 (expand_fold_extract_optab_fn): Likewise.
5947 (direct_fold_extract_optab_supported_p): Likewise.
5948 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5949 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5950 EXTRACT_LAST_REDUCTION.
5951 (get_initial_def_for_reduction): Do not create an initial vector
5952 for EXTRACT_LAST_REDUCTION reductions.
5953 (vectorizable_reduction): Leave the scalar phi in place for
5954 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5955 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5956 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5957 transform phase to vectorizable_condition.
5958 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5960 (vect_finish_stmt_generation): ...here.
5961 (vect_finish_replace_stmt): New function.
5962 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5963 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5965 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5968 Alan Hayward <alan.hayward@arm.com>
5969 David Sherwood <david.sherwood@arm.com>
5971 * doc/md.texi (extract_last_@var{m}): Document.
5972 * optabs.def (extract_last_optab): New optab.
5973 * internal-fn.def (EXTRACT_LAST): New internal function.
5974 * internal-fn.c (cond_unary_direct): New macro.
5975 (expand_cond_unary_optab_fn): Likewise.
5976 (direct_cond_unary_optab_supported_p): Likewise.
5977 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5978 loops using EXTRACT_LAST.
5979 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5980 (extract_last_<mode>): ...this optab.
5981 (vec_extract<mode><Vel>): Update accordingly.
5983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5984 Alan Hayward <alan.hayward@arm.com>
5985 David Sherwood <david.sherwood@arm.com>
5987 * target.def (empty_mask_is_expensive): New hook.
5988 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5989 * doc/tm.texi: Regenerate.
5990 * targhooks.h (default_empty_mask_is_expensive): Declare.
5991 * targhooks.c (default_empty_mask_is_expensive): New function.
5992 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5993 if the target says that empty masks are expensive.
5994 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5996 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5999 Alan Hayward <alan.hayward@arm.com>
6000 David Sherwood <david.sherwood@arm.com>
6002 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6003 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6004 (vect_use_loop_mask_for_alignment_p): New function.
6005 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6006 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6007 niters_skip argument. Make sure that the first niters_skip elements
6008 of the first iteration are inactive.
6009 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6010 Update call to vect_set_loop_masks_directly.
6011 (get_misalign_in_elems): New function, split out from...
6012 (vect_gen_prolog_loop_niters): ...here.
6013 (vect_update_init_of_dr): Take a code argument that specifies whether
6014 the adjustment should be added or subtracted.
6015 (vect_update_init_of_drs): Likewise.
6016 (vect_prepare_for_masked_peels): New function.
6017 (vect_do_peeling): Skip prologue peeling if we're using a mask
6018 instead. Update call to vect_update_inits_of_drs.
6019 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6021 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6022 alignment. Do not include the number of peeled iterations in
6023 the minimum threshold in that case.
6024 (vectorizable_induction): Adjust the start value down by
6025 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6026 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6027 Take the number of skipped iterations into account when calculating
6029 * tree-vect-stmts.c (vect_gen_while_not): New function.
6031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6032 Alan Hayward <alan.hayward@arm.com>
6033 David Sherwood <david.sherwood@arm.com>
6035 * doc/sourcebuild.texi (vect_fully_masked): Document.
6036 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6038 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6040 (vect_analyze_loop_2): ...here. Don't check the vectorization
6041 factor against the number of loop iterations if the loop is
6044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6045 Alan Hayward <alan.hayward@arm.com>
6046 David Sherwood <david.sherwood@arm.com>
6048 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6049 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6050 (dump_groups): Update accordingly.
6051 (iv_use::mem_type): New member variable.
6052 (address_p): New function.
6053 (record_use): Add a mem_type argument and initialize the new
6055 (record_group_use): Add a mem_type argument. Use address_p.
6056 Remove obsolete null checks of base_object. Update call to record_use.
6057 (find_interesting_uses_op): Update call to record_group_use.
6058 (find_interesting_uses_cond): Likewise.
6059 (find_interesting_uses_address): Likewise.
6060 (get_mem_type_for_internal_fn): New function.
6061 (find_address_like_use): Likewise.
6062 (find_interesting_uses_stmt): Try find_address_like_use before
6063 calling find_interesting_uses_op.
6064 (addr_offset_valid_p): Use the iv mem_type field as the type
6065 of the addressed memory.
6066 (add_autoinc_candidates): Likewise.
6067 (get_address_cost): Likewise.
6068 (split_small_address_groups_p): Use address_p.
6069 (split_address_groups): Likewise.
6070 (add_iv_candidate_for_use): Likewise.
6071 (autoinc_possible_for_pair): Likewise.
6072 (rewrite_groups): Likewise.
6073 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6074 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6075 (get_alias_ptr_type_for_ptr_address): New function.
6076 (rewrite_use_address): Rewrite address uses in calls that were
6077 identified by find_address_like_use.
6079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6080 Alan Hayward <alan.hayward@arm.com>
6081 David Sherwood <david.sherwood@arm.com>
6083 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6085 * gimple-expr.h (is_gimple_addressable: Likewise.
6086 * gimple-expr.c (is_gimple_address): Likewise.
6087 * internal-fn.c (expand_call_mem_ref): New function.
6088 (expand_mask_load_optab_fn): Use it.
6089 (expand_mask_store_optab_fn): Likewise.
6091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6092 Alan Hayward <alan.hayward@arm.com>
6093 David Sherwood <david.sherwood@arm.com>
6095 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6096 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6097 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6098 (cond_umax@var{mode}): Document.
6099 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6100 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6101 (cond_umin_optab, cond_umax_optab): New optabs.
6102 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6103 (COND_IOR, COND_XOR): New internal functions.
6104 * internal-fn.h (get_conditional_internal_fn): Declare.
6105 * internal-fn.c (cond_binary_direct): New macro.
6106 (expand_cond_binary_optab_fn): Likewise.
6107 (direct_cond_binary_optab_supported_p): Likewise.
6108 (get_conditional_internal_fn): New function.
6109 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6110 Cope with reduction statements that are vectorized as calls rather
6112 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6113 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6114 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6115 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6116 (UNSPEC_COND_EOR): New unspecs.
6117 (optab): Add mappings for them.
6118 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6119 (sve_int_op, sve_fp_op): New int attributes.
6121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6122 Alan Hayward <alan.hayward@arm.com>
6123 David Sherwood <david.sherwood@arm.com>
6125 * optabs.def (while_ult_optab): New optab.
6126 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6127 * internal-fn.def (WHILE_ULT): New internal function.
6128 * internal-fn.h (direct_internal_fn_supported_p): New override
6129 that takes two types as argument.
6130 * internal-fn.c (while_direct): New macro.
6131 (expand_while_optab_fn): New function.
6132 (convert_optab_supported_p): Likewise.
6133 (direct_while_optab_supported_p): New macro.
6134 * wide-int.h (wi::udiv_ceil): New function.
6135 * tree-vectorizer.h (rgroup_masks): New structure.
6136 (vec_loop_masks): New typedef.
6137 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6139 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6140 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6141 (vect_max_vf): New function.
6142 (slpeel_make_loop_iterate_ntimes): Delete.
6143 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6144 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6145 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6146 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6147 internal-fn.h, stor-layout.h and optabs-query.h.
6148 (vect_set_loop_mask): New function.
6149 (add_preheader_seq): Likewise.
6150 (add_header_seq): Likewise.
6151 (interleave_supported_p): Likewise.
6152 (vect_maybe_permute_loop_masks): Likewise.
6153 (vect_set_loop_masks_directly): Likewise.
6154 (vect_set_loop_condition_masked): Likewise.
6155 (vect_set_loop_condition_unmasked): New function, split out from
6156 slpeel_make_loop_iterate_ntimes.
6157 (slpeel_make_loop_iterate_ntimes): Rename to..
6158 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6159 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6160 (vect_do_peeling): Update call accordingly.
6161 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6163 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6164 mask_compare_type, can_fully_mask_p and fully_masked_p.
6165 (release_vec_loop_masks): New function.
6166 (_loop_vec_info): Use it to free the loop masks.
6167 (can_produce_all_loop_masks_p): New function.
6168 (vect_get_max_nscalars_per_iter): Likewise.
6169 (vect_verify_full_masking): Likewise.
6170 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6171 retries, and free the mask rgroups before retrying. Check loop-wide
6172 reasons for disallowing fully-masked loops. Make the final decision
6173 about whether use a fully-masked loop or not.
6174 (vect_estimate_min_profitable_iters): Do not assume that peeling
6175 for the number of iterations will be needed for fully-masked loops.
6176 (vectorizable_reduction): Disable fully-masked loops.
6177 (vectorizable_live_operation): Likewise.
6178 (vect_halve_mask_nunits): New function.
6179 (vect_double_mask_nunits): Likewise.
6180 (vect_record_loop_mask): Likewise.
6181 (vect_get_loop_mask): Likewise.
6182 (vect_transform_loop): Handle the case in which the final loop
6183 iteration might handle a partial vector. Call vect_set_loop_condition
6184 instead of slpeel_make_loop_iterate_ntimes.
6185 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6186 (check_load_store_masking): New function.
6187 (prepare_load_store_mask): Likewise.
6188 (vectorizable_store): Handle fully-masked loops.
6189 (vectorizable_load): Likewise.
6190 (supportable_widening_operation): Use vect_halve_mask_nunits for
6192 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6193 (vect_gen_while): New function.
6194 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6195 (aarch64_uqdec<mode>): New insn.
6197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6198 Alan Hayward <alan.hayward@arm.com>
6199 David Sherwood <david.sherwood@arm.com>
6201 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6202 (reduc_xor_scal_optab): New optabs.
6203 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6204 (reduc_xor_scal_@var{m}): Document.
6205 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6206 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6208 * fold-const-call.c (fold_const_call): Handle them.
6209 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6210 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6211 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6212 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6213 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6214 (UNSPEC_XORV): New unspecs.
6215 (optab): Add entries for them.
6216 (BITWISEV): New int iterator.
6217 (bit_reduc_op): New int attributes.
6219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6220 Alan Hayward <alan.hayward@arm.com>
6221 David Sherwood <david.sherwood@arm.com>
6223 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6224 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6225 * optabs.def (vec_shl_insert_optab): New optab.
6226 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6227 (duplicate_and_interleave): Likewise.
6228 * tree-vect-loop.c: Include internal-fn.h.
6229 (neutral_op_for_slp_reduction): New function, split out from
6230 get_initial_defs_for_reduction.
6231 (get_initial_def_for_reduction): Handle option 2 for variable-length
6232 vectors by loading the neutral value into a vector and then shifting
6233 the initial value into element 0.
6234 (get_initial_defs_for_reduction): Replace the code argument with
6235 the neutral value calculated by neutral_op_for_slp_reduction.
6236 Use gimple_build_vector for constant-length vectors.
6237 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6238 but the first group_size elements have a neutral value.
6239 Use duplicate_and_interleave otherwise.
6240 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6241 Update call to get_initial_defs_for_reduction. Handle SLP
6242 reductions for variable-length vectors by creating one vector
6243 result for each scalar result, with the elements associated
6244 with other scalar results stubbed out with the neutral value.
6245 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6246 Require IFN_VEC_SHL_INSERT for double reductions on
6247 variable-length vectors, or SLP reductions that have
6248 a neutral value. Require can_duplicate_and_interleave_p
6249 support for variable-length unchained SLP reductions if there
6250 is no neutral value, such as for MIN/MAX reductions. Also require
6251 the number of vector elements to be a multiple of the number of
6252 SLP statements when doing variable-length unchained SLP reductions.
6253 Update call to vect_create_epilog_for_reduction.
6254 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6255 and remove initial values.
6256 (duplicate_and_interleave): Make public.
6257 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6258 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6261 Alan Hayward <alan.hayward@arm.com>
6262 David Sherwood <david.sherwood@arm.com>
6264 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6265 (can_duplicate_and_interleave_p): New function.
6266 (vect_get_and_check_slp_defs): Take the vector of statements
6267 rather than just the current one. Remove excess parentheses.
6268 Restriction rejectinon of vect_constant_def and vect_external_def
6269 for variable-length vectors to boolean types, or types for which
6270 can_duplicate_and_interleave_p is false.
6271 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6272 (duplicate_and_interleave): New function.
6273 (vect_get_constant_vectors): Use gimple_build_vector for
6274 constant-length vectors and suitable variable-length constant
6275 vectors. Use duplicate_and_interleave for other variable-length
6276 vectors. Don't defer the update when inserting new statements.
6278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6279 Alan Hayward <alan.hayward@arm.com>
6280 David Sherwood <david.sherwood@arm.com>
6282 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6283 min_profitable_iters doesn't go negative.
6285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6286 Alan Hayward <alan.hayward@arm.com>
6287 David Sherwood <david.sherwood@arm.com>
6289 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6290 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6291 * optabs.def (vec_mask_load_lanes_optab): New optab.
6292 (vec_mask_store_lanes_optab): Likewise.
6293 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6294 (MASK_STORE_LANES): Likewise.
6295 * internal-fn.c (mask_load_lanes_direct): New macro.
6296 (mask_store_lanes_direct): Likewise.
6297 (expand_mask_load_optab_fn): Handle masked operations.
6298 (expand_mask_load_lanes_optab_fn): New macro.
6299 (expand_mask_store_optab_fn): Handle masked operations.
6300 (expand_mask_store_lanes_optab_fn): New macro.
6301 (direct_mask_load_lanes_optab_supported_p): Likewise.
6302 (direct_mask_store_lanes_optab_supported_p): Likewise.
6303 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6305 (vect_load_lanes_supported): Likewise.
6306 * tree-vect-data-refs.c (strip_conversion): New function.
6307 (can_group_stmts_p): Likewise.
6308 (vect_analyze_data_ref_accesses): Use it instead of checking
6309 for a pair of assignments.
6310 (vect_store_lanes_supported): Take a masked_p parameter.
6311 (vect_load_lanes_supported): Likewise.
6312 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6313 vect_store_lanes_supported and vect_load_lanes_supported.
6314 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6315 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6316 parameter. Don't allow gaps for masked accesses.
6317 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6318 and vect_load_lanes_supported.
6319 (get_load_store_type): Take a masked_p parameter and update
6320 call to get_group_load_store_type.
