| blob | 2b3369be703e71b7d37f6cba0fc55a33f86e4bb1 |
1 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2 Kito Cheng <kito.cheng@gmail.com>
4 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8 * config/nds32/nds32.md (sibcall_internal): New.
9 (sibcall_register): Remove.
10 (sibcall_immediate): Remove.
11 (sibcall_value_internal): New.
12 (sibcall_value_register): Remove.
13 (sibcall_value_immediate): Remove.
14 * config/nds32/predicates.md (nds32_general_register_operand): New.
15 (nds32_call_address_operand): New.
17 2018-04-03 Jakub Jelinek <jakub@redhat.com>
19 PR rtl-optimization/85167
20 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
21 bb_defs if *split_p, instead preinitialize it to NULL.
23 PR tree-optimization/85156
24 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
25 evaluating the argument multiple times.
27 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
29 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
30 than vector.
31 (_mm_cvtpd_ps): Likewise.
32 (_mm_cvttpd_epi32): Likewise.
33 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
34 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
35 vector, pixel, and bool following altivec.h include.
37 2018-04-03 Martin Sebor <msebor@redhat.com>
39 * doc/extend.texi (Common Function Attributes): Clarify.
40 (const attribute): Likewise.
41 (pure attribute): Likewise.
43 2018-04-03 Jakub Jelinek <jakub@redhat.com>
45 PR target/85169
46 * config/i386/i386.c (ix86_expand_vector_set): Use
47 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
49 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
51 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
52 instructions when changing rounding bits to preserve precision bits
53 in the x87 control word.
55 2018-04-03 Martin Liska <mliska@suse.cz>
57 PR tree-optimization/82491
58 * rtl.h (strip_offset_and_add): Replace += suboffset with
59 poly_uint64 () + suboffset.
61 2018-03-29 Martin Liska <mliska@suse.cz>
62 Martin Jambor <mjambor@suse.cz>
64 PR ipa/84947
65 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
66 param_type is not an integral or pointer type.
68 2018-04-03 Richard Biener <rguenther@suse.de>
70 * sese.h (recompute_all_dominators): Remove.
72 2018-04-02 Martin Sebor <msebor@redhat.com>
74 * doc/invoke.texi (-Wrestrict): Fix typos.
76 2018-04-02 Jim Wilson <jimw@sifive.com>
78 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
79 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
80 (<optab>di3, <optab>si3_extend): Likewise.
81 (<optab>si3_mask, <optab>si3_mask_1): New.
82 (<optab>di3_mask, <optab>di3_mask_1): New.
83 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
84 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
85 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
87 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
89 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
90 example.
92 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
94 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
95 (nds32_canonicalize_comparison): New function.
97 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
98 Kito Cheng <kito.cheng@gmail.com>
99 Kuan-Lin Chen <kuanlinchentw@gmail.com>
101 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
102 * config/nds32/constants.md (unspec_volatile_element): Add
103 UNSPEC_VOLATILE_RELAX_GROUP.
104 * config/nds32/nds32-relax-opt.c: New file.
105 * config/nds32/nds32-predicates.c
106 (nds32_symbol_load_store_p): New function.
107 * config/nds32/nds32-protos.h
108 (nds32_symbol_load_store_p): Declare function.
109 (make_pass_nds32_relax_opt): Declare new rtl pass function.
110 * config/nds32/nds32.c
111 (nds32_register_pass): New function to register pass.
112 (nds32_register_passes): New function to register passes.
113 * config/nds32/nds32.md (relax_group): New pattern.
114 * config/nds32/nds32.opt (mrelax-hint): New option.
115 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
117 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
119 * config/nds32/t-nds32: Modify files dependency.
121 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
123 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
124 (PROFILE_HOOK): Define its implementation.
126 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
128 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
129 type and 32-bit size.
131 2018-04-01 Jakub Jelinek <jakub@redhat.com>
133 PR middle-end/85090
134 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
135 (V_128_256): New mode iterator.
136 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
137 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
138 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
139 of V.
140 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
141 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
143 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
145 PR target/83315
146 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
147 NaN inputs correctly.
149 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
151 PR target/80546
152 * config/rs6000/vsx.md (??r): New mode attribute.
153 (*vsx_mov<mode>_64bit): Use it.
154 (*vsx_mov<mode>_32bit): Likewise.
156 2018-03-30 Martin Sebor <msebor@redhat.com>
158 PR tree-optimization/84818
159 * builtins.c (check_access): Use warning_n.
161 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
163 PR target/83822
164 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
165 condition.
166 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
167 condition.
169 2018-03-30 Julia Koval <julia.koval@intel.com>
171 PR target/84413
172 * x86-tune.def (movx, partial_reg_dependency): Enable for
173 m_SKYLAKE_AVX512.
175 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
177 PR inline-asm/84985
178 * lra-constraints.c (process_alt_operands): Move setting
179 this_alternative_matches below.
181 2018-03-29 Martin Liska <mliska@suse.cz>
183 PR lto/84995.
184 * doc/invoke.texi: Document how LTO works with debug info.
185 Describe auto-load support of binutils. Mention 'x86-64'
186 as valid option value of -march option.
188 2018-03-29 Jakub Jelinek <jakub@redhat.com>
190 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
192 PR c/85094
193 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
194 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
195 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
196 checking.
198 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
200 PR target/84912
201 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
202 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
203 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
204 for RS6000_BTM_POWERPC64.
205 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
206 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
207 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
208 definition.
209 (DIVDE): Use it.
210 (DIVDEU): Likewise.
212 2018-03-28 Carl Love <cel@us.ibm.com>
214 Revert
215 2017-09-27 Carl Love <cel@us.ibm.com>
217 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
218 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
219 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
220 fctiw instruction.
222 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
224 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
225 instead of __vector bool.
226 (_mm_max_pu8): Likewise.
227 (_mm_min_pi16): Likewise.
229 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
231 PR target/84912
232 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
233 (DIVWEUO): Likewise.
234 (DIVDEO): Likewise.
235 (DIVDEUO): Likewise.
236 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
237 DIVWEUO and DIVDEUO.
238 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
239 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
240 (div_extend): Likewise.
241 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
242 builtin function.
243 (__builtin_divweuo): Likewise.
244 (__builtin_divdeo): Likewise.
245 (__builtin_divdeuo): Likewise.
247 2018-03-28 Jakub Jelinek <jakub@redhat.com>
249 PR target/85095
250 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
251 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
253 PR tree-optimization/82004
254 * gimple-match-head.c (optimize_pow_to_exp): New function.
255 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
256 Don't fold to exp if optimize_pow_to_exp is false.
258 2018-03-28 Martin Liska <mliska@suse.cz>
260 PR other/84819
261 * calls.c (initialize_argument_information): Fix trailing space.
262 * common.opt: Fix typo and provide better explanation for
263 -fsanitize-coverage option.
264 * config/i386/i386.opt: Fix typo.
266 2018-03-28 Jakub Jelinek <jakub@redhat.com>
267 Martin Liska <mliska@suse.cz>
269 PR sanitizer/85081
270 * gimplify.c (asan_poison_variable): Don't do the check for
271 gimplify_omp_ctxp here.
272 (gimplify_decl_expr): Do it here.
273 (gimplify_target_expr): Likewise.
275 2018-03-28 Martin Liska <mliska@suse.cz>
277 PR target/84988
278 * config/i386/i386.c (ix86_function_arg_advance): Do not call
279 chkp_type_bounds_count if MPX is not enabled.
281 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
283 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
285 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
287 PR target/84914
288 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
289 function to create the function decl for complex long double
290 multiply and divide for -mabi=ieeelongdouble.
291 (init_float128_ieee): Call it.
293 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
295 PR target/85044
296 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
297 -fcf-protection=branch -mibt.
298 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
300 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
302 PR target/81863
303 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
305 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
307 PR target/85056
308 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
309 extern array declarations.
311 2018-03-27 Richard Biener <rguenther@suse.de>
313 PR middle-end/84067
314 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
315 explicit single_use checks.
317 2018-03-27 Richard Biener <rguenther@suse.de>
319 PR tree-optimization/85082
320 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
321 Valueize the VUSE.
323 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
325 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
326 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
327 Turn on fasynchronous-unwind-tables and funwind-tables.
329 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
331 PR target/85073
332 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
333 (*bmi_blsr_<mode>_ccz): Ditto.
335 2018-03-26 Tom de Vries <tom@codesourcery.com>
337 PR tree-optimization/85063
338 * omp-general.c (offloading_function_p): New function. Factor out
339 of ...
340 * omp-offload.c (pass_omp_target_link::gate): ... here.
341 * omp-general.h (offloading_function_p): Declare.
342 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
343 with attribute omp declare target for offloading functions.
345 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
347 PR tree-optimization/84005
348 * tree-data-ref.h (get_base_for_alignment): Declare.
349 * tree-data-ref.c (get_base_for_alignment_1): New function.
350 (get_base_for_alignment): Likewise.
351 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
352 get_base_for_alignment to find a suitable base object, instead
353 of always using drb->base_address.
355 2018-03-23 Jakub Jelinek <jakub@redhat.com>
357 PR inline-asm/85022
358 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
359 known size by default.
361 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
363 PR inline-asm/85030
364 * lra-constraints.c (process_alt_operands): Don't match BLKmode
365 and non BLKmode operands.
367 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
369 PR target/85026
370 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
371 Clean up attributes.
373 2018-03-23 Richard Biener <rguenther@suse.de>
375 PR debug/85020
376 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
377 we are going to emit early debug for LTO.
379 2018-03-23 Jakub Jelinek <jakub@redhat.com>
381 PR inline-asm/85034
382 * function.c (match_asm_constraints_1): Don't optimize if input
383 doesn't satisfy general_operand predicate for output's mode.
385 PR inline-asm/85022
386 * alias.c (write_dependence_p): Don't require for x_canonicalized
387 non-VOIDmode if x has VOIDmode.
389 PR sanitizer/85029
390 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
391 just don't try to optimize it rather than assert it never happens.
393 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
395 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
396 macro expansions for definition of ST_INTERNAL_<mode> and
397 LD_INTERNAL_<mode> builtins.
398 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
399 Remove prototype.
400 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
401 function.
402 (altivec_expand_st_builtin): Likewise.
403 (altivec_expand_builtin): Remove calls to deleted functions.
404 (rs6000_address_for_altivec): Delete this function.
405 * config/rs6000/vector.md: Remove expands for
406 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
408 2018-03-22 Sudakshina Das <sudi.das@arm.com>
410 PR target/84826
411 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
412 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
413 re-computing once computed.
414 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
415 (arm_init_machine_status): Initialize
416 machine->static_chain_stack_bytes.
418 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
420 PR target/84760
421 * doc/extend.texi: Add four new prototypes for vec_ld.
422 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
423 definitions for more logical presentation.
424 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
425 entries for V1TI variants of __builtin_altivec_ld builtin.
426 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
427 handling of V1TI variant of LVX icode pattern.
428 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
429 (rs6000_gimple_fold_builtin): Likewise.
430 (altivec_init_builtins): Add code to define
431 __builtin_altivec_lvx_v1ti function.
433 2018-03-22 Jakub Jelinek <jakub@redhat.com>
435 PR inline-asm/84941
436 * function.c (match_asm_constraints_1): Don't do the optimization
437 if input isn't a REG, SUBREG, MEM or constant.
439 2018-03-22 Tom de Vries <tom@codesourcery.com>
441 PR tree-optimization/84956
442 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
443 bb_has_abnormal_pred.
445 2018-03-22 Jakub Jelinek <jakub@redhat.com>
447 PR sanitizer/85018
448 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
449 DECL_INITIAL (decl) to decl at the end.
450 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
451 adjust the comment.
453 2018-03-21 Joseph Myers <joseph@codesourcery.com>
455 * doc/extend.texi (__builtin_tgmath): Document when complex
456 integer types are treated as _Complex _Float64.
458 2018-03-21 Tom de Vries <tom@codesourcery.com>
460 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
462 2018-03-21 Jakub Jelinek <jakub@redhat.com>
464 PR tree-optimization/84960
465 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
466 if it is ENTRY block, move them into single succ of ENTRY in that case.
468 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
470 PR tree-optimization/84811
471 * poly-int.h (poly_span_traits): Remove the T3 parameter and
472 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
473 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
474 (known_subrange_p): Update accordingly. Cast each value involved
475 in the size comparison, rather than casting the result of the
476 subtraction.
478 2018-03-21 Jakub Jelinek <jakub@redhat.com>
480 PR tree-optimization/84982
481 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
482 by flipping the least significant bit rather than all bits from
483 bitpos to bitpos + bitsize - 1.
485 2018-03-21 Nathan Sidwell <nathan@acm.org>
487 * doc/extend.texi (Deprecated Features): Remove mention of
488 long-deleted deprecations.
490 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
492 PR jit/84288
493 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
494 * configure: Regenerate.
496 2018-03-21 Tom de Vries <tom@codesourcery.com>
498 PR tree-optimization/83126
499 * tree-parloops.c (num_phis): New function.
500 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
502 2018-03-21 Nathan Sidwell <nathan@acm.org>
504 * doc/extend.texi (Deprecated Features): Update deprecated flags,
505 mention anon-struct/union members and trailing attributes.
507 2018-03-21 Bin Cheng <bin.cheng@arm.com>
509 PR tree-optimization/84969
510 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
511 builtin memset partitions if they set different rhs values.
513 2018-03-21 Jakub Jelinek <jakub@redhat.com>
515 PR rtl-optimization/84989
516 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
517 VEC_DUPLICATE with scalar result mode.
519 2018-03-21 Martin Liska <mliska@suse.cz>
521 PR ipa/84963
522 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
523 not intended return statement.
525 2018-03-21 Martin Liska <mliska@suse.cz>
527 PR target/84988
528 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
529 (chkp_find_bound_slots_1): Limit number of iterations.
531 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
533 PR target/84838
534 * Minor grammar fixes for x86 options.
536 2018-03-20 Jakub Jelinek <jakub@redhat.com>
538 PR debug/84875
539 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
540 holding REG_CFA_RESTORE notes, instead turn them into a USE.
542 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
544 PR target/83789
545 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
546 (altivec_lvx_<mode>_1op): Likewise.
547 (altivec_stvx_<mode>_2op): Likewise.
548 (altivec_stvx_<mode>_1op): Likewise.
549 (altivec_lvx_<VM2:mode>): New define_expand.
550 (altivec_stvx_<VM2:mode>): Likewise.
551 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
552 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
553 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
554 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
555 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
556 (rs6000_gen_lvx): Likewise.
557 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
558 (altivec_expand_stv_builtin): Likewise.
559 (altivec_expand_builtin): Likewise.
560 * config/rs6000/vector.md: Likewise.
562 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
564 PR target/82518
565 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
566 BYTES_BIG_ENDIAN.
568 2018-03-20 Richard Biener <rguenther@suse.de>
570 PR target/84986
571 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
572 sign-conversions as zero, fall back to standard scalar_stmt
573 cost for the rest.
575 2018-03-20 Martin Liska <mliska@suse.cz>
577 PR ipa/84825
578 * predict.c (rebuild_frequencies): Handle case when we have
579 PROFILE_ABSENT, but flag_guess_branch_prob is false.
581 2018-03-20 Jakub Jelinek <jakub@redhat.com>
583 PR target/84990
584 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
585 flag_section_anchors.
586 * varasm.c (use_blocks_for_decl_p): Remove hack for
587 dw2_force_const_mem.
589 PR target/84845
590 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
591 to ...
592 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
593 be created, use lowpart_subreg of operands[0] rather than operands[0]
594 itself.
595 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
596 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
597 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
598 and n constraint instead of aarch64_shift_imm_di and Usd.
599 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
600 (*aarch64_<optab>_reg_minus<mode>3): ... this.
602 2018-03-20 Sudakshina Das <sudi.das@arm.com>
604 PR target/82989
605 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
606 to favor GPR over NEON registers.
607 (<shift>di3_neon): Likewise.
609 2018-03-20 Tom de Vries <tom@codesourcery.com>
611 PR target/84952
612 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
613 (nvptx_process_pars): Emit bar.sync asap and alap.
615 2018-03-20 Tom de Vries <tom@codesourcery.com>
617 PR target/84954
618 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
619 seen_label if seen_label is already set.
621 2018-03-20 Jakub Jelinek <jakub@redhat.com>
623 PR target/84945
624 * config/i386/i386.c (fold_builtin_cpu): For features above 31
625 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
626 Use 1U instead of 1. Formatting fixes.
628 PR c/84953
629 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
630 instead of TREE_TYPE (s1) for the return value.
632 2018-03-19 Jakub Jelinek <jakub@redhat.com>
634 PR tree-optimization/84946
635 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
636 bitsize + bitsize in poly_uint64 rather than poly_int64.
638 PR sanitizer/78651
639 * dwarf2asm.c: Include fold-const.c.
640 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
641 of decl rather than decl itself.
643 PR rtl-optimization/84643
644 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
646 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
648 PR sanitizer/78651
649 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
650 calling assemble_variable.
652 2018-03-19 Sudakshina Das <sudi.das@arm.com>
654 PR target/81647
655 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
656 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
658 2018-03-19 Jim Wilson <jimw@sifive.com>
660 PR bootstrap/84856
661 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
662 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
663 (riscv_first_stack_step): Likewise.
664 (riscv_option_override): Use STACK_BOUNDARY instead of
665 MIN_STACK_BOUNDARY.
666 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
667 MIN_STACK_BOUNDARY.
668 (BIGGEST_ALIGNMENT): Set to 128.
669 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
670 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
671 STACK_BOUNDARY.
673 2018-03-19 Richard Biener <rguenther@suse.de>
675 PR tree-optimization/84933
676 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
677 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
679 2018-03-19 Richard Biener <rguenther@suse.de>
681 PR tree-optimization/84859
682 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
683 (cond_if_else_store_replacement): Perform sinking operation on
684 single-store BBs regardless of MAX_STORES_TO_SINK setting.
685 Generalize what a BB with a single eligible store is.
687 2018-03-19 Richard Biener <rguenther@suse.de>
689 PR tree-optimization/84929
690 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
691 chrec_is_positive against non-chrec arg.
693 2018-03-19 Tamar Christina <tamar.christina@arm.com>
695 PR target/84711
696 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
698 2018-03-18 Martin Liska <mliska@suse.cz>
700 PR rtl-optimization/84635
701 * regrename.c (build_def_use): Use matches_mode only when
702 matches >= 0.
704 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
706 PR tree-optimization/84913
707 * tree-vect-loop.c (vectorizable_reduction): Don't try to
708 vectorize chains of COND_EXPRs.
710 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
712 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
714 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
716 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
718 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
720 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
722 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
723 Kito Cheng <kito.cheng@gmail.com>
725 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
726 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
727 (nds32_adjust_reg_alloc_order): New function.
728 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
730 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
732 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
733 nds32_print_operand, nds32_print_operand_address): Use
734 HOST_WIDE_INT_PRINT_DEC instead.
736 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
738 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
740 2018-03-17 Jakub Jelinek <jakub@redhat.com>
742 PR target/84902
743 * config/i386/i386.c (initial_ix86_tune_features,
744 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
745 unsigned long long.
746 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
747 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
748 rather than 1u << ix86_tune. Formatting fix.
749 (ix86_option_override_internal): Change ix86_arch_mask from
750 unsigned int to unsigned HOST_WIDE_INT, initialize to
751 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
752 (ix86_function_specific_restore): Likewise.
754 2018-03-16 Jakub Jelinek <jakub@redhat.com>
756 PR target/84899
757 * postreload.c (reload_combine_recognize_pattern): Perform
758 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
759 truncate_int_for_mode the result for the destination's mode.
761 PR c/84909
762 * hsa-gen.c (mem_type_for_type): Fix comment typo.
763 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
764 Likewise.
765 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
766 Likewise.
768 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
770 PR target/84876
771 * lra-assigns.c (lra_split_hard_reg_for): Don't use
772 regno_allocno_class_array and sorted_pseudos.
773 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
774 insns where regno is used.
776 2018-03-16 Martin Liska <mliska@suse.cz>
778 PR ipa/84833
779 * multiple_target.c (create_dispatcher_calls): Redirect
780 reference in the symbol table.
782 2018-03-16 Martin Liska <mliska@suse.cz>
784 PR ipa/84722
785 * multiple_target.c (create_dispatcher_calls): Redirect also
786 an alias.
788 2018-03-16 Jakub Jelinek <jakub@redhat.com>
790 PR c++/79937
791 PR c++/82410
792 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
793 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
794 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
796 2018-03-16 Julia Koval <julia.koval@intel.com>
798 * doc/invoke.texi (Skylake Server): Add CLWB.
799 Cannonlake): Remove CLWB.
801 2018-03-16 Jakub Jelinek <jakub@redhat.com>
803 PR tree-optimization/84841
804 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
805 1 << 3.
806 (FLOAT_ONE_CONST_TYPE): Define.
807 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
808 (sort_by_operand_rank): Put entries with higher constant_type last
809 rather than first to match comments.
811 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
813 * config/nios2/nios2.md (movsi_internal): Fix thinko in
814 split predicate.
816 2018-03-15 Jakub Jelinek <jakub@redhat.com>
818 PR c++/79085
819 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
820 check and use address of target always.
822 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
824 PR target/84574
825 * config/i386/i386.c (indirect_thunk_needed): Update comments.
826 (indirect_thunk_bnd_needed): Likewise.
827 (indirect_thunks_used): Likewise.
828 (indirect_thunks_bnd_used): Likewise.
829 (indirect_return_needed): New.
830 (indirect_return_bnd_needed): Likewise.
831 (output_indirect_thunk_function): Add a bool argument for
832 function return.
833 (output_indirect_thunk_function): Don't generate alias for
834 function return thunk.
835 (ix86_code_end): Call output_indirect_thunk_function to generate
836 function return thunks.
837 (ix86_output_function_return): Set indirect_return_bnd_needed
838 and indirect_return_needed instead of indirect_thunk_bnd_needed
839 and indirect_thunk_needed.
841 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
843 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
844 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
845 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
847 2018-03-15 David Malcolm <dmalcolm@redhat.com>
848 Paul Hua <paul.hua.gm@gmail.com>
850 PR c/84852
851 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
853 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
855 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
856 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
857 resp. SFmode cases.
859 2018-03-15 Tamar Christina <tamar.christina@arm.com>
861 PR target/84711
862 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
863 instead of GET_MODE_SIZE when comparing Units.
865 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
867 PR target/68256
868 * varasm.c (hash_section): Return an unchangeble hash value
869 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
870 Return !aarch64_can_use_per_function_literal_pools_p ().
872 2018-03-15 Jakub Jelinek <jakub@redhat.com>
874 PR target/84860
875 * optabs.c (emit_conditional_move): Pass address of cmode's copy
876 rather than address of cmode as last argument to prepare_cmp_insn.
878 2018-03-15 Julia Koval <julia.koval@intel.com>
880 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
881 F_AVX512VNNI, F_AVX512BITALG): New.
883 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
885 PR target/83451
886 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
887 insn for floating-point loads and stores.
889 2018-03-14 Carl Love <cel@us.ibm.com>
891 * config/rs6000/rs6000-c.c: Add macro definitions for
892 ALTIVEC_BUILTIN_VEC_PERMXOR.
893 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
894 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
895 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
896 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
897 UNSPEC_VPERMXOR.
898 * config/doc/extend.texi: Add prototypes for vec_permxor.
900 2018-03-14 David Malcolm <dmalcolm@redhat.com>
902 PR c/84852
903 * diagnostic-show-locus.c (class layout_point): Convert m_line
904 from int to linenum_type.
905 (line_span::comparator): Use linenum "compare" function when
906 comparing line numbers.
907 (test_line_span): New function.
908 (layout_range::contains_point): Convert param "row" from int to
909 linenum_type.
910 (layout_range::intersects_line_p): Likewise.
911 (layout::will_show_line_p): Likewise.
912 (layout::print_source_line): Likewise.
913 (layout::should_print_annotation_line_p): Likewise.
914 (layout::print_annotation_line): Likewise.
915 (layout::print_leading_fixits): Likewise.
916 (layout::annotation_line_showed_range_p): Likewise.
917 (struct line_corrections): Likewise for field m_row.
918 (line_corrections::line_corrections): Likewise for param "row".
919 (layout::print_trailing_fixits): Likewise.
920 (layout::get_state_at_point): Likewise.
921 (layout::get_x_bound_for_row): Likewise.
922 (layout::print_line): Likewise.
923 (diagnostic_show_locus): Likewise for locals "last_line" and
924 "row".
925 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
926 * input.c (selftest::test_linenum_comparisons): New function.
927 (selftest::input_c_tests): Call it.
928 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
929 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
930 * selftest.h (ASSERT_GT): New macro.
931 (ASSERT_GT_AT): New macro.
932 (ASSERT_LT): New macro.
933 (ASSERT_LT_AT): New macro.
935 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
937 PR rtl-optimization/84780
938 * combine.c (distribute_links): Don't make a link based on pc_rtx.
940 2018-03-14 Martin Liska <mliska@suse.cz>
942 * tree.c (record_node_allocation_statistics): Use
943 get_stats_node_kind.
944 (get_stats_node_kind): New function extracted from
945 record_node_allocation_statistics.
946 (free_node): Use get_stats_node_kind.
948 2018-03-14 Richard Biener <rguenther@suse.de>
950 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
951 that the value-set of ANTIC_IN doesn't grow.
953 Revert
954 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
955 member.
956 (BB_VISITED_WITH_VISITED_SUCCS): New define.
957 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
959 2018-03-14 Julia Koval <julia.koval@intel.com>
961 * config.gcc (icelake-client, icelake-server): New.
962 (icelake): Remove.
963 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
964 (initial_ix86_arch_features): Ditto.
965 (PTA_SKYLAKE): Add SGX.
966 (PTA_ICELAKE): Remove.
967 (PTA_ICELAKE_CLIENT): New.
968 (PTA_ICELAKE_SERVER): New.
969 (ix86_option_override_internal): Split up icelake on icelake client and
970 icelake server.
971 (get_builtin_code_for_version): Ditto.
972 (fold_builtin_cpu): Ditto.
973 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
974 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
975 * config/i386/i386.h (processor_type): Ditto.
976 * doc/invoke.texi: Ditto.
978 2018-03-14 Jakub Jelinek <jakub@redhat.com>
980 PR sanitizer/83392
981 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
982 INTEGER_CST offset, add it together with bitpos / 8 and
983 sign extend based on POINTER_SIZE.
985 PR target/84844
986 Revert
987 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
989 PR target/78090
990 * config/i386/constraints.md (Yc): New register constraint.
991 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
992 Use Yc constraint for alternative 2 of operand 0. Remove
993 preferred_for_speed attribute.
995 2018-03-14 Richard Biener <rguenther@suse.de>
997 PR tree-optimization/84830
998 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
999 with the old one to avoid oscillations.
1001 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1003 PR target/83712
1004 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1005 pseudos.
1006 (assign_by_spills): Return a flag of reload assignment failure.
1007 Do not process the reload assignment failures. Do not spill other
1008 reload pseudos if they has the same reg class. Update n if
1009 necessary.
1010 (lra_assign): Add a return arg. Set up from the result of
1011 assign_by_spills call.
1012 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1013 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1014 usage_insns if it is not NULL.
1015 (spill_hard_reg_in_range): New function.
1016 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1017 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1018 function prototypes.
1019 (lra_assign): Change prototype.
1020 * lra.c (lra): Add code to deal with fails by splitting hard reg
1021 live ranges.
1023 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1025 * config/riscv/riscv.opt (mrelax): New option.
1026 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1027 "norelax" when riscv_mrelax is disabled.
1028 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1030 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1032 PR target/84743
1033 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1034 reassociation for int modes.
1036 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1038 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1039 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1040 for big-endian.
1041 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1042 * config/aarch64/aarch64-sve.md
1043 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1044 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1045 (*extend<mode><Vwide>2): Rename to...
1046 (aarch64_sve_extend<mode><Vwide>2): ...this.
1047 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1048 renaming the old pattern to...
1049 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1050 unsigned packs.
1051 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1052 define_expand, renaming the old pattern to...
1053 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1054 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1055 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1056 account when deciding which SVE instruction the optab should use.
1057 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1059 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1061 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1062 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1063 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1064 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1065 (tlsdesc_small_<mode>): Turn a define_expand and use
1066 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1067 (tlsdesc_small_advsimd_<mode>): ...this.
1068 (tlsdesc_small_sve_<mode>): New pattern.
1070 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1072 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1073 (UNSPEC_UMUL_HIGHPART): New constants.
1074 (MUL_HIGHPART): New int iteraor.
1075 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1076 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1077 define_expand.
1078 (*<su>mul<mode>3_highpart): New define_insn.
1080 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1082 PR lto/84805
1083 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1084 incomplete types.
1086 2018-03-13 Martin Liska <mliska@suse.cz>
1088 PR ipa/84658.
1089 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1090 vector.
1091 (sem_item_optimizer::~sem_item_optimizer): Release it.
1092 (sem_item_optimizer::merge_classes): Register variable aliases.
1093 (sem_item_optimizer::fixup_pt_set): New function.
1094 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1095 * ipa-icf.h: Declare new variables and functions.
1097 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1099 PR middle-end/84834
1100 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1101 integer_pow2p@2 and test integer_pow2p in condition.
1102 (A < 0 ? C : 0): Similarly for @1.
1104 PR middle-end/84831
1105 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1106 characters starting at p contain '\0' character, don't look beyond
1107 that.
1109 PR target/84827
1110 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1111 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1113 PR target/84828
1114 * reg-stack.c (change_stack): Change update_end var from int to
1115 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1116 also call set_block_for_insn on the newly added insns and rescan.
1118 PR target/84786
1119 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1120 on the last operand.
1122 PR c++/84704
1123 * tree.c (stabilize_reference_1): Return save_expr (e) for
1124 STATEMENT_LIST even if it doesn't have side-effects.
1126 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1128 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1130 2018-03-12 Renlin Li <renlin.li@arm.com>
1132 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1133 aarch64_output_scalar_simd_mov_immediate.
1135 2018-03-12 Martin Sebor <msebor@redhat.com>
1137 PR tree-optimization/83456
1138 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1139 for perfectly overlapping calls to memcpy.
1140 (gimple_fold_builtin_memory_chk): Same.
1141 (gimple_fold_builtin_strcpy): Handle no-warning.
1142 (gimple_fold_builtin_stxcpy_chk): Same.
1143 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1145 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1147 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1148 parameter. Use it for SFmode.
1149 (rs6000_function_arg_advance_1): Adjust.
1150 (rs6000_function_arg): Adjust.
1151 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1153 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1155 PR rtl-optimization/84169
1156 PR rtl-optimization/84780
1157 * combine.c (can_combine_p): Check for a 2-insn combination whether
1158 the destination register is used between the two insns, too.
1160 2018-03-12 Richard Biener <rguenther@suse.de>
1162 PR tree-optimization/84803
1163 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1164 for refs DR analysis didn't process.
1166 2018-03-12 Richard Biener <rguenther@suse.de>
1168 PR tree-optimization/84777
1169 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1170 force-vectorize loops ignore whether we are optimizing for size.
1172 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1174 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1175 (TARGET_MD_ASM_ADJUST): Define.
1177 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1178 Kito Cheng <kito.cheng@gmail.com>
1179 Chung-Ju Wu <jasonwucj@gmail.com>
1181 * config/nds32/nds32.c (nds32_compute_stack_frame,
1182 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1183 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1184 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1185 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1186 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1187 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1188 * config/nds32/nds32.md (prologue, epilogue): Use macro
1189 NDS32_V3PUSH_AVAILABLE_P to do checking.
1191 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1193 PR debug/58150
1194 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1195 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1196 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1197 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1198 addition of most attributes on !orig_type_die or the attribute not
1199 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1201 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1202 Chung-Ju Wu <jasonwucj@gmail.com>
1204 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1205 __NDS32_VH__ macro.
1206 * config/nds32/nds32.opt (mvh): New option.
1208 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1209 Chung-Ju Wu <jasonwucj@gmail.com>
1211 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1212 function.
1213 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1214 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1215 definition.
1217 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1218 Chung-Ju Wu <jasonwucj@gmail.com>
1220 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1221 function.
1222 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1223 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1225 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1226 Kito Cheng <kito.cheng@gmail.com>
1227 Chung-Ju Wu <jasonwucj@gmail.com>
1229 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1230 UNSPEC_FFMISM and UNSPEC_FLMISM.
1231 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1232 for ffb, ffmism and flmism.
1233 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1234 (unspec_ffmism): Ditto.
1235 (unspec_flmism): Ditto.
1236 (nds32_expand_builtin_impl): Check if string extension is available.
1237 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1238 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1240 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1242 Reverting patch:
1243 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1245 PR target/83712
1246 * lra-assigns.c (assign_by_spills): Return a flag of reload
1247 assignment failure. Do not process the reload assignment
1248 failures. Do not spill other reload pseudos if they has the same
1249 reg class.
1250 (lra_assign): Add a return arg. Set up from the result of
1251 assign_by_spills call.
1252 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1253 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1254 usage_insns if it is not NULL.
1255 (spill_hard_reg_in_range): New function.
1256 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1257 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1258 function prototypes.
1259 (lra_assign): Change prototype.
1260 * lra.c (lra): Add code to deal with fails by splitting hard reg
1261 live ranges.
1263 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1265 PR target/84807
1266 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1268 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1270 PR debug/84620
1271 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1272 (dw_val_node): Add val_symbolic_view.
1273 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1274 (symview_upper_bound): New.
1275 (new_line_info_table): Initialize symviews_since_reset.
1276 (dwarf2out_source_line): Count symviews_since_reset and set
1277 symview_upper_bound.
1278 (dw_val_equal_p): Handle symview.
1279 (add_AT_symview): New.
1280 (print_dw_val): Handle symview.
1281 (attr_checksum, attr_checksum_ordered): Likewise.
1282 (same_dw_val_p, size_of_die): Likewise.
1283 (value_format, output_die): Likewise.
1284 (add_high_low_attributes): Use add_AT_symview for entry_view.
1285 (dwarf2out_finish): Reset symview_upper_bound, clear
1286 zero_view_p.
1288 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1290 PR target/83969
1291 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1292 Add strict argument and use it.
1293 (rs6000_split_multireg_move): Update for new strict argument.
1294 (mem_operand_gpr): Disallow all non-offsettable addresses.
1295 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1297 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1299 PR target/84772
1300 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1301 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1302 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1304 PR c++/84767
1305 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1306 decl, use remap_type if we want to use the type.
1308 2018-03-09 Martin Sebor <msebor@redhat.com>
1310 PR tree-optimization/84526
1311 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1312 Remove dead code.
1313 (builtin_access::generic_overlap): Be prepared to handle non-array
1314 base objects.
1316 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1318 PR rtl-optimization/84682
1319 * lra-constraints.c (process_address_1): Check is_address flag
1320 for address constraints.
1321 (process_alt_operands): Likewise.
1322 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1323 preprocess_constraints.
1324 * recog.h (preprocess_constraints): Add oploc parameter.
1325 Adjust callers.
1326 * recog.c (preprocess_constraints): Test address_operand for
1327 CT_ADDRESS constraints.
1329 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1331 PR target/83712
1332 * lra-assigns.c (assign_by_spills): Return a flag of reload
1333 assignment failure. Do not process the reload assignment
1334 failures. Do not spill other reload pseudos if they has the same
1335 reg class.
1336 (lra_assign): Add a return arg. Set up from the result of
1337 assign_by_spills call.
1338 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1339 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1340 usage_insns if it is not NULL.
1341 (spill_hard_reg_in_range): New function.
1342 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1343 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1344 function prototypes.
1345 (lra_assign): Change prototype.
1346 * lra.c (lra): Add code to deal with fails by splitting hard reg
1347 live ranges.
1349 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1351 PR target/83193
1352 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1353 Accept complain bool parameter. Only emit errors if it is true.
1354 (arm_parse_cpu_option_name): Likewise.
1355 (arm_target_thumb_only): Adjust callers of the above.
1356 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1357 prototype to take a default true bool parameter.
1358 (arm_parse_arch_option_name): Likewise.
1360 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1361 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1363 PR jit/64089
1364 PR jit/84288
1365 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1366 * configure: Regenerate.
1367 * configure.ac ("linker --version-script option"): New.
1368 ("linker soname option"): New.
1370 2018-03-09 Richard Biener <rguenther@suse.de>
1372 PR tree-optimization/84775
1373 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1374 immediate uses of predicate stmts and mark them modified.
1376 Revert
1377 PR tree-optimization/84178
1378 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1379 to caller.
1380 (version_loop_for_if_conversion): Delay update_ssa call.
1381 (tree_if_conversion): Delay update_ssa until after predicate
1382 insertion.
1384 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1386 PR target/84763
1387 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1388 when the function accesses prior frames.
1390 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1392 PR debug/84456
1393 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1394 gen_llsym, otherwise call maybe_gen_llsym.
1396 PR inline-asm/84742
1397 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1398 has ',' character inside of it.
1400 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1402 PR target/84748
1403 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1404 as clobbering CC_REGNUM.
1406 2018-03-08 Richard Biener <rguenther@suse.de>
1408 PR middle-end/84552
1409 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1410 (follow_copies_to_constant): Do not follow SSA names registered
1411 for update.
1413 2018-03-08 Richard Biener <rguenther@suse.de>
1415 PR tree-optimization/84178
1416 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1417 to caller.
1418 (version_loop_for_if_conversion): Delay update_ssa call.
1419 (tree_if_conversion): Delay update_ssa until after predicate
1420 insertion.
1422 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1424 PR tree-optimization/84178
1425 * tree-if-conv.c (release_bb_predicate): Remove the
1426 the assertion that the stmts have NULL use_ops.
1427 Discard the statements, asserting that they haven't
1428 yet been added to a BB.
1430 2018-03-08 Richard Biener <rguenther@suse.de>
1432 PR tree-optimization/84746
1433 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1434 (phi_translate): Pass in destination ANTIC_OUT set.
1435 (phi_translate_1): Likewise. For a simplified result lookup
1436 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1437 (phi_translate_set): Adjust.
1438 (do_pre_regular_insertion): Likewise.
1439 (do_pre_partial_partial_insertion): Likewise.
1441 2018-03-08 Martin Liska <mliska@suse.cz>
1443 PR gcov-profile/84735
1444 * doc/gcov.texi: Document usage of profile files.
1445 * gcov-io.h: Document changes in the format.
1447 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1449 PR debug/84404
1450 PR debug/84408
1451 * dwarf2out.c (struct dw_line_info_table): Update comments for
1452 view == -1.
1453 (FORCE_RESET_NEXT_VIEW): New.
1454 (FORCE_RESETTING_VIEW_P): New.
1455 (RESETTING_VIEW_P): Check for -1 too.
1456 (ZERO_VIEW_P): Likewise.
1457 (new_line_info_table): Force-reset next view.
1458 (dwarf2out_begin_function): Likewise.
1459 (dwarf2out_source_line): Simplify zero_view_p initialization.
1460 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1461 view directly. Omit view when omitting .loc at line 0.
1463 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1465 PR tree-optimization/84740
1466 * tree-switch-conversion.c (process_switch): Call build_constructors
1467 only if info.phi_count is non-zero.
1469 PR tree-optimization/84739
1470 * tree-tailcall.c (find_tail_calls): Check call arguments against
1471 DECL_ARGUMENTS (current_function_decl) rather than
1472 DECL_ARGUMENTS (func) when checking for tail recursion.
1474 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1476 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1477 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1478 Volker Reichelt's entry and add entries for people that perform
1479 GCC fuzzy testing and report numerous bugs.
1481 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1483 PR target/82411
1484 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1485 readonly data in sdata, if that is disabled.
1486 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1487 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1488 -mreadonly-in-sdata option.
1490 2018-03-07 Martin Sebor <msebor@redhat.com>
1492 PR tree-optimization/84468
1493 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1494 basic block when looking for nul assignment.
1496 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1498 PR target/84277
1499 * except.h (output_function_exception_table): Adjust prototype.
1500 * except.c (output_function_exception_table): Remove FNNAME parameter
1501 and add SECTION parameter. Ouput one part of the table at a time.
1502 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1503 the first part of the exception table and emit unwind directives.
1504 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1505 (i386_pe_seh_cold_init): Likewise.
1506 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1507 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1508 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1509 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1510 * config/i386/winnt.c: Include except.h.
1511 (struct seh_frame_state): Add reg_offset, after_prologue and
1512 in_cold_section fields.
1513 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1514 (i386_pe_seh_cold_init): New function.
1515 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1516 to seh->in_cold_section.
1517 (seh_emit_push): Record the offset of the push.
1518 (seh_emit_save): Record the offet of the save.
1519 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1520 Test seh->after_prologue to disregard the epilogue.
1521 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1522 (i386_pe_end_cold_function): New function.
1524 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1526 PR fortran/84565
1527 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1528 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1530 PR c++/84704
1531 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1532 on tmp_var.
1533 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1534 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1536 PR middle-end/84723
1537 * multiple_target.c: Include tree-inline.h and intl.h.
1538 (expand_target_clones): Diagnose and fail if node->definition and
1539 !tree_versionable_function_p (node->decl).
1541 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1543 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1544 sprint_ul.
1545 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1546 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1547 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1549 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1551 PR target/84710
1552 * combine.c (try_combine): Use reg_or_subregno instead of handling
1553 just paradoxical SUBREGs and REGs.
1555 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1557 * config/arc/arc.c (arc_finalize_pic): Remove function.
1558 (arc_must_save_register): We use single base PIC register, remove
1559 checks to save/restore the PIC register.
1560 (arc_expand_prologue): Likewise.
1561 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1562 Remove.
1563 (arc_verify_short): Likewise.
1564 (arc_attr_type): Likewise.
1565 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1566 (walk_stores): Likewise.
1567 (arc_address_cost): Make it static.
1568 (arc_verify_short): Likewise.
1569 (branch_dest): Likewise.
1570 (arc_attr_type): Likewise.
1571 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1572 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1573 (arc_final_prescan_insn): Remove inserting the nops due to
1574 hardware hazards. It is done in reorg step.
1575 (insn_length_variant_t): Remove.
1576 (insn_length_parameters_t): Likewise.
1577 (arc_insn_length_parameters): Likewise.
1578 (arc_get_insn_variants): Likewise.
1579 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1581 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1583 PR inline-asm/84683
1584 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1585 assertion failure.
1587 PR tree-optimization/84687
1588 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1589 on new_node->decl.
1590 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1592 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1594 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1595 Rename to ppc_speculation_barrier.
1596 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1597 __builtin_ppc_speculation_barrier.
1599 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1601 PR target/84700
1602 * combine.c (combine_simplify_rtx): Don't try to simplify if
1603 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1604 are equal to x.
1606 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1608 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1609 to 32 bytes when compiling for POWER9.
1611 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1613 PR target/84564
1614 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1615 regparm >= 3 with no arg reg available also for calls with
1616 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1618 PR target/84524
1619 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1620 orig,vex.
1621 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1623 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1625 PR target/84264
1626 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1628 2018-03-05 Richard Biener <rguenther@suse.de>
1630 PR tree-optimization/84486
1631 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1632 When inserting a __builtin_assume_aligned call set the LHS
1633 SSA name alignment info accordingly.
1635 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1637 PR tree-optimization/84114
1638 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1639 Avoid reassociation of FLOAT_MODE addition.
1641 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1643 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1644 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1645 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1646 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1647 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1648 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1649 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1650 and -mwbnoinvd.
1651 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1652 __builtin_ia32_wbinvd): New builtins.
1653 (SPECIAL_ARGS2): New.
1654 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1655 (SPECIAL_ARGS2): New.
1656 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1657 (ix86_valid_target_attribute_inner_p): Ditto.
1658 (ix86_init_mmx_sse_builtins): Add special_args2.
1659 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1660 TARGET_WBNOINVD_P): New.
1661 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1662 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1663 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1664 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1665 * config/i386/pconfigintrin.h: New file.
1666 * config/i386/wbnoinvdintrin.h: Ditto.
1667 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1668 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1670 2018-03-05 Richard Biener <rguenther@suse.de>
1672 PR tree-optimization/84670
1673 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1674 member.
1675 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1676 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1677 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1678 doesn't grow if all successors (recursively) were visited at least
1679 once.
1681 2018-03-05 Richard Biener <rguenther@suse.de>
1683 PR tree-optimization/84650
1684 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1685 if executed in the loop pipeline.
1687 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1689 * doc/configfiles.texi (Configuration Files): Move info about
1690 conditionalizing $target-protos.h to...
1691 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1692 differs from $target-protos.h.
1694 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1695 Chung-Ju Wu <jasonwucj@gmail.com>
1697 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1698 * config/nds32/nds32-multiple.md (setmemsi): Define.
1699 * config/nds32/nds32-memory-manipulation.c
1700 (nds32_gen_dup_4_byte_to_word_value): New.
1701 (emit_setmem_word_loop): New.
1702 (emit_setmem_byte_loop): New.
1703 (nds32_expand_setmem_loop): New.
1704 (nds32_expand_setmem_loop_v3m): New.
1705 (nds32_expand_setmem_unroll): New.
1706 (nds32_expand_setmem): New.
1708 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1709 Chung-Ju Wu <jasonwucj@gmail.com>
1711 * config/nds32/nds32-memory-manipulation.c
1712 (nds32_emit_load_store): New.
1713 (nds32_emit_post_inc_load_store): New.
1714 (nds32_emit_mem_move): New.
1715 (nds32_emit_mem_move_block): New.
1716 (nds32_expand_movmemsi_loop_unknown_size): New.
1717 (nds32_expand_movmemsi_loop_known_size): New.
1718 (nds32_expand_movmemsi_loop): New.
1719 (nds32_expand_movmemsi_unroll): New.
1720 (nds32_expand_movmemqi): Rename ...
1721 (nds32_expand_movmemsi): ... to this.
1722 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1723 (movmemsi): ... to this.
1724 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1725 (nds32_expand_movmemsi): ... to this.
1727 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1728 Monk Chiang <sh.chiang04@gmail.com>
1729 Chung-Ju Wu <jasonwucj@gmail.com>
1731 * config/nds32/nds32-protos.h
1732 (nds32_expand_load_multiple): New arguments.
1733 (nds32_expand_store_multiple): Ditto.
1734 (nds32_valid_multiple_load_store): Rename ...
1735 (nds32_valid_multiple_load_store_p): ... to this.
1736 * config/nds32/nds32-memory-manipulation.c
1737 (nds32_expand_load_multiple): Refine implementation.
1738 (nds32_expand_store_multiple): Ditto.
1739 * config/nds32/nds32-multiple.md
1740 (load_multiple): Update nds32_expand_load_multiple interface.
1741 (store_multiple): Update nds32_expand_store_multiple interface.
1742 * config/nds32/nds32-predicates.c
1743 (nds32_valid_multiple_load_store): Rename ...
1744 (nds32_valid_multiple_load_store_p): ... to this and refine
1745 implementation.
1746 * config/nds32/predicates.md
1747 (nds32_load_multiple_and_update_address_operation): New predicate.
1748 (nds32_store_multiple_and_update_address_operation): New predicate.
1750 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1751 Chung-Ju Wu <jasonwucj@gmail.com>
1753 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1754 (combo): New attribute.
1755 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1757 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1759 * config/nds32/nds32.opt: Change -mcmodel= default value.
1761 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1762 Monk Chiang <sh.chiang04@gmail.com>
1763 Chung-Ju Wu <jasonwucj@gmail.com>
1765 * config/nds32/constants.md (unspec_element): New enum.
1766 * config/nds32/constraints.md (Umw): New constraint.
1767 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1768 * config/nds32/nds32-intrinsic.md: Likewise.
1769 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1770 (nds32_valid_smw_lwm_base_p): New.
1771 (nds32_output_smw_single_word): New.
1772 (nds32_output_lmw_single_word): New.
1773 (nds32_expand_unaligned_load): New.
1774 (nds32_expand_unaligned_store): New.
1775 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1776 (nds32_output_smw_single_word): Declare.
1777 (nds32_output_lmw_single_word): Declare.
1778 (nds32_expand_unaligned_load): Declare.
1779 (nds32_expand_unaligned_store): Declare.
1780 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1781 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1782 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1783 NDS32_BUILTIN_UASTORE_DW.
1784 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1785 predicate.
1787 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1788 Kito Cheng <kito.cheng@gmail.com>
1789 Chung-Ju Wu <jasonwucj@gmail.com>
1791 * config/nds32/nds32-intrinsic.c
1792 (nds32_expand_builtin_null_ftype_reg): Delete.
1793 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1794 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1795 (nds32_read_argument): New.
1796 (nds32_legitimize_target): Ditto.
1797 (nds32_legitimize_argument): Ditto.
1798 (nds32_check_constant_argument): Ditto.
1799 (nds32_expand_unop_builtin): Ditto.
1800 (nds32_expand_unopimm_builtin): Ditto.
1801 (nds32_expand_binop_builtin): Ditto.
1802 (nds32_builtin_decl_impl): Ditto.
1803 (builtin_description): Ditto.
1804 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1805 (nds32_init_builtins_impl): Ditto.
1806 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1807 (nds32_builtin_decl): New.
1808 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1809 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1811 2018-03-02 Jeff Law <law@redhat.com>
1813 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1814 (redundant_insn, fill_simple_delay_slots): Likewise.
1815 (fill_slots_from_thread): Likewise.
1816 * resource.c (mark_referenced_resources): Likewise.
1817 (mark_set_resources, find_dead_or_set_registers): Likewise.
1819 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1821 * substring-locations.h (format_warning_va): Formatting fix for
1822 ATTRIBUTE_GCC_DIAG.
1823 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1824 argument.
1825 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1826 * substring-locations.c: Include intl.h.
1827 (format_warning_va): Turned into small wrapper around
1828 format_warning_n_va, renamed to ...
1829 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1830 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1831 use ngettext.
1832 (format_warning_at_substring_n): New function.
1833 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1834 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1835 format_warning_at_substring with just a shorter name instead of
1836 const function pointer.
1837 (fmtwarn_n): New function.
1838 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1839 appropriate, get rid of all the fmtstr temporaries, move conditionals
1840 with G_() wrapped string literals directly into fmtwarn arguments,
1841 cast dir.len to (int), formatting fixes.
1843 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1845 * doc/invoke.texi: Remove "Cilk Plus" references.
1847 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1848 Richard Biener <rguenther@suse.de>
1850 PR ipa/84628
1851 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1852 for error or warning attributes if CALL_FROM_THUNK_P is set.
1853 Formatting fixes.
1855 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1857 PR target/56540
1858 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1859 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1861 PR target/56540
1862 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1863 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1865 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1866 instead of -1U in last predictors element's probability member.
1868 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1870 PR ipa/83983
1871 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1872 arguments if they are comparable.
1874 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1876 PR tree-optimization/84634
1877 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1878 masks and masked_loop_p with a single loop_masks, making sure it's
1879 null for bb vectorization.
1881 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1883 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1884 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1885 loop->force_vectorize to check whether there is no alias.
1887 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1889 PR target/84614
1890 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1891 prototypes.
1892 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1893 comments.
1894 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1895 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1896 instead of a loop around prev_real_insn.
1897 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1898 prev_real_insn.
1900 PR inline-asm/84625
1901 * config/i386/i386.c (ix86_print_operand): Use conditional
1902 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1903 zero vector.
1905 2018-03-02 Richard Biener <rguenther@suse.de>
1907 PR tree-optimization/84427
1908 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1909 (bitmap_set_subtract_values): Rewrite to handle multiple
1910 exprs per value.
1911 (clean): Likewise.
1912 (prune_clobbered_mems): Likewise.
1913 (phi_translate): Take edge instead of pred/phiblock.
1914 (phi_translate_1): Likewise.
1915 (phi_translate_set): Likewise. Insert all translated
1916 exprs for a value into the set, keeping possibly multiple
1917 expressions per value.
1918 (compute_antic_aux): Adjust for phi_translate changes.
1919 When intersecting union the expressions and prune those
1920 not in the final value set, keeping possibly multiple
1921 expressions per value. Do not use value-insertion
1922 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1923 all expressions. Add verification that the value-sets
1924 only shrink during iteration.
1925 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1926 (do_pre_regular_insertion): Likewise.
1927 (do_pre_partial_partial_insertion): Likewise.
1929 2018-03-02 Richard Biener <rguenther@suse.de>
1931 PR target/82005
1932 * config/darwin.c (saved_debug_info_level): New static global.
1933 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1934 (darwin_asm_lto_end): Restore debug info generation settings.
1936 2018-03-01 Martin Liska <mliska@suse.cz>
1938 PR sanitizer/82484
1939 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1940 volatile arguments.
1942 2018-03-01 Richard Biener <rguenther@suse.de>
1944 PR debug/84645
1945 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1946 type annotation with LTO when debug was disabled at compile-time.
1948 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1950 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1951 XINT with INTVAL.
1952 (mips_final_postscan_insn): Likewise.
1954 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1956 PR rtl-optimization/84528
1957 * alias.c (init_alias_target): Add commentary.
1958 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1959 a unique base value if the frame pointer is not eliminated
1960 to the stack pointer.
1962 2018-03-01 Tom de Vries <tom@codesourcery.com>
1964 PR rtl-optimization/83327
1965 * lra-int.h (hard_regs_spilled_into): Declare.
1966 * lra.c (hard_regs_spilled_into): Define.
1967 (init_reg_info): Init hard_regs_spilled_into.
1968 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1969 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1970 (process_bb_lives): Handle hard_regs_spilled_into.
1971 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1972 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1974 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1976 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1977 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1978 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1979 * config/rs6000/aix72.h: New file.
1981 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1983 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1984 instead of warning_at with conditional singular and plural messages
1985 where possible.
1987 PR target/52991
1988 * stor-layout.c (update_alignment_for_field): For
1989 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1990 && !DECL_PACKED (field), do the alignment update, just use
1991 only desired_align instead of MAX (type_align, desired_align)
1992 as the alignment.
1993 (place_field): Don't do known_align < desired_align handling
1994 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1995 is non-NULL, instead do it after rli->prev_field handling and
1996 only if not within a bitfield word. For DECL_PACKED (field)
1997 use type_align of BITS_PER_UNIT.
1999 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2001 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2002 superfluous parentheses and trailing spaces.
2004 2018-02-28 Richard Biener <rguenther@suse.de>
2006 PR tree-optimization/84584
2007 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2008 SCoPs with fake exit edge.
2010 2018-02-28 Martin Liska <mliska@suse.cz>
2012 PR testsuite/84597
2013 * timevar.c (timer::print): Fix format to properly print 100%
2014 values.
2016 2018-02-28 Richard Biener <rguenther@suse.de>
2018 PR middle-end/84607
2019 * genmatch.c (capture_info::walk_match): Do not mark
2020 captured expressions without operands as expr_p given
2021 they act more like predicates and should be subject to
2022 "lost tail" side-effect preserving.
2024 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2026 PR rtl-optimization/81611
2027 * auto-inc-dec.c (attempt_change): Move dead note from
2028 mem_insn if it's the next use of regno
2029 (find_address): Take address use of reg holding
2030 non-incremented value. Add parm to limit search to the named
2031 reg only.
2032 (merge_in_block): Attempt to use a mem insn that is the next
2033 use of the original regno.
2035 2018-02-27 Martin Sebor <msebor@redhat.com>
2037 PR c++/83871
2038 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2039 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2041 2018-02-27 Martin Sebor <msebor@redhat.com>
2043 PR translation/84207
2044 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2045 n argument to unsigned HOST_WIDE_INT.
2046 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2047 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2048 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2049 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2051 2018-02-27 Richard Biener <rguenther@suse.de>
2053 PR tree-optimization/84512
2054 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2055 Do not use the estimate returned from record_stmt_cost for
2056 the scalar iteration cost but sum properly using add_stmt_cost.
2058 2018-02-27 Richard Biener <rguenther@suse.de>
2060 PR tree-optimization/84466
2061 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2062 Adjust last change to less strictly validate use operands.
2064 2018-02-27 Martin Liska <mliska@suse.cz>
2066 PR gcov-profile/84548
2067 * gcov.c (process_file): Allow partial overlap and consider it
2068 also as group functions.
2069 (output_lines): Properly calculate range of lines for a group.
2071 2018-02-27 Martin Liska <mliska@suse.cz>
2073 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2074 'ggc' suffixes. Change first column width.
2075 (timer::print): Fix formatting of the column.
2077 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2079 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2080 preserve inline entry blocks for the sake of debug inline
2081 entry point markers alone.
2082 (remove_unused_locals): Suggest in comments a better place to
2083 force the preservation of inline entry blocks that are
2084 otherwise unused, but do not preserve them.
2086 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2088 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2090 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2092 PR target/84039
2093 * config/i386/constraints.md (Bs): Replace
2094 ix86_indirect_branch_register with
2095 TARGET_INDIRECT_BRANCH_REGISTER.
2096 (Bw): Likewise.
2097 * config/i386/i386.md (indirect_jump): Likewise.
2098 (tablejump): Likewise.
2099 (*sibcall_memory): Likewise.
2100 (*sibcall_value_memory): Likewise.
2101 Peepholes of indirect call and jump via memory: Likewise.
2102 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2103 (*sibcall_value_GOT_32): Likewise.
2104 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2105 (GOT_memory_operand): Likewise.
2106 (call_insn_operand): Likewise.
2107 (sibcall_insn_operand): Likewise.
2108 (GOT32_symbol_operand): Likewise.
2109 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2111 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2113 PR rtl-optimization/83496
2114 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2115 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2116 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2117 redundant insn, if any.
2118 (relax_delay_slots): Likewise.
2119 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2121 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2123 PR tree-optimization/83965
2124 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2125 that grouped statements are part of a reduction chain. Return
2126 true if the statement is not marked as a reduction itself but
2127 is part of a group.
2128 (vect_recog_dot_prod_pattern): Don't check whether the statement
2129 is part of a group here.
2130 (vect_recog_sad_pattern): Likewise.
2131 (vect_recog_widen_sum_pattern): Likewise.
2133 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2135 PR debug/84545
2136 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2138 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2140 PR target/84530
2141 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2142 the bool argument.
2143 (ix86_output_indirect_function_return): New prototype.
2144 (ix86_split_simple_return_pop_internal): Likewise.
2145 * config/i386/i386.c (indirect_return_via_cx): New.
2146 (indirect_return_via_cx_bnd): Likewise.
2147 (indirect_thunk_name): Handle return va CX_REG.
2148 (output_indirect_thunk_function): Create alias for
2149 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2150 (ix86_output_indirect_jmp): Remove the bool argument.
2151 (ix86_output_indirect_function_return): New function.
2152 (ix86_split_simple_return_pop_internal): Likewise.
2153 * config/i386/i386.md (*indirect_jump): Don't pass false
2154 to ix86_output_indirect_jmp.
2155 (*tablejump_1): Likewise.
2156 (simple_return_pop_internal): Change it to define_insn_and_split.
2157 Call ix86_split_simple_return_pop_internal to split it for
2158 -mfunction-return=.
2159 (simple_return_indirect_internal): Call
2160 ix86_output_indirect_function_return instead of
2161 ix86_output_indirect_jmp.
2163 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2165 PR bootstrap/84405
2166 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2167 memset and value initialization afterwards.
2169 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2171 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2173 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2175 PR target/84521
2176 * common/config/aarch64/aarch64-common.c
2177 (aarch_option_optimization_table[]): Switch
2178 off fomit-frame-pointer
2180 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2181 Chung-Ju Wu <jasonwucj@gmail.com>
2183 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2184 volatile memory.
2185 (store_multiple): Ditto.
2187 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2189 * config.gcc: Add --with-cpu support for nds32 target.
2190 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2191 * config/nds32/nds32.opt: Add -mcpu= option.
2193 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2195 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2196 isel=yes): Warn for these deprecated options.
2198 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2200 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2201 ISA_2_5_MASKS_EMBEDDED.
2203 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2205 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2206 p->max as pointers rather than using iterative_hash_expr.
2208 2018-02-23 Carl Love <cel@us.ibm.com>
2210 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2211 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2212 BU_P8V_OVERLOAD_2.
2213 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2214 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2215 P8V_BUILTIN_VEC_VUNSIGNED2.
2217 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2219 PR target/81572
2220 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2221 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2222 LRA_UNKNOWN_ALT.
2223 * lra-constraints.c (curr_insn_transform): Set up
2224 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2225 LRA_UNKNOWN_ALT.
2226 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2227 * lra-eliminations.c (spill_pseudos): Ditto.
2228 (process_insn_for_elimination): Ditto.
2229 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2230 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2231 LRA_NON_CLOBBERED_ALT.
2233 2018-02-22 Martin Sebor <msebor@redhat.com>
2235 PR tree-optimization/84480
2236 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2237 to maybe_diag_stxncpy_trunc. Call it.
2238 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2239 from gimple_fold_builtin_strcpy. Print inlining stack.
2240 (handle_builtin_stxncpy): Print inlining stack.
2241 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2243 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2245 PR target/84176
2246 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2247 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2248 and -fcheck-pointer-bounds are used together.
2249 (indirect_thunk_prefix): New enum.
2250 (indirect_thunk_need_prefix): New function.
2251 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2252 "_nt" instead of "_bnd" for NOTRACK prefix.
2253 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2254 (output_indirect_thunk_function): Likewise.
2255 (): Likewise.
2256 (ix86_code_end): Update output_indirect_thunk_function calls.
2257 (ix86_output_indirect_branch_via_reg): Replace
2258 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2259 (ix86_output_indirect_branch_via_push): Likewise.
2260 (ix86_output_function_return): Likewise.
2261 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2262 incompatible with -fcf-protection=branch and
2263 -fcheck-pointer-bounds.
2265 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2267 PR target/83335
2268 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2269 Change gcc_assert call to output_operand_lossage.
2271 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2273 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2275 2018-02-22 DJ Delorie <dj@redhat.com>
2276 Sebastian Perta <sebastian.perta@renesas.com>
2277 Oleg Endo <olegendo@gcc.gnu.org>
2279 * config/rx/rx.c (rx_rtx_costs): New function.
2280 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2282 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2284 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2286 2018-02-22 Martin Liska <mliska@suse.cz>
2288 PR driver/83193
2289 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2290 Add "native" as a possible value.
2292 2018-02-22 Martin Liska <mliska@suse.cz>
2294 PR driver/83193
2295 * config/i386/i386.c (ix86_option_override_internal):
2296 Add "native" as a possible value for -march and -mtune.
2298 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2300 PR target/84502
2301 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2302 to all type variants.
2304 PR tree-optimization/84503
2305 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2306 width as info->bitpos + info->bitsize - start.
2307 (merged_store_group::merge_overlapping): Simplify width computation.
2308 (check_no_overlap): New function.
2309 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2310 start + width and last_order of the group, fail if check_no_overlap
2311 fails.
2312 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2313 to group if check_no_overlap fails.
2315 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2317 * config/rs6000/altivec.md: Delete contraint arguments to
2318 define_expand, define_split, and define_peephole2, and in
2319 define_insn_and_split if always unused.
2320 * config/rs6000/darwin.md: Ditto.
2321 * config/rs6000/dfp.md: Ditto.
2322 * config/rs6000/rs6000.md: Ditto.
2323 * config/rs6000/sync.md: Ditto.
2324 * config/rs6000/vector.md: Ditto.
2325 * config/rs6000/vsx.md: Ditto.
2327 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2329 * config/rs6000/altivec.md: Write output control strings as braced
2330 blocks instead of double-quoted strings.
2331 * config/rs6000/darwin.md: Ditto.
2332 * config/rs6000/rs6000.md: Ditto.
2333 * config/rs6000/vector.md: Ditto.
2334 * config/rs6000/vsx.md: Ditto.
2336 2018-02-21 Jason Merrill <jason@redhat.com>
2338 PR c++/84314 - ICE with templates and fastcall attribute.
2339 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2341 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2343 * ipa-cp.c (determine_versionability): Fix comment typos.
2345 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2347 PR c/84229
2348 * ipa-cp.c (determine_versionability): Do not version functions caling
2349 va_arg_pack.
2351 2018-02-21 Martin Liska <mliska@suse.cz>
2353 PR driver/83193
2354 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2355 Add "native" as a possible value.
2356 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2357 the macro when native cpu detection is available.
2359 2018-02-21 Martin Liska <mliska@suse.cz>
2361 PR driver/83193
2362 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2363 Add "native" as a possible value.
2364 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2365 when native cpu detection is available.
2367 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2368 Martin Sebor <msebor@redhat.com>
2370 PR tree-optimization/84478
2371 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2372 false.
2373 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2374 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2375 support which is conservatively correct, for 2 only stay conservative
2376 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2377 argument to the 2 argument get_range_strlen, adjust 6 arg
2378 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2379 false.
2380 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2381 (gimple_fold_builtin_strlen): Pass true as last argument to
2382 get_range_strlen.
2384 2018-02-20 Martin Sebor <msebor@redhat.com>
2386 PR middle-end/84095
2387 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2388 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2389 (builtin_memref::builtin_memref): Factor out parts into
2390 set_base_and_offset and call it.
2392 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2394 PR middle-end/84406
2395 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2396 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2397 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2398 search at the associated MODE_INT.
2400 2018-02-20 Jeff Law <law@redhat.com>
2402 PR middle-end/82123
2403 PR tree-optimization/81592
2404 PR middle-end/79257
2405 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2406 for range data rather than using global data.
2407 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2408 range data rather than using global data.
2409 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2410 pass it to children as needed.
2411 (struct directive::fmtresult): Similarly.
2412 (struct directive::set_width): Similarly.
2413 (struct directive::set_precision): Similarly.
2414 (format_integer, format_directive, parse_directive): Similarly.
2415 (format_none): Accept unnamed vr_values parameter.
2416 (format_percent, format_floating, format_character): Similarly.
2417 (format_string, format_plain): Similarly.
2418 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2419 the EVRP range analyzer for range data rather than using global data.
2420 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2421 gimple-ssa-evrp-analyze.h
2422 (class sprintf_dom_walker): Add after_dom_children member function.
2423 Add evrp_range_analyzer member.
2424 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2425 range analyzer as needed.
2426 (sprintf_dom_walker::after_dom_children): New member function.
2427 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2428 if not optimizing.
2429 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2430 (evrp_range_analyzer::pop_to_marker): Likewise.
2432 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2434 PR tree-optimization/84419
2435 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2436 with the required type if its current type is compatible but
2437 different.