6321 (vectorizable_store): Update call to get_load_store_type.
6322 Handle IFN_MASK_STORE_LANES.
6323 (vectorizable_load): Update call to get_load_store_type.
6324 Handle IFN_MASK_LOAD_LANES.
6326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6327 Alan Hayward <alan.hayward@arm.com>
6328 David Sherwood <david.sherwood@arm.com>
6330 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6332 * config/aarch64/aarch64-protos.h
6333 (aarch64_sve_struct_memory_operand_p): Declare.
6334 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6335 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6336 (VPRED, vpred): Handle SVE structure modes.
6337 * config/aarch64/constraints.md (Utx): New constraint.
6338 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6339 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6340 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6341 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6342 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6343 structure modes. Split into pieces after RA.
6344 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6345 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6347 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6348 SVE structure modes.
6349 (aarch64_classify_address): Likewise.
6350 (sizetochar): Move earlier in file.
6351 (aarch64_print_operand): Handle SVE register lists.
6352 (aarch64_array_mode): New function.
6353 (aarch64_sve_struct_memory_operand_p): Likewise.
6354 (TARGET_ARRAY_MODE): Redefine.
6356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6357 Alan Hayward <alan.hayward@arm.com>
6358 David Sherwood <david.sherwood@arm.com>
6360 * target.def (array_mode): New target hook.
6361 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6362 * doc/tm.texi: Regenerate.
6363 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6364 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6365 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6367 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6371 Alan Hayward <alan.hayward@arm.com>
6372 David Sherwood <david.sherwood@arm.com>
6374 * fold-const.c (fold_binary_loc): Check the argument types
6375 rather than the result type when testing for a vector operation.
6377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6379 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6380 * doc/tm.texi: Regenerate.
6382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6383 Alan Hayward <alan.hayward@arm.com>
6384 David Sherwood <david.sherwood@arm.com>
6386 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6387 (sve): Document new AArch64 extension.
6388 * doc/md.texi (w): Extend the description of the AArch64
6389 constraint to include SVE vectors.
6390 (Upl, Upa): Document new AArch64 predicate constraints.
6391 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6393 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6394 (msve-vector-bits=): New option.
6395 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6396 SVE when these are disabled.
6397 (sve): New extension.
6398 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6399 modes. Adjust their number of units based on aarch64_sve_vg.
6400 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6401 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6402 aarch64_addr_query_type.
6403 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6404 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6405 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6406 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6407 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6408 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6409 (aarch64_simd_imm_zero_p): Delete.
6410 (aarch64_check_zero_based_sve_index_immediate): Declare.
6411 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6412 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6413 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6414 (aarch64_sve_float_mul_immediate_p): Likewise.
6415 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6417 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6418 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6419 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6420 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6421 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6422 (aarch64_regmode_natural_size): Likewise.
6423 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6424 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6426 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6427 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6428 for VG and the SVE predicate registers.
6429 (V_ALIASES): Add a "z"-prefixed alias.
6430 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6431 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6432 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6433 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6434 (REG_CLASS_NAMES): Add entries for them.
6435 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6436 and the predicate registers.
6437 (aarch64_sve_vg): Declare.
6438 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6439 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6440 (REGMODE_NATURAL_SIZE): Define.
6441 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6443 * config/aarch64/aarch64.c: Include cfgrtl.h.
6444 (simd_immediate_info): Add a constructor for series vectors,
6445 and an associated step field.
6446 (aarch64_sve_vg): New variable.
6447 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6448 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6449 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6450 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6451 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6452 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6453 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6454 (aarch64_get_mask_mode): New functions.
6455 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6456 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6457 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6458 predicate modes and predicate registers. Explicitly restrict
6459 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6460 to store a vector mode if it is recognized by
6461 aarch64_classify_vector_mode.
6462 (aarch64_regmode_natural_size): New function.
6463 (aarch64_hard_regno_caller_save_mode): Return the original mode
6465 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6466 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6467 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6468 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6470 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6471 does not overlap dest if the function is frame-related. Handle
6473 (aarch64_split_add_offset): New function.
6474 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6475 them aarch64_add_offset.
6476 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6477 and update call to aarch64_sub_sp.
6478 (aarch64_add_cfa_expression): New function.
6479 (aarch64_expand_prologue): Pass extra temporary registers to the
6480 functions above. Handle the case in which we need to emit new
6481 DW_CFA_expressions for registers that were originally saved
6482 relative to the stack pointer, but now have to be expressed
6483 relative to the frame pointer.
6484 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6486 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6487 IP0 and IP1 values for SVE frames.
6488 (aarch64_expand_vec_series): New function.
6489 (aarch64_expand_sve_widened_duplicate): Likewise.
6490 (aarch64_expand_sve_const_vector): Likewise.
6491 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6492 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6493 into the register, rather than emitting a SET directly.
6494 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6495 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6496 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6497 (offset_9bit_signed_scaled_p): New functions.
6498 (aarch64_replicate_bitmask_imm): New function.
6499 (aarch64_bitmask_imm): Use it.
6500 (aarch64_cannot_force_const_mem): Reject expressions involving
6501 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6502 (aarch64_classify_index): Handle SVE indices, by requiring
6503 a plain register index with a scale that matches the element size.
6504 (aarch64_classify_address): Handle SVE addresses. Assert that
6505 the mode of the address is VOIDmode or an integer mode.
6506 Update call to aarch64_classify_symbol.
6507 (aarch64_classify_symbolic_expression): Update call to
6508 aarch64_classify_symbol.
6509 (aarch64_const_vec_all_in_range_p): New function.
6510 (aarch64_print_vector_float_operand): Likewise.
6511 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6512 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6513 and the FP immediates 1.0 and 0.5.
6514 (aarch64_print_address_internal): Handle SVE addresses.
6515 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6516 (aarch64_regno_regclass): Handle predicate registers.
6517 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6519 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6520 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6521 (aarch64_convert_sve_vector_bits): New function.
6522 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6523 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6525 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6526 Handle SVE vector and predicate modes. Accept VL-based constants
6527 that need only one temporary register, and VL offsets that require
6528 no temporary registers.
6529 (aarch64_conditional_register_usage): Mark the predicate registers
6530 as fixed if SVE isn't available.
6531 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6532 Return true for SVE vector and predicate modes.
6533 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6534 rather than an unsigned int. Handle SVE modes.
6535 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6537 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6539 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6540 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6541 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6542 (aarch64_sve_float_mul_immediate_p): New functions.
6543 (aarch64_sve_valid_immediate): New function.
6544 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6545 Explicitly reject structure modes. Check for INDEX constants.
6546 Handle PTRUE and PFALSE constants.
6547 (aarch64_check_zero_based_sve_index_immediate): New function.
6548 (aarch64_simd_imm_zero_p): Delete.
6549 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6550 vector modes. Accept constants in the range of CNT[BHWD].
6551 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6552 ask for an Advanced SIMD mode.
6553 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6554 (aarch64_simd_vector_alignment): Handle SVE predicates.
6555 (aarch64_vectorize_preferred_vector_alignment): New function.
6556 (aarch64_simd_vector_alignment_reachable): Use it instead of
6558 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6559 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6561 (MAX_VECT_LEN): Delete.
6562 (expand_vec_perm_d): Add a vec_flags field.
6563 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6564 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6565 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6567 (aarch64_evpc_rev): Rename to...
6568 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6569 (aarch64_evpc_rev_global): New function.
6570 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6571 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6573 (aarch64_evpc_sve_tbl): New function.
6574 (aarch64_expand_vec_perm_const_1): Update after rename of
6575 aarch64_evpc_rev. Handle SVE permutes too, trying
6576 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6577 than aarch64_evpc_tbl.
6578 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6579 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6580 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6581 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6582 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6583 (aarch64_expand_sve_vcond): New functions.
6584 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6585 of aarch64_vector_mode_p.
6586 (aarch64_dwarf_poly_indeterminate_value): New function.
6587 (aarch64_compute_pressure_classes): Likewise.
6588 (aarch64_can_change_mode_class): Likewise.
6589 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6590 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6591 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6592 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6593 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6594 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6595 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6596 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6598 (Dn, Dl, Dr): Accept const as well as const_vector.
6599 (Dz): Likewise. Compare against CONST0_RTX.
6600 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6601 of "vector" where appropriate.
6602 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6603 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6604 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6605 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6606 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6607 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6608 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6609 (v_int_equiv): Extend to SVE modes.
6610 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6612 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6613 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6614 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6615 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6616 (SVE_COND_FP_CMP): New int iterators.
6617 (perm_hilo): Handle the new unpack unspecs.
6618 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6620 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6621 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6622 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6623 (aarch64_equality_operator, aarch64_constant_vector_operand)
6624 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6625 (aarch64_sve_nonimmediate_operand): Likewise.
6626 (aarch64_sve_general_operand): Likewise.
6627 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6628 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6629 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6630 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6631 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6632 (aarch64_sve_float_arith_immediate): Likewise.
6633 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6634 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6635 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6636 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6637 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6638 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6639 (aarch64_sve_float_arith_operand): Likewise.
6640 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6641 (aarch64_sve_float_mul_operand): Likewise.
6642 (aarch64_sve_vec_perm_operand): Likewise.
6643 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6644 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6645 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6646 as well as const_vector.
6647 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6648 in file. Use CONST0_RTX and CONSTM1_RTX.
6649 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6650 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6651 Use aarch64_simd_imm_zero.
6652 * config/aarch64/aarch64-sve.md: New file.
6653 * config/aarch64/aarch64.md: Include it.
6654 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6655 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6656 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6657 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6658 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6659 (sve): New attribute.
6660 (enabled): Disable instructions with the sve attribute unless
6662 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6663 aarch64_expand_mov_immediate.
6664 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6665 CNT[BHSD] immediates.
6666 (movti): Split CONST_POLY_INT moves into two halves.
6667 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6668 Split additions that need a temporary here if the destination
6669 is the stack pointer.
6670 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6671 (*add<mode>3_poly_1): New instruction.
6672 (set_clobber_cc): New expander.
6674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6676 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6677 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6678 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6679 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6680 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6681 Change innermode from fixed_mode_size to machine_mode.
6682 (simplify_subreg): Update call accordingly. Handle a constant-sized
6683 subreg of a variable-length CONST_VECTOR.
6685 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6686 Alan Hayward <alan.hayward@arm.com>
6687 David Sherwood <david.sherwood@arm.com>
6689 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6690 (add_offset_to_base): New function, split out from...
6691 (create_mem_ref): ...here. When handling a scale other than 1,
6692 check first whether the address is valid without the offset.
6693 Add it into the base if so, leaving the index and scale as-is.
6695 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6698 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6699 fold_for_warn before checking if arg2 is INTEGER_CST.
6701 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6703 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6704 (store_multiple_operation): Delete.
6705 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6706 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6707 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6708 guarded by TARGET_STRING.
6709 (rs6000_output_load_multiple): Delete.
6710 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6711 OPTION_MASK_STRING / TARGET_STRING handling.
6712 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6713 (const rs6000_opt_masks) <"string">: Change mask to 0.
6714 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6715 (MASK_STRING): Delete.
6716 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6718 (load_multiple): Delete.
6725 (store_multiple): Delete.
6732 (movmemsi_8reg): Delete.
6733 (corresponding unnamed define_insn): Delete.
6734 (movmemsi_6reg): Delete.
6735 (corresponding unnamed define_insn): Delete.
6736 (movmemsi_4reg): Delete.
6737 (corresponding unnamed define_insn): Delete.
6738 (movmemsi_2reg): Delete.
6739 (corresponding unnamed define_insn): Delete.
6740 (movmemsi_1reg): Delete.
6741 (corresponding unnamed define_insn): Delete.
6742 * config/rs6000/rs6000.opt (mno-string): New.
6743 (mstring): Replace by deprecation warning stub.
6744 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6746 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6748 * regrename.c (regrename_do_replace): If replacing the same
6749 reg multiple times, try to reuse last created gen_raw_REG.
6752 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6753 main to workaround a bug in GDB.
6755 2018-01-12 Tom de Vries <tom@codesourcery.com>
6758 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6760 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6762 PR rtl-optimization/80481
6763 * ira-color.c (get_cap_member): New function.
6764 (allocnos_conflict_by_live_ranges_p): Use it.
6765 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6766 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6768 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6771 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6772 (*saddl_se_1): Ditto.
6774 (*ssubl_se_1): Ditto.
6776 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6778 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6779 rather than wi::to_widest for DR_INITs.
6780 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6781 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6782 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6784 (vect_analyze_group_access_1): Note that here.
6786 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6788 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6789 polynomial type sizes.
6791 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6793 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6794 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6795 (gimple_add_tmp_var): Likewise.
6797 2018-01-12 Martin Liska <mliska@suse.cz>
6799 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6800 (gimple_alloc_sizes): Likewise.
6801 (dump_gimple_statistics): Use PRIu64 in printf format.
6802 * gimple.h: Change uint64_t to int.
6804 2018-01-12 Martin Liska <mliska@suse.cz>
6806 * tree-core.h: Use uint64_t instead of int.
6807 * tree.c (tree_node_counts): Likewise.
6808 (tree_node_sizes): Likewise.
6809 (dump_tree_statistics): Use PRIu64 in printf format.
6811 2018-01-12 Martin Liska <mliska@suse.cz>
6813 * Makefile.in: As qsort_chk is implemented in vec.c, add
6814 vec.o to linkage of gencfn-macros.
6815 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6816 passing the info to record_node_allocation_statistics.
6817 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6819 * ggc-common.c (struct ggc_usage): Add operator== and use
6820 it in operator< and compare function.
6821 * mem-stats.h (struct mem_usage): Likewise.
6822 * vec.c (struct vec_usage): Remove operator< and compare
6823 function. Can be simply inherited.
6825 2018-01-12 Martin Jambor <mjambor@suse.cz>
6828 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6829 * tree-ssa-math-opts.c: Include domwalk.h.
6830 (convert_mult_to_fma_1): New function.
6831 (fma_transformation_info): New type.
6832 (fma_deferring_state): Likewise.
6833 (cancel_fma_deferring): New function.
6834 (result_of_phi): Likewise.
6835 (last_fma_candidate_feeds_initial_phi): Likewise.