2439 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2441 PR middle-end/82004
2442 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2443 after vectorization.
2445 2018-02-20 Martin Liska <mliska@suse.cz>
2447 PR driver/83193
2448 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2449 possible values if we don't have a hint.
2451 2018-02-20 Martin Liska <mliska@suse.cz>
2453 PR c/84310
2454 PR target/79747
2455 * final.c (shorten_branches): Build align_tab array with one
2456 more element.
2457 * opts.c (finish_options): Add alignment option limit check.
2458 (MAX_CODE_ALIGN): Likewise.
2459 (MAX_CODE_ALIGN_VALUE): Likewise.
2460 * doc/invoke.texi: Document maximum allowed option value for
2461 all -falign-* options.
2463 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2465 PR target/84146
2466 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2467 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2468 * var-tracking.c (emit_note_insn_var_location): Remove all references
2469 to NOTE_INSN_CALL_ARG_LOCATION.
2470 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2471 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2472 Use copy_rtx_if_shared.
2473 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2474 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2475 (dwarf2out_var_location): Remove handling of
2476 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2477 on call_insn.
2478 * final.c (final_scan_insn): Remove all references to
2479 NOTE_INSN_CALL_ARG_LOCATION.
2480 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2481 before dumping final insns.
2482 * except.c (emit_note_eh_region_end): Remove all references to
2483 NOTE_INSN_CALL_ARG_LOCATION.
2484 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2485 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2486 * config/arc/arc.c (hwloop_optimize): Likewise.
2487 * config/arm/arm.c (create_fix_barrier): Likewise.
2488 * config/s390/s390.c (s390_chunkify_start): Likewise.
2489 * config/sh/sh.c (find_barrier): Likewise.
2490 * config/i386/i386.c (rest_of_insert_endbranch,
2491 ix86_seh_fixup_eh_fallthru): Likewise.
2492 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2493 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2494 * config/frv/frv.c (frv_function_prologue): Likewise.
2495 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2496 reg note.
2497 (note_outside_basic_block_p): Remove all references to
2498 NOTE_INSN_CALL_ARG_LOCATION.
2499 * gengtype.c (adjust_field_rtx_def): Likewise.
2500 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2501 Likewise.
2502 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2503 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2505 PR c++/84444
2506 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2507 is ADDR_EXPR.
2509 PR tree-optimization/84452
2510 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2511 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2512 is NULL.
2514 2018-02-19 Martin Liska <mliska@suse.cz>
2516 PR sanitizer/82183
2517 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2519 2018-02-19 Martin Liska <mliska@suse.cz>
2520 Richard Sandiford <richard.sandiford@linaro.org>
2522 PR tree-optimization/82491
2523 * gimple-fold.c (get_base_constructor): Make earlier bail out
2524 to prevent ubsan.
2526 2018-02-19 Carl Love <cel@us.ibm.com>
2528 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2529 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2530 BU_P8V_OVERLOAD_1.
2531 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2532 P8V_BUILTIN_VEC_NEG.
2534 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2536 * config/rl78/rl78.md (movdf): New define expand.
2538 2018-02-19 Martin Liska <mliska@suse.cz>
2540 PR other/80589
2541 * doc/invoke.texi: Fix typo.
2542 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2544 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2546 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2547 handle rs6000_single_float and rs6000_double_float specially for
2548 e500 family CPUs.
2550 2018-02-16 Jeff Law <law@redhat.com>
2552 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2553 (pop_regs): Use it.
2555 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2557 PR ipa/84425
2558 * ipa-inline.c (inline_small_functions): Fix a typo.
2560 2018-02-16 Nathan Sidwell <nathan@acm.org>
2562 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2564 2018-02-16 Carl Love <cel@us.ibm.com>
2566 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2567 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2568 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2569 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2570 expansion to P8V_BUILTIN_VEC_FLOAT2.
2572 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2574 PR rtl-optimization/70023
2575 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2576 src_regno into account.
2578 2018-02-16 Carl Love <cel@us.ibm.com>
2580 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2581 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2582 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2583 * config/rs6000/rs6000.c: Remove case statements for
2584 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2585 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2586 and P9V_BUILTIN_VEC_VINSERT4B.
2587 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2588 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2589 * config/rs6000/vsx.md:
2590 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2591 vec_insert4b.
2593 2018-02-16 Carl Love <cel@us.ibm.com>
2595 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2596 vec_insert4b.
2597 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2598 definitions.
2599 * config/rs6000/rs6000-c.c: Add the definitions for
2600 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2601 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2602 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2603 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2604 definition for insert4b and define insn *insert3b_internal.
2605 * doc/extend.texi: Add documentation for vec_extract4b.
2607 2018-02-16 Nathan Sidwell <nathan@acm.org>
2609 * doc/extend.texi (Backwards Compatibility): Mention friend
2610 injection. Note for-scope is deprecated.
2611 * doc/invoke.texi (-ffriend-injection): Deprecate.
2613 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2615 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2616 that moved to I2, also allow destinations that are a paradoxical
2617 subreg (instead of a normal reg).
2619 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2621 PR target/83831
2622 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2623 to QImode.
2625 2018-02-16 Richard Biener <rguenther@suse.de>
2627 PR tree-optimization/84037
2628 PR tree-optimization/84016
2629 PR target/82862
2630 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2631 Adjust vec_construct for the fact we need additional higher latency
2632 128bit inserts for AVX256 and AVX512 vector builds.
2633 (ix86_add_stmt_cost): Scale vector construction cost for
2634 elementwise loads.
2636 2018-02-16 Richard Biener <rguenther@suse.de>
2638 PR tree-optimization/84417
2639 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2640 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2641 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2643 2018-02-16 Martin Liska <mliska@suse.cz>
2645 PR sanitizer/84307
2646 * internal-fn.def (ASAN_CHECK): Set proper flags.
2647 (ASAN_MARK): Likewise.
2649 2018-02-16 Julia Koval <julia.koval@intel.com>
2651 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2652 from PTA_CANNONLAKE.
2654 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2656 PR target/84272
2657 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2658 Use ++iter rather than iter++ for std::list iterators.
2659 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2660 defer deleting them until all nodes in the forest are processed. Do
2661 free even leaf nodes. Change to_process into auto_vec.
2663 PR bootstrap/84405
2664 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2665 * vec.h (vec_default_construct): Use memset instead of placement new
2666 if BROKEN_VALUE_INITIALIZATION is defined.
2667 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2668 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2669 is defined.
2671 PR rtl-optimization/83723
2672 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2673 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2674 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2675 recursive calls.
2676 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2677 callers.
2678 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2680 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2682 PR rtl-optimization/81443
2683 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2684 from inner REGs to paradoxical SUBREGs.
2686 2018-02-16 Richard Biener <rguenther@suse.de>
2688 PR tree-optimization/84399
2689 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2690 For operands we can analyze at their definition make sure we can
2691 analyze them at each use as well.
2693 2018-02-16 Richard Biener <rguenther@suse.de>
2695 PR tree-optimization/84190
2696 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2697 volatile accesses if the decl isn't volatile.
2699 2018-02-15 Jason Merrill <jason@redhat.com>
2701 PR c++/84314 - ICE with templates and fastcall attribute.
2702 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2703 TYPE_CANONICAL on an existing type.
2705 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2707 PR tree-optimization/84383
2708 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2709 dstoff nor call operand_equal_p if dstbase is NULL.
2711 PR tree-optimization/84334
2712 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2713 also a CONSTANT_CLASS_P, punt.
2715 2018-02-14 Jim Wilson <jimw@sifive.com>
2717 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2718 first SMALL_OPERAND check. New local min_second_step. Move assert
2719 to where locals are set. Add TARGET_RVC support.
2720 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2722 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2724 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2726 2018-02-14 Martin Sebor <msebor@redhat.com>
2728 PR tree-optimization/83698
2729 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2730 arrays constrain the offset range to their bounds.
2731 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2732 (builtin_access::overlap): Avoid setting the size of overlap if it's
2733 already been set.
2734 (maybe_diag_overlap): Also consider arrays when deciding what values
2735 of offsets to include in diagnostics.
2737 2018-02-14 Martin Sebor <msebor@redhat.com>
2739 PR c/84108
2740 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2741 that correspond to the kind of a declaration.
2743 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2745 PR target/83984
2746 * config/pa/pa.md: Load address of PIC label using the linkage table
2747 if the label is nonlocal.
2749 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2751 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2752 warning message if user requests -maltivec=be.
2753 * doc/invoke.texi: Document deprecation of -maltivec=be.
2755 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2757 PR target/84220
2758 * config/rs6000/rs6000-c.c: Update definitions for
2759 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2760 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2762 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2764 PR target/84239
2765 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2766 add _get_ssp intrinsics. Remove argument from
2767 __builtin_ia32_rdssp[d|q].
2768 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2769 * config/i386/i386-builtin.def: Remove argument from
2770 __builtin_ia32_rdssp[d|q].
2771 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2772 ix86_expand_special_args_builtin for _rdssp[d|q].
2773 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2774 Clear register before usage.
2775 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2776 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2778 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2780 PR tree-optimization/84357
2781 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2782 operand 1 of an ARRAY_REF too.
2784 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2786 PR target/83831
2787 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2788 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2789 declarations.
2790 (set_of_reg): New struct.
2791 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2792 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2793 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2794 functions.
2795 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2796 Split into bitclr, bitset, bitinvert patterns if appropriate.
2797 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2798 use rx_fuse_in_memory_bitop.
2799 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2800 to named insn, correct maximum insn length.
2802 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2804 PR target/79242
2805 * machmode.def: Define a complex mode for PARTIAL_INT.
2806 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2807 MODE_PARTIAL_INT.
2808 * doc/rtl.texi: Document CSPImode.
2809 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2810 handling.
2811 (msp430_hard_regno_nregs_with_padding): Likewise.
2813 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2815 PR target/84279
2816 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2818 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2820 PR rtl-optimization/84169
2821 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2822 we generated a parallel as new i3 and we split that to new i2 and i3
2823 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2824 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2825 those to i2, not i1. Partially rewrite this scan code.
2827 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2829 PR c/82210
2830 * stor-layout.c (place_field): For variable length fields, adjust
2831 offset_align afterwards not just based on the field's alignment,
2832 but also on the size.
2834 PR middle-end/84309
2835 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2836 of exps and logs in the use_exp2 case.
2838 2018-02-13 Jeff Law <law@redhat.com>
2840 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2841 entry for "vector".
2843 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2844 ARGS as unused.
2846 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2848 PR debug/84342
2849 PR debug/84319
2850 * common.opt (gas-loc-support, gas-locview-support): New.
2851 (ginline-points, ginternal-reset-location-views): New.
2852 * doc/invoke.texi: Document them. Use @itemx where intended.
2853 (gvariable-location-views): Adjust.
2854 * target.def (reset_location_view): New.
2855 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2856 (TARGET_RESET_LOCATION_VIEW): New.
2857 * doc/tm.texi: Rebuilt.
2858 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2859 (dwarf2out_default_as_locview_support): New.
2860 (output_asm_line_debug_info): Use option variables.
2861 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2862 (output_loc_list): Likewise.
2863 (add_high_low_attributes): Check option variables.
2864 Don't output entry view attribute in strict mode.
2865 (gen_inlined_subroutine_die): Check option variables.
2866 (dwarf2out_inline_entry): Likewise.
2867 (init_sections_and_labels): Likewise.
2868 (dwarf2out_early_finish): Likewise.
2869 (maybe_reset_location_view): New, from...
2870 (dwarf2out_var_location): ... here. Call it.
2871 * debug.h (dwarf2out_default_as_loc_support): Declare.
2872 (dwarf2out_default_as_locview_support): Declare.
2873 * hooks.c (hook_int_rtx_insn_0): New.
2874 * hooks.h (hook_int_rtx_insn_0): Declare.
2875 * toplev.c (process_options): Take -gas-loc-support and
2876 -gas-locview-support from dwarf2out. Enable
2877 -gvariable-location-views by default only with locview
2878 assembler support. Enable -ginternal-reset-location-views by
2879 default only if the target defines the corresponding hook.
2880 Enable -ginline-points by default if location views are
2881 enabled; force it disabled if statement frontiers are
2882 disabled.
2883 * tree-inline.c (expand_call_inline): Check option variables.
2884 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2886 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2888 PR tree-optimization/84321
2889 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2890 handling. Also check whether the anti-range contains any values
2891 that satisfy the mask; switch to a VR_RANGE if not.
2893 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2895 PR sanitizer/84340
2896 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2898 2018-02-13 Martin Jambor <mjambor@suse.cz>
2900 PR c++/83990
2901 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2902 of call statements, also set location of a load to a temporary.
2904 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2906 * config/rl78/rl78.c (add_vector_labels): New function.
2907 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2908 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2909 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2910 which checks that no arguments are passed.
2911 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2912 * doc/extend.texi: Documentation for the new attribute.
2914 2018-02-13 Andreas Schwab <schwab@suse.de>
2916 * config/riscv/linux.h (CPP_SPEC): Define.
2918 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2920 PR target/84335
2921 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2922 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2923 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2924 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2925 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2926 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2927 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2928 temporarily for AES and PCLMUL builtins.
2930 PR tree-optimization/84339
2931 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2932 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2933 Formatting fixes.
2935 PR middle-end/84309
2936 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2937 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2938 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2939 inline function.
2940 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2941 inline function.
2942 * omp-simd-clone.h: New file.
2943 * omp-simd-clone.c: Include omp-simd-clone.h.
2944 (expand_simd_clones): No longer static.
2945 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2946 cgraph.h and omp-simd-clone.h.
2947 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2948 (vect_recog_widen_shift_pattern): Formatting fix.
2949 (vect_pattern_recog_1): Don't check optab for calls.
2951 PR target/84336
2952 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2953 operands[2] into a REG before using gen_lowpart on it.
2955 2018-02-12 Jeff Law <law@redhat.com>
2957 PR target/83760
2958 * config/sh/sh.c (find_barrier): Consider a sibling call
2959 a barrier as well.
2961 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2962 successfully back substituting a reg.
2964 2018-02-12 Richard Biener <rguenther@suse.de>
2966 PR tree-optimization/84037
2967 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2968 parameter, move visited init to caller.
2969 (vect_slp_analyze_operations): Separate cost from validity
2970 check, initialize visited once for all instances.
2971 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2972 for all instances.
2973 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2974 out an assert.
2975 (vect_model_promotion_demotion_cost): Likewise.
2976 (vectorizable_bswap): Guard cost modeling with !slp_node
2977 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2978 SLP stmts.
2979 (vectorizable_call): Likewise.
2980 (vectorizable_conversion): Likewise.
2981 (vectorizable_assignment): Likewise.
2982 (vectorizable_shift): Likewise.
2983 (vectorizable_operation): Likewise.
2984 (vectorizable_store): Likewise.
2985 (vectorizable_load): Likewise.
2986 (vectorizable_condition): Likewise.
2987 (vectorizable_comparison): Likewise.
2989 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2991 PR sanitizer/84307
2992 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2993 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2994 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2996 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2998 PR middle-end/83665
2999 * params.def (inline-min-speedup): Increase from 8 to 15.
3000 (max-inline-insns-auto): Decrease from 40 to 30.
3001 * ipa-split.c (consider_split): Add some buffer for function to
3002 be considered inlining candidate.
3003 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3004 default values.
3006 2018-02-12 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/84037
3009 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3010 matched stmts if we cannot swap the non-matched ones.
3012 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3014 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3015 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3016 _mm_maskz_scalef_round_ss): New intrinsics.
3017 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3018 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3019 __builtin_ia32_scalefss_round): Remove.
3020 (__builtin_ia32_scalefsd_mask_round,
3021 __builtin_ia32_scalefss_mask_round): New intrinsics.
3022 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3023 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3024 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3025 "<round_constraint>")): Changed to ...
3026 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3027 "<round_scalar_constraint>")): ... this.
3028 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3029 %0, %1, %2<round_op3>}"): Changed to ...
3030 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3031 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3032 %2<round_scalar_mask_op3>}"): ... this.
3033 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3035 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3037 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3038 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3039 (_mm_maskz_sqrt_round_ss): New intrinsics.
3040 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3041 (__builtin_ia32_sqrtsd_mask_round)
3042 (__builtin_ia32_sqrtss_mask_round): New builtins.
3043 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3044 (__builtin_ia32_sqrtss_round): Remove.
3045 (__builtin_ia32_sqrtsd_mask_round)
3046 (__builtin_ia32_sqrtss_mask_round): New builtins.
3047 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3048 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3049 ((match_operand:VF_128 1 "vector_operand"
3050 "xBm,<round_constraint>")): Changed to ...
3051 ((match_operand:VF_128 1 "vector_operand"
3052 "xBm,<round_scalar_constraint>")): ... this.
3053 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3054 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3055 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3056 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3057 %<iptr>1<round_scalar_mask_op3>}): ... this.
3058 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3059 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3061 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3063 PR target/84266
3064 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3065 Cast vec_cmpeq result to correct type.
3066 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3067 Cast vec_cmpgt result to correct type.
3069 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3071 * final.c (final_scan_insn_1): Renamed from...
3072 (final_scan_insn): ... this. New wrapper, to recover
3073 seen from the outermost call in recursive ones.
3074 * config/sparc/sparc.c (output_return): Drop seen from call.
3075 (output_sibcall): Likewise.
3076 * config/visium/visium.c (output_branch): Likewise.
3078 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3080 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3081 function label.
3083 2018-02-10 Alan Modra <amodra@gmail.com>
3085 PR target/84300
3086 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3087 Specify LR as an input.
3089 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3091 PR sanitizer/83987
3092 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3093 remove_member_access_dummy_vars): New functions.
3094 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3095 lower_omp_1, execute_lower_omp): Use them.
3097 PR rtl-optimization/84308
3098 * shrink-wrap.c (spread_components): Release todo vector.
3100 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3102 PR rtl-optimization/57193
3103 * ira-color.c (struct allocno_color_data): Add member
3104 conflict_allocno_hard_prefs.
3105 (update_conflict_allocno_hard_prefs): New.
3106 (bucket_allocno_compare_func): Add a preference based on
3107 conflict_allocno_hard_prefs.
3108 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3109 (color_allocnos): Remove a dead code. Initiate
3110 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3112 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3114 PR target/84226
3115 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3116 constraint from =wa to wa. Avoid a subreg on the output operand,
3117 instead use a pseudo and subreg it in a move.
3118 (p9_xxbrd_<mode>): Changed to ...
3119 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3120 (p9_xxbrd_v2df): New expander.
3121 (p9_xxbrw_<mode>): Changed to ...
3122 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3123 (p9_xxbrw_v4sf): New expander.
3125 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3127 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3129 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3131 PR target/83926
3132 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3133 multiply in 32-bit mode.
3134 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3135 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3136 mode.
3138 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3140 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3141 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3142 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3143 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3145 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3147 PR lto/84213
3148 * dwarf2out.c (is_trivial_indirect_ref): New function.
3149 (dwarf2out_late_global_decl): Do not generate a location
3150 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3151 and that are not defined in the current unit.
3153 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3155 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3156 instead of a libcall for UNORDERED.
3158 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3160 PR target/82641
3161 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3162 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3164 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3166 PR target/PR84295
3167 * config/s390/s390.c (s390_set_current_function): Invoke
3168 s390_indirect_branch_settings also if fndecl didn't change.
3170 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3172 * config/rs6000/rs6000.md (blockage): Set length to zero.
3174 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3176 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3178 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3180 PR sanitizer/84285
3181 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3182 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3183 -static-lib*san.
3185 PR debug/84252
3186 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3187 PARALLEL incoming that failed vt_get_decl_and_offset check.
3189 PR middle-end/84237
3190 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3191 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3192 TREE_READONLY bit.
3193 (get_variable_section): For decls in named .bss* sections pass true as
3194 second argument to bss_initializer_p.
3196 2018-02-09 Marek Polacek <polacek@redhat.com>
3197 Jakub Jelinek <jakub@redhat.com>
3199 PR c++/83659
3200 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3201 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3202 Sync some changes from cxx_fold_indirect_ref.
3204 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3206 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3207 markers.
3208 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3209 (BLOCK_INLINE_ENTRY_LABEL): New.
3210 (dwarf2out_var_location): Disregard inline entry markers.
3211 (inline_entry_data): New struct.
3212 (inline_entry_data_hasher): New hashtable type.
3213 (inline_entry_data_hasher::hash): New.
3214 (inline_entry_data_hasher::equal): New.
3215 (inline_entry_data_table): New variable.
3216 (add_high_low_attributes): Add DW_AT_entry_pc and
3217 DW_AT_GNU_entry_view attributes if a pending entry is found
3218 in inline_entry_data_table. Add old entry_pc attribute only
3219 if debug nonbinding markers are disabled.
3220 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3221 markers are enabled.
3222 (block_within_block_p, dwarf2out_inline_entry): New.
3223 (dwarf2out_finish): Check that no entries remained in
3224 inline_entry_data_table.
3225 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3226 (final_scan_insn, notice_source_line): Likewise.
3227 (rest_of_clean_state): Skip inline entry markers.
3228 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3229 markers.
3230 * gimple.c (gimple_build_debug_inline_entry): New.
3231 * gimple.h (enum gimple_debug_subcode): Add
3232 GIMPLE_DEBUG_INLINE_ENTRY.
3233 (gimple_build_debug_inline_entry): Declare.
3234 (gimple_debug_inline_entry_p): New.
3235 (gimple_debug_nonbind_marker_p): Adjust.
3236 * insn-notes.def (INLINE_ENTRY): New.
3237 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3238 inline entry marker notes.
3239 (print_insn): Likewise.
3240 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3241 (INSN_DEBUG_MARKER_KIND): Likewise.
3242 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3243 * tree-inline.c (expand_call_inline): Build and insert
3244 debug_inline_entry stmt.
3245 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3246 inline entry blocks early, if nonbind markers are enabled.
3247 (dump_scope_block): Dump fragment info.
3248 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3249 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3250 (gimple_build_debug_inline_entry): New.
3251 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3252 Enable/disable inline entry points too.
3253 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3254 (DEBUG_INSN): Describe inline entry markers.
3256 * common.opt (gvariable-location-views): New.
3257 (gvariable-location-views=incompat5): New.
3258 * config.in: Rebuilt.
3259 * configure: Rebuilt.
3260 * configure.ac: Test assembler for view support.
3261 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3262 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3263 * dwarf2out.c (var_loc_view): New typedef.
3264 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3265 (dwarf2out_locviews_in_attribute): New.
3266 (dwarf2out_locviews_in_loclist): New.
3267 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3268 (enum dw_line_info_opcode): Add LI_adv_address.
3269 (struct dw_line_info_table): Add view.
3270 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3271 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3272 (zero_view_p): New variable.
3273 (ZERO_VIEW_P): New macro.
3274 (output_asm_line_debug_info): New.
3275 (struct var_loc_node): Add view.
3276 (add_AT_view_list, AT_loc_list): New.
3277 (add_var_loc_to_decl): Add view param. Test it against last.
3278 (new_loc_list): Add view params. Record them.
3279 (AT_loc_list_ptr): Handle loc and view lists.
3280 (view_list_to_loc_list_val_node): New.
3281 (print_dw_val): Handle dw_val_class_view_list.
3282 (size_of_die): Likewise.
3283 (value_format): Likewise.
3284 (loc_list_has_views): New.
3285 (gen_llsym): Set vl_symbol too.
3286 (maybe_gen_llsym, skip_loc_list_entry): New.
3287 (dwarf2out_maybe_output_loclist_view_pair): New.
3288 (output_loc_list): Output view list or entries too.
3289 (output_view_list_offset): New.
3290 (output_die): Handle dw_val_class_view_list.
3291 (output_dwarf_version): New.
3292 (output_compilation_unit_header): Use it.
3293 (output_skeleton_debug_sections): Likewise.
3294 (output_rnglists, output_line_info): Likewise.
3295 (output_pubnames, output_aranges): Update version comments.
3296 (output_one_line_info_table): Output view numbers in asm comments.
3297 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3298 Call maybe_gen_llsym.
3299 (loc_list_from_tree_1): Adjust.
3300 (add_AT_location_description): Create view list attribute if
3301 needed, check it's absent otherwise.
3302 (convert_cfa_to_fb_loc_list): Adjust.
3303 (maybe_emit_file): Call output_asm_line_debug_info for test.
3304 (dwarf2out_var_location): Reset views as needed. Precompute
3305 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3306 attribute. Set view.
3307 (new_line_info_table): Reset next view.
3308 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3309 (dwarf2out_source_line): Likewise. Output view resets and labels to
3310 the assembler, or select appropriate line info opcodes.
3311 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3312 (optimize_string_length): Catch it. Adjust.
3313 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3314 dw_val_class_view_list, and remove it if no longer needed.
3315 (hash_loc_list): Hash view numbers.
3316 (loc_list_hasher::equal): Compare them.
3317 (optimize_location_lists): Check whether a view list symbol is
3318 needed, and whether the locview attribute is present, and
3319 whether they match. Remove the locview attribute if no longer
3320 needed.
3321 (index_location_lists): Call skip_loc_list_entry for test.
3322 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3323 Use output_dwarf_version.
3324 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3325 (struct dw_val_node): Add val_view_list.
3326 * final.c (SEEN_NEXT_VIEW): New.
3327 (set_next_view_needed): New.
3328 (clear_next_view_needed): New.
3329 (maybe_output_next_view): New.
3330 (final_start_function): Rename to...
3331 (final_start_function_1): ... this. Take pointer to FIRST,
3332 add SEEN parameter. Emit param bindings in the initial view.
3333 (final_start_function): Reintroduce SEEN-less interface.
3334 (final): Rename to...
3335 (final_1): ... this. Take SEEN parameter. Output final pending
3336 next view at the end.
3337 (final): Reintroduce seen-less interface.
3338 (final_scan_insn): Output pending next view before switching
3339 sections or ending a block. Mark the next view as needed when
3340 outputting variable locations. Notify debug backend of section
3341 changes, and of location view changes.
3342 (rest_of_handle_final): Adjust.
3343 * toplev.c (process_options): Autodetect value for debug variable
3344 location views option. Warn on incompat5 without -gdwarf-5.
3345 * doc/invoke.texi (gvariable-location-views): New.
3346 (gvariable-location-views=incompat5): New.
3347 (gno-variable-location-views): New.
3349 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3351 PR tree-optimization/84136
3352 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3353 that the result of find_edge is non-NULL.
3355 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3357 PR target/83008
3358 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3359 storing integer register in SImode. Fix cost of 256 and 512
3360 byte aligned SSE register store.
3362 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3364 * config/i386/i386.c (ix86_multiplication_cost): Fix
3365 multiplication cost for TARGET_AVX512DQ.
3367 2018-02-08 Marek Polacek <polacek@redhat.com>
3369 PR tree-optimization/84238
3370 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3371 get_range_strlen.
3373 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3375 PR tree-optimization/84265
3376 * tree-vect-stmts.c (vectorizable_store): Don't treat
3377 VMAT_CONTIGUOUS accesses as grouped.
3378 (vectorizable_load): Likewise.
3380 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3382 PR tree-optimization/81635
3383 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3384 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3385 (test_round_for_mask): New functions.
3386 (wide_int_cc_tests): Call test_round_for_mask.
3387 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3388 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3389 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3390 range returned by get_range_info.
3392 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3394 PR ipa/81360
3395 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3396 * symtab.c: Include builtins.h
3397 (symtab_node::output_to_lto_symbol_table_p): Move here
3398 from lto-streamer-out.c:output_symbol_p.
3399 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3400 (output_symbol_p): Move all logic to symtab.c
3401 (produce_symtab): Update.
3403 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3405 * config/s390/s390-opts.h (enum indirect_branch): Define.
3406 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3407 (s390_indirect_branch_via_thunk)
3408 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3409 (enum s390_indirect_branch_type): Define.
3410 * config/s390/s390.c (struct s390_frame_layout, struct
3411 machine_function): Remove.
3412 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3413 (indirect_branch_table_label_no, indirect_branch_table_name):
3414 Define variables.
3415 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3416 (enum s390_indirect_branch_option): Define.
3417 (s390_return_addr_from_memory): New function.
3418 (s390_handle_string_attribute): New function.
3419 (s390_attribute_table): Add new attribute handler.
3420 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3421 (s390_indirect_branch_via_thunk): New function.
3422 (s390_indirect_branch_via_inline_thunk): New function.
3423 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3424 sibling call optimization for non z10 compiles.
3425 (s390_emit_call): Force indirect branch target to be a single
3426 register. Add r1 clobber for non-z10 compiles.
3427 (s390_emit_epilogue): Emit return jump via return_use expander.
3428 (s390_reorg): Handle JUMP_INSNs as execute targets.
3429 (s390_option_override_internal): Perform validity checks for the
3430 new command line options.
3431 (s390_indirect_branch_attrvalue): New function.
3432 (s390_indirect_branch_settings): New function.
3433 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3434 (s390_output_indirect_thunk_function): New function.
3435 (s390_code_end): Implement target hook.
3436 (s390_case_values_threshold): Implement target hook.
3437 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3438 macros.
3439 * config/s390/s390.h (struct s390_frame_layout)
3440 (struct machine_function): Move here from s390.c.
3441 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3442 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3443 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3444 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3445 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3446 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3447 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3448 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3449 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3450 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3451 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3452 (mnemonic attribute): Add values which aren't recognized
3453 automatically.
3454 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3455 pattern for branch conversion. Fix mnemonic attribute.
3456 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3457 indirect branch via thunk if requested.
3458 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3459 ("*indirect_jump"): Disable for branch conversion using out of
3460 line thunks.
3461 ("indirect_jump_via_thunk<mode>_z10")
3462 ("indirect_jump_via_thunk<mode>")
3463 ("indirect_jump_via_inlinethunk<mode>_z10")
3464 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3465 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3466 ("casesi_jump_via_inlinethunk<mode>_z10")
3467 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3468 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3469 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3470 ("*indirect2_jump"): Disable for branch conversion.
3471 ("casesi_jump"): Turn into expander and expand patterns for branch
3472 conversion.
3473 ("return_use"): New expander.
3474 ("*return"): Emit return via thunk and rename it to ...
3475 ("*return<mode>"): ... this one.
3476 * config/s390/s390.opt: Add new options and and enum for the
3477 option values.
3479 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3481 * lra-constraints.c (match_reload): Unconditionally use
3482 gen_lowpart_SUBREG, rather than selecting between that
3483 and equivalent gen_rtx_SUBREG code.
3485 2018-02-08 Richard Biener <rguenther@suse.de>
3487 PR tree-optimization/84233
3488 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3489 changed flag instead of boguously re-using phi_inserted.
3491 2018-02-08 Martin Jambor <mjambor@suse.cz>
3493 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3494 static local variables.
3496 2018-02-08 Richard Biener <rguenther@suse.de>
3498 PR tree-optimization/84278
3499 * tree-vect-stmts.c (vectorizable_store): When looking for
3500 smaller vector types to perform grouped strided loads/stores
3501 make sure the mode is supported by the target.
3502 (vectorizable_load): Likewise.
3504 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3506 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3507 Increase LDP/STP opportunities by adding adjacent callee-saves.
3509 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3511 PR rtl-optimization/84068
3512 PR rtl-optimization/83459
3513 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3515 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3517 PR tree-optimization/84224
3518 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3519 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3520 non-zero arguments.
3522 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3524 PR target/84113
3525 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3526 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3527 count, remove one USE.
3529 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3531 * doc/install.texi (Configuration): Document the
3532 --with-long-double-format={ibm,ieee} PowerPC configuration
3533 options.
3535 PR target/84154
3536 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3537 Convert from define_expand to be define_insn_and_split. Rework
3538 float/double/_Float128 conversions to QI/HI/SImode to work with
3539 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3540 conversions to QI/HImode types did a store and then a load to
3541 truncate the value. For conversions to VSX registers, don't split
3542 the insn, instead emit the code directly. Use the code iterator
3543 any_fix to combine signed and unsigned conversions.
3544 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3545 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3546 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3547 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3548 (fix_<mode>di2_hw): Likewise.
3549 (fixuns_<mode>di2_hw): Likewise.
3550 (fix_<mode>si2_hw): Likewise.
3551 (fixuns_<mode>si2_hw): Likewise.
3552 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3553 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3554 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3555 fix<uns>_trunc<SFDF:mode>si2_p8.
3556 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3557 used.
3558 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3559 (fix<uns>_<mode>_mem): Likewise.
3560 (fctiw<u>z_<mode>_mem): Likewise.
3561 (fix<uns>_<mode>_mem): Likewise.
3562 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3563 the register allocator from doing a direct move to the GPRs to do
3564 a store, and instead use the ISA 3.0 store byte/half-word from
3565 vector register instruction. For IEEE 128-bit floating point,
3566 also optimize stores of 32-bit ints.
3567 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3569 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3571 * genextract.c (push_pathstr_operand): New function to support
3572 [a-zA-Z].
3573 (walk_rtx): Call push_pathstr_operand.
3574 (print_path): Support [a-zA-Z].
3576 2018-02-07 Richard Biener <rguenther@suse.de>
3578 PR tree-optimization/84037
3579 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3580 (cse_and_gimplify_to_preheader): Declare.
3581 (vect_get_place_in_interleaving_chain): Likewise.
3582 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3583 ivexpr_map.