6836 (convert_mult_to_fma): Added deferring logic, split actual
6837 transformation to convert_mult_to_fma_1.
6838 (math_opts_dom_walker): New type.
6839 (math_opts_dom_walker::after_dom_children): New method, body moved
6840 here from pass_optimize_widening_mul::execute, added deferring logic
6842 (pass_optimize_widening_mul::execute): Moved most of code to
6843 math_opts_dom_walker::after_dom_children.
6844 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6845 * config/i386/i386.c (ix86_option_override_internal): Added
6846 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6848 2018-01-12 Richard Biener <rguenther@suse.de>
6851 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6852 inline instance vars.
6854 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6857 * config/rx/rx.c (rx_is_restricted_memory_address):
6860 2018-01-12 Richard Biener <rguenther@suse.de>
6862 PR tree-optimization/80846
6863 * target.def (split_reduction): New target hook.
6864 * targhooks.c (default_split_reduction): New function.
6865 * targhooks.h (default_split_reduction): Declare.
6866 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6867 target requests first reduce vectors by combining low and high
6869 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6870 (get_vectype_for_scalar_type_and_size): Export.
6871 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6872 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6873 * doc/tm.texi: Regenerate.
6874 * config/i386/i386.c (ix86_split_reduction): Implement
6875 TARGET_VECTORIZE_SPLIT_REDUCTION.
6877 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6880 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6881 in PIC mode except for TARGET_VXWORKS_RTP.
6882 * config/sparc/sparc.c: Include cfgrtl.h.
6883 (TARGET_INIT_PIC_REG): Define.
6884 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6885 (sparc_pic_register_p): New predicate.
6886 (sparc_legitimate_address_p): Use it.
6887 (sparc_legitimize_pic_address): Likewise.
6888 (sparc_delegitimize_address): Likewise.
6889 (sparc_mode_dependent_address_p): Likewise.
6890 (gen_load_pcrel_sym): Remove 4th parameter.
6891 (load_got_register): Adjust call to above. Remove obsolete stuff.
6892 (sparc_expand_prologue): Do not call load_got_register here.
6893 (sparc_flat_expand_prologue): Likewise.
6894 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6895 (sparc_use_pseudo_pic_reg): New function.
6896 (sparc_init_pic_reg): Likewise.
6897 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6898 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6900 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6902 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6903 Add item for branch_cost.
6905 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6907 PR rtl-optimization/83565
6908 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6909 not extend the result to a larger mode for rotate operations.
6910 (num_sign_bit_copies1): Likewise.
6912 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6915 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6917 Use values-Xc.o for -pedantic.
6918 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6920 2018-01-12 Martin Liska <mliska@suse.cz>
6923 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6925 (possible_polymorphic_call_targets): Use it.
6926 (ipa_devirt): Likewise.
6928 2018-01-12 Martin Liska <mliska@suse.cz>
6930 * profile-count.h (enum profile_quality): Use 0 as invalid
6931 enum value of profile_quality.
6933 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6935 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6936 -mext-string options.
6938 2018-01-12 Richard Biener <rguenther@suse.de>
6940 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6941 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6942 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6944 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6946 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6948 * configure.ac (--with-long-double-format): Add support for the
6949 configuration option to change the default long double format on
6951 * config.gcc (powerpc*-linux*-*): Likewise.
6952 * configure: Regenerate.
6953 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6954 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6955 used without modification.
6957 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6959 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6960 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6961 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6962 MISC_BUILTIN_SPEC_BARRIER.
6963 (rs6000_init_builtins): Likewise.
6964 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6966 (speculation_barrier): New define_insn.
6967 * doc/extend.texi: Document __builtin_speculation_barrier.
6969 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6972 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6973 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6974 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6976 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6977 integral modes instead of "ss" and "sd".
6978 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6979 vectors with 32-bit and 64-bit elements.
6980 (vecdupssescalarmodesuffix): New mode attribute.
6981 (vec_dup<mode>): Use it.
6983 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6986 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6987 frame if argument is passed on stack.
6989 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6992 * ree.c (combine_reaching_defs): Optimize also
6993 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6994 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6996 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6999 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7001 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7004 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7005 after they are computed.
7007 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7009 PR tree-optimization/83695
7010 * gimple-loop-linterchange.cc
7011 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7012 reset cached scev information after interchange.
7013 (pass_linterchange::execute): Remove call to scev_reset_htab.
7015 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7017 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7018 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7019 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7020 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7021 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7022 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7023 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7024 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7025 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7026 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7027 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7028 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7029 (V_lane_reg): Likewise.
7030 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7032 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7033 (vfmal_lane_low<mode>_intrinsic,
7034 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7035 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7036 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7037 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7038 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7039 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7041 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7043 * config/arm/arm-cpus.in (fp16fml): New feature.
7044 (ALL_SIMD): Add fp16fml.
7045 (armv8.2-a): Add fp16fml as an option.
7046 (armv8.3-a): Likewise.
7047 (armv8.4-a): Add fp16fml as part of fp16.
7048 * config/arm/arm.h (TARGET_FP16FML): Define.
7049 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7051 * config/arm/arm-modes.def (V2HF): Define.
7052 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7053 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7054 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7055 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7056 vfmsl_low, vfmsl_high): New set of builtins.
7057 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7058 (vfml_op): New code attribute.
7059 (VFMLHALVES): New int iterator.
7060 (VFML, VFMLSEL): New mode attributes.
7061 (V_reg): Define mapping for V2HF.
7062 (V_hi, V_lo): New mode attributes.
7063 (VF_constraint): Likewise.
7064 (vfml_half, vfml_half_selector): New int attributes.
7065 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7067 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7068 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7070 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7071 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7072 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7073 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7075 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7076 Document new effective target and option set.
7078 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7080 * config/arm/arm-cpus.in (armv8_4): New feature.
7081 (ARMv8_4a): New fgroup.
7082 (armv8.4-a): New arch.
7083 * config/arm/arm-tables.opt: Regenerate.
7084 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7085 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7086 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7087 Add matching rules for -march=armv8.4-a and extensions.
7088 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7090 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7093 * config/rx/rx.md (BW): New mode attribute.
7094 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7096 2018-01-11 Richard Biener <rguenther@suse.de>
7098 PR tree-optimization/83435
7099 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7100 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7101 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7103 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7104 Alan Hayward <alan.hayward@arm.com>
7105 David Sherwood <david.sherwood@arm.com>
7107 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7109 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7110 (aarch64_print_address_internal): Use it to check for a zero offset.
7112 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7113 Alan Hayward <alan.hayward@arm.com>
7114 David Sherwood <david.sherwood@arm.com>
7116 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7117 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7118 Return a poly_int64 rather than a HOST_WIDE_INT.
7119 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7120 rather than a HOST_WIDE_INT.
7121 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7122 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7123 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7124 final_offset from HOST_WIDE_INT to poly_int64.
7125 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7126 to_constant when getting the number of units in an Advanced SIMD
7128 (aarch64_builtin_vectorized_function): Check for a constant number
7130 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7132 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7133 attribute instead of GET_MODE_NUNITS.
7134 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7135 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7136 GET_MODE_SIZE for fixed-size registers.
7137 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7138 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7139 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7140 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7141 (aarch64_print_operand, aarch64_print_address_internal)
7142 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7143 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7144 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7145 Handle polynomial GET_MODE_SIZE.
7146 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7147 wider than SImode without modification.
7148 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7149 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7150 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7151 passing and returning SVE modes.
7152 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7153 rather than GEN_INT.
7154 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7155 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7156 (aarch64_allocate_and_probe_stack_space): Likewise.
7157 (aarch64_layout_frame): Cope with polynomial offsets.
7158 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7159 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7161 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7162 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7163 poly_int64 rather than a HOST_WIDE_INT.
7164 (aarch64_get_separate_components, aarch64_process_components)
7165 (aarch64_expand_prologue, aarch64_expand_epilogue)
7166 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7167 (aarch64_anchor_offset): New function, split out from...
7168 (aarch64_legitimize_address): ...here.
7169 (aarch64_builtin_vectorization_cost): Handle polynomial
7170 TYPE_VECTOR_SUBPARTS.
7171 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7173 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7174 number of elements from the PARALLEL rather than the mode.
7175 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7176 rather than GET_MODE_BITSIZE.
7177 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7178 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7179 (aarch64_expand_vec_perm_const_1): Handle polynomial
7180 d->perm.length () and d->perm elements.
7181 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7182 Apply to_constant to d->perm elements.
7183 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7184 polynomial CONST_VECTOR_NUNITS.
7185 (aarch64_move_pointer): Take amount as a poly_int64 rather
7187 (aarch64_progress_pointer): Avoid temporary variable.
7188 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7189 the mode attribute instead of GET_MODE.
7191 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7192 Alan Hayward <alan.hayward@arm.com>
7193 David Sherwood <david.sherwood@arm.com>
7195 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7196 x exists before using it.
7197 (aarch64_add_constant_internal): Rename to...
7198 (aarch64_add_offset_1): ...this. Replace regnum with separate
7199 src and dest rtxes. Handle the case in which they're different,
7200 including when the offset is zero. Replace scratchreg with an rtx.
7201 Use 2 additions if there is no spare register into which we can
7202 move a 16-bit constant.
7203 (aarch64_add_constant): Delete.
7204 (aarch64_add_offset): Replace reg with separate src and dest
7205 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7206 Use aarch64_add_offset_1.
7207 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7208 an rtx rather than an int. Take the delta as a poly_int64
7209 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7210 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7211 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7212 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7213 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7215 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7216 aarch64_add_constant.
7218 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7220 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7221 Use scalar_float_mode.
7223 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7225 * config/aarch64/aarch64-simd.md
7226 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7227 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7228 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7229 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7230 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7231 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7232 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7233 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7234 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7235 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7237 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7240 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7241 targ_options->x_arm_arch_string is non NULL.
7243 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7245 * config/aarch64/aarch64.h
7246 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7248 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7251 * expmed.c (emit_store_flag_force): Swap if const op0
7252 and change VOIDmode to mode of op0.
7254 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7256 PR rtl-optimization/83761
7257 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7258 than bytes to mode_for_size.
7260 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7263 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7264 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7267 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7270 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7271 when in layout mode.
7272 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7273 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7276 2018-01-10 Michael Collison <michael.collison@arm.com>
7278 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7279 * config/aarch64/aarch64-option-extension.def: Add
7280 AARCH64_OPT_EXTENSION of 'fp16fml'.
7281 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7282 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7283 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7284 * config/aarch64/constraints.md (Ui7): New constraint.
7285 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7286 (VFMLA_SEL_W): Ditto.
7289 (VFMLA16_LOW): New int iterator.
7290 (VFMLA16_HIGH): Ditto.
7291 (UNSPEC_FMLAL): New unspec.
7292 (UNSPEC_FMLSL): Ditto.
7293 (UNSPEC_FMLAL2): Ditto.
7294 (UNSPEC_FMLSL2): Ditto.
7295 (f16mac): New code attribute.
7296 * config/aarch64/aarch64-simd-builtins.def
7297 (aarch64_fmlal_lowv2sf): Ditto.
7298 (aarch64_fmlsl_lowv2sf): Ditto.
7299 (aarch64_fmlalq_lowv4sf): Ditto.
7300 (aarch64_fmlslq_lowv4sf): Ditto.
7301 (aarch64_fmlal_highv2sf): Ditto.
7302 (aarch64_fmlsl_highv2sf): Ditto.
7303 (aarch64_fmlalq_highv4sf): Ditto.
7304 (aarch64_fmlslq_highv4sf): Ditto.
7305 (aarch64_fmlal_lane_lowv2sf): Ditto.
7306 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7307 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7308 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7309 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7310 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7311 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7312 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7313 (aarch64_fmlal_lane_highv2sf): Ditto.
7314 (aarch64_fmlsl_lane_highv2sf): Ditto.
7315 (aarch64_fmlal_laneq_highv2sf): Ditto.
7316 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7317 (aarch64_fmlalq_lane_highv4sf): Ditto.
7318 (aarch64_fmlsl_lane_highv4sf): Ditto.
7319 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7320 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7321 * config/aarch64/aarch64-simd.md:
7322 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7323 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7324 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7325 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7326 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7327 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7328 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7329 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7330 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7331 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7332 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7333 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7334 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7335 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7336 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7337 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7338 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7339 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7340 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7341 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7342 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7343 (vfmlsl_low_u32): Ditto.
7344 (vfmlalq_low_u32): Ditto.
7345 (vfmlslq_low_u32): Ditto.
7346 (vfmlal_high_u32): Ditto.
7347 (vfmlsl_high_u32): Ditto.
7348 (vfmlalq_high_u32): Ditto.
7349 (vfmlslq_high_u32): Ditto.
7350 (vfmlal_lane_low_u32): Ditto.
7351 (vfmlsl_lane_low_u32): Ditto.
7352 (vfmlal_laneq_low_u32): Ditto.
7353 (vfmlsl_laneq_low_u32): Ditto.
7354 (vfmlalq_lane_low_u32): Ditto.
7355 (vfmlslq_lane_low_u32): Ditto.
7356 (vfmlalq_laneq_low_u32): Ditto.
7357 (vfmlslq_laneq_low_u32): Ditto.
7358 (vfmlal_lane_high_u32): Ditto.
7359 (vfmlsl_lane_high_u32): Ditto.
7360 (vfmlal_laneq_high_u32): Ditto.
7361 (vfmlsl_laneq_high_u32): Ditto.
7362 (vfmlalq_lane_high_u32): Ditto.
7363 (vfmlslq_lane_high_u32): Ditto.
7364 (vfmlalq_laneq_high_u32): Ditto.
7365 (vfmlslq_laneq_high_u32): Ditto.
7366 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7367 (AARCH64_FL_FOR_ARCH8_4): New.
7368 (AARCH64_ISA_F16FML): New ISA flag.
7369 (TARGET_F16FML): New feature flag for fp16fml.
7370 (doc/invoke.texi): Document new fp16fml option.
7372 2018-01-10 Michael Collison <michael.collison@arm.com>
7374 * config/aarch64/aarch64-builtins.c:
7375 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7376 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7377 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7378 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7379 (AARCH64_ISA_SHA3): New ISA flag.