3584 (_loop_vec_info::~_loop_vec_info): Delete it.
3585 (cse_and_gimplify_to_preheader): New function.
3586 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3587 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3588 (vectorizable_load): Likewise. For grouped stores always base
3589 the IV on the first element.
3590 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3591 condition before gimplifying.
3593 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3595 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3596 *DIV_EXPR and *MOD_EXPR.
3598 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3600 PR target/84248
3601 * config/i386/i386.c (ix86_option_override_internal): Mask out
3602 the CF_SET bit when checking -fcf-protection.
3604 2018-02-07 Tom de Vries <tom@codesourcery.com>
3606 PR libgomp/84217
3607 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3608 enough.
3610 2018-02-07 Richard Biener <rguenther@suse.de>
3612 PR tree-optimization/84204
3613 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3614 this place.
3616 PR tree-optimization/84205
3617 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3618 special-case isl_ast_op_zdiv_r.
3620 PR tree-optimization/84223
3621 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3622 Only add conditions from within the region.
3623 (gather_bbs::after_dom_children): Adjust.
3625 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3627 PR target/84209
3628 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3629 * config/avr/avr.md: Only post-reload split REG-REG moves if
3630 either register is GENERAL_REG_P.
3632 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3634 PR tree-optimization/84235
3635 * tree-ssa-scopedtables.c
3636 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3637 if the subtraction is performed in floating point type where NaNs are
3638 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3639 build 1. Formatting fix.
3641 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3643 PR target/84146
3644 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3645 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3646 and skip it regardless of bb boundaries. Use CALL_P macro,
3647 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3648 unnecessarily, formatting fix.
3650 2018-02-06 Michael Collison <michael.collison@arm.com>
3652 * config/arm/thumb2.md:
3653 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3654 (*thumb_mov_notscc): Ditto.
3656 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3658 PR target/84154
3659 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3660 unsigned_fix, not "s".
3662 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3664 * configure.ac (gcc_fn_eh_frame_ro): New function.
3665 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3666 correct .eh_frame permissions.
3667 * configure: Regenerate.
3669 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3671 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3672 irrelevant options.
3674 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3676 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3677 Display warning message for -mno-speculate-indirect-jumps.
3679 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3681 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3682 Undocumented.
3683 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3685 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3687 PR tree-optimization/84225
3688 * tree-eh.c (find_trapping_overflow): Only call
3689 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3691 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3693 PR target/84145
3694 * config/i386/i386.c: Reimplement the check of possible options
3695 -mibt/-mshstk conbination. Change error messages.
3696 * doc/invoke.texi: Fix a typo: remove extra '='.
3698 2018-02-06 Marek Polacek <polacek@redhat.com>
3700 PR tree-optimization/84228
3701 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3703 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3705 PR target/82641
3706 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3707 emitted arch directives.
3708 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3709 __ARM_FEATURE_COPROC before changing architectures.
3711 2018-02-06 Richard Biener <rguenther@suse.de>
3713 * config/i386/i386.c (print_reg): Fix typo.
3714 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3716 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3718 * configure: Regenerate.
3720 2018-02-05 Martin Sebor <msebor@redhat.com>
3722 PR tree-optimization/83369
3723 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3724 inlining context.
3726 2018-02-05 Martin Liska <mliska@suse.cz>
3728 * doc/invoke.texi: Cherry-pick upstream r323995.
3730 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3732 * ira.c (ira_init_register_move_cost): Adjust comment.
3734 2018-02-05 Martin Liska <mliska@suse.cz>
3736 PR gcov-profile/84137
3737 * doc/gcov.texi: Fix typo in documentation.
3739 2018-02-05 Martin Liska <mliska@suse.cz>
3741 PR gcov-profile/83879
3742 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3743 using dlopen functionality.
3745 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3747 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3748 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3749 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3750 _mm_maskz_range_round_ss): New intrinsics.
3751 (__builtin_ia32_rangesd128_round)
3752 (__builtin_ia32_rangess128_round): Remove.
3753 (__builtin_ia32_rangesd128_mask_round,
3754 __builtin_ia32_rangess128_mask_round): New builtins.
3755 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3756 __builtin_ia32_rangess128_round): Remove.
3757 (__builtin_ia32_rangesd128_mask_round,
3758 __builtin_ia32_rangess128_mask_round): New builtins.
3759 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3760 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3761 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3762 "<round_saeonly_constraint>")): Changed to ...
3763 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3764 "<round_saeonly_scalar_constraint>")): ... this.
3765 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3766 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3767 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3768 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3769 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3771 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3773 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3774 options.
3775 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3776 Remove all values except native, 8540 and 8548.
3778 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3780 * config/i386/i386.c (ix86_output_function_return): Pass
3781 INVALID_REGNUM, instead of -1, as invalid register number to
3782 indirect_thunk_name and output_indirect_thunk.
3784 2018-02-02 Julia Koval <julia.koval@intel.com>
3786 * config.gcc: Add -march=icelake.
3787 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3788 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3789 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3790 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3791 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3792 (processor_target_table): Add icelake.
3793 (ix86_option_override_internal): Handle new PTAs.
3794 (get_builtin_code_for_version): Handle icelake.
3795 (M_INTEL_COREI7_ICELAKE): New.
3796 (fold_builtin_cpu): Handle icelake.
3797 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3798 * doc/invoke.texi: Add -march=icelake.
3800 2018-02-02 Julia Koval <julia.koval@intel.com>
3802 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3803 to wide_int_bitmask.
3804 * wide-int-bitmask.h: New.
3806 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3808 PR target/84066
3809 * config/i386/i386.md: Replace Pmode with word_mode in
3810 builtin_setjmp_setup and builtin_longjmp to support x32.
3812 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3814 PR target/56010
3815 PR target/83743
3816 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3817 #include "opts.h".
3818 (rs6000_supported_cpu_names): New static variable.
3819 (linux_cpu_translation_table): Likewise.
3820 (elf_platform) <cpu>: Define new static variable and use it.
3821 Translate kernel AT_PLATFORM name to canonical name if needed.
3822 Error if platform name is unknown.
3824 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3826 PR target/84089
3827 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3829 2018-02-01 Jeff Law <law@redhat.com>
3831 PR target/84128
3832 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3833 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3834 the scratch if RELEASE_VIA_POP is false.
3835 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3836 If we have to save a temporary register, decrement SIZE appropriately.
3837 Pass new arguments to release_scratch_register_on_entry.
3838 (ix86_adjust_stack_and_probe): Likewise.
3839 (ix86_emit_probe_stack_range): Pass new arguments to
3840 release_scratch_register_on_entry.
3842 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3844 PR rtl-optimization/84157
3845 * combine.c (change_zero_ext): Use REG_P predicate in
3846 front of HARD_REGISTER_P predicate.
3848 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3850 * config/avr/avr.c (avr_option_override): Move disabling of
3851 -fdelete-null-pointer-checks to...
3852 * common/config/avr/avr-common.c (avr_option_optimization_table):
3853 ...here.
3855 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3857 PR tree-optimization/81635
3858 * tree-data-ref.c (split_constant_offset_1): For types that
3859 wrap on overflow, try to use range info to prove that wrapping
3860 cannot occur.
3862 2018-02-01 Renlin Li <renlin.li@arm.com>
3864 PR target/83370
3865 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3866 TAILCALL_ADDR_REGS.
3867 (aarch64_register_move_cost): Likewise.
3868 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3869 TAILCALL_ADDR_REGS.
3870 (REG_CLASS_NAMES): Likewise.
3871 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3872 TAILCALL_ADDR_REGS. Remove IP registers.
3873 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3875 2018-02-01 Richard Biener <rguenther@suse.de>
3877 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3878 for specifying RPO order and allow NULL for that.
3879 * domwalk.c (dom_walker::dom_walker): Likewise.
3880 (dom_walker::walk): Handle NULL RPO order.
3881 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3882 in RPO order.
3883 (rewrite_update_dom_walker): Likewise.
3884 (mark_def_dom_walker): Likewise.
3886 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3888 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3889 (aarch64_maybe_expand_sve_subreg_move): Declare.
3890 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3891 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3892 predicate.
3893 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3894 that are semantically a reverse operation.
3895 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3896 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3897 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3898 functions.
3899 (aarch64_can_change_mode_class): For big-endian, forbid changes
3900 between two SVE modes if they have different element sizes.
3902 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3904 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3905 the TImode handling for big-endian targets.
3907 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3909 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3910 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3911 not just bytes.
3912 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3913 Remove BSWAP handing for big-endian targets and use the form of
3914 LD1RQ appropariate for the mode.
3916 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3918 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3919 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3920 duplicated element.
3922 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3924 PR tearget/83845
3925 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3926 check for operands that need to go through aarch64_sve_reload_be.
3928 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3930 PR tree-optimization/81661
3931 PR tree-optimization/84117
3932 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3933 * tree-eh.c: Include gimplify.h.
3934 (find_trapping_overflow, replace_trapping_overflow,
3935 rewrite_to_non_trapping_overflow): New functions.
3936 * tree-vect-loop.c: Include tree-eh.h.
3937 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3938 * tree-data-ref.c: Include tree-eh.h.
3939 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3941 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3943 PR rtl-optimization/84123
3944 * combine.c (change_zero_ext): Check if hard register satisfies
3945 can_change_dest_mode before calling gen_lowpart_SUBREG.
3947 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3949 PR target/82444
3950 * ira.c (ira_init_register_move_cost): Remove assert.
3952 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3954 PR rtl-optimization/84071
3955 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3956 * doc/tm.texi: Regenerate.
3958 2018-01-31 Richard Biener <rguenther@suse.de>
3960 PR tree-optimization/84132
3961 * tree-data-ref.c (analyze_miv_subscript): Properly
3962 check whether evolution_function_is_affine_multivariate_p
3963 before calling gcd_of_steps_may_divide_p.
3965 2018-01-31 Julia Koval <julia.koval@intel.com>
3967 PR target/83618
3968 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3969 * config/i386/i386.md (rdpid_rex64) New.
3970 (rdpid): Make 32bit only.
3972 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3974 PR lto/84105
3975 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3976 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3978 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3980 Revert
3981 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3983 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3985 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3987 PR rtl-optimization/84071
3988 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3989 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3991 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3993 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3994 (arc_attribute_table): Add 'aux' attribute.
3995 (arc_in_small_data_p): Consider aux like variables.
3996 (arc_is_aux_reg_p): New function.
3997 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3998 (arc_get_aux_arg): New function.
3999 (prepare_move_operands): Handle aux-register access.
4000 (arc_handle_aux_attribute): New function.
4001 * doc/extend.texi (ARC Variable attributes): Add subsection.
4003 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4005 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4006 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4007 (arc_attribute_table): Add 'uncached' attribute.
4008 (arc_print_operand): Print '.di' flag for uncached memory
4009 accesses.
4010 (arc_in_small_data_p): Do not consider for small data the uncached
4011 types.
4012 (arc_is_uncached_mem_p): New function.
4013 * config/arc/predicates.md (compact_store_memory_operand): Check
4014 for uncached memory accesses.
4015 (nonvol_nonimm_operand): Likewise.
4016 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4018 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4020 PR c/84100
4021 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4022 falign-loops=): Add Optimization flag.
4024 2018-01-30 Jeff Law <law@redhat.com>
4026 PR target/84064
4027 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4028 INT_REGISTERS_SAVED. Check it prior to calling
4029 get_scratch_register_on_entry.
4030 (ix86_adjust_stack_and_probe): Similarly.
4031 (ix86_emit_probe_stack_range): Similarly.
4032 (ix86_expand_prologue): Corresponding changes.
4034 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4036 PR target/40411
4037 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4038 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4040 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4042 PR target/84112
4043 * lra-constraints.c (curr_insn_transform): Process AND in the
4044 address.
4046 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4048 PR rtl-optimization/83986
4049 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4050 dependence against last_pending_memory_flush in addition to
4051 pending_jump_insns.
4053 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4055 PR tree-optimization/81611
4056 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4057 copies.
4059 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4061 PR target/83758
4062 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4063 a reg rtx.
4065 2018-01-30 Richard Biener <rguenther@suse.de>
4066 Jakub Jelinek <jakub@redhat.com>
4068 PR tree-optimization/84111
4069 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4070 inner loops added during recursion, as they don't have up-to-date
4071 SSA form.
4073 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4075 PR ipa/81360
4076 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4077 (can_inline_edge_by_limits_p): ... here.
4078 (can_early_inline_edge_p, check_callers,
4079 update_caller_keys, update_callee_keys, recursive_inlining,
4080 add_new_edges_to_heap, speculation_useful_p,
4081 inline_small_functions,
4082 inline_small_functions, flatten_function,
4083 inline_to_all_callers_1): Update.
4085 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4087 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4088 zeros correctly.
4090 2018-01-30 Richard Biener <rguenther@suse.de>
4092 PR tree-optimization/83008
4093 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4094 invariant and constant vector uses in stmts when they need
4095 more than one stmt.
4097 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4099 PR bootstrap/84017
4100 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4101 * configure: Regenerate.
4103 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4105 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4106 pattern.
4107 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4108 Use gen_rtx_REG rather than gen_lowpart.
4110 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4112 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4113 rather than 0 when creating partial subregs.
4115 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4117 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4118 of usage.
4120 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4122 PR target/81550
4123 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4124 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4125 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4126 flags. This restores the settings used before the 2017-07-24.
4127 Turning off pre increment/decrement/modify allows IVOPTS to
4128 optimize DF/SF loops where the index is an int.
4130 2018-01-29 Richard Biener <rguenther@suse.de>
4131 Kelvin Nilsen <kelvin@gcc.gnu.org>
4133 PR bootstrap/80867
4134 * tree-vect-stmts.c (vectorizable_call): Don't call
4135 targetm.vectorize_builtin_md_vectorized_function if callee is
4136 NULL.
4138 2018-01-22 Carl Love <cel@us.ibm.com>
4140 * doc/extend.tex: Fix typo in second arg in
4141 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4143 2018-01-29 Richard Biener <rguenther@suse.de>
4145 PR tree-optimization/84086
4146 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4147 (flush_ssaname_freelist): When SSA names were released reset
4148 the SCEV hash table.
4150 2018-01-29 Richard Biener <rguenther@suse.de>
4152 PR tree-optimization/84057
4153 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4154 removed paths when removing edges.
4156 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4158 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4159 -mfunction-return=@var{choice}.
4161 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4163 PR diagnostic/84034
4164 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4165 Handle CR like TAB.
4166 (layout::print_source_line): Likewise.
4167 (test_get_line_width_without_trailing_whitespace): Add test cases.
4169 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4171 PR middle-end/84040
4172 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4173 debug insns.
4175 2018-01-26 Jim Wilson <jimw@sifive.com>
4177 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4179 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4180 specified.
4182 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4184 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4185 and CMP + SUB-immediate -> SUBS.
4187 2018-01-26 Martin Sebor <msebor@redhat.com>
4189 PR tree-optimization/83896
4190 * tree-ssa-strlen.c (get_string_len): Rename...
4191 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4192 Avoid assuming length is constant.
4193 (handle_char_store): Use HOST_WIDE_INT for string length.
4195 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4197 PR target/81763
4198 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4199 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4201 2018-01-26 Richard Biener <rguenther@suse.de>
4203 PR rtl-optimization/84003
4204 * dse.c (record_store): Only record redundant stores when
4205 the earlier store aliases at least all accesses the later one does.
4207 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4209 PR rtl-optimization/83985
4210 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4211 REG_CFA_RESTORE insns.
4212 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4213 REG_CFA_RESTORE insns here.
4215 PR c/83989
4216 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4217 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4219 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4221 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4222 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4223 (arc_init): Likewise.
4224 (arc_override_options): Likewise.
4225 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4226 value.
4227 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4228 support.
4229 * config/arc/arc.h (TARGET_DBNZ): Define.
4230 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4231 properly set the tune attribute.
4232 (dbnz): Use TARGET_DBNZ guard.
4233 * config/arc/arc.opt (mtune): Add core3 option.
4235 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4237 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4238 recognize new pic like addresses.
4239 (arc_delegitimize_address): Clean up.
4241 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4243 * config/arc/arc-arches.def: Option mrf16 valid for all
4244 architectures.
4245 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4246 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4247 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4248 * config/arc/arc-tables.opt: Regenerate.
4249 * config/arc/arc.c (arc_conditional_register_usage): Handle
4250 reduced register file case.
4251 (arc_file_start): Set must have build attributes.
4252 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4253 mrf16 option value.
4254 * config/arc/arc.opt (mrf16): Add new option.
4255 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4256 * config/arc/genmultilib.awk: Handle new mrf16 option.
4257 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4258 * config/arc/t-multilib: Regenerate.
4259 * doc/invoke.texi (ARC Options): Document mrf16 option.
4261 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4263 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4264 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4265 (arc_attribute_table): Add 'secure_call' attribute.
4266 (arc_print_operand): Print secure call operand.
4267 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4268 secure.
4269 (arc_is_secure_call_p): New function. * config/arc/arc.md
4270 (call_i): Add support for sjli instruction.
4271 (call_value_i): Likewise.
4272 * config/arc/constraints.md (Csc): New constraint.
4274 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4275 John Eric Martin <John.Martin@emmicro-us.com>
4277 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4278 * config/arc/arc.c (_arc_jli_section): New struct.
4279 (arc_jli_section): New type.
4280 (rc_jli_sections): New static variable.
4281 (arc_handle_jli_attribute): New function.
4282 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4283 (arc_file_end): New function.
4284 (TARGET_ASM_FILE_END): Define.
4285 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4286 (arc_add_jli_section): New function.
4287 (jli_call_scan): Likewise.
4288 (arc_reorg): Call jli_call_scan.
4289 (arc_output_addsi): Remove 'S' from printing asm operand.
4290 (arc_is_jli_call_p): New function.
4291 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4292 operand.
4293 (movhi_insn): Likewise.
4294 (movsi_insn): Likewise.
4295 (movsi_set_cc_insn): Likewise.
4296 (loadqi_update): Likewise.
4297 (load_zeroextendqisi_update): Likewise.
4298 (load_signextendqisi_update): Likewise.
4299 (loadhi_update): Likewise.
4300 (load_zeroextendhisi_update): Likewise.
4301 (load_signextendhisi_update): Likewise.
4302 (loadsi_update): Likewise.
4303 (loadsf_update): Likewise.
4304 (movsicc_insn): Likewise.
4305 (bset_insn): Likewise.
4306 (bxor_insn): Likewise.
4307 (bclr_insn): Likewise.
4308 (bmsk_insn): Likewise.
4309 (bicsi3_insn): Likewise.
4310 (cmpsi_cc_c_insn): Likewise.
4311 (movsi_ne): Likewise.
4312 (movsi_cond_exec): Likewise.
4313 (clrsbsi2): Likewise.
4314 (norm_f): Likewise.
4315 (normw): Likewise.
4316 (swap): Likewise.
4317 (divaw): Likewise.
4318 (flag): Likewise.
4319 (sr): Likewise.
4320 (kflag): Likewise.
4321 (ffs): Likewise.
4322 (ffs_f): Likewise.
4323 (fls): Likewise.
4324 (call_i): Remove 'S' asm letter, add jli instruction.
4325 (call_value_i): Likewise.
4326 * config/arc/arc.op (mjli-always): New option.
4327 * config/arc/constraints.md (Cji): New constraint.
4328 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4329 operand.
4330 (subsf3_fpx): Likewise.
4331 (mulsf3_fpx): Likewise.
4332 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4333 asm operand.
4334 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4335 function attrbutes.
4336 * doc/invoke.texi (ARC): Document mjli-always option.
4338 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4340 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4341 avoid addition with 0 and use incw and decw where possible.
4343 2018-01-26 Richard Biener <rguenther@suse.de>
4345 PR tree-optimization/81082
4346 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4347 association if it requires casting to unsigned.
4348 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4349 from fold_plusminus_mult_expr to catch important cases late when
4350 range info is available.
4352 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4354 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4355 * configure.ac (hidden_linkonce): New test.
4356 * configure: Regenerate.
4357 * config.in: Regenerate.
4359 2018-01-26 Julia Koval <julia.koval@intel.com>
4361 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4362 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4363 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4364 _mm_mask_bitshuffle_epi64_mask): Fix type.
4365 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4366 USI_FTYPE_V4DI_V4DI_USI): Remove.
4367 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4368 __builtin_ia32_vpshufbitqmb256_mask,
4369 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4370 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4371 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4373 2018-01-26 Alan Modra <amodra@gmail.com>
4375 PR target/84033
4376 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4377 UNSPEC_VBPERMQ. Sort other unspecs.
4379 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4381 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4383 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4385 PR middle-end/83055
4386 * predict.c (drop_profile): Do not push/pop cfun; update also
4387 node->count.
4388 (handle_missing_profiles): Fix logic looking for zero profiles.
4390 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4392 PR middle-end/83977
4393 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4394 on functions with #pragma omp declare simd or functions with simd
4395 attribute.
4396 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4397 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4398 Remove trailing \n from warning_at calls.
4400 2018-01-25 Tom de Vries <tom@codesourcery.com>
4402 PR target/84028
4403 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4404 for neutered workers.
4406 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4408 PR target/68467
4409 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4410 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4412 2018-01-24 Jeff Law <law@redhat.com>
4414 PR target/83994
4415 * i386.c (get_probe_interval): Move to earlier point.
4416 (ix86_compute_frame_layout): If -fstack-clash-protection and
4417 the frame is larger than the probe interval, then use pushes
4418 to save registers rather than reg->mem moves.
4419 (ix86_expand_prologue): Remove conditional for int_registers_saved
4420 assertion.
4422 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4424 PR target/84014
4425 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4426 min/max for never referenced object.
4428 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4430 PR middle-end/83977
4431 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4432 here.
4433 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4434 attributes from DECL_ATTRIBUTES (decl) without affecting
4435 DECL_ATTRIBUTES (current_function_decl).
4436 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4437 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4439 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4441 PR tree-optimization/83979
4442 * fold-const.c (fold_comparison): Use constant_boolean_node
4443 instead of boolean_{true,false}_node.
4445 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4447 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4448 with zero counts.
4450 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4452 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4453 Simplify the clause that sets the length attribute.
4454 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4455 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4456 clause that sets the length attribute.
4457 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4459 2018-01-24 Tom de Vries <tom@codesourcery.com>
4461 PR target/83589
4462 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4463 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4464 Add strict parameter.
4465 (prevent_branch_around_nothing): Insert dummy insn between branch to
4466 label and label with no ptx insn inbetween.
4467 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4469 2018-01-24 Tom de Vries <tom@codesourcery.com>
4471 PR target/81352
4472 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4473 for neutered threads in warp.
4474 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4476 2018-01-24 Richard Biener <rguenther@suse.de>
4478 PR tree-optimization/83176
4479 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4480 operands.
4482 2018-01-24 Richard Biener <rguenther@suse.de>
4484 PR tree-optimization/82819
4485 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4486 code generating pluses that are no-ops in the target precision.
4488 2018-01-24 Richard Biener <rguenther@suse.de>
4490 PR middle-end/84000
4491 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4493 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4495 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4496 to merge probabilities.
4497 * predict.c (probably_never_executed): Also mark as cold functions
4498 with global 0 profile and guessed local profile.
4499 * profile-count.c (profile_probability::combine_with_count): New
4500 member function.
4501 * profile-count.h (profile_probability::operator*,
4502 profile_probability::operator*=, profile_probability::operator/,
4503 profile_probability::operator/=): Reduce precision to adjusted
4504 and set value to guessed on contradictory divisions.
4505 (profile_probability::combine_with_freq): Remove.
4506 (profile_probability::combine_wiht_count): Declare.
4507 (profile_count::force_nonzero):: Set to adjusted.
4508 (profile_count::probability_in):: Set quality to adjusted.
4509 * tree-ssa-tail-merge.c (replace_block_by): Use
4510 combine_with_count.
4512 2018-01-23 Andrew Waterman <andrew@sifive.com>
4513 Jim Wilson <jimw@sifive.com>
4515 * config/riscv/riscv.c (riscv_stack_boundary): New.
4516 (riscv_option_override): Set riscv_stack_boundary. Handle
4517 riscv_preferred_stack_boundary_arg.
4518 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4519 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4520 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4521 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4522 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4523 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4525 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4527 PR target/83905
4528 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4529 of struct ix86_frame.
4530 (ix86_expand_epilogue): Likewise. Add a local variable for
4531 the reg_save_offset field in struct ix86_frame.
4533 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4535 PR tree-optimization/82604
4536 * tree-loop-distribution.c (enum partition_kind): New enum item
4537 PKIND_PARTIAL_MEMSET.
4538 (partition_builtin_p): Support above new enum item.
4539 (generate_code_for_partition): Ditto.
4540 (compute_access_range): Differentiate cases that equality can be
4541 proven at all loops, the innermost loops or no loops.
4542 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4543 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4544 (finalize_partitions, distribute_loop): Don't fuse partition of
4545 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4546 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4547 parloop is enabled.
4549 2018-01-23 Martin Liska <mliska@suse.cz>
4551 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4552 order to ignore the predictor.
4553 (PRED_POLYMORPHIC_CALL): Likewise.
4554 (PRED_RECURSIVE_CALL): Likewise.
4556 2018-01-23 Martin Liska <mliska@suse.cz>
4558 * tree-profile.c (tree_profiling): Print function header to
4559 aware reader which function we are working on.
4560 * value-prof.c (gimple_find_values_to_profile): Do not print
4561 not interesting value histograms.
4563 2018-01-23 Martin Liska <mliska@suse.cz>
4565 * profile-count.h (enum profile_quality): Add
4566 profile_uninitialized as the first value. Do not number values
4567 as they are zero based.
4568 (profile_count::verify): Update sanity check.
4569 (profile_probability::verify): Likewise.
4571 2018-01-23 Nathan Sidwell <nathan@acm.org>
4573 * doc/invoke.texi (ffor-scope): Deprecate.
4575 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4577 PR tree-optimization/83510
4578 * domwalk.c (set_all_edges_as_executable): New function.
4579 (dom_walker::dom_walker): Convert bool param
4580 "skip_unreachable_blocks" to enum reachability. Move setup of
4581 edge flags to set_all_edges_as_executable and only do it when
4582 reachability is REACHABLE_BLOCKS.
4583 * domwalk.h (enum dom_walker::reachability): New enum.
4584 (dom_walker::dom_walker): Convert bool param
4585 "skip_unreachable_blocks" to enum reachability.
4586 (set_all_edges_as_executable): New decl.
4587 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4588 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4589 "reachability".
4590 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4591 but converting true to REACHABLE_BLOCKS.
4592 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4593 * tree-vrp.c
4594 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4595 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4596 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4597 REACHABLE_BLOCKS.
4598 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4599 if check_all_array_refs will be called.
4601 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4603 * tree.c (selftest::test_location_wrappers): Add more test
4604 coverage.
4606 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4608 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4609 (selftest::test_bit_in_range): Likewise.
4611 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4613 PR testsuite/83888
4614 * doc/sourcebuild.texi (vect_float): Say that the selector
4615 only describes the situation when -funsafe-math-optimizations is on.
4616 (vect_float_strict): Document.
4618 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4620 PR tree-optimization/83965
4621 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4622 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4623 instead of checking only for a reduction.
4624 (vect_recog_widen_sum_pattern): Likewise.
4626 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4628 * predict.c (probably_never_executed): Only use precise profile info.
4629 (compute_function_frequency): Skip after inlining hack since we now
4630 have quality checking.
4632 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4634 * profile-count.h (profile_probability::very_unlikely,
4635 profile_probability::unlikely, profile_probability::even): Set
4636 precision to guessed.
4638 2018-01-23 Richard Biener <rguenther@suse.de>
4640 PR tree-optimization/83963
4641 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4642 Properly terminate dominator walk when crossing the exit edge not
4643 when visiting its source block.
4645 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4647 PR c++/83918
4648 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4649 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4651 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4653 PR tree-optimization/83957
4654 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4655 semicolon after for body surrounded by braces.
4657 PR tree-optimization/83081
4658 * profile-count.h (profile_probability::split): New method.
4659 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4660 Use profile_probability::split.
4661 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4662 when splitting a single conditional jump into 2.
4664 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4666 PR tree-optimization/69452
4667 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4668 decl.
4670 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4672 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4673 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4674 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4676 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4678 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4679 * config/rl78/rl78.md (movdi): New define_expand.
4680 * config/rl78/rl78.c (rl78_split_movdi): New function.
4682 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4684 PR target/83862
4685 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4686 no longer used.
4687 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4688 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4689 128-bit to produce an UNSPEC move to get the double word with the
4690 signbit and then a shift directly to do signbit.
4691 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4692 implementation with a new version that just does either a direct
4693 move or a regular move. Move memory interface to separate insns.
4694 Move insns so they are next to the expander.
4695 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4696 with signbit move. Split big and little endian case.
4697 (signbit<mode>2_dm_mem_le): Likewise.
4698 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4699 (signbit<mode>2_dm2): Likewise.
4701 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4703 * config/rl78/rl78.md (anddi3): New define_expand.
4705 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4707 * config/rl78/rl78.md (umindi3): New define_expand.
4709 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4711 * config/rl78/rl78.md (smindi3): New define_expand.
4713 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4715 * config/rl78/rl78.md (smaxdi3): New define_expand.
4717 2018-01-22 Carl Love <cel@us.ibm.com>
4719 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4720 LVX_V1TI): Add macro expansion.
4721 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4722 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4723 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4724 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4725 Change check to determine if the instruction is a byte reversing
4726 entry. Fix typo in comment.
4727 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4728 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4729 Add def_builtin calls for new builtins.
4730 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4731 Add define_insn expansion.
4733 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4735 * config/rl78/rl78.md (umaxdi3): New define_expand.
4737 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4739 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4740 for non-QImode registers.
4742 2018-01-22 Richard Biener <rguenther@suse.de>
4744 PR tree-optimization/83963
4745 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4746 including the loop exit block.
4747 (scop_detection::merge_sese): Likewise.
4748 (scop_detection::add_scop): Do it here instead.
4750 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4752 * doc/sourcebuild.texi (arm_softfloat): Document.
4754 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4756 PR gcc/77734
4757 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4758 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4759 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4761 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4762 David Edelsohn <dje.gcc@gmail.com>
4764 PR target/83946
4765 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4766 Change "crset eq" to "crset 2".
4767 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4768 (*call_indirect_aix<mode>_nospec): Likewise.
4769 (*call_value_indirect_aix<mode>_nospec): Likewise.
4770 (*call_indirect_elfv2<mode>_nospec): Likewise.
4771 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4772 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4773 change assembly output from . to $.
4774 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4775 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4776 (*tablejump<mode>_internal1_nospec): Likewise.
4778 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4780 PR target/80870
4781 * config/sh/sh_optimize_sett_clrt.cc:
4782 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4784 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4786 PR tree-optimization/83940
4787 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4788 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4789 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4790 use it to pass back the definition type.
4791 (vect_check_store_rhs): Likewise rhs_dt_out.
4792 (vect_build_gather_load_calls): Add a mask_dt argument and use
4793 it instead of a call to vect_is_simple_use.
4794 (vectorizable_store): Update calls to vect_check_load_store_mask
4795 and vect_check_store_rhs. Use the dt returned by the latter instead
4796 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4797 instead of calls to vect_is_simple_use. Pass the scalar rather
4798 than the vector operand to vect_is_simple_use when handling
4799 second and subsequent copies of an rhs value.
4800 (vectorizable_load): Update calls to vect_check_load_store_mask
4801 and vect_build_gather_load_calls. Use the cached mask_dt and
4802 gs_info.offset_dt instead of calls to vect_is_simple_use.
4804 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4806 PR middle-end/83945
4807 * tree-emutls.c: Include gimplify.h.
4808 (lower_emutls_2): New function.
4809 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4810 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4811 it before further processing.
4813 PR target/83930
4814 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4815 UINTVAL (trueop1) instead of INTVAL (op1).
4817 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4819 PR debug/81570
4820 PR debug/83728
4821 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4822 INCOMING_FRAME_SP_OFFSET if not defined.
4823 (scan_trace): Add ENTRY argument. If true and
4824 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4825 emit a note to adjust the CFA offset.
4826 (create_cfi_notes): Adjust scan_trace callers.
4827 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4828 INCOMING_FRAME_SP_OFFSET in the CIE.
4829 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4830 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4831 Likewise.
4832 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4833 * doc/tm.texi: Regenerated.
4835 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4837 PR rtl-optimization/83147
4838 * lra-constraints.c (remove_inheritance_pseudos): Use
4839 lra_substitute_pseudo_within_insn.
4841 2018-01-19 Tom de Vries <tom@codesourcery.com>
4842 Cesar Philippidis <cesar@codesourcery.com>
4844 PR target/83920
4845 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4847 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4849 PR target/83790
4850 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4851 spaces for function labels.
4853 2018-01-19 Martin Liska <mliska@suse.cz>
4855 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4856 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4857 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4858 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4859 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4860 (PRED_CONST_RETURN): Change from 69 to 65.
4861 (PRED_NULL_RETURN): Change from 91 to 71.
4862 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4863 (PRED_LOOP_GUARD): Change from 66 to 73.
4865 2018-01-19 Martin Liska <mliska@suse.cz>
4867 * predict.c (predict_insn_def): Add new assert.