7380 (TARGET_SHA3): New feature flag for sha3.
7381 * config/aarch64/iterators.md (sha512_op): New int attribute.
7382 (CRYPTO_SHA512): New int iterator.
7383 (UNSPEC_SHA512H): New unspec.
7384 (UNSPEC_SHA512H2): Ditto.
7385 (UNSPEC_SHA512SU0): Ditto.
7386 (UNSPEC_SHA512SU1): Ditto.
7387 * config/aarch64/aarch64-simd-builtins.def
7388 (aarch64_crypto_sha512hqv2di): New builtin.
7389 (aarch64_crypto_sha512h2qv2di): Ditto.
7390 (aarch64_crypto_sha512su0qv2di): Ditto.
7391 (aarch64_crypto_sha512su1qv2di): Ditto.
7392 (aarch64_eor3qv8hi): Ditto.
7393 (aarch64_rax1qv2di): Ditto.
7394 (aarch64_xarqv2di): Ditto.
7395 (aarch64_bcaxqv8hi): Ditto.
7396 * config/aarch64/aarch64-simd.md:
7397 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7398 (aarch64_crypto_sha512su0qv2di): Ditto.
7399 (aarch64_crypto_sha512su1qv2di): Ditto.
7400 (aarch64_eor3qv8hi): Ditto.
7401 (aarch64_rax1qv2di): Ditto.
7402 (aarch64_xarqv2di): Ditto.
7403 (aarch64_bcaxqv8hi): Ditto.
7404 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7405 (vsha512h2q_u64): Ditto.
7406 (vsha512su0q_u64): Ditto.
7407 (vsha512su1q_u64): Ditto.
7408 (veor3q_u16): Ditto.
7409 (vrax1q_u64): Ditto.
7411 (vbcaxq_u16): Ditto.
7412 * config/arm/types.md (crypto_sha512): New type attribute.
7413 (crypto_sha3): Ditto.
7414 (doc/invoke.texi): Document new sha3 option.
7416 2018-01-10 Michael Collison <michael.collison@arm.com>
7418 * config/aarch64/aarch64-builtins.c:
7419 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7421 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7422 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7423 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7424 (AARCH64_ISA_SM4): New ISA flag.
7425 (TARGET_SM4): New feature flag for sm4.
7426 * config/aarch64/aarch64-simd-builtins.def
7427 (aarch64_sm3ss1qv4si): Ditto.
7428 (aarch64_sm3tt1aq4si): Ditto.
7429 (aarch64_sm3tt1bq4si): Ditto.
7430 (aarch64_sm3tt2aq4si): Ditto.
7431 (aarch64_sm3tt2bq4si): Ditto.
7432 (aarch64_sm3partw1qv4si): Ditto.
7433 (aarch64_sm3partw2qv4si): Ditto.
7434 (aarch64_sm4eqv4si): Ditto.
7435 (aarch64_sm4ekeyqv4si): Ditto.
7436 * config/aarch64/aarch64-simd.md:
7437 (aarch64_sm3ss1qv4si): Ditto.
7438 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7439 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7440 (aarch64_sm4eqv4si): Ditto.
7441 (aarch64_sm4ekeyqv4si): Ditto.
7442 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7443 (sm3part_op): Ditto.
7444 (CRYPTO_SM3TT): Ditto.
7445 (CRYPTO_SM3PART): Ditto.
7446 (UNSPEC_SM3SS1): New unspec.
7447 (UNSPEC_SM3TT1A): Ditto.
7448 (UNSPEC_SM3TT1B): Ditto.
7449 (UNSPEC_SM3TT2A): Ditto.
7450 (UNSPEC_SM3TT2B): Ditto.
7451 (UNSPEC_SM3PARTW1): Ditto.
7452 (UNSPEC_SM3PARTW2): Ditto.
7453 (UNSPEC_SM4E): Ditto.
7454 (UNSPEC_SM4EKEY): Ditto.
7455 * config/aarch64/constraints.md (Ui2): New constraint.
7456 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7457 * config/arm/types.md (crypto_sm3): New type attribute.
7458 (crypto_sm4): Ditto.
7459 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7460 (vsm3tt1aq_u32): Ditto.
7461 (vsm3tt1bq_u32): Ditto.
7462 (vsm3tt2aq_u32): Ditto.
7463 (vsm3tt2bq_u32): Ditto.
7464 (vsm3partw1q_u32): Ditto.
7465 (vsm3partw2q_u32): Ditto.
7466 (vsm4eq_u32): Ditto.
7467 (vsm4ekeyq_u32): Ditto.
7468 (doc/invoke.texi): Document new sm4 option.
7470 2018-01-10 Michael Collison <michael.collison@arm.com>
7472 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7473 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7474 (AARCH64_FL_FOR_ARCH8_4): New.
7475 (AARCH64_FL_V8_4): New flag.
7476 (doc/invoke.texi): Document new armv8.4-a option.
7478 2018-01-10 Michael Collison <michael.collison@arm.com>
7480 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7481 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7482 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7483 * config/aarch64/aarch64-option-extension.def: Add
7484 AARCH64_OPT_EXTENSION of 'sha2'.
7485 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7486 (crypto): Disable sha2 and aes if crypto disabled.
7487 (crypto): Enable aes and sha2 if enabled.
7488 (simd): Disable sha2 and aes if simd disabled.
7489 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7491 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7492 (TARGET_SHA2): New feature flag for sha2.
7493 (TARGET_AES): New feature flag for aes.
7494 * config/aarch64/aarch64-simd.md:
7495 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7496 conditional on TARGET_AES.
7497 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7498 (aarch64_crypto_sha1hsi): Make pattern conditional
7500 (aarch64_crypto_sha1hv4si): Ditto.
7501 (aarch64_be_crypto_sha1hv4si): Ditto.
7502 (aarch64_crypto_sha1su1v4si): Ditto.
7503 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7504 (aarch64_crypto_sha1su0v4si): Ditto.
7505 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7506 (aarch64_crypto_sha256su0v4si): Ditto.
7507 (aarch64_crypto_sha256su1v4si): Ditto.
7508 (doc/invoke.texi): Document new aes and sha2 options.
7510 2018-01-10 Martin Sebor <msebor@redhat.com>
7512 PR tree-optimization/83781
7513 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7516 2018-01-11 Martin Sebor <msebor@gmail.com>
7517 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7519 PR tree-optimization/83501
7520 PR tree-optimization/81703
7522 * tree-ssa-strlen.c (get_string_cst): Rename...
7523 (get_string_len): ...to this. Handle global constants.
7524 (handle_char_store): Adjust.
7526 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7527 Jim Wilson <jimw@sifive.com>
7529 * config/riscv/riscv-protos.h (riscv_output_return): New.
7530 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7531 (riscv_attribute_table, riscv_output_return),
7532 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7533 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7534 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7536 (riscv_expand_prologue): Add early return for naked function.
7537 (riscv_expand_epilogue): Likewise.
7538 (riscv_function_ok_for_sibcall): Return false for naked function.
7539 (riscv_set_current_function): New.
7540 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7541 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7542 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7543 * doc/extend.texi (RISC-V Function Attributes): New.
7545 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7547 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7548 check for 128-bit long double before checking TCmode.
7549 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7550 128-bit long doubles before checking TFmode or TCmode.
7551 (FLOAT128_IBM_P): Likewise.
7553 2018-01-10 Martin Sebor <msebor@redhat.com>
7555 PR tree-optimization/83671
7556 * builtins.c (c_strlen): Unconditionally return zero for the empty
7558 Use -Warray-bounds for warnings.
7559 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7560 for non-constant array indices with COMPONENT_REF, arrays of
7561 arrays, and pointers to arrays.
7562 (gimple_fold_builtin_strlen): Determine and set length range for
7563 non-constant character arrays.
7565 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7568 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7571 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7573 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7575 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7578 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7579 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7580 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7581 indexed_or_indirect_operand predicate.
7582 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7583 (*vsx_le_perm_load_v8hi): Likewise.
7584 (*vsx_le_perm_load_v16qi): Likewise.
7585 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7586 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7587 (*vsx_le_perm_store_v8hi): Likewise.
7588 (*vsx_le_perm_store_v16qi): Likewise.
7589 (eight unnamed splitters): Likewise.
7591 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7593 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7594 * config/rs6000/emmintrin.h: Likewise.
7595 * config/rs6000/mmintrin.h: Likewise.
7596 * config/rs6000/xmmintrin.h: Likewise.
7598 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7601 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7603 * tree.c (tree_nop_conversion): Return true for location wrapper
7605 (maybe_wrap_with_location): New function.
7606 (selftest::check_strip_nops): New function.
7607 (selftest::test_location_wrappers): New function.
7608 (selftest::tree_c_tests): Call it.
7609 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7610 (maybe_wrap_with_location): New decl.
7611 (EXPR_LOCATION_WRAPPER_P): New macro.
7612 (location_wrapper_p): New inline function.
7613 (tree_strip_any_location_wrapper): New inline function.
7615 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7618 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7619 stack_realign_offset for the largest alignment of stack slot
7621 (ix86_find_max_used_stack_alignment): New function.
7622 (ix86_finalize_stack_frame_flags): Use it. Set
7623 max_used_stack_alignment if we don't realign stack.
7624 * config/i386/i386.h (machine_function): Add
7625 max_used_stack_alignment.
7627 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7629 * config/arm/arm.opt (-mbranch-cost): New option.
7630 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7633 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7636 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7637 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7639 2018-01-10 Richard Biener <rguenther@suse.de>
7642 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7643 early out so it also covers the case where we have a non-NULL
7646 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7648 PR tree-optimization/83753
7649 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7650 for non-strided grouped accesses if the number of elements is 1.
7652 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7655 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7656 * i386.h (TARGET_USE_GATHER): Define.
7657 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7659 2018-01-10 Martin Liska <mliska@suse.cz>
7662 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7663 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7665 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7666 CLEANUP_NO_PARTITIONING is not set.
7668 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7670 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7671 for vectors, as a partial revert of r254296.
7672 * rtl.h (const_vec_p): Delete.
7673 (const_vec_duplicate_p): Don't test for vector CONSTs.
7674 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7675 * expmed.c (make_tree): Likewise.
7678 * common.md (E, F): Use CONSTANT_P instead of checking for
7680 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7681 checking for CONST_VECTOR.
7683 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7686 * predict.c (force_edge_cold): Handle in more sane way edges
7689 2018-01-09 Carl Love <cel@us.ibm.com>
7691 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7693 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7694 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7695 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7696 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7697 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7698 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7699 * config/rs6000/rs6000-protos.h: Add extern defition for
7700 rs6000_generate_float2_double_code.
7701 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7703 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7704 (float2_v2df): Add define_expand.
7706 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7709 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7710 op_mode in the force_to_mode call.
7712 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7714 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7715 instead of checking each element individually.
7716 (aarch64_evpc_uzp): Likewise.
7717 (aarch64_evpc_zip): Likewise.
7718 (aarch64_evpc_ext): Likewise.
7719 (aarch64_evpc_rev): Likewise.
7720 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7721 instead of checking each element individually. Return true without
7723 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7724 whether all selected elements come from the same input, instead of
7725 checking each element individually. Remove calls to gen_rtx_REG,
7726 start_sequence and end_sequence and instead assert that no rtl is
7729 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7731 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7732 order of HIGH and CONST checks.
7734 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7736 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7737 if the destination isn't an SSA_NAME.
7739 2018-01-09 Richard Biener <rguenther@suse.de>
7741 PR tree-optimization/83668
7742 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7744 (canonicalize_loop_form): ... here, renamed from ...
7745 (canonicalize_loop_closed_ssa_form): ... this and amended to
7746 swap successor edges for loop exit blocks to make us use
7747 the RPO order we need for initial schedule generation.
7749 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7751 PR tree-optimization/64811
7752 * match.pd: When optimizing comparisons with Inf, avoid
7753 introducing or losing exceptions from comparisons with NaN.
7755 2018-01-09 Martin Liska <mliska@suse.cz>
7758 * asan.c (shadow_mem_size): Add gcc_assert.
7760 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7762 Don't save registers in main().
7765 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7766 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7767 * config/avr/avr.c (avr_set_current_function): Don't error if
7768 naked, OS_task or OS_main are specified at the same time.
7769 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7771 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7773 * common/config/avr/avr-common.c (avr_option_optimization_table):
7774 Switch on -mmain-is-OS_task for optimizing compilations.
7776 2018-01-09 Richard Biener <rguenther@suse.de>
7778 PR tree-optimization/83572
7779 * graphite.c: Include cfganal.h.
7780 (graphite_transform_loops): Connect infinite loops to exit
7781 and remove fake edges at the end.
7783 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7785 * ipa-inline.c (edge_badness): Revert accidental checkin.
7787 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7790 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7791 symbols; not inline clones.
7793 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7796 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7797 hard registers. Formatting fixes.
7799 PR preprocessor/83722
7800 * gcc.c (try_generate_repro): Pass
7801 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7802 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7805 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7806 Kito Cheng <kito.cheng@gmail.com>
7808 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7809 (riscv_leaf_function_p): Delete.
7810 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7812 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7814 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7816 (do_ifelse): New function.
7817 (do_isel): New function.
7818 (do_sub3): New function.
7819 (do_add3): New function.
7820 (do_load_mask_compare): New function.
7821 (do_overlap_load_compare): New function.
7822 (expand_compare_loop): New function.
7823 (expand_block_compare): Call expand_compare_loop() when appropriate.
7824 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7826 (-mblock-compare-inline-loop-limit): New option.
7828 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7831 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7832 Reverse order of second and third operands in first alternative.
7833 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7834 of first and second elements in UNSPEC_VPERMR vector.
7835 (altivec_expand_vec_perm_le): Likewise.
7837 2018-01-08 Jeff Law <law@redhat.com>
7839 PR rtl-optimizatin/81308
7840 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7841 (process_switch): If group_case_labels makes a change, then set
7843 (pass_convert_switch::execute): If a switch is converted, then
7844 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7846 PR rtl-optimization/81308
7847 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7850 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7852 PR target/83663 - Revert r255946
7853 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7854 generation for cases where splatting a value is not useful.