4868 (struct branch_predictor): Change type to signed integer.
4869 (test_prediction_value_range): Amend test to cover
4870 PROB_UNINITIALIZED.
4871 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4872 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4873 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4874 (PRED_LOOP_IV_COMPARE): Likewise.
4875 * predict.h (PROB_UNINITIALIZED): Define new constant.
4877 2018-01-19 Martin Liska <mliska@suse.cz>
4879 * predict.c (dump_prediction): Add new format for
4880 analyze_brprob.py script which is enabled with -details
4881 suboption.
4882 * profile-count.h (precise_p): New function.
4884 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4886 PR tree-optimization/83922
4887 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4888 there are no statements that need masking.
4889 (vect_active_double_reduction_p): New function.
4890 (vect_analyze_loop_operations): Use it when handling phis that
4891 are not in the loop header.
4893 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4895 PR tree-optimization/83914
4896 * tree-vect-loop.c (vectorizable_induction): Don't convert
4897 init_expr or apply the peeling adjustment for inductions
4898 that are nested within the vectorized loop.
4900 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4902 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4903 instead of NEG.
4905 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4907 PR sanitizer/81715
4908 PR testsuite/83882
4909 * function.h (gimplify_parameters): Add gimple_seq * argument.
4910 * function.c: Include gimple.h and options.h.
4911 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4912 for the added local temporaries if needed.
4913 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4914 if there are any parameter cleanups, wrap whole body into a
4915 try/finally with the cleanups.
4917 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4919 PR target/82964
4920 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4921 Use GET_MODE_CLASS for scalar floating point.
4923 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4925 PR ipa/82256
4926 patch by PaX Team
4927 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4928 Fix call of call_cgraph_insertion_hooks.
4930 2018-01-18 Martin Sebor <msebor@redhat.com>
4932 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4934 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4936 PR ipa/83619
4937 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4938 frequencies.
4940 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4942 PR other/70268
4943 * common.opt: (-ffile-prefix-map): New option.
4944 * opts.c (common_handle_option): Defer it.
4945 * opts-global.c (handle_common_deferred_options): Handle it.
4946 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4947 * file-prefix-map.h: New file.
4948 (remap_debug_filename, add_debug_prefix_map): ...here.
4949 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4950 * final.c (debug_prefix_map, add_debug_prefix_map
4951 remap_debug_filename): Move to...
4952 * file-prefix-map.c: New file.
4953 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4954 generalize, get rid of alloca(), use strrchr() instead of strchr().
4955 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4956 Implement in terms of add_prefix_map().
4957 (remap_macro_filename, remap_debug_filename): Implement in term of
4958 remap_filename().
4959 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4960 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4961 * dbxout.c: Include file-prefix-map.h.
4962 * varasm.c: Likewise.
4963 * vmsdbgout.c: Likewise.
4964 * xcoffout.c: Likewise.
4965 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4966 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4967 * doc/invoke.texi (-ffile-prefix-map): Document.
4968 (-fdebug-prefix-map): Update description.
4970 2018-01-18 Martin Liska <mliska@suse.cz>
4972 * config/i386/i386.c (indirect_thunk_name): Document that also
4973 lfence is emitted.
4974 (output_indirect_thunk): Document why both instructions
4975 (pause and lfence) are generated.
4977 2018-01-18 Richard Biener <rguenther@suse.de>
4979 PR tree-optimization/83887
4980 * graphite-scop-detection.c
4981 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4982 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4983 (scop_detection::merge_sese): Re-implement with a flood-fill
4984 algorithm that properly finds a SESE region if it exists.
4986 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4988 PR c/61240
4989 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4990 pointer_diff optimizations use view_convert instead of convert.
4992 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4994 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4995 Generate different code for -mno-speculate-indirect-jumps.
4996 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4997 (*call_indirect_aix<mode>): Disable for
4998 -mno-speculate-indirect-jumps.
4999 (*call_indirect_aix<mode>_nospec): New define_insn.
5000 (*call_value_indirect_aix<mode>): Disable for
5001 -mno-speculate-indirect-jumps.
5002 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5003 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5004 -mno-speculate-indirect-jumps.
5005 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5007 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5009 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5010 long double type, set the flags for noting the default long double
5011 type, even if we don't pass or return a long double type.
5013 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5015 PR ipa/83051
5016 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5017 failure.
5019 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5021 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5022 support for merge[hl].
5023 (fold_mergehl_helper): New helper function.
5024 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5025 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5026 (altivec_vmrglw_direct): Add xxmrglw insn.
5028 2018-01-17 Andrew Waterman <andrew@sifive.com>
5030 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5031 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5033 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5035 PR lto/83121
5036 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5037 call the lto_location_cache before reading the
5038 DECL_SOURCE_LOCATION of the types.
5040 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5041 Richard Sandiford <richard.sandiford@linaro.org>
5043 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5044 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5045 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5046 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5047 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5048 Add declaration.
5049 * config/aarch64/constraints.md (aarch64_movti_operand):
5050 Limit immediates.
5051 * config/aarch64/predicates.md (Uti): Add new constraint.
5053 2018-01-17 Carl Love <cel@us.ibm.com>
5055 * config/rs6000/vsx.md (define_expand xl_len_r,
5056 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5057 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5058 lxvll.
5059 (define_expand, define_insn): Move the shift left from the
5060 define_insn to the define_expand for lxvl and stxvl instructions.
5061 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5062 and XL_LEN_R definitions to PURE.
5064 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5066 * config/i386/i386.c (indirect_thunk_name): Declare regno
5067 as unsigned int. Compare regno with INVALID_REGNUM.
5068 (output_indirect_thunk): Ditto.
5069 (output_indirect_thunk_function): Ditto.
5070 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5071 in the call to output_indirect_thunk_function.
5073 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5075 PR middle-end/83884
5076 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5077 rather than the size of inner_type to determine the stack slot size
5078 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5080 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5082 PR target/83546
5083 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5084 to PTA_SILVERMONT.
5086 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5088 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5089 endian Linux systems to optionally enable multilibs for selecting
5090 the long double type if the user configured an explicit type.
5091 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5092 have no long double multilibs if not defined.
5093 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5094 warn if the user used -mabi={ieee,ibm}longdouble and we built
5095 multilibs for long double.
5096 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5097 appropriate multilib option.
5098 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5099 multilib options.
5100 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5101 for building long double multilibs.
5102 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5104 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5106 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5107 copies.
5109 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5110 64 bits.
5111 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5112 128 bits.
5114 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5115 variables.
5117 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5118 return value.
5120 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5122 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5123 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5125 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5127 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5128 different rtl trees depending on TARGET_64BIT.
5129 (rs6000_gen_lvx): Likewise.
5131 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5133 * config/visium/visium.md (nop): Tweak comment.
5134 (hazard_nop): Likewise.
5136 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5138 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5139 -mspeculate-indirect-jumps.
5140 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5141 for -mno-speculate-indirect-jumps.
5142 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5143 (*call_value_indirect_elfv2<mode>): Disable for
5144 -mno-speculate-indirect-jumps.
5145 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5146 (indirect_jump): Emit different RTL for
5147 -mno-speculate-indirect-jumps.
5148 (*indirect_jump<mode>): Disable for
5149 -mno-speculate-indirect-jumps.
5150 (*indirect_jump<mode>_nospec): New define_insn.
5151 (tablejump): Emit different RTL for
5152 -mno-speculate-indirect-jumps.
5153 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5154 (tablejumpsi_nospec): New define_expand.
5155 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5156 (tablejumpdi_nospec): New define_expand.
5157 (*tablejump<mode>_internal1): Disable for
5158 -mno-speculate-indirect-jumps.
5159 (*tablejump<mode>_internal1_nospec): New define_insn.
5160 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5161 option.
5163 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5165 * caller-save.c (insert_save): Drop unnecessary parameter. All
5166 callers updated.
5168 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5169 Richard Biener <rguenth@suse.de>
5171 PR libgomp/83590
5172 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5173 return early, inline manually is_gimple_sizepos. Make sure if we
5174 call gimplify_expr we don't end up with a gimple constant.
5175 * tree.c (variably_modified_type_p): Don't return true for
5176 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5177 * gimplify.h (is_gimple_sizepos): Remove.
5179 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5181 PR tree-optimization/83857
5182 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5183 vectorizable_live_operation for pure SLP statements.
5184 (vectorizable_live_operation): Handle PHIs.
5186 2018-01-16 Richard Biener <rguenther@suse.de>
5188 PR tree-optimization/83867
5189 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5190 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5192 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5194 PR c/83844
5195 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5196 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5197 If off is not INTEGER_CST, issue a may not be aligned warning
5198 rather than isn't aligned. Use isn%'t rather than isn't.
5199 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5200 into MULT_EXPR.
5201 <case MULT_EXPR>: Improve the case when bottom and one of the
5202 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5203 operand, in that case check if the other operand is multiple of
5204 bottom divided by the INTEGER_CST operand.
5206 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5208 PR target/83858
5209 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5210 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5211 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5212 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5213 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5214 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5215 (pa_function_arg_size): New function.
5217 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5219 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5220 in a separate statement.
5222 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5224 PR tree-optimization/83847
5225 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5226 group gathers and scatters.
5228 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5230 PR rtl-optimization/86620
5231 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5233 PR rtl-optimization/83213
5234 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5235 to last if both are JUMP_INSNs.
5237 PR tree-optimization/83843
5238 * gimple-ssa-store-merging.c
5239 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5240 store_immediate_info for bswap/nop orig_stores.
5242 2018-01-15 Andrew Waterman <andrew@sifive.com>
5244 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5245 !TARGET_MUL.
5246 <UDIV>: Increase cost if !TARGET_DIV.
5248 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5250 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5251 (define_attr "cr_logical_3op"): New.
5252 (cceq_ior_compare): Adjust.
5253 (cceq_ior_compare_complement): Adjust.
5254 (*cceq_rev_compare): Adjust.
5255 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5256 (is_cracked_insn): Adjust.
5257 (insn_must_be_first_in_group): Adjust.
5258 * config/rs6000/40x.md: Adjust.
5259 * config/rs6000/440.md: Adjust.
5260 * config/rs6000/476.md: Adjust.
5261 * config/rs6000/601.md: Adjust.
5262 * config/rs6000/603.md: Adjust.
5263 * config/rs6000/6xx.md: Adjust.
5264 * config/rs6000/7450.md: Adjust.
5265 * config/rs6000/7xx.md: Adjust.
5266 * config/rs6000/8540.md: Adjust.
5267 * config/rs6000/cell.md: Adjust.
5268 * config/rs6000/e300c2c3.md: Adjust.
5269 * config/rs6000/e500mc.md: Adjust.
5270 * config/rs6000/e500mc64.md: Adjust.
5271 * config/rs6000/e5500.md: Adjust.
5272 * config/rs6000/e6500.md: Adjust.
5273 * config/rs6000/mpc.md: Adjust.
5274 * config/rs6000/power4.md: Adjust.
5275 * config/rs6000/power5.md: Adjust.
5276 * config/rs6000/power6.md: Adjust.
5277 * config/rs6000/power7.md: Adjust.
5278 * config/rs6000/power8.md: Adjust.
5279 * config/rs6000/power9.md: Adjust.
5280 * config/rs6000/rs64.md: Adjust.
5281 * config/rs6000/titan.md: Adjust.
5283 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5285 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5286 ix86_indirect_branch_register logic.
5288 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5290 * config/i386/constraints.md (Bs): Update
5291 ix86_indirect_branch_register check. Don't check
5292 ix86_indirect_branch_register with GOT_memory_operand.
5293 (Bw): Likewise.
5294 * config/i386/predicates.md (GOT_memory_operand): Don't check
5295 ix86_indirect_branch_register here.
5296 (GOT32_symbol_operand): Likewise.
5298 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5300 * config/i386/predicates.md (constant_call_address_operand):
5301 Rewrite ix86_indirect_branch_register logic.
5302 (sibcall_insn_operand): Likewise.
5304 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5306 * config/i386/constraints.md (Bs): Replace
5307 ix86_indirect_branch_thunk_register with
5308 ix86_indirect_branch_register.
5309 (Bw): Likewise.
5310 * config/i386/i386.md (indirect_jump): Likewise.
5311 (tablejump): Likewise.
5312 (*sibcall_memory): Likewise.
5313 (*sibcall_value_memory): Likewise.
5314 Peepholes of indirect call and jump via memory: Likewise.
5315 * config/i386/i386.opt: Likewise.
5316 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5317 (GOT_memory_operand): Likewise.
5318 (call_insn_operand): Likewise.
5319 (sibcall_insn_operand): Likewise.
5320 (GOT32_symbol_operand): Likewise.
5322 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5324 PR middle-end/83837
5325 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5326 type rather than type addr's type points to.
5327 (expand_omp_atomic_mutex): Likewise.
5328 (expand_omp_atomic): Likewise.
5330 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5332 PR target/83839
5333 * config/i386/i386.c (output_indirect_thunk_function): Use
5334 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5335 for __x86_return_thunk.
5337 2018-01-15 Richard Biener <rguenther@suse.de>
5339 PR middle-end/83850
5340 * expmed.c (extract_bit_field_1): Fix typo.
5342 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5344 PR target/83687
5345 * config/arm/iterators.md (VF): New mode iterator.
5346 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5347 Remove integer-related logic from pattern.
5348 (neon_vabd<mode>_3): Likewise.
5350 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5352 PR middle-end/82694
5353 * common.opt (fstrict-overflow): No longer an alias.
5354 (fwrapv-pointer): New option.
5355 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5356 also for pointer types based on flag_wrapv_pointer.
5357 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5358 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5359 opts->x_flag_wrapv got set.
5360 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5361 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5362 POINTER_TYPE_OVERFLOW_UNDEFINED.
5363 * match.pd: Likewise in address comparison pattern.
5364 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5366 2018-01-15 Richard Biener <rguenther@suse.de>
5368 PR lto/83804
5369 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5370 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5371 Reset type names to their identifier if their TYPE_DECL doesn't
5372 have linkage (and thus is used for ODR and devirt).
5373 (save_debug_info_for_decl): Remove.
5374 (save_debug_info_for_type): Likewise.
5375 (add_tree_to_fld_list): Adjust.
5376 * tree-pretty-print.c (dump_generic_node): Make dumping of
5377 type names more robust.
5379 2018-01-15 Richard Biener <rguenther@suse.de>
5381 * BASE-VER: Bump to 8.0.1.
5383 2018-01-14 Martin Sebor <msebor@redhat.com>
5385 PR other/83508
5386 * builtins.c (check_access): Avoid warning when the no-warning bit
5387 is set.
5389 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5391 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5392 * ira-color (allocno_hard_regs_compare): Likewise.
5394 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5396 PR target/83013
5397 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5398 Use .pushsection/.popsection.
5400 2018-01-14 Martin Sebor <msebor@redhat.com>
5402 PR c++/81327
5403 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5405 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5407 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5408 entry from extra_headers.
5409 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5410 extra_headers, make the list bitwise identical to the i?86-*-* one.
5412 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5414 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5415 -mcmodel=large with -mindirect-branch=thunk,
5416 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5417 -mfunction-return=thunk-extern.
5418 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5419 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5420 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5422 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5424 * config/i386/i386.c (print_reg): Print the name of the full
5425 integer register without '%'.
5426 (ix86_print_operand): Handle 'V'.
5427 * doc/extend.texi: Document 'V' modifier.
5429 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5431 * config/i386/constraints.md (Bs): Disallow memory operand for
5432 -mindirect-branch-register.
5433 (Bw): Likewise.
5434 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5435 (GOT_memory_operand): Likewise.
5436 (call_insn_operand): Likewise.
5437 (sibcall_insn_operand): Likewise.
5438 (GOT32_symbol_operand): Likewise.
5439 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5440 for -mindirect-branch-register.
5441 (tablejump): Likewise.
5442 (*sibcall_memory): Likewise.
5443 (*sibcall_value_memory): Likewise.
5444 Disallow peepholes of indirect call and jump via memory for
5445 -mindirect-branch-register.
5446 (*call_pop): Replace m with Bw.
5447 (*call_value_pop): Likewise.
5448 (*sibcall_pop_memory): Replace m with Bs.
5449 * config/i386/i386.opt (mindirect-branch-register): New option.
5450 * doc/invoke.texi: Document -mindirect-branch-register option.
5452 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5454 * config/i386/i386-protos.h (ix86_output_function_return): New.
5455 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5456 set function_return_type.
5457 (indirect_thunk_name): Add ret_p to indicate thunk for function
5458 return.
5459 (output_indirect_thunk_function): Pass false to
5460 indirect_thunk_name.
5461 (ix86_output_indirect_branch_via_reg): Likewise.
5462 (ix86_output_indirect_branch_via_push): Likewise.
5463 (output_indirect_thunk_function): Create alias for function
5464 return thunk if regno < 0.
5465 (ix86_output_function_return): New function.
5466 (ix86_handle_fndecl_attribute): Handle function_return.
5467 (ix86_attribute_table): Add function_return.
5468 * config/i386/i386.h (machine_function): Add
5469 function_return_type.
5470 * config/i386/i386.md (simple_return_internal): Use
5471 ix86_output_function_return.
5472 (simple_return_internal_long): Likewise.
5473 * config/i386/i386.opt (mfunction-return=): New option.
5474 (indirect_branch): Mention -mfunction-return=.
5475 * doc/extend.texi: Document function_return function attribute.
5476 * doc/invoke.texi: Document -mfunction-return= option.
5478 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5480 * config/i386/i386-opts.h (indirect_branch): New.
5481 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5482 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5483 with local indirect jump when converting indirect call and jump.
5484 (ix86_set_indirect_branch_type): New.
5485 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5486 (indirectlabelno): New.
5487 (indirect_thunk_needed): Likewise.
5488 (indirect_thunk_bnd_needed): Likewise.
5489 (indirect_thunks_used): Likewise.
5490 (indirect_thunks_bnd_used): Likewise.
5491 (INDIRECT_LABEL): Likewise.
5492 (indirect_thunk_name): Likewise.
5493 (output_indirect_thunk): Likewise.
5494 (output_indirect_thunk_function): Likewise.
5495 (ix86_output_indirect_branch_via_reg): Likewise.
5496 (ix86_output_indirect_branch_via_push): Likewise.
5497 (ix86_output_indirect_branch): Likewise.
5498 (ix86_output_indirect_jmp): Likewise.
5499 (ix86_code_end): Call output_indirect_thunk_function if needed.
5500 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5501 needed.
5502 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5503 (ix86_attribute_table): Add indirect_branch.
5504 * config/i386/i386.h (machine_function): Add indirect_branch_type
5505 and has_local_indirect_jump.
5506 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5507 to true.
5508 (tablejump): Likewise.
5509 (*indirect_jump): Use ix86_output_indirect_jmp.
5510 (*tablejump_1): Likewise.
5511 (simple_return_indirect_internal): Likewise.
5512 * config/i386/i386.opt (mindirect-branch=): New option.
5513 (indirect_branch): New.
5514 (keep): Likewise.
5515 (thunk): Likewise.
5516 (thunk-inline): Likewise.
5517 (thunk-extern): Likewise.
5518 * doc/extend.texi: Document indirect_branch function attribute.
5519 * doc/invoke.texi: Document -mindirect-branch= option.
5521 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5523 PR ipa/83051
5524 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5526 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5528 * ipa-inline.c (want_inline_small_function_p): Return false if
5529 inlining has already failed with CIF_FINAL_ERROR.
5530 (update_caller_keys): Call want_inline_small_function_p before
5531 can_inline_edge_p.
5532 (update_callee_keys): Likewise.
5534 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5536 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5537 New function.
5538 (rs6000_quadword_masked_address_p): Likewise.
5539 (quad_aligned_load_p): Likewise.
5540 (quad_aligned_store_p): Likewise.
5541 (const_load_sequence_p): Add comment to describe the outer-most loop.
5542 (mimic_memory_attributes_and_flags): New function.
5543 (rs6000_gen_stvx): Likewise.
5544 (replace_swapped_aligned_store): Likewise.
5545 (rs6000_gen_lvx): Likewise.
5546 (replace_swapped_aligned_load): Likewise.
5547 (replace_swapped_load_constant): Capitalize argument name in
5548 comment describing this function.
5549 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5550 and stores that access quad-word aligned addresses and replace
5551 with stvx or lvx instructions when appropriate.
5552 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5553 New function prototype.
5554 (rs6000_quadword_masked_address_p): Likewise.
5555 (rs6000_gen_lvx): Likewise.
5556 (rs6000_gen_stvx): Likewise.
5557 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5558 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5559 when memory address is aligned.
5560 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5561 this split to select lvx instruction when memory address is aligned.
5562 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5563 instruction when memory address is aligned.
5564 (*vsx_le_perm_load_v16qi): Likewise.
5565 (four unnamed splitters): Modify to select the stvx instruction
5566 when memory is aligned.
5568 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5570 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5571 which appears in the queue multiple times.
5573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5574 Alan Hayward <alan.hayward@arm.com>
5575 David Sherwood <david.sherwood@arm.com>
5577 * tree-vectorizer.h (vec_lower_bound): New structure.
5578 (_loop_vec_info): Add check_nonzero and lower_bounds.
5579 (LOOP_VINFO_CHECK_NONZERO): New macro.
5580 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5581 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5582 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5583 fields. Make seg_len the distance travelled, not including the
5584 access size.
5585 (dr_direction_indicator): Declare.
5586 (dr_zero_step_indicator): Likewise.
5587 (dr_known_forward_stride_p): Likewise.
5588 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5589 tree-ssanames.h.
5590 (runtime_alias_check_p): Allow runtime alias checks with
5591 variable strides.
5592 (operator ==): Compare access_size and align.
5593 (prune_runtime_alias_test_list): Rework for new distinction between
5594 the access_size and seg_len.
5595 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5596 segment lengths.
5597 (get_segment_min_max): New function.
5598 (create_intersect_range_checks): Use it.
5599 (dr_step_indicator): New function.
5600 (dr_direction_indicator): Likewise.
5601 (dr_zero_step_indicator): Likewise.
5602 (dr_known_forward_stride_p): Likewise.
5603 * tree-loop-distribution.c (data_ref_segment_size): Return
5604 DR_STEP * (niters - 1).
5605 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5606 constructor.
5607 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5608 (vect_preserves_scalar_order_p): New function, split out from...
5609 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5610 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5611 (vect_vfa_access_size): New function.
5612 (vect_vfa_align): Likewise.
5613 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5614 (dump_lower_bound): New function.
5615 (vect_check_lower_bound): Likewise.
5616 (vect_small_gap_p): Likewise.
5617 (vectorizable_with_step_bound_p): Likewise.
5618 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5619 depencies if the vectorization factor is 1. Convert the checks
5620 for nonzero steps into checks on the bounds of DR_STEP. Try using
5621 a bunds check for variable steps if the minimum required step is
5622 relatively small. Update calls to the dr_with_seg_len
5623 constructor and to vect_compile_time_alias.
5624 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5625 function.
5626 (vect_loop_versioning): Call it.
5627 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5628 when retrying.
5629 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5632 Alan Hayward <alan.hayward@arm.com>
5633 David Sherwood <david.sherwood@arm.com>
5635 * doc/sourcebuild.texi (vect_scatter_store): Document.
5636 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5637 optabs.
5638 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5639 Document.
5640 * genopinit.c (main): Add supports_vec_scatter_store and
5641 supports_vec_scatter_store_cached to target_optabs.
5642 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5643 IFN_MASK_SCATTER_STORE.
5644 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5645 functions.
5646 * internal-fn.h (internal_store_fn_p): Declare.
5647 (internal_fn_stored_value_index): Likewise.
5648 * internal-fn.c (scatter_store_direct): New macro.
5649 (expand_scatter_store_optab_fn): New function.
5650 (direct_scatter_store_optab_supported_p): New macro.
5651 (internal_store_fn_p): New function.
5652 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5653 IFN_MASK_SCATTER_STORE.
5654 (internal_fn_mask_index): Likewise.
5655 (internal_fn_stored_value_index): New function.
5656 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5657 for scatter stores.
5658 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5659 * optabs-query.c (supports_vec_scatter_store_p): New function.
5660 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5661 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5662 true for scatter stores.
5663 (vect_gather_scatter_fn_p): Handle scatter stores too.
5664 (vect_check_gather_scatter): Consider using scatter stores if
5665 supports_vec_scatter_store_p.
5666 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5667 scatter stores too.
5668 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5669 internal_fn_stored_value_index.
5670 (check_load_store_masking): Handle scatter stores too.
5671 (vect_get_store_rhs): Make public.
5672 (vectorizable_call): Use internal_store_fn_p.
5673 (vectorizable_store): Handle scatter store internal functions.
5674 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5675 when deciding whether the end of the group has been reached.
5676 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5677 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5678 (mask_scatter_store<mode>): New insns.
5680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5681 Alan Hayward <alan.hayward@arm.com>
5682 David Sherwood <david.sherwood@arm.com>
5684 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5685 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5686 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5687 function.
5688 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5689 Use vect_truncate_gather_scatter_offset if we can't treat the
5690 operation as a normal gather load or scatter store.
5691 (get_group_load_store_type): Take the gather_scatter_info
5692 as argument. Try using a gather load or scatter store for
5693 single-element groups.
5694 (get_load_store_type): Update calls to get_group_load_store_type
5695 and vect_use_strided_gather_scatters_p.
5697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5698 Alan Hayward <alan.hayward@arm.com>
5699 David Sherwood <david.sherwood@arm.com>
5701 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5702 optional tree argument.
5703 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5704 null target hooks.
5705 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5706 but continue to use the current value as a fallback.
5707 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5708 to compare the updates.
5709 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5710 (get_load_store_type): Use it when handling a strided access.
5711 (vect_get_strided_load_store_ops): New function.
5712 (vect_get_data_ptr_increment): Likewise.
5713 (vectorizable_load): Handle strided gather loads. Always pass
5714 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5717 Alan Hayward <alan.hayward@arm.com>
5718 David Sherwood <david.sherwood@arm.com>
5720 * doc/md.texi (gather_load@var{m}): Document.
5721 (mask_gather_load@var{m}): Likewise.
5722 * genopinit.c (main): Add supports_vec_gather_load and
5723 supports_vec_gather_load_cached to target_optabs.
5724 * optabs-tree.c (init_tree_optimization_optabs): Use
5725 ggc_cleared_alloc to allocate target_optabs.
5726 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5727 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5728 functions.
5729 * internal-fn.h (internal_load_fn_p): Declare.
5730 (internal_gather_scatter_fn_p): Likewise.
5731 (internal_fn_mask_index): Likewise.
5732 (internal_gather_scatter_fn_supported_p): Likewise.
5733 * internal-fn.c (gather_load_direct): New macro.
5734 (expand_gather_load_optab_fn): New function.
5735 (direct_gather_load_optab_supported_p): New macro.
5736 (direct_internal_fn_optab): New function.
5737 (internal_load_fn_p): Likewise.
5738 (internal_gather_scatter_fn_p): Likewise.
5739 (internal_fn_mask_index): Likewise.
5740 (internal_gather_scatter_fn_supported_p): Likewise.
5741 * optabs-query.c (supports_at_least_one_mode_p): New function.
5742 (supports_vec_gather_load_p): Likewise.
5743 * optabs-query.h (supports_vec_gather_load_p): Declare.
5744 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5745 and memory_type field.
5746 (NUM_PATTERNS): Bump to 15.
5747 * tree-vect-data-refs.c: Include internal-fn.h.
5748 (vect_gather_scatter_fn_p): New function.
5749 (vect_describe_gather_scatter_call): Likewise.
5750 (vect_check_gather_scatter): Try using internal functions for
5751 gather loads. Recognize existing calls to a gather load function.
5752 (vect_analyze_data_refs): Consider using gather loads if
5753 supports_vec_gather_load_p.
5754 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5755 (vect_get_gather_scatter_offset_type): Likewise.
5756 (vect_convert_mask_for_vectype): Likewise.
5757 (vect_add_conversion_to_patterm): Likewise.
5758 (vect_try_gather_scatter_pattern): Likewise.
5759 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5760 (vect_vect_recog_func_ptrs): Add it.
5761 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5762 internal_fn_mask_index and internal_gather_scatter_fn_p.
5763 (check_load_store_masking): Take the gather_scatter_info as an
5764 argument and handle gather loads.
5765 (vect_get_gather_scatter_ops): New function.
5766 (vectorizable_call): Check internal_load_fn_p.
5767 (vectorizable_load): Likewise. Handle gather load internal
5768 functions.
5769 (vectorizable_store): Update call to check_load_store_masking.
5770 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5771 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5772 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5773 (aarch64_gather_scale_operand_d): New predicates.
5774 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5775 (mask_gather_load<mode>): New insns.
5777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5778 Alan Hayward <alan.hayward@arm.com>
5779 David Sherwood <david.sherwood@arm.com>
5781 * optabs.def (fold_left_plus_optab): New optab.
5782 * doc/md.texi (fold_left_plus_@var{m}): Document.
5783 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5784 * internal-fn.c (fold_left_direct): Define.
5785 (expand_fold_left_optab_fn): Likewise.
5786 (direct_fold_left_optab_supported_p): Likewise.
5787 * fold-const-call.c (fold_const_fold_left): New function.
5788 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5789 * tree-parloops.c (valid_reduction_p): New function.
5790 (gather_scalar_reductions): Use it.
5791 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5792 (vect_finish_replace_stmt): Declare.
5793 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5794 (needs_fold_left_reduction_p): New function, split out from...
5795 (vect_is_simple_reduction): ...here. Accept reductions that
5796 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5797 (vect_force_simple_reduction): Also store the reduction type in
5798 the assignment's STMT_VINFO_REDUC_TYPE.
5799 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5800 (merge_with_identity): New function.
5801 (vect_expand_fold_left): Likewise.
5802 (vectorize_fold_left_reduction): Likewise.
5803 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5804 scalar phi in place for it. Check for target support and reject
5805 cases that would reassociate the operation. Defer the transform
5806 phase to vectorize_fold_left_reduction.
5807 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5808 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5809 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5813 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5814 call to ifc_temp_var.
5816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5817 Alan Hayward <alan.hayward@arm.com>
5818 David Sherwood <david.sherwood@arm.com>
5820 * target.def (legitimize_address_displacement): Take the original
5821 offset as a poly_int.
5822 * targhooks.h (default_legitimize_address_displacement): Update
5823 accordingly.
5824 * targhooks.c (default_legitimize_address_displacement): Likewise.
5825 * doc/tm.texi: Regenerate.
5826 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5827 as an argument, moving assert of ad->disp == ad->disp_term to...
5828 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5829 Try calling targetm.legitimize_address_displacement before expanding
5830 the address rather than afterwards, and adjust for the new interface.
5831 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5832 Match the new hook interface. Handle SVE addresses.
5833 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5834 new hook interface.
5836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5838 * Makefile.in (OBJS): Add early-remat.o.
5839 * target.def (select_early_remat_modes): New hook.
5840 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5841 * doc/tm.texi: Regenerate.
5842 * targhooks.h (default_select_early_remat_modes): Declare.
5843 * targhooks.c (default_select_early_remat_modes): New function.
5844 * timevar.def (TV_EARLY_REMAT): New timevar.
5845 * passes.def (pass_early_remat): New pass.
5846 * tree-pass.h (make_pass_early_remat): Declare.
5847 * early-remat.c: New file.
5848 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5849 function.
5850 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5853 Alan Hayward <alan.hayward@arm.com>
5854 David Sherwood <david.sherwood@arm.com>
5856 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5857 vfm1 with a bound_epilog parameter.
5858 (vect_do_peeling): Update calls accordingly, and move the prologue
5859 call earlier in the function. Treat the base bound_epilog as 0 for
5860 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5861 this base when peeling for gaps.
5862 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5863 with fully-masked loops.
5864 (vect_estimate_min_profitable_iters): Handle the single peeled
5865 iteration in that case.
5867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5868 Alan Hayward <alan.hayward@arm.com>
5869 David Sherwood <david.sherwood@arm.com>
5871 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5872 single-element interleaving even if the size is not a power of 2.
5873 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5874 accesses for single-element interleaving if the group size is
5875 not a power of 2.
5877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5878 Alan Hayward <alan.hayward@arm.com>
5879 David Sherwood <david.sherwood@arm.com>
5881 * doc/md.texi (fold_extract_last_@var{m}): Document.
5882 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5883 * optabs.def (fold_extract_last_optab): New optab.
5884 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5885 * internal-fn.c (fold_extract_direct): New macro.
5886 (expand_fold_extract_optab_fn): Likewise.
5887 (direct_fold_extract_optab_supported_p): Likewise.
5888 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5889 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5890 EXTRACT_LAST_REDUCTION.
5891 (get_initial_def_for_reduction): Do not create an initial vector
5892 for EXTRACT_LAST_REDUCTION reductions.
5893 (vectorizable_reduction): Leave the scalar phi in place for
5894 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5895 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5896 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5897 transform phase to vectorizable_condition.
5898 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5899 split out from...
5900 (vect_finish_stmt_generation): ...here.
5901 (vect_finish_replace_stmt): New function.
5902 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5903 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5904 pattern.
5905 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5908 Alan Hayward <alan.hayward@arm.com>
5909 David Sherwood <david.sherwood@arm.com>
5911 * doc/md.texi (extract_last_@var{m}): Document.