7855 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7856 across a vec_duplicate and a paradoxical subreg forming a vector
7857 mode to a vec_concat.
7859 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7861 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7862 -march=armv8.3-a variants.
7863 * config/arm/t-multilib: Likewise.
7864 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7866 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7868 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7870 (cceq_ior_compare_complement): Give it a name so I can use it, and
7871 change boolean_or_operator predicate to boolean_operator so it can
7872 be used to generate a crand.
7873 (eqne): New code iterator.
7874 (bd/bd_neg): New code_attrs.
7875 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7876 a single define_insn.
7877 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7878 decrement (bdnzt/bdnzf/bdzt/bdzf).
7879 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7880 with the new names of the branch decrement patterns, and added the
7881 names of the branch decrement conditional patterns.
7883 2018-01-08 Richard Biener <rguenther@suse.de>
7885 PR tree-optimization/83563
7886 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7889 2018-01-08 Richard Biener <rguenther@suse.de>
7892 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7894 2018-01-08 Richard Biener <rguenther@suse.de>
7896 PR tree-optimization/83685
7897 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7898 references to abnormals.
7900 2018-01-08 Richard Biener <rguenther@suse.de>
7903 * dwarf2out.c (output_indirect_strings): Handle empty
7904 skeleton_debug_str_hash.
7905 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7907 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7909 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7910 (emit_store_direct): Likewise.
7911 (arc_trampoline_adjust_address): Likewise.
7912 (arc_asm_trampoline_template): New function.
7913 (arc_initialize_trampoline): Use asm_trampoline_template.
7914 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7915 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7916 * config/arc/arc.md (flush_icache): Delete pattern.
7918 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7920 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7921 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7924 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7927 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7928 by not USED_FOR_TARGET.
7929 (make_pass_resolve_sw_modes): Likewise.
7931 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7933 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7936 2018-01-08 Richard Biener <rguenther@suse.de>
7939 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7941 2018-01-08 Richard Biener <rguenther@suse.de>
7944 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7946 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7949 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7950 basic blocks with a small number of successors.
7951 (convert_control_dep_chain_into_preds): Improve handling of
7953 (dump_predicates): Split apart into...
7954 (dump_pred_chain): ...here...
7955 (dump_pred_info): ...and here.
7956 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7957 (can_chain_union_be_invalidated_p): Improve check for invalidation
7959 (uninit_uses_cannot_happen): Avoid unnecessary if
7960 convert_control_dep_chain_into_preds yielded nothing.
7962 2018-01-06 Martin Sebor <msebor@redhat.com>
7964 PR tree-optimization/83640
7965 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7966 subtracting negative offset from size.
7967 (builtin_access::overlap): Adjust offset bounds of the access to fall
7968 within the size of the object if possible.
7970 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7972 PR rtl-optimization/83699
7973 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7974 extract_bit_field_as_subreg to cases in which the extracted
7975 value is also a vector.
7977 * lra-constraints.c (process_alt_operands): Test for the equivalence
7978 substitutions when detecting a possible reload cycle.
7980 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7983 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7984 by default if flag_selective_schedling{,2}. Formatting fixes.
7986 PR rtl-optimization/83682
7987 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7988 if it has non-VECTOR_MODE element mode.
7989 (vec_duplicate_p): Likewise.
7992 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7993 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7995 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7998 * config/i386/i386-builtin.def
7999 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8000 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8001 Require also OPTION_MASK_ISA_AVX512F in addition to
8002 OPTION_MASK_ISA_GFNI.
8003 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8004 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8005 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8006 to OPTION_MASK_ISA_GFNI.
8007 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8008 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8009 OPTION_MASK_ISA_AVX512BW.
8010 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8011 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8012 addition to OPTION_MASK_ISA_GFNI.
8013 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8014 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8015 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8016 to OPTION_MASK_ISA_GFNI.
8017 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8018 a requirement for all ISAs rather than any of them with a few
8020 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8022 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8023 bitmasks to be enabled with 3 exceptions, instead of requiring any
8024 enabled ISA with lots of exceptions.
8025 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8026 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8027 Change avx512bw in isa attribute to avx512f.
8028 * config/i386/sgxintrin.h: Add license boilerplate.
8029 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8030 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8031 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8032 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8034 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8035 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8036 temporarily sse2 rather than sse if not enabled already.
8039 * config/i386/sse.md (VI248_VLBW): Rename to ...
8040 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8041 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8042 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8043 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8044 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8045 mode iterator instead of VI248_VLBW.
8047 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8049 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8050 (record_modified): Skip clobbers; add debug output.
8051 (param_change_prob): Use sreal frequencies.
8053 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8055 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8056 punt for user-aligned variables.
8058 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8060 * tree-chrec.c (chrec_contains_symbols): Return true for
8063 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8066 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8067 of (x|y) == x for BICS pattern.
8069 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8071 PR tree-optimization/83605
8072 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8073 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8076 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8078 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8079 * config/epiphany/rtems.h: New file.
8081 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8082 Uros Bizjak <ubizjak@gmail.com>
8085 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8086 QIreg_operand instead of register_operand predicate.
8087 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8088 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8089 comments instead of -fmitigate[-_]rop.
8091 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8094 * cgraphunit.c (symbol_table::compile): Switch to text_section
8095 before calling assembly_start debug hook.
8096 * run-rtl-passes.c (run_rtl_passes): Likewise.
8099 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8101 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8102 range_int_cst_p rather than !symbolic_range_p before calling
8103 extract_range_from_multiplicative_op_1.
8105 2018-01-04 Jeff Law <law@redhat.com>
8107 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8108 redundant test in assertion.
8110 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8112 * doc/rtl.texi: Document machine_mode wrapper classes.
8114 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8116 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8119 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8121 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8122 the VEC_PERM_EXPR fold to fail.
8124 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8127 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8128 to switched_sections.
8130 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8133 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8136 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8139 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8140 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8142 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8145 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8146 is BLKmode and bitpos not zero or mode change is needed.
8148 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8151 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8154 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8157 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8158 instead of MULT rtx. Update all corresponding splitters.
8160 (*ssub<modesuffix>): Ditto.
8162 (*cmp_sadd_di): Update split patterns.
8163 (*cmp_sadd_si): Ditto.
8164 (*cmp_sadd_sidi): Ditto.
8165 (*cmp_ssub_di): Ditto.
8166 (*cmp_ssub_si): Ditto.
8167 (*cmp_ssub_sidi): Ditto.
8168 * config/alpha/predicates.md (const23_operand): New predicate.
8169 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8170 Look for ASHIFT, not MULT inner operand.
8171 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8173 2018-01-04 Martin Liska <mliska@suse.cz>
8175 PR gcov-profile/83669
8176 * gcov.c (output_intermediate_file): Add version to intermediate
8178 * doc/gcov.texi: Document new field 'version' in intermediate
8179 file format. Fix location of '-k' option of gcov command.
8181 2018-01-04 Martin Liska <mliska@suse.cz>
8184 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8186 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8188 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8190 2018-01-03 Martin Sebor <msebor@redhat.com>
8192 PR tree-optimization/83655
8193 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8194 checking calls with invalid arguments.
8196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8198 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8199 (vectorizable_mask_load_store): Delete.
8200 (vectorizable_call): Return false for masked loads and stores.
8201 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8202 instead of gimple_assign_rhs1.
8203 (vectorizable_load): Handle IFN_MASK_LOAD.
8204 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8208 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8210 (vectorizable_mask_load_store): ...here.
8211 (vectorizable_load): ...and here.
8213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8215 * tree-vect-stmts.c (vect_build_all_ones_mask)
8216 (vect_build_zero_merge_argument): New functions, split out from...
8217 (vectorizable_load): ...here.
8219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8221 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8223 (vectorizable_mask_load_store): ...here.
8224 (vectorizable_store): ...and here.
8226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8228 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8230 (vectorizable_mask_load_store): ...here.
8232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8234 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8235 (vect_model_store_cost): Take a vec_load_store_type instead of a
8237 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8238 (vect_model_store_cost): Take a vec_load_store_type instead of a
8240 (vectorizable_mask_load_store): Update accordingly.
8241 (vectorizable_store): Likewise.
8242 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8246 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8247 IFN_MASK_LOAD calls here rather than...
8248 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8251 Alan Hayward <alan.hayward@arm.com>
8252 David Sherwood <david.sherwood@arm.com>
8254 * expmed.c (extract_bit_field_1): For vector extracts,
8255 fall back to extract_bit_field_as_subreg if vec_extract
8258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8259 Alan Hayward <alan.hayward@arm.com>
8260 David Sherwood <david.sherwood@arm.com>
8262 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8263 they are variable or constant sized.
8264 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8265 slots for constant-sized data.
8267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8268 Alan Hayward <alan.hayward@arm.com>
8269 David Sherwood <david.sherwood@arm.com>
8271 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8272 handling COND_EXPRs with boolean comparisons, try to find a better
8273 basis for the mask type than the boolean itself.
8275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8277 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8278 is calculated and how it can be overridden.
8279 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8280 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8282 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8286 Alan Hayward <alan.hayward@arm.com>
8287 David Sherwood <david.sherwood@arm.com>
8289 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8290 Remove the mode argument.
8291 (aarch64_simd_valid_immediate): Remove the mode and inverse
8293 * config/aarch64/iterators.md (bitsize): New iterator.
8294 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8295 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8296 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8297 aarch64_simd_valid_immediate.
8298 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8299 (aarch64_reg_or_bic_imm): Likewise.
8300 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8301 with an insn_type enum and msl with a modifier_type enum.
8302 Replace element_width with a scalar_mode. Change the shift
8303 to unsigned int. Add constructors for scalar_float_mode and
8304 scalar_int_mode elements.
8305 (aarch64_vect_float_const_representable_p): Delete.
8306 (aarch64_can_const_movi_rtx_p)
8307 (aarch64_simd_scalar_immediate_valid_for_move)
8308 (aarch64_simd_make_constant): Update call to
8309 aarch64_simd_valid_immediate.
8310 (aarch64_advsimd_valid_immediate_hs): New function.
8311 (aarch64_advsimd_valid_immediate): Likewise.
8312 (aarch64_simd_valid_immediate): Remove mode and inverse
8313 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8314 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8315 and aarch64_float_const_representable_p on the result.
8316 (aarch64_output_simd_mov_immediate): Remove mode argument.
8317 Update call to aarch64_simd_valid_immediate and use of
8318 simd_immediate_info.
8319 (aarch64_output_scalar_simd_mov_immediate): Update call
8322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8323 Alan Hayward <alan.hayward@arm.com>
8324 David Sherwood <david.sherwood@arm.com>
8326 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8327 (mode_nunits): Likewise CONST_MODE_NUNITS.
8328 * machmode.def (ADJUST_NUNITS): Document.
8329 * genmodes.c (mode_data::need_nunits_adj): New field.
8330 (blank_mode): Update accordingly.
8331 (adj_nunits): New variable.
8332 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8334 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8335 listed in adj_nunits.
8336 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8337 listed in adj_nunits. Don't emit case statements for such modes.
8338 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8339 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8340 nothing if adj_nunits is nonnull.
8341 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8342 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8343 (emit_mode_fbit): Update use of print_maybe_const_decl.
8344 (emit_move_size): Likewise. Treat the array as non-const
8346 (emit_mode_adjustments): Handle adj_nunits.
8348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8350 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8351 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8352 (VECTOR_MODES): Use it.
8353 (make_vector_modes): Take the prefix as an argument.
8355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8356 Alan Hayward <alan.hayward@arm.com>
8357 David Sherwood <david.sherwood@arm.com>
8359 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8360 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8361 for MODE_VECTOR_BOOL.
8362 * machmode.def (VECTOR_BOOL_MODE): Document.
8363 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8364 (make_vector_bool_mode): New function.
8365 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8367 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8368 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8370 * stor-layout.c (int_mode_for_mode): Likewise.
8371 * tree.c (build_vector_type_for_mode): Likewise.
8372 * varasm.c (output_constant_pool_2): Likewise.
8373 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8374 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8375 for MODE_VECTOR_BOOL.
8376 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8377 of mode class checks.
8378 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8379 instead of a list of mode class checks.
8380 (expand_vector_scalar_condition): Likewise.
8381 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8384 Alan Hayward <alan.hayward@arm.com>
8385 David Sherwood <david.sherwood@arm.com>
8387 * machmode.h (mode_size): Change from unsigned short to
8389 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8390 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8391 or if measurement_type is not polynomial.
8392 (fixed_size_mode::includes_p): Check for constant-sized modes.
8393 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8394 return a poly_uint16 rather than an unsigned short.
8395 (emit_mode_size): Change the type of mode_size from unsigned short
8396 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8397 (emit_mode_adjustments): Cope with polynomial vector sizes.
8398 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8400 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8402 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8403 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8404 * caller-save.c (setup_save_areas): Likewise.
8405 (replace_reg_with_saved_mem): Likewise.
8406 * calls.c (emit_library_call_value_1): Likewise.
8407 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8408 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8409 (gen_lowpart_for_combine): Likewise.
8410 * convert.c (convert_to_integer_1): Likewise.
8411 * cse.c (equiv_constant, cse_insn): Likewise.
8412 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8413 (cselib_subst_to_values): Likewise.
8414 * dce.c (word_dce_process_block): Likewise.
8415 * df-problems.c (df_word_lr_mark_ref): Likewise.
8416 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8417 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8418 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8419 (rtl_for_decl_location): Likewise.
8420 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8421 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8422 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8423 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8424 (expand_expr_real_1): Likewise.
8425 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8426 (pad_below): Likewise.
8427 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8428 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8429 * ira.c (get_subreg_tracking_sizes): Likewise.
8430 * ira-build.c (ira_create_allocno_objects): Likewise.
8431 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8432 (ira_sort_regnos_for_alter_reg): Likewise.
8433 * ira-costs.c (record_operand_costs): Likewise.
8434 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8435 (resolve_simple_move): Likewise.
8436 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8437 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8438 (lra_constraints): Likewise.
8439 (CONST_POOL_OK_P): Reject variable-sized modes.
8440 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8441 (add_pseudo_to_slot, lra_spill): Likewise.
8442 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8443 * optabs-query.c (get_best_extraction_insn): Likewise.
8444 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8445 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8446 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8447 * recog.c (offsettable_address_addr_space_p): Likewise.
8448 * regcprop.c (maybe_mode_change): Likewise.
8449 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8450 * regrename.c (build_def_use): Likewise.
8451 * regstat.c (dump_reg_info): Likewise.
8452 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8453 (find_reloads, find_reloads_subreg_address): Likewise.
8454 * reload1.c (eliminate_regs_1): Likewise.
8455 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8456 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8457 (simplify_binary_operation_1, simplify_subreg): Likewise.
8458 * targhooks.c (default_function_arg_padding): Likewise.
8459 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8460 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8461 (verify_gimple_assign_ternary): Likewise.
8462 * tree-inline.c (estimate_move_cost): Likewise.
8463 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8464 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8465 (get_address_cost_ainc): Likewise.
8466 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8467 (vect_supportable_dr_alignment): Likewise.
8468 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8469 (vectorizable_reduction): Likewise.
8470 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8471 (vectorizable_operation, vectorizable_load): Likewise.
8472 * tree.c (build_same_sized_truth_vector_type): Likewise.
8473 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8474 * var-tracking.c (emit_note_insn_var_location): Likewise.
8475 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8476 (ADDR_VEC_ALIGN): Likewise.
8478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8479 Alan Hayward <alan.hayward@arm.com>
8480 David Sherwood <david.sherwood@arm.com>
8482 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8484 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8485 or if measurement_type is polynomial.
8486 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8487 * combine.c (make_extraction): Likewise.
8488 * dse.c (find_shift_sequence): Likewise.
8489 * dwarf2out.c (mem_loc_descriptor): Likewise.
8490 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8491 (extract_bit_field, extract_low_bits): Likewise.
8492 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8493 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8494 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8495 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8496 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8497 * reload.c (find_reloads): Likewise.
8498 * reload1.c (alter_reg): Likewise.
8499 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8500 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8501 * tree-if-conv.c (predicate_mem_writes): Likewise.
8502 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8503 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8504 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8505 * valtrack.c (dead_debug_insert_temp): Likewise.
8506 * varasm.c (mergeable_constant_section): Likewise.
8507 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8510 Alan Hayward <alan.hayward@arm.com>
8511 David Sherwood <david.sherwood@arm.com>
8513 * expr.c (expand_assignment): Cope with polynomial mode sizes
8514 when assigning to a CONCAT.
8516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8517 Alan Hayward <alan.hayward@arm.com>
8518 David Sherwood <david.sherwood@arm.com>
8520 * machmode.h (mode_precision): Change from unsigned short to
8522 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8524 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8525 or if measurement_type is not polynomial.
8526 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8527 in which the mode is already known to be a scalar_int_mode.
8528 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8529 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8531 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8532 for GET_MODE_PRECISION.
8533 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8534 for GET_MODE_PRECISION.
8535 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8537 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8538 (expand_field_assignment, make_extraction): Likewise.
8539 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8540 (get_last_value): Likewise.
8541 * convert.c (convert_to_integer_1): Likewise.
8542 * cse.c (cse_insn): Likewise.
8543 * expr.c (expand_expr_real_1): Likewise.
8544 * lra-constraints.c (simplify_operand_subreg): Likewise.
8545 * optabs-query.c (can_atomic_load_p): Likewise.
8546 * optabs.c (expand_atomic_load): Likewise.
8547 (expand_atomic_store): Likewise.
8548 * ree.c (combine_reaching_defs): Likewise.
8549 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8550 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8551 * tree.h (type_has_mode_precision_p): Likewise.
8552 * ubsan.c (instrument_si_overflow): Likewise.
8554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8555 Alan Hayward <alan.hayward@arm.com>
8556 David Sherwood <david.sherwood@arm.com>
8558 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8559 polynomial numbers of units.
8560 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8561 (valid_vector_subparts_p): New function.
8562 (build_vector_type): Remove temporary shim and take the number
8563 of units as a poly_uint64 rather than an int.
8564 (build_opaque_vector_type): Take the number of units as a
8565 poly_uint64 rather than an int.
8566 * tree.c (build_vector_from_ctor): Handle polynomial
8567 TYPE_VECTOR_SUBPARTS.
8568 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8569 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8570 (build_vector_from_val): If the number of units is variable,
8571 use build_vec_duplicate_cst for constant operands and
8572 VEC_DUPLICATE_EXPR otherwise.
8573 (make_vector_type): Remove temporary is_constant ().
8574 (build_vector_type, build_opaque_vector_type): Take the number of
8575 units as a poly_uint64 rather than an int.
8576 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8578 * cfgexpand.c (expand_debug_expr): Likewise.
8579 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8580 (store_constructor, expand_expr_real_1): Likewise.
8581 (const_scalar_mask_from_tree): Likewise.
8582 * fold-const-call.c (fold_const_reduction): Likewise.
8583 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8584 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8585 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8586 (fold_relational_const): Likewise.
8587 (native_interpret_vector): Likewise. Change the size from an
8588 int to an unsigned int.
8589 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8590 TYPE_VECTOR_SUBPARTS.
8591 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8592 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8593 duplicating a non-constant operand into a variable-length vector.
8594 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8595 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8596 * ipa-icf.c (sem_variable::equals): Likewise.
8597 * match.pd: Likewise.
8598 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8599 * print-tree.c (print_node): Likewise.
8600 * stor-layout.c (layout_type): Likewise.
8601 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8602 * tree-cfg.c (verify_gimple_comparison): Likewise.
8603 (verify_gimple_assign_binary): Likewise.
8604 (verify_gimple_assign_ternary): Likewise.
8605 (verify_gimple_assign_single): Likewise.
8606 * tree-pretty-print.c (dump_generic_node): Likewise.
8607 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8608 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8609 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8610 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8611 (vect_shift_permute_load_chain): Likewise.
8612 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8613 (expand_vector_condition, optimize_vector_constructor): Likewise.
8614 (lower_vec_perm, get_compute_type): Likewise.
8615 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8616 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8617 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8618 (vect_recog_mask_conversion_pattern): Likewise.
8619 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8620 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8621 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8622 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8623 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8624 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8625 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8626 (supportable_widening_operation): Likewise.
8627 (supportable_narrowing_operation): Likewise.
8628 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8630 * varasm.c (output_constant): Likewise.
8632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8633 Alan Hayward <alan.hayward@arm.com>
8634 David Sherwood <david.sherwood@arm.com>
8636 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8637 so that both the length == 3 and length != 3 cases set up their
8638 own permute vectors. Add comments explaining why we know the
8639 number of elements is constant.
8640 (vect_permute_load_chain): Likewise.
8642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8643 Alan Hayward <alan.hayward@arm.com>
8644 David Sherwood <david.sherwood@arm.com>
8646 * machmode.h (mode_nunits): Change from unsigned char to
8648 (ONLY_FIXED_SIZE_MODES): New macro.
8649 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8650 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8651 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8653 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8654 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8655 or if measurement_type is not polynomial.
8656 * genmodes.c (ZERO_COEFFS): New macro.
8657 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8659 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8660 Use ZERO_COEFFS when emitting initializers.
8661 * data-streamer.h (bp_pack_poly_value): New function.
8662 (bp_unpack_poly_value): Likewise.
8663 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8664 for GET_MODE_NUNITS.
8665 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8666 for GET_MODE_NUNITS.
8667 * tree.c (make_vector_type): Remove temporary shim and make
8668 the real function take the number of units as a poly_uint64
8670 (build_vector_type_for_mode): Handle polynomial nunits.
8671 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8672 * emit-rtl.c (const_vec_series_p_1): Likewise.
8673 (gen_rtx_CONST_VECTOR): Likewise.
8674 * fold-const.c (test_vec_duplicate_folding): Likewise.
8675 * genrecog.c (validate_pattern): Likewise.
8676 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8677 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8678 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8679 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8680 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8681 * rtlanal.c (subreg_get_info): Likewise.
8682 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8683 (vect_grouped_load_supported): Likewise.
8684 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8685 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8686 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8687 (simplify_const_unary_operation, simplify_binary_operation_1)
8688 (simplify_const_binary_operation, simplify_ternary_operation)
8689 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8690 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8691 instead of CONST_VECTOR_NUNITS.
8692 * varasm.c (output_constant_pool_2): Likewise.
8693 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8694 explicit-encoded elements in the XVEC for variable-length vectors.
8696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8698 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8701 Alan Hayward <alan.hayward@arm.com>
8702 David Sherwood <david.sherwood@arm.com>
8704 * coretypes.h (fixed_size_mode): Declare.
8705 (fixed_size_mode_pod): New typedef.
8706 * builtins.h (target_builtins::x_apply_args_mode)
8707 (target_builtins::x_apply_result_mode): Change type to
8708 fixed_size_mode_pod.
8709 * builtins.c (apply_args_size, apply_result_size, result_vector)
8710 (expand_builtin_apply_args_1, expand_builtin_apply)
8711 (expand_builtin_return): Update accordingly.
8713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8715 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8716 * cselib.c (cselib_hash_rtx): Likewise.
8717 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8718 CONST_VECTOR encoding.
8720 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8721 Jeff Law <law@redhat.com>
8724 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8725 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8726 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8727 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8730 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8731 explicitly probe *sp in a noreturn function if there were any callee
8732 register saves or frame pointer is needed.
8734 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8737 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8738 BLKmode for ternary, binary or unary expressions.
8741 * var-tracking.c (delete_vta_debug_insn): New inline function.
8742 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8743 insns from get_insns () to NULL instead of each bb separately.
8744 Use delete_vta_debug_insn. No longer static.
8745 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8746 delete_vta_debug_insns callers.
8747 * rtl.h (delete_vta_debug_insns): Declare.
8748 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8749 instead of variable_tracking_main.
8751 2018-01-03 Martin Sebor <msebor@redhat.com>
8753 PR tree-optimization/83603
8754 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8755 arguments past the endof the argument list in functions declared
8756 without a prototype.
8757 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8758 Avoid checking when arguments are null.
8760 2018-01-03 Martin Sebor <msebor@redhat.com>
8763 * doc/extend.texi (attribute const): Fix a typo.
8764 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8765 issuing -Wsuggest-attribute for void functions.
8767 2018-01-03 Martin Sebor <msebor@redhat.com>
8769 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8770 offset_int::from instead of wide_int::to_shwi.
8771 (maybe_diag_overlap): Remove assertion.
8772 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8773 * gimple-ssa-sprintf.c (format_directive): Same.
8774 (parse_directive): Same.
8775 (sprintf_dom_walker::compute_format_length): Same.
8776 (try_substitute_return_value): Same.
8778 2018-01-03 Jeff Law <law@redhat.com>
8781 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8782 non-constant residual for zero at runtime and avoid probing in
8783 that case. Reorganize code for trailing problem to mirror handling
8786 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8788 PR tree-optimization/83501
8789 * tree-ssa-strlen.c (get_string_cst): New.
8790 (handle_char_store): Call get_string_cst.
8792 2018-01-03 Martin Liska <mliska@suse.cz>
8794 PR tree-optimization/83593
8795 * tree-ssa-strlen.c: Include tree-cfg.h.
8796 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8797 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8798 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8800 (strlen_dom_walker::before_dom_children): Call
8801 gimple_purge_dead_eh_edges. Dump tranformation with details
8803 (strlen_dom_walker::before_dom_children): Update call by adding
8804 new argument cleanup_eh.
8805 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8807 2018-01-03 Martin Liska <mliska@suse.cz>
8810 * cif-code.def (VARIADIC_THUNK): New enum value.
8811 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8814 2018-01-03 Jan Beulich <jbeulich@suse.com>
8816 * sse.md (mov<mode>_internal): Tighten condition for when to use
8817 vmovdqu<ssescalarsize> for TI and OI modes.
8819 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8821 Update copyright years.
8823 2018-01-03 Martin Liska <mliska@suse.cz>
8826 * ipa-visibility.c (function_and_variable_visibility): Skip
8827 functions with noipa attribure.
8829 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8831 * gcc.c (process_command): Update copyright notice dates.
8832 * gcov-dump.c (print_version): Ditto.
8833 * gcov.c (print_version): Ditto.
8834 * gcov-tool.c (print_version): Ditto.
8835 * gengtype.c (create_file): Ditto.
8836 * doc/cpp.texi: Bump @copying's copyright year.
8837 * doc/cppinternals.texi: Ditto.
8838 * doc/gcc.texi: Ditto.
8839 * doc/gccint.texi: Ditto.
8840 * doc/gcov.texi: Ditto.
8841 * doc/install.texi: Ditto.
8842 * doc/invoke.texi: Ditto.
8844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8846 * vector-builder.h (vector_builder::m_full_nelts): Change from
8847 unsigned int to poly_uint64.
8848 (vector_builder::full_nelts): Update prototype accordingly.
8849 (vector_builder::new_vector): Likewise.
8850 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8851 (vector_builder::operator ==): Likewise.
8852 (vector_builder::finalize): Likewise.
8853 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8854 Take the number of elements as a poly_uint64 rather than an
8856 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8857 from unsigned int to poly_uint64.
8858 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8859 (vec_perm_indices::new_vector): Likewise.
8860 (vec_perm_indices::length): Likewise.
8861 (vec_perm_indices::nelts_per_input): Likewise.
8862 (vec_perm_indices::input_nelts): Likewise.
8863 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8864 number of elements per input as a poly_uint64 rather than an
8865 unsigned int. Use the original encoding for variable-length
8866 vectors, rather than clamping each individual element.
8867 For the second and subsequent elements in each pattern,
8868 clamp the step and base before clamping their sum.
8869 (vec_perm_indices::series_p): Handle polynomial element counts.
8870 (vec_perm_indices::all_in_range_p): Likewise.
8871 (vec_perm_indices_to_tree): Likewise.
8872 (vec_perm_indices_to_rtx): Likewise.
8873 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8874 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8875 (tree_vector_builder::new_binary_operation): Handle polynomial
8876 element counts. Return false if we need to know the number
8877 of elements at compile time.
8878 * fold-const.c (fold_vec_perm): Punt if the number of elements
8879 isn't known at compile time.
8881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8883 * vec-perm-indices.h (vec_perm_builder): Change element type
8884 from HOST_WIDE_INT to poly_int64.