5912 * optabs.def (extract_last_optab): New optab.
5913 * internal-fn.def (EXTRACT_LAST): New internal function.
5914 * internal-fn.c (cond_unary_direct): New macro.
5915 (expand_cond_unary_optab_fn): Likewise.
5916 (direct_cond_unary_optab_supported_p): Likewise.
5917 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5918 loops using EXTRACT_LAST.
5919 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5920 (extract_last_<mode>): ...this optab.
5921 (vec_extract<mode><Vel>): Update accordingly.
5923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5924 Alan Hayward <alan.hayward@arm.com>
5925 David Sherwood <david.sherwood@arm.com>
5927 * target.def (empty_mask_is_expensive): New hook.
5928 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5929 * doc/tm.texi: Regenerate.
5930 * targhooks.h (default_empty_mask_is_expensive): Declare.
5931 * targhooks.c (default_empty_mask_is_expensive): New function.
5932 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5933 if the target says that empty masks are expensive.
5934 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5935 New function.
5936 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5939 Alan Hayward <alan.hayward@arm.com>
5940 David Sherwood <david.sherwood@arm.com>
5942 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5943 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5944 (vect_use_loop_mask_for_alignment_p): New function.
5945 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5946 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5947 niters_skip argument. Make sure that the first niters_skip elements
5948 of the first iteration are inactive.
5949 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5950 Update call to vect_set_loop_masks_directly.
5951 (get_misalign_in_elems): New function, split out from...
5952 (vect_gen_prolog_loop_niters): ...here.
5953 (vect_update_init_of_dr): Take a code argument that specifies whether
5954 the adjustment should be added or subtracted.
5955 (vect_update_init_of_drs): Likewise.
5956 (vect_prepare_for_masked_peels): New function.
5957 (vect_do_peeling): Skip prologue peeling if we're using a mask
5958 instead. Update call to vect_update_inits_of_drs.
5959 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5960 mask_skip_niters.
5961 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5962 alignment. Do not include the number of peeled iterations in
5963 the minimum threshold in that case.
5964 (vectorizable_induction): Adjust the start value down by
5965 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5966 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5967 Take the number of skipped iterations into account when calculating
5968 the loop bounds.
5969 * tree-vect-stmts.c (vect_gen_while_not): New function.
5971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5972 Alan Hayward <alan.hayward@arm.com>
5973 David Sherwood <david.sherwood@arm.com>
5975 * doc/sourcebuild.texi (vect_fully_masked): Document.
5976 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5977 default value to 0.
5978 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5979 split out from...
5980 (vect_analyze_loop_2): ...here. Don't check the vectorization
5981 factor against the number of loop iterations if the loop is
5982 fully-masked.
5984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5985 Alan Hayward <alan.hayward@arm.com>
5986 David Sherwood <david.sherwood@arm.com>
5988 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5989 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5990 (dump_groups): Update accordingly.
5991 (iv_use::mem_type): New member variable.
5992 (address_p): New function.
5993 (record_use): Add a mem_type argument and initialize the new
5994 mem_type field.
5995 (record_group_use): Add a mem_type argument. Use address_p.
5996 Remove obsolete null checks of base_object. Update call to record_use.
5997 (find_interesting_uses_op): Update call to record_group_use.
5998 (find_interesting_uses_cond): Likewise.
5999 (find_interesting_uses_address): Likewise.
6000 (get_mem_type_for_internal_fn): New function.
6001 (find_address_like_use): Likewise.
6002 (find_interesting_uses_stmt): Try find_address_like_use before
6003 calling find_interesting_uses_op.
6004 (addr_offset_valid_p): Use the iv mem_type field as the type
6005 of the addressed memory.
6006 (add_autoinc_candidates): Likewise.
6007 (get_address_cost): Likewise.
6008 (split_small_address_groups_p): Use address_p.
6009 (split_address_groups): Likewise.
6010 (add_iv_candidate_for_use): Likewise.
6011 (autoinc_possible_for_pair): Likewise.
6012 (rewrite_groups): Likewise.
6013 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6014 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6015 (get_alias_ptr_type_for_ptr_address): New function.
6016 (rewrite_use_address): Rewrite address uses in calls that were
6017 identified by find_address_like_use.
6019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6020 Alan Hayward <alan.hayward@arm.com>
6021 David Sherwood <david.sherwood@arm.com>
6023 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6024 TARGET_MEM_REFs.
6025 * gimple-expr.h (is_gimple_addressable: Likewise.
6026 * gimple-expr.c (is_gimple_address): Likewise.
6027 * internal-fn.c (expand_call_mem_ref): New function.
6028 (expand_mask_load_optab_fn): Use it.
6029 (expand_mask_store_optab_fn): Likewise.
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/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6036 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6037 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6038 (cond_umax@var{mode}): Document.
6039 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6040 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6041 (cond_umin_optab, cond_umax_optab): New optabs.
6042 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6043 (COND_IOR, COND_XOR): New internal functions.
6044 * internal-fn.h (get_conditional_internal_fn): Declare.
6045 * internal-fn.c (cond_binary_direct): New macro.
6046 (expand_cond_binary_optab_fn): Likewise.
6047 (direct_cond_binary_optab_supported_p): Likewise.
6048 (get_conditional_internal_fn): New function.
6049 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6050 Cope with reduction statements that are vectorized as calls rather
6051 than assignments.
6052 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6053 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6054 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6055 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6056 (UNSPEC_COND_EOR): New unspecs.
6057 (optab): Add mappings for them.
6058 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6059 (sve_int_op, sve_fp_op): New int attributes.
6061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6062 Alan Hayward <alan.hayward@arm.com>
6063 David Sherwood <david.sherwood@arm.com>
6065 * optabs.def (while_ult_optab): New optab.
6066 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6067 * internal-fn.def (WHILE_ULT): New internal function.
6068 * internal-fn.h (direct_internal_fn_supported_p): New override
6069 that takes two types as argument.
6070 * internal-fn.c (while_direct): New macro.
6071 (expand_while_optab_fn): New function.
6072 (convert_optab_supported_p): Likewise.
6073 (direct_while_optab_supported_p): New macro.
6074 * wide-int.h (wi::udiv_ceil): New function.
6075 * tree-vectorizer.h (rgroup_masks): New structure.
6076 (vec_loop_masks): New typedef.
6077 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6078 and fully_masked_p.
6079 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6080 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6081 (vect_max_vf): New function.
6082 (slpeel_make_loop_iterate_ntimes): Delete.
6083 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6084 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6085 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6086 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6087 internal-fn.h, stor-layout.h and optabs-query.h.
6088 (vect_set_loop_mask): New function.
6089 (add_preheader_seq): Likewise.
6090 (add_header_seq): Likewise.
6091 (interleave_supported_p): Likewise.
6092 (vect_maybe_permute_loop_masks): Likewise.
6093 (vect_set_loop_masks_directly): Likewise.
6094 (vect_set_loop_condition_masked): Likewise.
6095 (vect_set_loop_condition_unmasked): New function, split out from
6096 slpeel_make_loop_iterate_ntimes.
6097 (slpeel_make_loop_iterate_ntimes): Rename to..
6098 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6099 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6100 (vect_do_peeling): Update call accordingly.
6101 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6102 loops.
6103 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6104 mask_compare_type, can_fully_mask_p and fully_masked_p.
6105 (release_vec_loop_masks): New function.
6106 (_loop_vec_info): Use it to free the loop masks.
6107 (can_produce_all_loop_masks_p): New function.
6108 (vect_get_max_nscalars_per_iter): Likewise.
6109 (vect_verify_full_masking): Likewise.
6110 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6111 retries, and free the mask rgroups before retrying. Check loop-wide
6112 reasons for disallowing fully-masked loops. Make the final decision
6113 about whether use a fully-masked loop or not.
6114 (vect_estimate_min_profitable_iters): Do not assume that peeling
6115 for the number of iterations will be needed for fully-masked loops.
6116 (vectorizable_reduction): Disable fully-masked loops.
6117 (vectorizable_live_operation): Likewise.
6118 (vect_halve_mask_nunits): New function.
6119 (vect_double_mask_nunits): Likewise.
6120 (vect_record_loop_mask): Likewise.
6121 (vect_get_loop_mask): Likewise.
6122 (vect_transform_loop): Handle the case in which the final loop
6123 iteration might handle a partial vector. Call vect_set_loop_condition
6124 instead of slpeel_make_loop_iterate_ntimes.
6125 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6126 (check_load_store_masking): New function.
6127 (prepare_load_store_mask): Likewise.
6128 (vectorizable_store): Handle fully-masked loops.
6129 (vectorizable_load): Likewise.
6130 (supportable_widening_operation): Use vect_halve_mask_nunits for
6131 booleans.
6132 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6133 (vect_gen_while): New function.
6134 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6135 (aarch64_uqdec<mode>): New insn.
6137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6138 Alan Hayward <alan.hayward@arm.com>
6139 David Sherwood <david.sherwood@arm.com>
6141 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6142 (reduc_xor_scal_optab): New optabs.
6143 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6144 (reduc_xor_scal_@var{m}): Document.
6145 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6146 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6147 internal functions.
6148 * fold-const-call.c (fold_const_call): Handle them.
6149 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6150 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6151 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6152 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6153 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6154 (UNSPEC_XORV): New unspecs.
6155 (optab): Add entries for them.
6156 (BITWISEV): New int iterator.
6157 (bit_reduc_op): New int attributes.
6159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6160 Alan Hayward <alan.hayward@arm.com>
6161 David Sherwood <david.sherwood@arm.com>
6163 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6164 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6165 * optabs.def (vec_shl_insert_optab): New optab.
6166 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6167 (duplicate_and_interleave): Likewise.
6168 * tree-vect-loop.c: Include internal-fn.h.
6169 (neutral_op_for_slp_reduction): New function, split out from
6170 get_initial_defs_for_reduction.
6171 (get_initial_def_for_reduction): Handle option 2 for variable-length
6172 vectors by loading the neutral value into a vector and then shifting
6173 the initial value into element 0.
6174 (get_initial_defs_for_reduction): Replace the code argument with
6175 the neutral value calculated by neutral_op_for_slp_reduction.
6176 Use gimple_build_vector for constant-length vectors.
6177 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6178 but the first group_size elements have a neutral value.
6179 Use duplicate_and_interleave otherwise.
6180 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6181 Update call to get_initial_defs_for_reduction. Handle SLP
6182 reductions for variable-length vectors by creating one vector
6183 result for each scalar result, with the elements associated
6184 with other scalar results stubbed out with the neutral value.
6185 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6186 Require IFN_VEC_SHL_INSERT for double reductions on
6187 variable-length vectors, or SLP reductions that have
6188 a neutral value. Require can_duplicate_and_interleave_p
6189 support for variable-length unchained SLP reductions if there
6190 is no neutral value, such as for MIN/MAX reductions. Also require
6191 the number of vector elements to be a multiple of the number of
6192 SLP statements when doing variable-length unchained SLP reductions.
6193 Update call to vect_create_epilog_for_reduction.
6194 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6195 and remove initial values.
6196 (duplicate_and_interleave): Make public.
6197 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6198 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6201 Alan Hayward <alan.hayward@arm.com>
6202 David Sherwood <david.sherwood@arm.com>
6204 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6205 (can_duplicate_and_interleave_p): New function.
6206 (vect_get_and_check_slp_defs): Take the vector of statements
6207 rather than just the current one. Remove excess parentheses.
6208 Restriction rejectinon of vect_constant_def and vect_external_def
6209 for variable-length vectors to boolean types, or types for which
6210 can_duplicate_and_interleave_p is false.
6211 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6212 (duplicate_and_interleave): New function.
6213 (vect_get_constant_vectors): Use gimple_build_vector for
6214 constant-length vectors and suitable variable-length constant
6215 vectors. Use duplicate_and_interleave for other variable-length
6216 vectors. Don't defer the update when inserting new statements.
6218 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6219 Alan Hayward <alan.hayward@arm.com>
6220 David Sherwood <david.sherwood@arm.com>
6222 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6223 min_profitable_iters doesn't go negative.
6225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6226 Alan Hayward <alan.hayward@arm.com>
6227 David Sherwood <david.sherwood@arm.com>
6229 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6230 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6231 * optabs.def (vec_mask_load_lanes_optab): New optab.
6232 (vec_mask_store_lanes_optab): Likewise.
6233 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6234 (MASK_STORE_LANES): Likewise.
6235 * internal-fn.c (mask_load_lanes_direct): New macro.
6236 (mask_store_lanes_direct): Likewise.
6237 (expand_mask_load_optab_fn): Handle masked operations.
6238 (expand_mask_load_lanes_optab_fn): New macro.
6239 (expand_mask_store_optab_fn): Handle masked operations.
6240 (expand_mask_store_lanes_optab_fn): New macro.
6241 (direct_mask_load_lanes_optab_supported_p): Likewise.
6242 (direct_mask_store_lanes_optab_supported_p): Likewise.
6243 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6244 parameter.
6245 (vect_load_lanes_supported): Likewise.
6246 * tree-vect-data-refs.c (strip_conversion): New function.
6247 (can_group_stmts_p): Likewise.
6248 (vect_analyze_data_ref_accesses): Use it instead of checking
6249 for a pair of assignments.
6250 (vect_store_lanes_supported): Take a masked_p parameter.
6251 (vect_load_lanes_supported): Likewise.
6252 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6253 vect_store_lanes_supported and vect_load_lanes_supported.
6254 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6255 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6256 parameter. Don't allow gaps for masked accesses.
6257 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6258 and vect_load_lanes_supported.
6259 (get_load_store_type): Take a masked_p parameter and update
6260 call to get_group_load_store_type.
6261 (vectorizable_store): Update call to get_load_store_type.
6262 Handle IFN_MASK_STORE_LANES.
6263 (vectorizable_load): Update call to get_load_store_type.
6264 Handle IFN_MASK_LOAD_LANES.
6266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6267 Alan Hayward <alan.hayward@arm.com>
6268 David Sherwood <david.sherwood@arm.com>
6270 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6271 modes for SVE.
6272 * config/aarch64/aarch64-protos.h
6273 (aarch64_sve_struct_memory_operand_p): Declare.
6274 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6275 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6276 (VPRED, vpred): Handle SVE structure modes.
6277 * config/aarch64/constraints.md (Utx): New constraint.
6278 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6279 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6280 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6281 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6282 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6283 structure modes. Split into pieces after RA.
6284 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6285 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6286 New patterns.
6287 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6288 SVE structure modes.
6289 (aarch64_classify_address): Likewise.
6290 (sizetochar): Move earlier in file.
6291 (aarch64_print_operand): Handle SVE register lists.
6292 (aarch64_array_mode): New function.
6293 (aarch64_sve_struct_memory_operand_p): Likewise.
6294 (TARGET_ARRAY_MODE): Redefine.
6296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6297 Alan Hayward <alan.hayward@arm.com>
6298 David Sherwood <david.sherwood@arm.com>
6300 * target.def (array_mode): New target hook.
6301 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6302 * doc/tm.texi: Regenerate.
6303 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6304 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6305 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6306 targetm.array_mode.
6307 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6308 type sizes.
6310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6311 Alan Hayward <alan.hayward@arm.com>
6312 David Sherwood <david.sherwood@arm.com>
6314 * fold-const.c (fold_binary_loc): Check the argument types
6315 rather than the result type when testing for a vector operation.
6317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6319 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6320 * doc/tm.texi: Regenerate.
6322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6323 Alan Hayward <alan.hayward@arm.com>
6324 David Sherwood <david.sherwood@arm.com>
6326 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6327 (sve): Document new AArch64 extension.
6328 * doc/md.texi (w): Extend the description of the AArch64
6329 constraint to include SVE vectors.
6330 (Upl, Upa): Document new AArch64 predicate constraints.
6331 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6332 enum.
6333 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6334 (msve-vector-bits=): New option.
6335 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6336 SVE when these are disabled.
6337 (sve): New extension.
6338 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6339 modes. Adjust their number of units based on aarch64_sve_vg.
6340 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6341 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6342 aarch64_addr_query_type.
6343 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6344 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6345 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6346 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6347 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6348 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6349 (aarch64_simd_imm_zero_p): Delete.
6350 (aarch64_check_zero_based_sve_index_immediate): Declare.
6351 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6352 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6353 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6354 (aarch64_sve_float_mul_immediate_p): Likewise.
6355 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6356 rather than an rtx.
6357 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6358 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6359 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6360 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6361 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6362 (aarch64_regmode_natural_size): Likewise.
6363 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6364 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6365 left one place.
6366 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6367 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6368 for VG and the SVE predicate registers.
6369 (V_ALIASES): Add a "z"-prefixed alias.
6370 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6371 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6372 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6373 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6374 (REG_CLASS_NAMES): Add entries for them.
6375 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6376 and the predicate registers.
6377 (aarch64_sve_vg): Declare.
6378 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6379 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6380 (REGMODE_NATURAL_SIZE): Define.
6381 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6382 SVE macros.
6383 * config/aarch64/aarch64.c: Include cfgrtl.h.
6384 (simd_immediate_info): Add a constructor for series vectors,
6385 and an associated step field.
6386 (aarch64_sve_vg): New variable.
6387 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6388 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6389 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6390 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6391 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6392 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6393 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6394 (aarch64_get_mask_mode): New functions.
6395 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6396 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6397 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6398 predicate modes and predicate registers. Explicitly restrict
6399 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6400 to store a vector mode if it is recognized by
6401 aarch64_classify_vector_mode.
6402 (aarch64_regmode_natural_size): New function.
6403 (aarch64_hard_regno_caller_save_mode): Return the original mode
6404 for predicates.
6405 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6406 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6407 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6408 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6409 functions.
6410 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6411 does not overlap dest if the function is frame-related. Handle
6412 SVE constants.
6413 (aarch64_split_add_offset): New function.
6414 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6415 them aarch64_add_offset.
6416 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6417 and update call to aarch64_sub_sp.
6418 (aarch64_add_cfa_expression): New function.
6419 (aarch64_expand_prologue): Pass extra temporary registers to the
6420 functions above. Handle the case in which we need to emit new
6421 DW_CFA_expressions for registers that were originally saved
6422 relative to the stack pointer, but now have to be expressed
6423 relative to the frame pointer.
6424 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6425 functions above.
6426 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6427 IP0 and IP1 values for SVE frames.
6428 (aarch64_expand_vec_series): New function.
6429 (aarch64_expand_sve_widened_duplicate): Likewise.
6430 (aarch64_expand_sve_const_vector): Likewise.
6431 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6432 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6433 into the register, rather than emitting a SET directly.
6434 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6435 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6436 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6437 (offset_9bit_signed_scaled_p): New functions.
6438 (aarch64_replicate_bitmask_imm): New function.
6439 (aarch64_bitmask_imm): Use it.
6440 (aarch64_cannot_force_const_mem): Reject expressions involving
6441 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6442 (aarch64_classify_index): Handle SVE indices, by requiring
6443 a plain register index with a scale that matches the element size.
6444 (aarch64_classify_address): Handle SVE addresses. Assert that
6445 the mode of the address is VOIDmode or an integer mode.
6446 Update call to aarch64_classify_symbol.
6447 (aarch64_classify_symbolic_expression): Update call to
6448 aarch64_classify_symbol.
6449 (aarch64_const_vec_all_in_range_p): New function.
6450 (aarch64_print_vector_float_operand): Likewise.
6451 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6452 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6453 and the FP immediates 1.0 and 0.5.
6454 (aarch64_print_address_internal): Handle SVE addresses.
6455 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6456 (aarch64_regno_regclass): Handle predicate registers.
6457 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6458 data modes.
6459 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6460 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6461 (aarch64_convert_sve_vector_bits): New function.
6462 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6463 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6464 rather than an rtx.
6465 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6466 Handle SVE vector and predicate modes. Accept VL-based constants
6467 that need only one temporary register, and VL offsets that require
6468 no temporary registers.
6469 (aarch64_conditional_register_usage): Mark the predicate registers
6470 as fixed if SVE isn't available.
6471 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6472 Return true for SVE vector and predicate modes.
6473 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6474 rather than an unsigned int. Handle SVE modes.
6475 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6476 SVE modes.
6477 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6478 if SVE is enabled.
6479 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6480 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6481 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6482 (aarch64_sve_float_mul_immediate_p): New functions.
6483 (aarch64_sve_valid_immediate): New function.
6484 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6485 Explicitly reject structure modes. Check for INDEX constants.
6486 Handle PTRUE and PFALSE constants.
6487 (aarch64_check_zero_based_sve_index_immediate): New function.
6488 (aarch64_simd_imm_zero_p): Delete.
6489 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6490 vector modes. Accept constants in the range of CNT[BHWD].
6491 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6492 ask for an Advanced SIMD mode.
6493 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6494 (aarch64_simd_vector_alignment): Handle SVE predicates.
6495 (aarch64_vectorize_preferred_vector_alignment): New function.
6496 (aarch64_simd_vector_alignment_reachable): Use it instead of
6497 the vector size.
6498 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6499 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6500 functions.
6501 (MAX_VECT_LEN): Delete.
6502 (expand_vec_perm_d): Add a vec_flags field.
6503 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6504 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6505 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6506 for SVE modes.
6507 (aarch64_evpc_rev): Rename to...
6508 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6509 (aarch64_evpc_rev_global): New function.
6510 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6511 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6512 MAX_VECT_LEN.
6513 (aarch64_evpc_sve_tbl): New function.
6514 (aarch64_expand_vec_perm_const_1): Update after rename of
6515 aarch64_evpc_rev. Handle SVE permutes too, trying
6516 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6517 than aarch64_evpc_tbl.
6518 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6519 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6520 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6521 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6522 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6523 (aarch64_expand_sve_vcond): New functions.
6524 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6525 of aarch64_vector_mode_p.
6526 (aarch64_dwarf_poly_indeterminate_value): New function.
6527 (aarch64_compute_pressure_classes): Likewise.
6528 (aarch64_can_change_mode_class): Likewise.
6529 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6530 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6531 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6532 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6533 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6534 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6535 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6536 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6537 constraints.
6538 (Dn, Dl, Dr): Accept const as well as const_vector.
6539 (Dz): Likewise. Compare against CONST0_RTX.
6540 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6541 of "vector" where appropriate.
6542 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6543 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6544 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6545 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6546 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6547 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6548 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6549 (v_int_equiv): Extend to SVE modes.
6550 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6551 mode attributes.
6552 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6553 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6554 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6555 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6556 (SVE_COND_FP_CMP): New int iterators.
6557 (perm_hilo): Handle the new unpack unspecs.
6558 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6559 attributes.
6560 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6561 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6562 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6563 (aarch64_equality_operator, aarch64_constant_vector_operand)
6564 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6565 (aarch64_sve_nonimmediate_operand): Likewise.
6566 (aarch64_sve_general_operand): Likewise.
6567 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6568 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6569 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6570 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6571 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6572 (aarch64_sve_float_arith_immediate): Likewise.
6573 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6574 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6575 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6576 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6577 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6578 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6579 (aarch64_sve_float_arith_operand): Likewise.
6580 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6581 (aarch64_sve_float_mul_operand): Likewise.
6582 (aarch64_sve_vec_perm_operand): Likewise.
6583 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6584 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6585 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6586 as well as const_vector.
6587 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6588 in file. Use CONST0_RTX and CONSTM1_RTX.
6589 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6590 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6591 Use aarch64_simd_imm_zero.
6592 * config/aarch64/aarch64-sve.md: New file.
6593 * config/aarch64/aarch64.md: Include it.
6594 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6595 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6596 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6597 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6598 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6599 (sve): New attribute.
6600 (enabled): Disable instructions with the sve attribute unless
6601 TARGET_SVE.
6602 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6603 aarch64_expand_mov_immediate.
6604 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6605 CNT[BHSD] immediates.
6606 (movti): Split CONST_POLY_INT moves into two halves.
6607 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6608 Split additions that need a temporary here if the destination
6609 is the stack pointer.
6610 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6611 (*add<mode>3_poly_1): New instruction.
6612 (set_clobber_cc): New expander.
6614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6616 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6617 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6618 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6619 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6620 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6621 Change innermode from fixed_mode_size to machine_mode.
6622 (simplify_subreg): Update call accordingly. Handle a constant-sized
6623 subreg of a variable-length CONST_VECTOR.
6625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6626 Alan Hayward <alan.hayward@arm.com>
6627 David Sherwood <david.sherwood@arm.com>
6629 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6630 (add_offset_to_base): New function, split out from...
6631 (create_mem_ref): ...here. When handling a scale other than 1,
6632 check first whether the address is valid without the offset.
6633 Add it into the base if so, leaving the index and scale as-is.
6635 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6637 PR c++/83778
6638 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6639 fold_for_warn before checking if arg2 is INTEGER_CST.
6641 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6643 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6644 (store_multiple_operation): Delete.
6645 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6646 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6647 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6648 guarded by TARGET_STRING.
6649 (rs6000_output_load_multiple): Delete.
6650 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6651 OPTION_MASK_STRING / TARGET_STRING handling.
6652 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6653 (const rs6000_opt_masks) <"string">: Change mask to 0.
6654 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6655 (MASK_STRING): Delete.
6656 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6657 parts. Simplify.
6658 (load_multiple): Delete.
6659 (*ldmsi8): Delete.
6660 (*ldmsi7): Delete.
6661 (*ldmsi6): Delete.
6662 (*ldmsi5): Delete.
6663 (*ldmsi4): Delete.
6664 (*ldmsi3): Delete.
6665 (store_multiple): Delete.
6666 (*stmsi8): Delete.
6667 (*stmsi7): Delete.
6668 (*stmsi6): Delete.
6669 (*stmsi5): Delete.
6670 (*stmsi4): Delete.
6671 (*stmsi3): Delete.
6672 (movmemsi_8reg): Delete.
6673 (corresponding unnamed define_insn): Delete.
6674 (movmemsi_6reg): Delete.
6675 (corresponding unnamed define_insn): Delete.
6676 (movmemsi_4reg): Delete.
6677 (corresponding unnamed define_insn): Delete.
6678 (movmemsi_2reg): Delete.
6679 (corresponding unnamed define_insn): Delete.
6680 (movmemsi_1reg): Delete.
6681 (corresponding unnamed define_insn): Delete.
6682 * config/rs6000/rs6000.opt (mno-string): New.
6683 (mstring): Replace by deprecation warning stub.
6684 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6686 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6688 * regrename.c (regrename_do_replace): If replacing the same
6689 reg multiple times, try to reuse last created gen_raw_REG.
6691 PR debug/81155
6692 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6693 main to workaround a bug in GDB.
6695 2018-01-12 Tom de Vries <tom@codesourcery.com>
6697 PR target/83737
6698 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6700 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6702 PR rtl-optimization/80481
6703 * ira-color.c (get_cap_member): New function.
6704 (allocnos_conflict_by_live_ranges_p): Use it.
6705 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6706 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6708 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6710 PR target/83628
6711 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6712 (*saddl_se_1): Ditto.
6713 (*ssubsi_1): Ditto.
6714 (*ssubl_se_1): Ditto.
6716 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6718 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6719 rather than wi::to_widest for DR_INITs.
6720 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6721 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6722 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6723 INTEGER_CSTs.
6724 (vect_analyze_group_access_1): Note that here.
6726 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6728 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6729 polynomial type sizes.
6731 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6733 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6734 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6735 (gimple_add_tmp_var): Likewise.
6737 2018-01-12 Martin Liska <mliska@suse.cz>
6739 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6740 (gimple_alloc_sizes): Likewise.
6741 (dump_gimple_statistics): Use PRIu64 in printf format.
6742 * gimple.h: Change uint64_t to int.
6744 2018-01-12 Martin Liska <mliska@suse.cz>
6746 * tree-core.h: Use uint64_t instead of int.
6747 * tree.c (tree_node_counts): Likewise.
6748 (tree_node_sizes): Likewise.
6749 (dump_tree_statistics): Use PRIu64 in printf format.
6751 2018-01-12 Martin Liska <mliska@suse.cz>
6753 * Makefile.in: As qsort_chk is implemented in vec.c, add
6754 vec.o to linkage of gencfn-macros.
6755 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6756 passing the info to record_node_allocation_statistics.
6757 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6758 and pass the info.
6759 * ggc-common.c (struct ggc_usage): Add operator== and use
6760 it in operator< and compare function.
6761 * mem-stats.h (struct mem_usage): Likewise.
6762 * vec.c (struct vec_usage): Remove operator< and compare
6763 function. Can be simply inherited.
6765 2018-01-12 Martin Jambor <mjambor@suse.cz>
6767 PR target/81616
6768 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6769 * tree-ssa-math-opts.c: Include domwalk.h.
6770 (convert_mult_to_fma_1): New function.
6771 (fma_transformation_info): New type.
6772 (fma_deferring_state): Likewise.
6773 (cancel_fma_deferring): New function.
6774 (result_of_phi): Likewise.
6775 (last_fma_candidate_feeds_initial_phi): Likewise.
6776 (convert_mult_to_fma): Added deferring logic, split actual
6777 transformation to convert_mult_to_fma_1.
6778 (math_opts_dom_walker): New type.
6779 (math_opts_dom_walker::after_dom_children): New method, body moved
6780 here from pass_optimize_widening_mul::execute, added deferring logic
6781 bits.
6782 (pass_optimize_widening_mul::execute): Moved most of code to
6783 math_opts_dom_walker::after_dom_children.
6784 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6785 * config/i386/i386.c (ix86_option_override_internal): Added
6786 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6788 2018-01-12 Richard Biener <rguenther@suse.de>
6790 PR debug/83157
6791 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6792 inline instance vars.
6794 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6796 PR target/81819
6797 * config/rx/rx.c (rx_is_restricted_memory_address):
6798 Handle SUBREG case.
6800 2018-01-12 Richard Biener <rguenther@suse.de>
6802 PR tree-optimization/80846
6803 * target.def (split_reduction): New target hook.
6804 * targhooks.c (default_split_reduction): New function.
6805 * targhooks.h (default_split_reduction): Declare.
6806 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6807 target requests first reduce vectors by combining low and high
6808 parts.
6809 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6810 (get_vectype_for_scalar_type_and_size): Export.
6811 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6812 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6813 * doc/tm.texi: Regenerate.
6814 * config/i386/i386.c (ix86_split_reduction): Implement
6815 TARGET_VECTORIZE_SPLIT_REDUCTION.
6817 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6819 PR target/83368
6820 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6821 in PIC mode except for TARGET_VXWORKS_RTP.
6822 * config/sparc/sparc.c: Include cfgrtl.h.
6823 (TARGET_INIT_PIC_REG): Define.
6824 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6825 (sparc_pic_register_p): New predicate.
6826 (sparc_legitimate_address_p): Use it.
6827 (sparc_legitimize_pic_address): Likewise.
6828 (sparc_delegitimize_address): Likewise.
6829 (sparc_mode_dependent_address_p): Likewise.
6830 (gen_load_pcrel_sym): Remove 4th parameter.
6831 (load_got_register): Adjust call to above. Remove obsolete stuff.
6832 (sparc_expand_prologue): Do not call load_got_register here.
6833 (sparc_flat_expand_prologue): Likewise.
6834 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6835 (sparc_use_pseudo_pic_reg): New function.
6836 (sparc_init_pic_reg): Likewise.
6837 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6838 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6840 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6842 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6843 Add item for branch_cost.
6845 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6847 PR rtl-optimization/83565
6848 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6849 not extend the result to a larger mode for rotate operations.
6850 (num_sign_bit_copies1): Likewise.
6852 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6854 PR target/40411
6855 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6856 -symbolic.
6857 Use values-Xc.o for -pedantic.
6858 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6860 2018-01-12 Martin Liska <mliska@suse.cz>
6862 PR ipa/83054
6863 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6864 New function.
6865 (possible_polymorphic_call_targets): Use it.
6866 (ipa_devirt): Likewise.
6868 2018-01-12 Martin Liska <mliska@suse.cz>
6870 * profile-count.h (enum profile_quality): Use 0 as invalid
6871 enum value of profile_quality.
6873 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6875 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6876 -mext-string options.
6878 2018-01-12 Richard Biener <rguenther@suse.de>
6880 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6881 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6882 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6883 Likewise.
6884 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6886 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6888 * configure.ac (--with-long-double-format): Add support for the
6889 configuration option to change the default long double format on
6890 PowerPC systems.
6891 * config.gcc (powerpc*-linux*-*): Likewise.
6892 * configure: Regenerate.
6893 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6894 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6895 used without modification.
6897 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6899 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6900 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6901 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6902 MISC_BUILTIN_SPEC_BARRIER.
6903 (rs6000_init_builtins): Likewise.
6904 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6905 enum value.
6906 (speculation_barrier): New define_insn.
6907 * doc/extend.texi: Document __builtin_speculation_barrier.
6909 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6911 PR target/83203
6912 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6913 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6914 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6915 iterators.
6916 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6917 integral modes instead of "ss" and "sd".
6918 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6919 vectors with 32-bit and 64-bit elements.
6920 (vecdupssescalarmodesuffix): New mode attribute.
6921 (vec_dup<mode>): Use it.
6923 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6925 PR target/83330
6926 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6927 frame if argument is passed on stack.
6929 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6931 PR target/82682
6932 * ree.c (combine_reaching_defs): Optimize also
6933 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6934 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6936 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6938 PR middle-end/83189
6939 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6941 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6943 PR middle-end/83718
6944 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6945 after they are computed.