8885 (vec_perm_indices::element_type): Update accordingly.
8886 (vec_perm_indices::clamp): Handle polynomial element_types.
8887 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8888 (vec_perm_indices::all_in_range_p): Likewise.
8889 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8891 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8892 polynomial vec_perm_indices element types.
8893 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8894 * fold-const.c (fold_vec_perm): Likewise.
8895 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8896 * tree-vect-generic.c (lower_vec_perm): Likewise.
8897 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8898 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8899 element type to HOST_WIDE_INT.
8901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8902 Alan Hayward <alan.hayward@arm.com>
8903 David Sherwood <david.sherwood@arm.com>
8905 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8906 rather than an int. Use plus_constant.
8907 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8908 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8911 Alan Hayward <alan.hayward@arm.com>
8912 David Sherwood <david.sherwood@arm.com>
8914 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8915 a HOST_WIDE_INT to a poly_int64.
8917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8918 Alan Hayward <alan.hayward@arm.com>
8919 David Sherwood <david.sherwood@arm.com>
8921 * calls.c (load_register_parameters): Cope with polynomial
8922 mode sizes. Require a constant size for BLKmode parameters
8923 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8924 forces a parameter to be padded at the lsb end in order to
8925 fill a complete number of words, require the parameter size
8926 to be ordered wrt UNITS_PER_WORD.
8928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8929 Alan Hayward <alan.hayward@arm.com>
8930 David Sherwood <david.sherwood@arm.com>
8932 * reload1.c (spill_stack_slot_width): Change element type
8933 from unsigned int to poly_uint64_pod.
8934 (alter_reg): Treat mode sizes as polynomial.
8936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8937 Alan Hayward <alan.hayward@arm.com>
8938 David Sherwood <david.sherwood@arm.com>
8940 * reload.c (complex_word_subreg_p): New function.
8941 (reload_inner_reg_of_subreg, push_reload): Use it.
8943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8944 Alan Hayward <alan.hayward@arm.com>
8945 David Sherwood <david.sherwood@arm.com>
8947 * lra-constraints.c (process_alt_operands): Reject matched
8948 operands whose sizes aren't ordered.
8949 (match_reload): Refer to this check here.
8951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8952 Alan Hayward <alan.hayward@arm.com>
8953 David Sherwood <david.sherwood@arm.com>
8955 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8956 that the mode size is in the set {1, 2, 4, 8, 16}.
8958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8959 Alan Hayward <alan.hayward@arm.com>
8960 David Sherwood <david.sherwood@arm.com>
8962 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8963 Use plus_constant instead of gen_rtx_PLUS.
8965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8966 Alan Hayward <alan.hayward@arm.com>
8967 David Sherwood <david.sherwood@arm.com>
8969 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8970 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8971 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8972 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8973 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8974 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8975 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8976 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8977 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8978 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8980 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8981 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8982 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8983 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8984 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8985 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8986 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8987 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8988 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8989 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8991 * expr.c (emit_move_resolve_push): Treat the input and result
8992 of PUSH_ROUNDING as a poly_int64.
8993 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8994 (emit_push_insn): Likewise.
8995 * lra-eliminations.c (mark_not_eliminable): Likewise.
8996 * recog.c (push_operand): Likewise.
8997 * reload1.c (elimination_effects): Likewise.
8998 * rtlanal.c (nonzero_bits1): Likewise.
8999 * calls.c (store_one_arg): Likewise. Require the padding to be
9000 known at compile time.
9002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9003 Alan Hayward <alan.hayward@arm.com>
9004 David Sherwood <david.sherwood@arm.com>
9006 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9007 Use plus_constant instead of gen_rtx_PLUS.
9009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9017 Alan Hayward <alan.hayward@arm.com>
9018 David Sherwood <david.sherwood@arm.com>
9020 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9021 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9022 via stack temporaries. Treat the mode size as polynomial too.
9024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9025 Alan Hayward <alan.hayward@arm.com>
9026 David Sherwood <david.sherwood@arm.com>
9028 * expr.c (expand_expr_real_2): When handling conversions involving
9029 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9030 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9031 as a poly_uint64 too.
9033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9034 Alan Hayward <alan.hayward@arm.com>
9035 David Sherwood <david.sherwood@arm.com>
9037 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
9044 REGMODE_NATURAL_SIZE.
9045 * expmed.c (store_bit_field_1): Likewise.
9046 * expr.c (store_constructor): Likewise.
9047 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9048 and polynomial REGMODE_NATURAL_SIZE.
9049 (gen_lowpart_common): Likewise.
9050 * reginfo.c (record_subregs_of_mode): Likewise.
9051 * rtlanal.c (read_modify_subreg_p): Likewise.
9053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9054 Alan Hayward <alan.hayward@arm.com>
9055 David Sherwood <david.sherwood@arm.com>
9057 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9058 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
9066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9067 Alan Hayward <alan.hayward@arm.com>
9068 David Sherwood <david.sherwood@arm.com>
9070 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9071 in a POINTER_PLUS_EXPR.
9073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9074 Alan Hayward <alan.hayward@arm.com>
9075 David Sherwood <david.sherwood@arm.com>
9077 * omp-simd-clone.c (simd_clone_subparts): New function.
9078 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9079 (ipa_simd_modify_function_body): Likewise.
9081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9082 Alan Hayward <alan.hayward@arm.com>
9083 David Sherwood <david.sherwood@arm.com>
9085 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9086 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9087 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9088 (expand_vector_condition, vector_element): Likewise.
9089 (subparts_gt): New function.
9090 (get_compute_type): Use subparts_gt.
9091 (count_type_subparts): Delete.
9092 (expand_vector_operations_1): Use subparts_gt instead of
9093 count_type_subparts.
9095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9096 Alan Hayward <alan.hayward@arm.com>
9097 David Sherwood <david.sherwood@arm.com>
9099 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9100 (vect_compile_time_alias): ...this new function. Do the calculation
9101 on poly_ints rather than trees.
9102 (vect_prune_runtime_alias_test_list): Update call accordingly.
9104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9105 Alan Hayward <alan.hayward@arm.com>
9106 David Sherwood <david.sherwood@arm.com>
9108 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9110 (vect_schedule_slp_instance): Likewise.
9112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9113 Alan Hayward <alan.hayward@arm.com>
9114 David Sherwood <david.sherwood@arm.com>
9116 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9117 constant and extern definitions for variable-length vectors.
9118 (vect_get_constant_vectors): Note that the number of units
9119 is known to be constant.
9121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9122 Alan Hayward <alan.hayward@arm.com>
9123 David Sherwood <david.sherwood@arm.com>
9125 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9126 of units as polynomial. Choose between WIDE and NARROW based
9129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9130 Alan Hayward <alan.hayward@arm.com>
9131 David Sherwood <david.sherwood@arm.com>
9133 * tree-vect-stmts.c (simd_clone_subparts): New function.
9134 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9137 Alan Hayward <alan.hayward@arm.com>
9138 David Sherwood <david.sherwood@arm.com>
9140 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9141 vectors as polynomial. Use build_index_vector for
9144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9145 Alan Hayward <alan.hayward@arm.com>
9146 David Sherwood <david.sherwood@arm.com>
9148 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9149 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9150 for variable-length vectors.
9151 (vectorizable_mask_load_store): Treat the number of units as
9152 polynomial, asserting that it is constant if the condition has
9153 already been enforced.
9154 (vectorizable_store, vectorizable_load): Likewise.
9156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9157 Alan Hayward <alan.hayward@arm.com>
9158 David Sherwood <david.sherwood@arm.com>
9160 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9161 of units as polynomial. Punt if we can't tell at compile time
9162 which vector contains the final result.
9164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9165 Alan Hayward <alan.hayward@arm.com>
9166 David Sherwood <david.sherwood@arm.com>
9168 * tree-vect-loop.c (vectorizable_induction): Treat the number
9169 of units as polynomial. Punt on SLP inductions. Use an integer
9170 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9171 cast of such a series for variable-length floating-point
9174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9175 Alan Hayward <alan.hayward@arm.com>
9176 David Sherwood <david.sherwood@arm.com>
9178 * tree.h (build_index_vector): Declare.
9179 * tree.c (build_index_vector): New function.
9180 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9181 of units as polynomial, forcibly converting it to a constant if
9182 vectorizable_reduction has already enforced the condition.
9183 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9184 to create a {1,2,3,...} vector.
9185 (vectorizable_reduction): Treat the number of units as polynomial.
9186 Choose vectype_in based on the largest scalar element size rather
9187 than the smallest number of units. Enforce the restrictions
9190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9191 Alan Hayward <alan.hayward@arm.com>
9192 David Sherwood <david.sherwood@arm.com>
9194 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9195 number of units as polynomial.
9197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9198 Alan Hayward <alan.hayward@arm.com>
9199 David Sherwood <david.sherwood@arm.com>
9201 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9202 * target.def (autovectorize_vector_sizes): Return the vector sizes
9203 by pointer, using vector_sizes rather than a bitmask.
9204 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9205 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9206 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9208 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9209 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9210 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9211 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9212 * omp-general.c (omp_max_vf): Likewise.
9213 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9214 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9215 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9216 * tree-vect-slp.c (vect_slp_bb): Likewise.
9217 * doc/tm.texi: Regenerate.
9218 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9220 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9221 the vector size as a poly_uint64 rather than an unsigned int.
9222 (current_vector_size): Change from an unsigned int to a poly_uint64.
9223 (get_vectype_for_scalar_type): Update accordingly.
9224 * tree.h (build_truth_vector_type): Take the size and number of
9225 units as a poly_uint64 rather than an unsigned int.
9226 (build_vector_type): Add a temporary overload that takes
9227 the number of units as a poly_uint64 rather than an unsigned int.
9228 * tree.c (make_vector_type): Likewise.
9229 (build_truth_vector_type): Take the number of units as a poly_uint64
9230 rather than an unsigned int.
9232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9233 Alan Hayward <alan.hayward@arm.com>
9234 David Sherwood <david.sherwood@arm.com>
9236 * target.def (get_mask_mode): Take the number of units and length
9237 as poly_uint64s rather than unsigned ints.
9238 * targhooks.h (default_get_mask_mode): Update accordingly.
9239 * targhooks.c (default_get_mask_mode): Likewise.
9240 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9241 * doc/tm.texi: Regenerate.
9243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9244 Alan Hayward <alan.hayward@arm.com>
9245 David Sherwood <david.sherwood@arm.com>
9247 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9248 * omp-general.c (omp_max_vf): Likewise.
9249 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9250 (expand_omp_simd): Handle polynomial safelen.
9251 * omp-low.c (omplow_simd_context): Add a default constructor.
9252 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9253 (lower_rec_simd_input_clauses): Update accordingly.
9254 (lower_rec_input_clauses): Likewise.
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 (vect_nunits_for_cost): New function.
9261 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9262 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9263 (vect_analyze_slp_cost): Likewise.
9264 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9265 (vect_model_load_cost): Likewise.
9267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9268 Alan Hayward <alan.hayward@arm.com>
9269 David Sherwood <david.sherwood@arm.com>
9271 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9272 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9273 from an unsigned int * to a poly_uint64_pod *.
9274 (calculate_unrolling_factor): New function.
9275 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9278 Alan Hayward <alan.hayward@arm.com>
9279 David Sherwood <david.sherwood@arm.com>
9281 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9282 from an unsigned int to a poly_uint64.
9283 (_loop_vec_info::slp_unrolling_factor): Likewise.
9284 (_loop_vec_info::vectorization_factor): Change from an int
9286 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9287 (vect_get_num_vectors): New function.
9288 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9289 (vect_get_num_copies): Use vect_get_num_vectors.
9290 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9291 to an unsigned int *.
9292 (vect_analyze_data_refs): Change min_vf from an int * to a
9294 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9295 than an unsigned HOST_WIDE_INT.
9296 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9297 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9298 to an unsigned int *.
9299 (vect_analyze_data_ref_dependences): Likewise.
9300 (vect_compute_data_ref_alignment): Handle polynomial vf.
9301 (vect_enhance_data_refs_alignment): Likewise.
9302 (vect_prune_runtime_alias_test_list): Likewise.
9303 (vect_shift_permute_load_chain): Likewise.
9304 (vect_supportable_dr_alignment): Likewise.
9305 (dependence_distance_ge_vf): Take the vectorization factor as a
9306 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9307 (vect_analyze_data_refs): Change min_vf from an int * to a
9309 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9310 vfm1 as a poly_uint64 rather than an int. Make the same change
9311 for the returned bound_scalar.
9312 (vect_gen_vector_loop_niters): Handle polynomial vf.
9313 (vect_do_peeling): Likewise. Update call to
9314 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9315 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9317 * tree-vect-loop.c (vect_determine_vectorization_factor)
9318 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9319 (vect_get_known_peeling_cost): Likewise.
9320 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9321 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9322 (vect_transform_loop): Likewise. Use the lowest possible VF when
9323 updating the upper bounds of the loop.
9324 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9326 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9327 polynomial unroll factors.
9328 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9329 (vect_make_slp_decision): Likewise.
9330 (vect_supported_load_permutation_p): Likewise, and polynomial
9332 (vect_analyze_slp_cost): Handle polynomial vf.
9333 (vect_slp_analyze_node_operations): Likewise.
9334 (vect_slp_analyze_bb_1): Likewise.
9335 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9336 than an unsigned HOST_WIDE_INT.
9337 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9338 (vectorizable_load): Handle polynomial vf.
9339 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9341 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9344 Alan Hayward <alan.hayward@arm.com>
9345 David Sherwood <david.sherwood@arm.com>
9347 * match.pd: Handle bit operations involving three constants
9348 and try to fold one pair.
9350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9352 * tree-vect-loop-manip.c: Include gimple-fold.h.
9353 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9354 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9355 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9356 Add a path that uses a step of VF instead of 1, but disable it
9358 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9359 and niters_no_overflow parameters. Update calls to
9360 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9361 Create a new SSA name if the latter choses to use a ste other
9362 than zero, and return it via niters_vector_mult_vf_var.
9363 * tree-vect-loop.c (vect_transform_loop): Update calls to
9364 vect_do_peeling, vect_gen_vector_loop_niters and
9365 slpeel_make_loop_iterate_ntimes.