6947 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6949 PR tree-optimization/83695
6950 * gimple-loop-linterchange.cc
6951 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6952 reset cached scev information after interchange.
6953 (pass_linterchange::execute): Remove call to scev_reset_htab.
6955 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6957 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6958 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6959 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6960 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6961 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6962 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6963 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6964 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6965 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6966 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6967 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6968 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6969 (V_lane_reg): Likewise.
6970 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6971 New define_expand.
6972 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6973 (vfmal_lane_low<mode>_intrinsic,
6974 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6975 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6976 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6977 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6978 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6979 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6981 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6983 * config/arm/arm-cpus.in (fp16fml): New feature.
6984 (ALL_SIMD): Add fp16fml.
6985 (armv8.2-a): Add fp16fml as an option.
6986 (armv8.3-a): Likewise.
6987 (armv8.4-a): Add fp16fml as part of fp16.
6988 * config/arm/arm.h (TARGET_FP16FML): Define.
6989 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6990 when appropriate.
6991 * config/arm/arm-modes.def (V2HF): Define.
6992 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6993 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6994 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6995 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6996 vfmsl_low, vfmsl_high): New set of builtins.
6997 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6998 (vfml_op): New code attribute.
6999 (VFMLHALVES): New int iterator.
7000 (VFML, VFMLSEL): New mode attributes.
7001 (V_reg): Define mapping for V2HF.
7002 (V_hi, V_lo): New mode attributes.
7003 (VF_constraint): Likewise.
7004 (vfml_half, vfml_half_selector): New int attributes.
7005 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7006 define_expand.
7007 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7008 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7009 New define_insn.
7010 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7011 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7012 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7013 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7014 documentation.
7015 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7016 Document new effective target and option set.
7018 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7020 * config/arm/arm-cpus.in (armv8_4): New feature.
7021 (ARMv8_4a): New fgroup.
7022 (armv8.4-a): New arch.
7023 * config/arm/arm-tables.opt: Regenerate.
7024 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7025 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7026 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7027 Add matching rules for -march=armv8.4-a and extensions.
7028 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7030 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7032 PR target/81821
7033 * config/rx/rx.md (BW): New mode attribute.
7034 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7036 2018-01-11 Richard Biener <rguenther@suse.de>
7038 PR tree-optimization/83435
7039 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7040 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7041 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7043 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7044 Alan Hayward <alan.hayward@arm.com>
7045 David Sherwood <david.sherwood@arm.com>
7047 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7048 field.
7049 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7050 (aarch64_print_address_internal): Use it to check for a zero offset.
7052 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7053 Alan Hayward <alan.hayward@arm.com>
7054 David Sherwood <david.sherwood@arm.com>
7056 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7057 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7058 Return a poly_int64 rather than a HOST_WIDE_INT.
7059 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7060 rather than a HOST_WIDE_INT.
7061 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7062 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7063 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7064 final_offset from HOST_WIDE_INT to poly_int64.
7065 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7066 to_constant when getting the number of units in an Advanced SIMD
7067 mode.
7068 (aarch64_builtin_vectorized_function): Check for a constant number
7069 of units.
7070 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7071 GET_MODE_SIZE.
7072 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7073 attribute instead of GET_MODE_NUNITS.
7074 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7075 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7076 GET_MODE_SIZE for fixed-size registers.
7077 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7078 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7079 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7080 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7081 (aarch64_print_operand, aarch64_print_address_internal)
7082 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7083 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7084 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7085 Handle polynomial GET_MODE_SIZE.
7086 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7087 wider than SImode without modification.
7088 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7089 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7090 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7091 passing and returning SVE modes.
7092 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7093 rather than GEN_INT.
7094 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7095 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7096 (aarch64_allocate_and_probe_stack_space): Likewise.
7097 (aarch64_layout_frame): Cope with polynomial offsets.
7098 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7099 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7100 polynomial offsets.
7101 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7102 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7103 poly_int64 rather than a HOST_WIDE_INT.
7104 (aarch64_get_separate_components, aarch64_process_components)
7105 (aarch64_expand_prologue, aarch64_expand_epilogue)
7106 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7107 (aarch64_anchor_offset): New function, split out from...
7108 (aarch64_legitimize_address): ...here.
7109 (aarch64_builtin_vectorization_cost): Handle polynomial
7110 TYPE_VECTOR_SUBPARTS.
7111 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7112 GET_MODE_NUNITS.
7113 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7114 number of elements from the PARALLEL rather than the mode.
7115 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7116 rather than GET_MODE_BITSIZE.
7117 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7118 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7119 (aarch64_expand_vec_perm_const_1): Handle polynomial
7120 d->perm.length () and d->perm elements.
7121 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7122 Apply to_constant to d->perm elements.
7123 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7124 polynomial CONST_VECTOR_NUNITS.
7125 (aarch64_move_pointer): Take amount as a poly_int64 rather
7126 than an int.
7127 (aarch64_progress_pointer): Avoid temporary variable.
7128 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7129 the mode attribute instead of GET_MODE.
7131 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7132 Alan Hayward <alan.hayward@arm.com>
7133 David Sherwood <david.sherwood@arm.com>
7135 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7136 x exists before using it.
7137 (aarch64_add_constant_internal): Rename to...
7138 (aarch64_add_offset_1): ...this. Replace regnum with separate
7139 src and dest rtxes. Handle the case in which they're different,
7140 including when the offset is zero. Replace scratchreg with an rtx.
7141 Use 2 additions if there is no spare register into which we can
7142 move a 16-bit constant.
7143 (aarch64_add_constant): Delete.
7144 (aarch64_add_offset): Replace reg with separate src and dest
7145 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7146 Use aarch64_add_offset_1.
7147 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7148 an rtx rather than an int. Take the delta as a poly_int64
7149 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7150 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7151 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7152 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7153 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7154 and aarch64_add_sp.
7155 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7156 aarch64_add_constant.
7158 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7160 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7161 Use scalar_float_mode.
7163 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7165 * config/aarch64/aarch64-simd.md
7166 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7167 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7168 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7169 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7170 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7171 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7172 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7173 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7174 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7175 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7177 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7179 PR target/83514
7180 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7181 targ_options->x_arm_arch_string is non NULL.
7183 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7185 * config/aarch64/aarch64.h
7186 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7188 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7190 PR target/82096
7191 * expmed.c (emit_store_flag_force): Swap if const op0
7192 and change VOIDmode to mode of op0.
7194 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7196 PR rtl-optimization/83761
7197 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7198 than bytes to mode_for_size.
7200 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7202 PR middle-end/83189
7203 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7204 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7205 profile.
7207 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7209 PR middle-end/83575
7210 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7211 when in layout mode.
7212 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7213 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7214 partition fixup.
7216 2018-01-10 Michael Collison <michael.collison@arm.com>
7218 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7219 * config/aarch64/aarch64-option-extension.def: Add
7220 AARCH64_OPT_EXTENSION of 'fp16fml'.
7221 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7222 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7223 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7224 * config/aarch64/constraints.md (Ui7): New constraint.
7225 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7226 (VFMLA_SEL_W): Ditto.
7227 (f16quad): Ditto.
7228 (f16mac1): Ditto.
7229 (VFMLA16_LOW): New int iterator.
7230 (VFMLA16_HIGH): Ditto.
7231 (UNSPEC_FMLAL): New unspec.
7232 (UNSPEC_FMLSL): Ditto.
7233 (UNSPEC_FMLAL2): Ditto.
7234 (UNSPEC_FMLSL2): Ditto.
7235 (f16mac): New code attribute.
7236 * config/aarch64/aarch64-simd-builtins.def
7237 (aarch64_fmlal_lowv2sf): Ditto.
7238 (aarch64_fmlsl_lowv2sf): Ditto.
7239 (aarch64_fmlalq_lowv4sf): Ditto.
7240 (aarch64_fmlslq_lowv4sf): Ditto.
7241 (aarch64_fmlal_highv2sf): Ditto.
7242 (aarch64_fmlsl_highv2sf): Ditto.
7243 (aarch64_fmlalq_highv4sf): Ditto.
7244 (aarch64_fmlslq_highv4sf): Ditto.
7245 (aarch64_fmlal_lane_lowv2sf): Ditto.
7246 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7247 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7248 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7249 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7250 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7251 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7252 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7253 (aarch64_fmlal_lane_highv2sf): Ditto.
7254 (aarch64_fmlsl_lane_highv2sf): Ditto.
7255 (aarch64_fmlal_laneq_highv2sf): Ditto.
7256 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7257 (aarch64_fmlalq_lane_highv4sf): Ditto.
7258 (aarch64_fmlsl_lane_highv4sf): Ditto.
7259 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7260 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7261 * config/aarch64/aarch64-simd.md:
7262 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7263 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7264 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7265 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7266 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7267 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7268 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7269 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7270 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7271 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7272 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7273 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7274 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7275 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7276 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7277 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7278 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7279 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7280 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7281 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7282 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7283 (vfmlsl_low_u32): Ditto.
7284 (vfmlalq_low_u32): Ditto.
7285 (vfmlslq_low_u32): Ditto.
7286 (vfmlal_high_u32): Ditto.
7287 (vfmlsl_high_u32): Ditto.
7288 (vfmlalq_high_u32): Ditto.
7289 (vfmlslq_high_u32): Ditto.
7290 (vfmlal_lane_low_u32): Ditto.
7291 (vfmlsl_lane_low_u32): Ditto.
7292 (vfmlal_laneq_low_u32): Ditto.
7293 (vfmlsl_laneq_low_u32): Ditto.
7294 (vfmlalq_lane_low_u32): Ditto.
7295 (vfmlslq_lane_low_u32): Ditto.
7296 (vfmlalq_laneq_low_u32): Ditto.
7297 (vfmlslq_laneq_low_u32): Ditto.
7298 (vfmlal_lane_high_u32): Ditto.
7299 (vfmlsl_lane_high_u32): Ditto.
7300 (vfmlal_laneq_high_u32): Ditto.
7301 (vfmlsl_laneq_high_u32): Ditto.
7302 (vfmlalq_lane_high_u32): Ditto.
7303 (vfmlslq_lane_high_u32): Ditto.
7304 (vfmlalq_laneq_high_u32): Ditto.
7305 (vfmlslq_laneq_high_u32): Ditto.
7306 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7307 (AARCH64_FL_FOR_ARCH8_4): New.
7308 (AARCH64_ISA_F16FML): New ISA flag.
7309 (TARGET_F16FML): New feature flag for fp16fml.
7310 (doc/invoke.texi): Document new fp16fml option.
7312 2018-01-10 Michael Collison <michael.collison@arm.com>
7314 * config/aarch64/aarch64-builtins.c:
7315 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7316 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7317 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7318 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7319 (AARCH64_ISA_SHA3): New ISA flag.
7320 (TARGET_SHA3): New feature flag for sha3.
7321 * config/aarch64/iterators.md (sha512_op): New int attribute.
7322 (CRYPTO_SHA512): New int iterator.
7323 (UNSPEC_SHA512H): New unspec.
7324 (UNSPEC_SHA512H2): Ditto.
7325 (UNSPEC_SHA512SU0): Ditto.
7326 (UNSPEC_SHA512SU1): Ditto.
7327 * config/aarch64/aarch64-simd-builtins.def
7328 (aarch64_crypto_sha512hqv2di): New builtin.
7329 (aarch64_crypto_sha512h2qv2di): Ditto.
7330 (aarch64_crypto_sha512su0qv2di): Ditto.
7331 (aarch64_crypto_sha512su1qv2di): Ditto.
7332 (aarch64_eor3qv8hi): Ditto.
7333 (aarch64_rax1qv2di): Ditto.
7334 (aarch64_xarqv2di): Ditto.
7335 (aarch64_bcaxqv8hi): Ditto.
7336 * config/aarch64/aarch64-simd.md:
7337 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7338 (aarch64_crypto_sha512su0qv2di): Ditto.
7339 (aarch64_crypto_sha512su1qv2di): Ditto.
7340 (aarch64_eor3qv8hi): Ditto.
7341 (aarch64_rax1qv2di): Ditto.
7342 (aarch64_xarqv2di): Ditto.
7343 (aarch64_bcaxqv8hi): Ditto.
7344 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7345 (vsha512h2q_u64): Ditto.
7346 (vsha512su0q_u64): Ditto.
7347 (vsha512su1q_u64): Ditto.
7348 (veor3q_u16): Ditto.
7349 (vrax1q_u64): Ditto.
7350 (vxarq_u64): Ditto.
7351 (vbcaxq_u16): Ditto.
7352 * config/arm/types.md (crypto_sha512): New type attribute.
7353 (crypto_sha3): Ditto.
7354 (doc/invoke.texi): Document new sha3 option.
7356 2018-01-10 Michael Collison <michael.collison@arm.com>
7358 * config/aarch64/aarch64-builtins.c:
7359 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7360 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7361 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7362 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7363 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7364 (AARCH64_ISA_SM4): New ISA flag.
7365 (TARGET_SM4): New feature flag for sm4.
7366 * config/aarch64/aarch64-simd-builtins.def
7367 (aarch64_sm3ss1qv4si): Ditto.
7368 (aarch64_sm3tt1aq4si): Ditto.
7369 (aarch64_sm3tt1bq4si): Ditto.
7370 (aarch64_sm3tt2aq4si): Ditto.
7371 (aarch64_sm3tt2bq4si): Ditto.
7372 (aarch64_sm3partw1qv4si): Ditto.
7373 (aarch64_sm3partw2qv4si): Ditto.
7374 (aarch64_sm4eqv4si): Ditto.
7375 (aarch64_sm4ekeyqv4si): Ditto.
7376 * config/aarch64/aarch64-simd.md:
7377 (aarch64_sm3ss1qv4si): Ditto.
7378 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7379 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7380 (aarch64_sm4eqv4si): Ditto.
7381 (aarch64_sm4ekeyqv4si): Ditto.
7382 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7383 (sm3part_op): Ditto.
7384 (CRYPTO_SM3TT): Ditto.
7385 (CRYPTO_SM3PART): Ditto.
7386 (UNSPEC_SM3SS1): New unspec.
7387 (UNSPEC_SM3TT1A): Ditto.
7388 (UNSPEC_SM3TT1B): Ditto.
7389 (UNSPEC_SM3TT2A): Ditto.
7390 (UNSPEC_SM3TT2B): Ditto.
7391 (UNSPEC_SM3PARTW1): Ditto.
7392 (UNSPEC_SM3PARTW2): Ditto.
7393 (UNSPEC_SM4E): Ditto.
7394 (UNSPEC_SM4EKEY): Ditto.
7395 * config/aarch64/constraints.md (Ui2): New constraint.
7396 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7397 * config/arm/types.md (crypto_sm3): New type attribute.
7398 (crypto_sm4): Ditto.
7399 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7400 (vsm3tt1aq_u32): Ditto.
7401 (vsm3tt1bq_u32): Ditto.
7402 (vsm3tt2aq_u32): Ditto.
7403 (vsm3tt2bq_u32): Ditto.
7404 (vsm3partw1q_u32): Ditto.
7405 (vsm3partw2q_u32): Ditto.
7406 (vsm4eq_u32): Ditto.
7407 (vsm4ekeyq_u32): Ditto.
7408 (doc/invoke.texi): Document new sm4 option.
7410 2018-01-10 Michael Collison <michael.collison@arm.com>
7412 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7413 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7414 (AARCH64_FL_FOR_ARCH8_4): New.
7415 (AARCH64_FL_V8_4): New flag.
7416 (doc/invoke.texi): Document new armv8.4-a option.
7418 2018-01-10 Michael Collison <michael.collison@arm.com>
7420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7421 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7422 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7423 * config/aarch64/aarch64-option-extension.def: Add
7424 AARCH64_OPT_EXTENSION of 'sha2'.
7425 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7426 (crypto): Disable sha2 and aes if crypto disabled.
7427 (crypto): Enable aes and sha2 if enabled.
7428 (simd): Disable sha2 and aes if simd disabled.
7429 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7430 New flags.
7431 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7432 (TARGET_SHA2): New feature flag for sha2.
7433 (TARGET_AES): New feature flag for aes.
7434 * config/aarch64/aarch64-simd.md:
7435 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7436 conditional on TARGET_AES.
7437 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7438 (aarch64_crypto_sha1hsi): Make pattern conditional
7439 on TARGET_SHA2.
7440 (aarch64_crypto_sha1hv4si): Ditto.
7441 (aarch64_be_crypto_sha1hv4si): Ditto.
7442 (aarch64_crypto_sha1su1v4si): Ditto.
7443 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7444 (aarch64_crypto_sha1su0v4si): Ditto.
7445 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7446 (aarch64_crypto_sha256su0v4si): Ditto.
7447 (aarch64_crypto_sha256su1v4si): Ditto.
7448 (doc/invoke.texi): Document new aes and sha2 options.
7450 2018-01-10 Martin Sebor <msebor@redhat.com>
7452 PR tree-optimization/83781
7453 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7454 as string arrays.
7456 2018-01-11 Martin Sebor <msebor@gmail.com>
7457 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7459 PR tree-optimization/83501
7460 PR tree-optimization/81703
7462 * tree-ssa-strlen.c (get_string_cst): Rename...
7463 (get_string_len): ...to this. Handle global constants.
7464 (handle_char_store): Adjust.
7466 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7467 Jim Wilson <jimw@sifive.com>
7469 * config/riscv/riscv-protos.h (riscv_output_return): New.
7470 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7471 (riscv_attribute_table, riscv_output_return),
7472 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7473 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7474 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7475 function.
7476 (riscv_expand_prologue): Add early return for naked function.
7477 (riscv_expand_epilogue): Likewise.
7478 (riscv_function_ok_for_sibcall): Return false for naked function.
7479 (riscv_set_current_function): New.
7480 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7481 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7482 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7483 * doc/extend.texi (RISC-V Function Attributes): New.
7485 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7487 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7488 check for 128-bit long double before checking TCmode.
7489 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7490 128-bit long doubles before checking TFmode or TCmode.
7491 (FLOAT128_IBM_P): Likewise.
7493 2018-01-10 Martin Sebor <msebor@redhat.com>
7495 PR tree-optimization/83671
7496 * builtins.c (c_strlen): Unconditionally return zero for the empty
7497 string.
7498 Use -Warray-bounds for warnings.
7499 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7500 for non-constant array indices with COMPONENT_REF, arrays of
7501 arrays, and pointers to arrays.
7502 (gimple_fold_builtin_strlen): Determine and set length range for
7503 non-constant character arrays.
7505 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7507 PR middle-end/81897
7508 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7509 empty blocks.
7511 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7513 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7515 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7517 PR target/83399
7518 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7519 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7520 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7521 indexed_or_indirect_operand predicate.
7522 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7523 (*vsx_le_perm_load_v8hi): Likewise.
7524 (*vsx_le_perm_load_v16qi): Likewise.
7525 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7526 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7527 (*vsx_le_perm_store_v8hi): Likewise.
7528 (*vsx_le_perm_store_v16qi): Likewise.
7529 (eight unnamed splitters): Likewise.
7531 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7533 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7534 * config/rs6000/emmintrin.h: Likewise.
7535 * config/rs6000/mmintrin.h: Likewise.
7536 * config/rs6000/xmmintrin.h: Likewise.
7538 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7540 PR c++/43486
7541 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7542 "public_flag".
7543 * tree.c (tree_nop_conversion): Return true for location wrapper
7544 nodes.
7545 (maybe_wrap_with_location): New function.
7546 (selftest::check_strip_nops): New function.
7547 (selftest::test_location_wrappers): New function.
7548 (selftest::tree_c_tests): Call it.
7549 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7550 (maybe_wrap_with_location): New decl.
7551 (EXPR_LOCATION_WRAPPER_P): New macro.
7552 (location_wrapper_p): New inline function.
7553 (tree_strip_any_location_wrapper): New inline function.
7555 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7557 PR target/83735
7558 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7559 stack_realign_offset for the largest alignment of stack slot
7560 actually used.
7561 (ix86_find_max_used_stack_alignment): New function.
7562 (ix86_finalize_stack_frame_flags): Use it. Set
7563 max_used_stack_alignment if we don't realign stack.
7564 * config/i386/i386.h (machine_function): Add
7565 max_used_stack_alignment.
7567 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7569 * config/arm/arm.opt (-mbranch-cost): New option.
7570 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7571 account.
7573 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7575 PR target/83629
7576 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7577 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7579 2018-01-10 Richard Biener <rguenther@suse.de>
7581 PR debug/83765
7582 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7583 early out so it also covers the case where we have a non-NULL
7584 origin.
7586 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7588 PR tree-optimization/83753
7589 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7590 for non-strided grouped accesses if the number of elements is 1.
7592 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7594 PR target/81616
7595 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7596 * i386.h (TARGET_USE_GATHER): Define.
7597 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7599 2018-01-10 Martin Liska <mliska@suse.cz>
7601 PR bootstrap/82831
7602 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7603 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7604 partitioning.
7605 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7606 CLEANUP_NO_PARTITIONING is not set.
7608 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7610 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7611 for vectors, as a partial revert of r254296.
7612 * rtl.h (const_vec_p): Delete.
7613 (const_vec_duplicate_p): Don't test for vector CONSTs.
7614 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7615 * expmed.c (make_tree): Likewise.
7617 Revert:
7618 * common.md (E, F): Use CONSTANT_P instead of checking for
7619 CONST_VECTOR.
7620 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7621 checking for CONST_VECTOR.
7623 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7625 PR middle-end/83575
7626 * predict.c (force_edge_cold): Handle in more sane way edges
7627 with no prediction.
7629 2018-01-09 Carl Love <cel@us.ibm.com>
7631 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7632 V4SI, V4SF types.
7633 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7634 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7635 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7636 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7637 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7638 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7639 * config/rs6000/rs6000-protos.h: Add extern defition for
7640 rs6000_generate_float2_double_code.
7641 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7642 function.
7643 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7644 (float2_v2df): Add define_expand.
7646 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7648 PR target/83628
7649 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7650 op_mode in the force_to_mode call.
7652 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7654 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7655 instead of checking each element individually.
7656 (aarch64_evpc_uzp): Likewise.
7657 (aarch64_evpc_zip): Likewise.
7658 (aarch64_evpc_ext): Likewise.
7659 (aarch64_evpc_rev): Likewise.
7660 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7661 instead of checking each element individually. Return true without
7662 generating rtl if
7663 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7664 whether all selected elements come from the same input, instead of
7665 checking each element individually. Remove calls to gen_rtx_REG,
7666 start_sequence and end_sequence and instead assert that no rtl is
7667 generated.
7669 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7671 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7672 order of HIGH and CONST checks.
7674 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7676 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7677 if the destination isn't an SSA_NAME.
7679 2018-01-09 Richard Biener <rguenther@suse.de>
7681 PR tree-optimization/83668
7682 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7683 move prologue...
7684 (canonicalize_loop_form): ... here, renamed from ...
7685 (canonicalize_loop_closed_ssa_form): ... this and amended to
7686 swap successor edges for loop exit blocks to make us use
7687 the RPO order we need for initial schedule generation.
7689 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7691 PR tree-optimization/64811
7692 * match.pd: When optimizing comparisons with Inf, avoid
7693 introducing or losing exceptions from comparisons with NaN.
7695 2018-01-09 Martin Liska <mliska@suse.cz>
7697 PR sanitizer/82517
7698 * asan.c (shadow_mem_size): Add gcc_assert.
7700 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7702 Don't save registers in main().
7704 PR target/83738
7705 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7706 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7707 * config/avr/avr.c (avr_set_current_function): Don't error if
7708 naked, OS_task or OS_main are specified at the same time.
7709 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7710 OS_main.
7711 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7712 attribute.
7713 * common/config/avr/avr-common.c (avr_option_optimization_table):
7714 Switch on -mmain-is-OS_task for optimizing compilations.
7716 2018-01-09 Richard Biener <rguenther@suse.de>
7718 PR tree-optimization/83572
7719 * graphite.c: Include cfganal.h.
7720 (graphite_transform_loops): Connect infinite loops to exit
7721 and remove fake edges at the end.
7723 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7725 * ipa-inline.c (edge_badness): Revert accidental checkin.
7727 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7729 PR ipa/80763
7730 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7731 symbols; not inline clones.
7733 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7735 PR target/83507
7736 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7737 hard registers. Formatting fixes.
7739 PR preprocessor/83722
7740 * gcc.c (try_generate_repro): Pass
7741 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7742 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7743 do_report_bug.
7745 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7746 Kito Cheng <kito.cheng@gmail.com>
7748 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7749 (riscv_leaf_function_p): Delete.
7750 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7752 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7754 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7755 function.
7756 (do_ifelse): New function.
7757 (do_isel): New function.
7758 (do_sub3): New function.
7759 (do_add3): New function.
7760 (do_load_mask_compare): New function.
7761 (do_overlap_load_compare): New function.
7762 (expand_compare_loop): New function.
7763 (expand_block_compare): Call expand_compare_loop() when appropriate.
7764 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7765 option description.
7766 (-mblock-compare-inline-loop-limit): New option.
7768 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7770 PR target/83677
7771 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7772 Reverse order of second and third operands in first alternative.
7773 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7774 of first and second elements in UNSPEC_VPERMR vector.
7775 (altivec_expand_vec_perm_le): Likewise.
7777 2018-01-08 Jeff Law <law@redhat.com>
7779 PR rtl-optimizatin/81308
7780 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7781 (process_switch): If group_case_labels makes a change, then set
7782 cfg_altered.
7783 (pass_convert_switch::execute): If a switch is converted, then
7784 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7786 PR rtl-optimization/81308
7787 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7788 splitting insns.
7790 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7792 PR target/83663 - Revert r255946
7793 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7794 generation for cases where splatting a value is not useful.
7795 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7796 across a vec_duplicate and a paradoxical subreg forming a vector
7797 mode to a vec_concat.
7799 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7801 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7802 -march=armv8.3-a variants.
7803 * config/arm/t-multilib: Likewise.
7804 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7806 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7808 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7809 to generate rtl.
7810 (cceq_ior_compare_complement): Give it a name so I can use it, and
7811 change boolean_or_operator predicate to boolean_operator so it can
7812 be used to generate a crand.
7813 (eqne): New code iterator.
7814 (bd/bd_neg): New code_attrs.
7815 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7816 a single define_insn.
7817 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7818 decrement (bdnzt/bdnzf/bdzt/bdzf).
7819 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7820 with the new names of the branch decrement patterns, and added the
7821 names of the branch decrement conditional patterns.
7823 2018-01-08 Richard Biener <rguenther@suse.de>
7825 PR tree-optimization/83563
7826 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7827 cache.
7829 2018-01-08 Richard Biener <rguenther@suse.de>
7831 PR middle-end/83713
7832 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7834 2018-01-08 Richard Biener <rguenther@suse.de>
7836 PR tree-optimization/83685
7837 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7838 references to abnormals.
7840 2018-01-08 Richard Biener <rguenther@suse.de>
7842 PR lto/83719
7843 * dwarf2out.c (output_indirect_strings): Handle empty
7844 skeleton_debug_str_hash.
7845 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7847 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7849 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7850 (emit_store_direct): Likewise.
7851 (arc_trampoline_adjust_address): Likewise.
7852 (arc_asm_trampoline_template): New function.
7853 (arc_initialize_trampoline): Use asm_trampoline_template.
7854 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7855 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7856 * config/arc/arc.md (flush_icache): Delete pattern.
7858 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7860 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7861 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7862 munaligned-access.
7864 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7866 PR target/83681
7867 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7868 by not USED_FOR_TARGET.
7869 (make_pass_resolve_sw_modes): Likewise.
7871 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7873 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7874 USED_FOR_TARGET.
7876 2018-01-08 Richard Biener <rguenther@suse.de>
7878 PR middle-end/83580
7879 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7881 2018-01-08 Richard Biener <rguenther@suse.de>
7883 PR middle-end/83517
7884 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7886 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7888 PR middle-end/81897
7889 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7890 basic blocks with a small number of successors.
7891 (convert_control_dep_chain_into_preds): Improve handling of
7892 forwarder blocks.
7893 (dump_predicates): Split apart into...
7894 (dump_pred_chain): ...here...
7895 (dump_pred_info): ...and here.
7896 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7897 (can_chain_union_be_invalidated_p): Improve check for invalidation
7898 of paths.
7899 (uninit_uses_cannot_happen): Avoid unnecessary if
7900 convert_control_dep_chain_into_preds yielded nothing.
7902 2018-01-06 Martin Sebor <msebor@redhat.com>
7904 PR tree-optimization/83640
7905 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7906 subtracting negative offset from size.
7907 (builtin_access::overlap): Adjust offset bounds of the access to fall
7908 within the size of the object if possible.
7910 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7912 PR rtl-optimization/83699
7913 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7914 extract_bit_field_as_subreg to cases in which the extracted
7915 value is also a vector.
7917 * lra-constraints.c (process_alt_operands): Test for the equivalence
7918 substitutions when detecting a possible reload cycle.
7920 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7922 PR debug/83480
7923 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7924 by default if flag_selective_schedling{,2}. Formatting fixes.
7926 PR rtl-optimization/83682
7927 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7928 if it has non-VECTOR_MODE element mode.
7929 (vec_duplicate_p): Likewise.
7931 PR middle-end/83694
7932 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7933 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7935 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7937 PR target/83604
7938 * config/i386/i386-builtin.def
7939 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7940 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7941 Require also OPTION_MASK_ISA_AVX512F in addition to
7942 OPTION_MASK_ISA_GFNI.
7943 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7944 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7945 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7946 to OPTION_MASK_ISA_GFNI.
7947 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7948 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7949 OPTION_MASK_ISA_AVX512BW.
7950 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7951 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7952 addition to OPTION_MASK_ISA_GFNI.
7953 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7954 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7955 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7956 to OPTION_MASK_ISA_GFNI.
7957 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7958 a requirement for all ISAs rather than any of them with a few
7959 exceptions.
7960 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7961 processing.
7962 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7963 bitmasks to be enabled with 3 exceptions, instead of requiring any
7964 enabled ISA with lots of exceptions.
7965 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7966 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7967 Change avx512bw in isa attribute to avx512f.
7968 * config/i386/sgxintrin.h: Add license boilerplate.
7969 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7970 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7971 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7972 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7973 defined.
7974 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7975 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7976 temporarily sse2 rather than sse if not enabled already.
7978 PR target/83604
7979 * config/i386/sse.md (VI248_VLBW): Rename to ...
7980 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7981 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7982 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7983 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7984 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7985 mode iterator instead of VI248_VLBW.
7987 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7989 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7990 (record_modified): Skip clobbers; add debug output.
7991 (param_change_prob): Use sreal frequencies.
7993 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7995 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7996 punt for user-aligned variables.
7998 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8000 * tree-chrec.c (chrec_contains_symbols): Return true for
8001 POLY_INT_CST.
8003 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8005 PR target/82439
8006 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8007 of (x|y) == x for BICS pattern.
8009 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8011 PR tree-optimization/83605
8012 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8013 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8014 can throw.
8016 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8018 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8019 * config/epiphany/rtems.h: New file.
8021 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8022 Uros Bizjak <ubizjak@gmail.com>
8024 PR target/83554
8025 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8026 QIreg_operand instead of register_operand predicate.
8027 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8028 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8029 comments instead of -fmitigate[-_]rop.
8031 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8033 PR bootstrap/81926
8034 * cgraphunit.c (symbol_table::compile): Switch to text_section
8035 before calling assembly_start debug hook.
8036 * run-rtl-passes.c (run_rtl_passes): Likewise.
8037 Include output.h.
8039 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8041 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8042 range_int_cst_p rather than !symbolic_range_p before calling
8043 extract_range_from_multiplicative_op_1.
8045 2018-01-04 Jeff Law <law@redhat.com>
8047 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8048 redundant test in assertion.
8050 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8052 * doc/rtl.texi: Document machine_mode wrapper classes.
8054 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8056 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8057 using tree_to_uhwi.
8059 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8061 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8062 the VEC_PERM_EXPR fold to fail.
8064 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8066 PR debug/83585
8067 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8068 to switched_sections.
8070 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8072 PR target/83680
8073 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8074 test for d.testing.
8076 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8078 PR target/83387
8079 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8080 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8082 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8084 PR debug/83666
8085 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8086 is BLKmode and bitpos not zero or mode change is needed.
8088 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8090 PR target/83675
8091 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8092 TARGET_VIS2.
8094 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8096 PR target/83628
8097 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8098 instead of MULT rtx. Update all corresponding splitters.
8099 (*saddl_se): Ditto.
8100 (*ssub<modesuffix>): Ditto.
8101 (*ssubl_se): Ditto.
8102 (*cmp_sadd_di): Update split patterns.
8103 (*cmp_sadd_si): Ditto.
8104 (*cmp_sadd_sidi): Ditto.
8105 (*cmp_ssub_di): Ditto.
8106 (*cmp_ssub_si): Ditto.
8107 (*cmp_ssub_sidi): Ditto.
8108 * config/alpha/predicates.md (const23_operand): New predicate.
8109 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8110 Look for ASHIFT, not MULT inner operand.
8111 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8113 2018-01-04 Martin Liska <mliska@suse.cz>
8115 PR gcov-profile/83669
8116 * gcov.c (output_intermediate_file): Add version to intermediate
8117 gcov file.