9366 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9367 (vect_gen_vector_loop_niters): Update declarations after above changes.
9369 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9371 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9372 128-bit round to integer instructions.
9373 (ceil<mode>2): Likewise.
9374 (btrunc<mode>2): Likewise.
9375 (round<mode>2): Likewise.
9377 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9379 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9380 unaligned VSX load/store on P8/P9.
9381 (expand_block_clear): Allow the use of unaligned VSX
9382 load/store on P8/P9.
9384 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9386 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9388 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9389 swap associated with both a load and a store.
9391 2018-01-02 Andrew Waterman <andrew@sifive.com>
9393 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9394 * config/riscv/riscv.md (clear_cache): Use it.
9396 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9398 * web.c: Remove out-of-date comment.
9400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9402 * expr.c (fixup_args_size_notes): Check that any existing
9403 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9404 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9405 (emit_single_push_insn): ...here.
9407 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9409 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9410 (const_vector_encoded_nelts): New function.
9411 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9412 (const_vector_int_elt, const_vector_elt): Declare.
9413 * emit-rtl.c (const_vector_int_elt_1): New function.
9414 (const_vector_elt): Likewise.
9415 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9416 of CONST_VECTOR_ELT.
9418 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9420 * expr.c: Include rtx-vector-builder.h.
9421 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9422 directly on the tree encoding.
9423 (const_vector_from_tree): Likewise.
9424 * optabs.c: Include rtx-vector-builder.h.
9425 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9426 sequence of "u" values.
9427 * vec-perm-indices.c: Include rtx-vector-builder.h.
9428 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9429 directly on the vec_perm_indices encoding.
9431 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9433 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9434 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9435 * rtx-vector-builder.h: New file.
9436 * rtx-vector-builder.c: Likewise.
9437 * rtl.h (rtx_def::u2): Add a const_vector field.
9438 (CONST_VECTOR_NPATTERNS): New macro.
9439 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9440 (CONST_VECTOR_DUPLICATE_P): Likewise.
9441 (CONST_VECTOR_STEPPED_P): Likewise.
9442 (CONST_VECTOR_ENCODED_ELT): Likewise.
9443 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9444 (unwrap_const_vec_duplicate): Likewise.
9445 (const_vec_series_p): Check for a non-duplicated vector encoding.
9446 Say that the function only returns true for integer vectors.
9447 * emit-rtl.c: Include rtx-vector-builder.h.
9448 (gen_const_vec_duplicate_1): Delete.
9449 (gen_const_vector): Call gen_const_vec_duplicate instead of
9450 gen_const_vec_duplicate_1.
9451 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9452 (gen_const_vec_duplicate): Use rtx_vector_builder.
9453 (gen_const_vec_series): Likewise.
9454 (gen_rtx_CONST_VECTOR): Likewise.
9455 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9456 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9457 Build a new vector rather than modifying a CONST_VECTOR in-place.
9458 (handle_special_swappables): Update call accordingly.
9459 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9460 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9461 Build a new vector rather than modifying a CONST_VECTOR in-place.
9462 (handle_special_swappables): Update call accordingly.
9464 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9466 * simplify-rtx.c (simplify_const_binary_operation): Use
9467 CONST_VECTOR_ELT instead of XVECEXP.
9469 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9471 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9472 the selector elements to be different from the data elements
9473 if the selector is a VECTOR_CST.
9474 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9475 ssizetype for the selector.
9477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9479 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9480 before testing each element individually.
9481 * tree-vect-generic.c (lower_vec_perm): Likewise.
9483 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9485 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9486 * selftest-run-tests.c (selftest::run_tests): Call it.
9487 * vector-builder.h (vector_builder::operator ==): New function.
9488 (vector_builder::operator !=): Likewise.
9489 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9490 (vec_perm_indices::all_from_input_p): New function.
9491 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9492 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9493 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9494 instead of reading the VECTOR_CST directly. Detect whether both
9495 vector inputs are the same before constructing the vec_perm_indices,
9496 and update the number of inputs argument accordingly. Use the
9497 utility functions added above. Only construct sel2 if we need to.
9499 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9501 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9502 the broadcast of the low byte.
9503 (expand_mult_highpart): Use an explicit encoding for the permutes.
9504 * optabs-query.c (can_mult_highpart_p): Likewise.
9505 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9506 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9507 (vectorizable_bswap): Likewise.
9508 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9509 explicit encoding for the power-of-2 permutes.
9510 (vect_permute_store_chain): Likewise.
9511 (vect_grouped_load_supported): Likewise.
9512 (vect_permute_load_chain): Likewise.
9514 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9516 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9517 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9518 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9519 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9520 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9521 (vect_gen_perm_mask_any): Likewise.
9523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9525 * int-vector-builder.h: New file.
9526 * vec-perm-indices.h: Include int-vector-builder.h.
9527 (vec_perm_indices): Redefine as an int_vector_builder.
9528 (auto_vec_perm_indices): Delete.
9529 (vec_perm_builder): Redefine as a stand-alone class.
9530 (vec_perm_indices::vec_perm_indices): New function.
9531 (vec_perm_indices::clamp): Likewise.
9532 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9533 (vec_perm_indices::new_vector): New function.
9534 (vec_perm_indices::new_expanded_vector): Update for new
9535 vec_perm_indices class.
9536 (vec_perm_indices::rotate_inputs): New function.
9537 (vec_perm_indices::all_in_range_p): Operate directly on the
9538 encoded form, without computing elided elements.
9539 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9540 encoding. Update for new vec_perm_indices class.
9541 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9542 the given vec_perm_builder.
9543 (expand_vec_perm_var): Update vec_perm_builder constructor.
9544 (expand_mult_highpart): Use vec_perm_builder instead of
9545 auto_vec_perm_indices.
9546 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9547 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9548 or double series encoding as appropriate.
9549 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9550 vec_perm_indices instead of auto_vec_perm_indices.
9551 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9552 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9553 (vect_permute_store_chain): Likewise.
9554 (vect_grouped_load_supported): Likewise.
9555 (vect_permute_load_chain): Likewise.
9556 (vect_shift_permute_load_chain): Likewise.
9557 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9558 (vect_transform_slp_perm_load): Likewise.
9559 (vect_schedule_slp_instance): Likewise.
9560 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9561 (vectorizable_mask_load_store): Likewise.
9562 (vectorizable_bswap): Likewise.
9563 (vectorizable_store): Likewise.
9564 (vectorizable_load): Likewise.
9565 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9566 vec_perm_indices instead of auto_vec_perm_indices. Use
9567 tree_to_vec_perm_builder to read the vector from a tree.
9568 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9569 vec_perm_builder instead of a vec_perm_indices.
9570 (have_whole_vector_shift): Use vec_perm_builder and
9571 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9572 truncation to calc_vec_perm_mask_for_shift.
9573 (vect_create_epilog_for_reduction): Likewise.
9574 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9575 from auto_vec_perm_indices to vec_perm_indices.
9576 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9577 instead of changing individual elements.
9578 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9579 the vector in d.perm.
9580 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9581 from auto_vec_perm_indices to vec_perm_indices.
9582 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9583 instead of changing individual elements.
9584 (arm_vectorize_vec_perm_const): Use new_vector to install
9585 the vector in d.perm.
9586 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9587 Update vec_perm_builder constructor.
9588 (rs6000_expand_interleave): Likewise.
9589 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9590 (rs6000_expand_interleave): Likewise.
9592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9594 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9595 to qimode could truncate the indices.
9596 * optabs.c (expand_vec_perm_var): Likewise.
9598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9600 * Makefile.in (OBJS): Add vec-perm-indices.o.
9601 * vec-perm-indices.h: New file.
9602 * vec-perm-indices.c: Likewise.
9603 * target.h (vec_perm_indices): Replace with a forward class
9605 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9606 * optabs.h: Include vec-perm-indices.h.
9607 (expand_vec_perm): Delete.
9608 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9609 (expand_vec_perm_const): Declare.
9610 * target.def (vec_perm_const_ok): Replace with...
9611 (vec_perm_const): ...this new hook.
9612 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9613 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9614 * doc/tm.texi: Regenerate.
9615 * optabs.def (vec_perm_const): Delete.
9616 * doc/md.texi (vec_perm_const): Likewise.
9617 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9618 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9619 expand_vec_perm for constant permutation vectors. Assert that
9620 the mode of variable permutation vectors is the integer equivalent
9621 of the mode that is being permuted.
9622 * optabs-query.h (selector_fits_mode_p): Declare.
9623 * optabs-query.c: Include vec-perm-indices.h.
9624 (selector_fits_mode_p): New function.
9625 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9626 is defined, instead of checking whether the vec_perm_const_optab
9627 exists. Use targetm.vectorize.vec_perm_const instead of
9628 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9629 fit in the vector mode before using a variable permute.
9630 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9631 vec_perm_indices instead of an rtx.
9632 (expand_vec_perm): Replace with...
9633 (expand_vec_perm_const): ...this new function. Take the selector
9634 as a vec_perm_indices rather than an rtx. Also take the mode of
9635 the selector. Update call to shift_amt_for_vec_perm_mask.
9636 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9637 Use vec_perm_indices::new_expanded_vector to expand the original
9638 selector into bytes. Check whether the indices fit in the vector
9639 mode before using a variable permute.
9640 (expand_vec_perm_var): Make global.
9641 (expand_mult_highpart): Use expand_vec_perm_const.
9642 * fold-const.c: Includes vec-perm-indices.h.
9643 * tree-ssa-forwprop.c: Likewise.
9644 * tree-vect-data-refs.c: Likewise.
9645 * tree-vect-generic.c: Likewise.
9646 * tree-vect-loop.c: Likewise.
9647 * tree-vect-slp.c: Likewise.
9648 * tree-vect-stmts.c: Likewise.
9649 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9651 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9652 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9653 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9654 (aarch64_vectorize_vec_perm_const): ...this new function.
9655 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9656 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9657 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9658 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9659 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9660 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9661 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9663 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9664 check for NEON modes.
9665 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9666 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9667 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9668 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9670 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9671 the old VEC_PERM_CONST conditions.
9672 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9673 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9674 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9675 (ia64_vectorize_vec_perm_const_ok): Merge into...
9676 (ia64_vectorize_vec_perm_const): ...this new function.
9677 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9678 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9679 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9680 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9681 * config/mips/mips.c (mips_expand_vec_perm_const)
9682 (mips_vectorize_vec_perm_const_ok): Merge into...
9683 (mips_vectorize_vec_perm_const): ...this new function.
9684 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9685 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9686 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9687 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9688 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9689 (rs6000_expand_vec_perm_const): Delete.
9690 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9692 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9693 (altivec_expand_vec_perm_const_le): Take each operand individually.
9694 Operate on constant selectors rather than rtxes.
9695 (altivec_expand_vec_perm_const): Likewise. Update call to
9696 altivec_expand_vec_perm_const_le.
9697 (rs6000_expand_vec_perm_const): Delete.
9698 (rs6000_vectorize_vec_perm_const_ok): Delete.
9699 (rs6000_vectorize_vec_perm_const): New function.
9700 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9701 an element count and rtx array.
9702 (rs6000_expand_extract_even): Update call accordingly.
9703 (rs6000_expand_interleave): Likewise.
9704 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9705 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9706 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9707 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9708 (rs6000_expand_vec_perm_const): Delete.
9709 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9710 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9711 (altivec_expand_vec_perm_const_le): Take each operand individually.
9712 Operate on constant selectors rather than rtxes.
9713 (altivec_expand_vec_perm_const): Likewise. Update call to
9714 altivec_expand_vec_perm_const_le.
9715 (rs6000_expand_vec_perm_const): Delete.
9716 (rs6000_vectorize_vec_perm_const_ok): Delete.
9717 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9718 reference to the SPE evmerge intructions.
9719 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9720 an element count and rtx array.
9721 (rs6000_expand_extract_even): Update call accordingly.
9722 (rs6000_expand_interleave): Likewise.
9723 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9724 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9726 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9728 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9730 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9731 vector mode and that that mode matches the mode of the data
9733 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9734 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9735 directly using expand_vec_perm_1 when forcing selectors into
9737 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9739 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9741 * optabs-query.h (can_vec_perm_p): Delete.
9742 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9743 * optabs-query.c (can_vec_perm_p): Split into...
9744 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9745 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9746 particular selector is valid.
9747 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9748 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9749 (vect_grouped_load_supported): Likewise.
9750 (vect_shift_permute_load_chain): Likewise.
9751 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9752 (vect_transform_slp_perm_load): Likewise.
9753 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9754 (vectorizable_bswap): Likewise.
9755 (vect_gen_perm_mask_checked): Likewise.
9756 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9757 implementations of variable permutation vectors into account
9758 when deciding which selector to use.
9759 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9760 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9761 with a false third argument.
9762 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9763 to test whether the constant selector is valid and can_vec_perm_var_p
9764 to test whether a variable selector is valid.
9766 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9768 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9769 * optabs-query.c (can_vec_perm_p): Likewise.
9770 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9771 instead of vec_perm_indices.
9772 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9773 (vect_gen_perm_mask_checked): Likewise,
9774 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9775 (vect_gen_perm_mask_checked): Likewise,
9777 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9779 * optabs-query.h (qimode_for_vec_perm): Declare.
9780 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9781 (qimode_for_vec_perm): ...this new function.
9782 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9784 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9786 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9787 does not have a conditional at the top.
9789 2018-01-02 Richard Biener <rguenther@suse.de>
9791 * ipa-inline.c (big_speedup_p): Fix expression.
9793 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9796 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9799 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9803 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9804 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9805 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9806 cond_taken_branch_cost 3->4.
9808 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9810 PR tree-optimization/83581
9811 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9812 TODO_cleanup_cfg if any changes have been made.
9815 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9816 convert_modes if target mode has the right side, but different mode
9820 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9821 last argument when extracting from CONCAT. If either from_real or
9822 from_imag is NULL, use expansion through memory. If result is not
9823 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9824 the parts directly to inner mode, if even that fails, use expansion
9828 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9829 check for bswap in mode rather than HImode and use that in expand_unop
9832 Copyright (C) 2018 Free Software Foundation, Inc.
9834 Copying and distribu