8118 * doc/gcov.texi: Document new field 'version' in intermediate
8119 file format. Fix location of '-k' option of gcov command.
8121 2018-01-04 Martin Liska <mliska@suse.cz>
8123 PR ipa/82352
8124 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8126 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8128 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8130 2018-01-03 Martin Sebor <msebor@redhat.com>
8132 PR tree-optimization/83655
8133 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8134 checking calls with invalid arguments.
8136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8138 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8139 (vectorizable_mask_load_store): Delete.
8140 (vectorizable_call): Return false for masked loads and stores.
8141 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8142 instead of gimple_assign_rhs1.
8143 (vectorizable_load): Handle IFN_MASK_LOAD.
8144 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8148 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8149 split out from..,
8150 (vectorizable_mask_load_store): ...here.
8151 (vectorizable_load): ...and here.
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8155 * tree-vect-stmts.c (vect_build_all_ones_mask)
8156 (vect_build_zero_merge_argument): New functions, split out from...
8157 (vectorizable_load): ...here.
8159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8161 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8162 split out from...
8163 (vectorizable_mask_load_store): ...here.
8164 (vectorizable_store): ...and here.
8166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8168 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8169 split out from...
8170 (vectorizable_mask_load_store): ...here.
8172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8174 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8175 (vect_model_store_cost): Take a vec_load_store_type instead of a
8176 vect_def_type.
8177 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8178 (vect_model_store_cost): Take a vec_load_store_type instead of a
8179 vect_def_type.
8180 (vectorizable_mask_load_store): Update accordingly.
8181 (vectorizable_store): Likewise.
8182 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8186 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8187 IFN_MASK_LOAD calls here rather than...
8188 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8191 Alan Hayward <alan.hayward@arm.com>
8192 David Sherwood <david.sherwood@arm.com>
8194 * expmed.c (extract_bit_field_1): For vector extracts,
8195 fall back to extract_bit_field_as_subreg if vec_extract
8196 isn't available.
8198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8199 Alan Hayward <alan.hayward@arm.com>
8200 David Sherwood <david.sherwood@arm.com>
8202 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8203 they are variable or constant sized.
8204 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8205 slots for constant-sized data.
8207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8208 Alan Hayward <alan.hayward@arm.com>
8209 David Sherwood <david.sherwood@arm.com>
8211 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8212 handling COND_EXPRs with boolean comparisons, try to find a better
8213 basis for the mask type than the boolean itself.
8215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8217 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8218 is calculated and how it can be overridden.
8219 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8220 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8221 if defined.
8222 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8223 if nonzero.
8225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8226 Alan Hayward <alan.hayward@arm.com>
8227 David Sherwood <david.sherwood@arm.com>
8229 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8230 Remove the mode argument.
8231 (aarch64_simd_valid_immediate): Remove the mode and inverse
8232 arguments.
8233 * config/aarch64/iterators.md (bitsize): New iterator.
8234 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8235 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8236 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8237 aarch64_simd_valid_immediate.
8238 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8239 (aarch64_reg_or_bic_imm): Likewise.
8240 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8241 with an insn_type enum and msl with a modifier_type enum.
8242 Replace element_width with a scalar_mode. Change the shift
8243 to unsigned int. Add constructors for scalar_float_mode and
8244 scalar_int_mode elements.
8245 (aarch64_vect_float_const_representable_p): Delete.
8246 (aarch64_can_const_movi_rtx_p)
8247 (aarch64_simd_scalar_immediate_valid_for_move)
8248 (aarch64_simd_make_constant): Update call to
8249 aarch64_simd_valid_immediate.
8250 (aarch64_advsimd_valid_immediate_hs): New function.
8251 (aarch64_advsimd_valid_immediate): Likewise.
8252 (aarch64_simd_valid_immediate): Remove mode and inverse
8253 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8254 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8255 and aarch64_float_const_representable_p on the result.
8256 (aarch64_output_simd_mov_immediate): Remove mode argument.
8257 Update call to aarch64_simd_valid_immediate and use of
8258 simd_immediate_info.
8259 (aarch64_output_scalar_simd_mov_immediate): Update call
8260 accordingly.
8262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8263 Alan Hayward <alan.hayward@arm.com>
8264 David Sherwood <david.sherwood@arm.com>
8266 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8267 (mode_nunits): Likewise CONST_MODE_NUNITS.
8268 * machmode.def (ADJUST_NUNITS): Document.
8269 * genmodes.c (mode_data::need_nunits_adj): New field.
8270 (blank_mode): Update accordingly.
8271 (adj_nunits): New variable.
8272 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8273 parameter.
8274 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8275 listed in adj_nunits.
8276 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8277 listed in adj_nunits. Don't emit case statements for such modes.
8278 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8279 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8280 nothing if adj_nunits is nonnull.
8281 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8282 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8283 (emit_mode_fbit): Update use of print_maybe_const_decl.
8284 (emit_move_size): Likewise. Treat the array as non-const
8285 if adj_nunits.
8286 (emit_mode_adjustments): Handle adj_nunits.
8288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8290 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8291 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8292 (VECTOR_MODES): Use it.
8293 (make_vector_modes): Take the prefix as an argument.
8295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8296 Alan Hayward <alan.hayward@arm.com>
8297 David Sherwood <david.sherwood@arm.com>
8299 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8300 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8301 for MODE_VECTOR_BOOL.
8302 * machmode.def (VECTOR_BOOL_MODE): Document.
8303 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8304 (make_vector_bool_mode): New function.
8305 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8306 MODE_VECTOR_BOOL.
8307 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8308 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8309 Likewise.
8310 * stor-layout.c (int_mode_for_mode): Likewise.
8311 * tree.c (build_vector_type_for_mode): Likewise.
8312 * varasm.c (output_constant_pool_2): Likewise.
8313 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8314 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8315 for MODE_VECTOR_BOOL.
8316 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8317 of mode class checks.
8318 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8319 instead of a list of mode class checks.
8320 (expand_vector_scalar_condition): Likewise.
8321 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8324 Alan Hayward <alan.hayward@arm.com>
8325 David Sherwood <david.sherwood@arm.com>
8327 * machmode.h (mode_size): Change from unsigned short to
8328 poly_uint16_pod.
8329 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8330 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8331 or if measurement_type is not polynomial.
8332 (fixed_size_mode::includes_p): Check for constant-sized modes.
8333 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8334 return a poly_uint16 rather than an unsigned short.
8335 (emit_mode_size): Change the type of mode_size from unsigned short
8336 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8337 (emit_mode_adjustments): Cope with polynomial vector sizes.
8338 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8339 for GET_MODE_SIZE.
8340 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8341 for GET_MODE_SIZE.
8342 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8343 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8344 * caller-save.c (setup_save_areas): Likewise.
8345 (replace_reg_with_saved_mem): Likewise.
8346 * calls.c (emit_library_call_value_1): Likewise.
8347 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8348 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8349 (gen_lowpart_for_combine): Likewise.
8350 * convert.c (convert_to_integer_1): Likewise.
8351 * cse.c (equiv_constant, cse_insn): Likewise.
8352 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8353 (cselib_subst_to_values): Likewise.
8354 * dce.c (word_dce_process_block): Likewise.
8355 * df-problems.c (df_word_lr_mark_ref): Likewise.
8356 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8357 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8358 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8359 (rtl_for_decl_location): Likewise.
8360 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8361 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8362 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8363 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8364 (expand_expr_real_1): Likewise.
8365 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8366 (pad_below): Likewise.
8367 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8368 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8369 * ira.c (get_subreg_tracking_sizes): Likewise.
8370 * ira-build.c (ira_create_allocno_objects): Likewise.
8371 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8372 (ira_sort_regnos_for_alter_reg): Likewise.
8373 * ira-costs.c (record_operand_costs): Likewise.
8374 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8375 (resolve_simple_move): Likewise.
8376 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8377 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8378 (lra_constraints): Likewise.
8379 (CONST_POOL_OK_P): Reject variable-sized modes.
8380 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8381 (add_pseudo_to_slot, lra_spill): Likewise.
8382 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8383 * optabs-query.c (get_best_extraction_insn): Likewise.
8384 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8385 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8386 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8387 * recog.c (offsettable_address_addr_space_p): Likewise.
8388 * regcprop.c (maybe_mode_change): Likewise.
8389 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8390 * regrename.c (build_def_use): Likewise.
8391 * regstat.c (dump_reg_info): Likewise.
8392 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8393 (find_reloads, find_reloads_subreg_address): Likewise.
8394 * reload1.c (eliminate_regs_1): Likewise.
8395 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8396 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8397 (simplify_binary_operation_1, simplify_subreg): Likewise.
8398 * targhooks.c (default_function_arg_padding): Likewise.
8399 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8400 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8401 (verify_gimple_assign_ternary): Likewise.
8402 * tree-inline.c (estimate_move_cost): Likewise.
8403 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8404 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8405 (get_address_cost_ainc): Likewise.
8406 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8407 (vect_supportable_dr_alignment): Likewise.
8408 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8409 (vectorizable_reduction): Likewise.
8410 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8411 (vectorizable_operation, vectorizable_load): Likewise.
8412 * tree.c (build_same_sized_truth_vector_type): Likewise.
8413 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8414 * var-tracking.c (emit_note_insn_var_location): Likewise.
8415 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8416 (ADDR_VEC_ALIGN): Likewise.
8418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8419 Alan Hayward <alan.hayward@arm.com>
8420 David Sherwood <david.sherwood@arm.com>
8422 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8423 unsigned short.
8424 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8425 or if measurement_type is polynomial.
8426 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8427 * combine.c (make_extraction): Likewise.
8428 * dse.c (find_shift_sequence): Likewise.
8429 * dwarf2out.c (mem_loc_descriptor): Likewise.
8430 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8431 (extract_bit_field, extract_low_bits): Likewise.
8432 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8433 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8434 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8435 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8436 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8437 * reload.c (find_reloads): Likewise.
8438 * reload1.c (alter_reg): Likewise.
8439 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8440 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8441 * tree-if-conv.c (predicate_mem_writes): Likewise.
8442 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8443 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8444 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8445 * valtrack.c (dead_debug_insert_temp): Likewise.
8446 * varasm.c (mergeable_constant_section): Likewise.
8447 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8450 Alan Hayward <alan.hayward@arm.com>
8451 David Sherwood <david.sherwood@arm.com>
8453 * expr.c (expand_assignment): Cope with polynomial mode sizes
8454 when assigning to a CONCAT.
8456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8457 Alan Hayward <alan.hayward@arm.com>
8458 David Sherwood <david.sherwood@arm.com>
8460 * machmode.h (mode_precision): Change from unsigned short to
8461 poly_uint16_pod.
8462 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8463 short.
8464 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8465 or if measurement_type is not polynomial.
8466 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8467 in which the mode is already known to be a scalar_int_mode.
8468 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8469 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8470 initializer.
8471 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8472 for GET_MODE_PRECISION.
8473 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8474 for GET_MODE_PRECISION.
8475 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8476 as polynomial.
8477 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8478 (expand_field_assignment, make_extraction): Likewise.
8479 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8480 (get_last_value): Likewise.
8481 * convert.c (convert_to_integer_1): Likewise.
8482 * cse.c (cse_insn): Likewise.
8483 * expr.c (expand_expr_real_1): Likewise.
8484 * lra-constraints.c (simplify_operand_subreg): Likewise.
8485 * optabs-query.c (can_atomic_load_p): Likewise.
8486 * optabs.c (expand_atomic_load): Likewise.
8487 (expand_atomic_store): Likewise.
8488 * ree.c (combine_reaching_defs): Likewise.
8489 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8490 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8491 * tree.h (type_has_mode_precision_p): Likewise.
8492 * ubsan.c (instrument_si_overflow): Likewise.
8494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8495 Alan Hayward <alan.hayward@arm.com>
8496 David Sherwood <david.sherwood@arm.com>
8498 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8499 polynomial numbers of units.
8500 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8501 (valid_vector_subparts_p): New function.
8502 (build_vector_type): Remove temporary shim and take the number
8503 of units as a poly_uint64 rather than an int.
8504 (build_opaque_vector_type): Take the number of units as a
8505 poly_uint64 rather than an int.
8506 * tree.c (build_vector_from_ctor): Handle polynomial
8507 TYPE_VECTOR_SUBPARTS.
8508 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8509 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8510 (build_vector_from_val): If the number of units is variable,
8511 use build_vec_duplicate_cst for constant operands and
8512 VEC_DUPLICATE_EXPR otherwise.
8513 (make_vector_type): Remove temporary is_constant ().
8514 (build_vector_type, build_opaque_vector_type): Take the number of
8515 units as a poly_uint64 rather than an int.
8516 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8517 VECTOR_CST_NELTS.
8518 * cfgexpand.c (expand_debug_expr): Likewise.
8519 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8520 (store_constructor, expand_expr_real_1): Likewise.
8521 (const_scalar_mask_from_tree): Likewise.
8522 * fold-const-call.c (fold_const_reduction): Likewise.
8523 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8524 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8525 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8526 (fold_relational_const): Likewise.
8527 (native_interpret_vector): Likewise. Change the size from an
8528 int to an unsigned int.
8529 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8530 TYPE_VECTOR_SUBPARTS.
8531 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8532 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8533 duplicating a non-constant operand into a variable-length vector.
8534 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8535 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8536 * ipa-icf.c (sem_variable::equals): Likewise.
8537 * match.pd: Likewise.
8538 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8539 * print-tree.c (print_node): Likewise.
8540 * stor-layout.c (layout_type): Likewise.
8541 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8542 * tree-cfg.c (verify_gimple_comparison): Likewise.
8543 (verify_gimple_assign_binary): Likewise.
8544 (verify_gimple_assign_ternary): Likewise.
8545 (verify_gimple_assign_single): Likewise.
8546 * tree-pretty-print.c (dump_generic_node): Likewise.
8547 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8548 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8549 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8550 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8551 (vect_shift_permute_load_chain): Likewise.
8552 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8553 (expand_vector_condition, optimize_vector_constructor): Likewise.
8554 (lower_vec_perm, get_compute_type): Likewise.
8555 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8556 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8557 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8558 (vect_recog_mask_conversion_pattern): Likewise.
8559 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8560 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8561 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8562 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8563 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8564 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8565 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8566 (supportable_widening_operation): Likewise.
8567 (supportable_narrowing_operation): Likewise.
8568 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8569 Likewise.
8570 * varasm.c (output_constant): Likewise.
8572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8573 Alan Hayward <alan.hayward@arm.com>
8574 David Sherwood <david.sherwood@arm.com>
8576 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8577 so that both the length == 3 and length != 3 cases set up their
8578 own permute vectors. Add comments explaining why we know the
8579 number of elements is constant.
8580 (vect_permute_load_chain): Likewise.
8582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8583 Alan Hayward <alan.hayward@arm.com>
8584 David Sherwood <david.sherwood@arm.com>
8586 * machmode.h (mode_nunits): Change from unsigned char to
8587 poly_uint16_pod.
8588 (ONLY_FIXED_SIZE_MODES): New macro.
8589 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8590 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8591 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8592 New typedefs.
8593 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8594 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8595 or if measurement_type is not polynomial.
8596 * genmodes.c (ZERO_COEFFS): New macro.
8597 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8598 poly_uint16.
8599 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8600 Use ZERO_COEFFS when emitting initializers.
8601 * data-streamer.h (bp_pack_poly_value): New function.
8602 (bp_unpack_poly_value): Likewise.
8603 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8604 for GET_MODE_NUNITS.
8605 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8606 for GET_MODE_NUNITS.
8607 * tree.c (make_vector_type): Remove temporary shim and make
8608 the real function take the number of units as a poly_uint64
8609 rather than an int.
8610 (build_vector_type_for_mode): Handle polynomial nunits.
8611 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8612 * emit-rtl.c (const_vec_series_p_1): Likewise.
8613 (gen_rtx_CONST_VECTOR): Likewise.
8614 * fold-const.c (test_vec_duplicate_folding): Likewise.
8615 * genrecog.c (validate_pattern): Likewise.
8616 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8617 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8618 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8619 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8620 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8621 * rtlanal.c (subreg_get_info): Likewise.
8622 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8623 (vect_grouped_load_supported): Likewise.
8624 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8625 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8626 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8627 (simplify_const_unary_operation, simplify_binary_operation_1)
8628 (simplify_const_binary_operation, simplify_ternary_operation)
8629 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8630 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8631 instead of CONST_VECTOR_NUNITS.
8632 * varasm.c (output_constant_pool_2): Likewise.
8633 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8634 explicit-encoded elements in the XVEC for variable-length vectors.
8636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8638 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8641 Alan Hayward <alan.hayward@arm.com>
8642 David Sherwood <david.sherwood@arm.com>
8644 * coretypes.h (fixed_size_mode): Declare.
8645 (fixed_size_mode_pod): New typedef.
8646 * builtins.h (target_builtins::x_apply_args_mode)
8647 (target_builtins::x_apply_result_mode): Change type to
8648 fixed_size_mode_pod.
8649 * builtins.c (apply_args_size, apply_result_size, result_vector)
8650 (expand_builtin_apply_args_1, expand_builtin_apply)
8651 (expand_builtin_return): Update accordingly.
8653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8655 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8656 * cselib.c (cselib_hash_rtx): Likewise.
8657 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8658 CONST_VECTOR encoding.
8660 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8661 Jeff Law <law@redhat.com>
8663 PR target/83641
8664 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8665 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8666 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8667 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8669 PR target/83641
8670 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8671 explicitly probe *sp in a noreturn function if there were any callee
8672 register saves or frame pointer is needed.
8674 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8676 PR debug/83621
8677 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8678 BLKmode for ternary, binary or unary expressions.
8680 PR debug/83645
8681 * var-tracking.c (delete_vta_debug_insn): New inline function.
8682 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8683 insns from get_insns () to NULL instead of each bb separately.
8684 Use delete_vta_debug_insn. No longer static.
8685 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8686 delete_vta_debug_insns callers.
8687 * rtl.h (delete_vta_debug_insns): Declare.
8688 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8689 instead of variable_tracking_main.
8691 2018-01-03 Martin Sebor <msebor@redhat.com>
8693 PR tree-optimization/83603
8694 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8695 arguments past the endof the argument list in functions declared
8696 without a prototype.
8697 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8698 Avoid checking when arguments are null.
8700 2018-01-03 Martin Sebor <msebor@redhat.com>
8702 PR c/83559
8703 * doc/extend.texi (attribute const): Fix a typo.
8704 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8705 issuing -Wsuggest-attribute for void functions.
8707 2018-01-03 Martin Sebor <msebor@redhat.com>
8709 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8710 offset_int::from instead of wide_int::to_shwi.
8711 (maybe_diag_overlap): Remove assertion.
8712 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8713 * gimple-ssa-sprintf.c (format_directive): Same.
8714 (parse_directive): Same.
8715 (sprintf_dom_walker::compute_format_length): Same.
8716 (try_substitute_return_value): Same.
8718 2018-01-03 Jeff Law <law@redhat.com>
8720 PR middle-end/83654
8721 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8722 non-constant residual for zero at runtime and avoid probing in
8723 that case. Reorganize code for trailing problem to mirror handling
8724 of the residual.
8726 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8728 PR tree-optimization/83501
8729 * tree-ssa-strlen.c (get_string_cst): New.
8730 (handle_char_store): Call get_string_cst.
8732 2018-01-03 Martin Liska <mliska@suse.cz>
8734 PR tree-optimization/83593
8735 * tree-ssa-strlen.c: Include tree-cfg.h.
8736 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8737 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8738 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8739 to false.
8740 (strlen_dom_walker::before_dom_children): Call
8741 gimple_purge_dead_eh_edges. Dump tranformation with details
8742 dump flags.
8743 (strlen_dom_walker::before_dom_children): Update call by adding
8744 new argument cleanup_eh.
8745 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8747 2018-01-03 Martin Liska <mliska@suse.cz>
8749 PR ipa/83549
8750 * cif-code.def (VARIADIC_THUNK): New enum value.
8751 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8752 thunks.
8754 2018-01-03 Jan Beulich <jbeulich@suse.com>
8756 * sse.md (mov<mode>_internal): Tighten condition for when to use
8757 vmovdqu<ssescalarsize> for TI and OI modes.
8759 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8761 Update copyright years.
8763 2018-01-03 Martin Liska <mliska@suse.cz>
8765 PR ipa/83594
8766 * ipa-visibility.c (function_and_variable_visibility): Skip
8767 functions with noipa attribure.
8769 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8771 * gcc.c (process_command): Update copyright notice dates.
8772 * gcov-dump.c (print_version): Ditto.
8773 * gcov.c (print_version): Ditto.
8774 * gcov-tool.c (print_version): Ditto.
8775 * gengtype.c (create_file): Ditto.
8776 * doc/cpp.texi: Bump @copying's copyright year.
8777 * doc/cppinternals.texi: Ditto.
8778 * doc/gcc.texi: Ditto.
8779 * doc/gccint.texi: Ditto.
8780 * doc/gcov.texi: Ditto.
8781 * doc/install.texi: Ditto.
8782 * doc/invoke.texi: Ditto.
8784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8786 * vector-builder.h (vector_builder::m_full_nelts): Change from
8787 unsigned int to poly_uint64.
8788 (vector_builder::full_nelts): Update prototype accordingly.
8789 (vector_builder::new_vector): Likewise.
8790 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8791 (vector_builder::operator ==): Likewise.
8792 (vector_builder::finalize): Likewise.
8793 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8794 Take the number of elements as a poly_uint64 rather than an
8795 unsigned int.
8796 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8797 from unsigned int to poly_uint64.
8798 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8799 (vec_perm_indices::new_vector): Likewise.
8800 (vec_perm_indices::length): Likewise.
8801 (vec_perm_indices::nelts_per_input): Likewise.
8802 (vec_perm_indices::input_nelts): Likewise.
8803 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8804 number of elements per input as a poly_uint64 rather than an
8805 unsigned int. Use the original encoding for variable-length
8806 vectors, rather than clamping each individual element.
8807 For the second and subsequent elements in each pattern,
8808 clamp the step and base before clamping their sum.
8809 (vec_perm_indices::series_p): Handle polynomial element counts.
8810 (vec_perm_indices::all_in_range_p): Likewise.
8811 (vec_perm_indices_to_tree): Likewise.
8812 (vec_perm_indices_to_rtx): Likewise.
8813 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8814 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8815 (tree_vector_builder::new_binary_operation): Handle polynomial
8816 element counts. Return false if we need to know the number
8817 of elements at compile time.
8818 * fold-const.c (fold_vec_perm): Punt if the number of elements
8819 isn't known at compile time.
8821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8823 * vec-perm-indices.h (vec_perm_builder): Change element type
8824 from HOST_WIDE_INT to poly_int64.
8825 (vec_perm_indices::element_type): Update accordingly.
8826 (vec_perm_indices::clamp): Handle polynomial element_types.
8827 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8828 (vec_perm_indices::all_in_range_p): Likewise.
8829 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8830 than shwi trees.
8831 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8832 polynomial vec_perm_indices element types.
8833 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8834 * fold-const.c (fold_vec_perm): Likewise.
8835 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8836 * tree-vect-generic.c (lower_vec_perm): Likewise.
8837 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8838 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8839 element type to HOST_WIDE_INT.
8841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8842 Alan Hayward <alan.hayward@arm.com>
8843 David Sherwood <david.sherwood@arm.com>
8845 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8846 rather than an int. Use plus_constant.
8847 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8848 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8851 Alan Hayward <alan.hayward@arm.com>
8852 David Sherwood <david.sherwood@arm.com>
8854 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8855 a HOST_WIDE_INT to a poly_int64.
8857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8858 Alan Hayward <alan.hayward@arm.com>
8859 David Sherwood <david.sherwood@arm.com>
8861 * calls.c (load_register_parameters): Cope with polynomial
8862 mode sizes. Require a constant size for BLKmode parameters
8863 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8864 forces a parameter to be padded at the lsb end in order to
8865 fill a complete number of words, require the parameter size
8866 to be ordered wrt UNITS_PER_WORD.
8868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8869 Alan Hayward <alan.hayward@arm.com>
8870 David Sherwood <david.sherwood@arm.com>
8872 * reload1.c (spill_stack_slot_width): Change element type
8873 from unsigned int to poly_uint64_pod.
8874 (alter_reg): Treat mode sizes as polynomial.
8876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8877 Alan Hayward <alan.hayward@arm.com>
8878 David Sherwood <david.sherwood@arm.com>
8880 * reload.c (complex_word_subreg_p): New function.
8881 (reload_inner_reg_of_subreg, push_reload): Use it.
8883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8884 Alan Hayward <alan.hayward@arm.com>
8885 David Sherwood <david.sherwood@arm.com>
8887 * lra-constraints.c (process_alt_operands): Reject matched
8888 operands whose sizes aren't ordered.
8889 (match_reload): Refer to this check here.
8891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8892 Alan Hayward <alan.hayward@arm.com>
8893 David Sherwood <david.sherwood@arm.com>
8895 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8896 that the mode size is in the set {1, 2, 4, 8, 16}.
8898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8899 Alan Hayward <alan.hayward@arm.com>
8900 David Sherwood <david.sherwood@arm.com>
8902 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8903 Use plus_constant instead of gen_rtx_PLUS.
8905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8906 Alan Hayward <alan.hayward@arm.com>
8907 David Sherwood <david.sherwood@arm.com>
8909 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8910 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8911 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8912 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8913 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8914 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8915 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8916 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8917 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8918 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8919 a poly_int64.
8920 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8921 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8922 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8923 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8924 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8925 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8926 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8927 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8928 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8929 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8930 function.
8931 * expr.c (emit_move_resolve_push): Treat the input and result
8932 of PUSH_ROUNDING as a poly_int64.
8933 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8934 (emit_push_insn): Likewise.
8935 * lra-eliminations.c (mark_not_eliminable): Likewise.
8936 * recog.c (push_operand): Likewise.
8937 * reload1.c (elimination_effects): Likewise.
8938 * rtlanal.c (nonzero_bits1): Likewise.
8939 * calls.c (store_one_arg): Likewise. Require the padding to be
8940 known at compile time.
8942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8943 Alan Hayward <alan.hayward@arm.com>
8944 David Sherwood <david.sherwood@arm.com>
8946 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8947 Use plus_constant instead of gen_rtx_PLUS.
8949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8950 Alan Hayward <alan.hayward@arm.com>
8951 David Sherwood <david.sherwood@arm.com>
8953 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8954 rather than an int.
8956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8957 Alan Hayward <alan.hayward@arm.com>
8958 David Sherwood <david.sherwood@arm.com>
8960 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8961 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8962 via stack temporaries. Treat the mode size as polynomial too.
8964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8965 Alan Hayward <alan.hayward@arm.com>
8966 David Sherwood <david.sherwood@arm.com>
8968 * expr.c (expand_expr_real_2): When handling conversions involving
8969 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8970 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8971 as a poly_uint64 too.
8973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8974 Alan Hayward <alan.hayward@arm.com>
8975 David Sherwood <david.sherwood@arm.com>
8977 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8980 Alan Hayward <alan.hayward@arm.com>
8981 David Sherwood <david.sherwood@arm.com>
8983 * combine.c (can_change_dest_mode): Handle polynomial
8984 REGMODE_NATURAL_SIZE.
8985 * expmed.c (store_bit_field_1): Likewise.
8986 * expr.c (store_constructor): Likewise.
8987 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8988 and polynomial REGMODE_NATURAL_SIZE.
8989 (gen_lowpart_common): Likewise.
8990 * reginfo.c (record_subregs_of_mode): Likewise.
8991 * rtlanal.c (read_modify_subreg_p): Likewise.
8993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8994 Alan Hayward <alan.hayward@arm.com>
8995 David Sherwood <david.sherwood@arm.com>
8997 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8998 numbers of elements.
9000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9001 Alan Hayward <alan.hayward@arm.com>
9002 David Sherwood <david.sherwood@arm.com>
9004 * match.pd: Cope with polynomial numbers of vector elements.
9006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9007 Alan Hayward <alan.hayward@arm.com>
9008 David Sherwood <david.sherwood@arm.com>
9010 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9011 in a POINTER_PLUS_EXPR.
9013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9014 Alan Hayward <alan.hayward@arm.com>
9015 David Sherwood <david.sherwood@arm.com>
9017 * omp-simd-clone.c (simd_clone_subparts): New function.
9018 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9019 (ipa_simd_modify_function_body): Likewise.
9021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9022 Alan Hayward <alan.hayward@arm.com>
9023 David Sherwood <david.sherwood@arm.com>
9025 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9026 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9027 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9028 (expand_vector_condition, vector_element): Likewise.
9029 (subparts_gt): New function.
9030 (get_compute_type): Use subparts_gt.
9031 (count_type_subparts): Delete.
9032 (expand_vector_operations_1): Use subparts_gt instead of
9033 count_type_subparts.
9035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9036 Alan Hayward <alan.hayward@arm.com>
9037 David Sherwood <david.sherwood@arm.com>
9039 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9040 (vect_compile_time_alias): ...this new function. Do the calculation
9041 on poly_ints rather than trees.
9042 (vect_prune_runtime_alias_test_list): Update call accordingly.
9044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9045 Alan Hayward <alan.hayward@arm.com>
9046 David Sherwood <david.sherwood@arm.com>
9048 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9049 numbers of units.
9050 (vect_schedule_slp_instance): Likewise.
9052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9053 Alan Hayward <alan.hayward@arm.com>
9054 David Sherwood <david.sherwood@arm.com>
9056 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9057 constant and extern definitions for variable-length vectors.
9058 (vect_get_constant_vectors): Note that the number of units
9059 is known to be constant.
9061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9062 Alan Hayward <alan.hayward@arm.com>
9063 David Sherwood <david.sherwood@arm.com>
9065 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9066 of units as polynomial. Choose between WIDE and NARROW based
9067 on multiple_p.
9069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9070 Alan Hayward <alan.hayward@arm.com>
9071 David Sherwood <david.sherwood@arm.com>
9073 * tree-vect-stmts.c (simd_clone_subparts): New function.
9074 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9077 Alan Hayward <alan.hayward@arm.com>
9078 David Sherwood <david.sherwood@arm.com>
9080 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9081 vectors as polynomial. Use build_index_vector for
9082 IFN_GOMP_SIMD_LANE.
9084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9085 Alan Hayward <alan.hayward@arm.com>
9086 David Sherwood <david.sherwood@arm.com>
9088 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9089 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9090 for variable-length vectors.
9091 (vectorizable_mask_load_store): Treat the number of units as
9092 polynomial, asserting that it is constant if the condition has
9093 already been enforced.
9094 (vectorizable_store, vectorizable_load): Likewise.
9096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9097 Alan Hayward <alan.hayward@arm.com>
9098 David Sherwood <david.sherwood@arm.com>
9100 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9101 of units as polynomial. Punt if we can't tell at compile time
9102 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
9109 of units as polynomial. Punt on SLP inductions. Use an integer
9110 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9111 cast of such a series for variable-length floating-point
9112 reductions.
9114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9115 Alan Hayward <alan.hayward@arm.com>
9116 David Sherwood <david.sherwood@arm.com>
9118 * tree.h (build_index_vector): Declare.
9119 * tree.c (build_index_vector): New function.
9120 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9121 of units as polynomial, forcibly converting it to a constant if
9122 vectorizable_reduction has already enforced the condition.
9123 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9124 to create a {1,2,3,...} vector.
9125 (vectorizable_reduction): Treat the number of units as polynomial.
9126 Choose vectype_in based on the largest scalar element size rather
9127 than the smallest number of units. Enforce the restrictions
9128 relied on above.
9130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9131 Alan Hayward <alan.hayward@arm.com>
9132 David Sherwood <david.sherwood@arm.com>
9134 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9135 number of units as polynomial.
9137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9138 Alan Hayward <alan.hayward@arm.com>
9139 David Sherwood <david.sherwood@arm.com>
9141 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9142 * target.def (autovectorize_vector_sizes): Return the vector sizes
9143 by pointer, using vector_sizes rather than a bitmask.
9144 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9145 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9146 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9147 Likewise.
9148 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9149 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9150 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9151 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9152 * omp-general.c (omp_max_vf): Likewise.
9153 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9154 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9155 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9156 * tree-vect-slp.c (vect_slp_bb): Likewise.
9157 * doc/tm.texi: Regenerate.
9158 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9159 to a poly_uint64.
9160 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9161 the vector size as a poly_uint64 rather than an unsigned int.
9162 (current_vector_size): Change from an unsigned int to a poly_uint64.
9163 (get_vectype_for_scalar_type): Update accordingly.
9164 * tree.h (build_truth_vector_type): Take the size and number of
9165 units as a poly_uint64 rather than an unsigned int.
9166 (build_vector_type): Add a temporary overload that takes
9167 the number of units as a poly_uint64 rather than an unsigned int.
9168 * tree.c (make_vector_type): Likewise.
9169 (build_truth_vector_type): Take the number of units as a poly_uint64
9170 rather than an unsigned int.
9172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9173 Alan Hayward <alan.hayward@arm.com>
9174 David Sherwood <david.sherwood@arm.com>
9176 * target.def (get_mask_mode): Take the number of units and length
9177 as poly_uint64s rather than unsigned ints.
9178 * targhooks.h (default_get_mask_mode): Update accordingly.
9179 * targhooks.c (default_get_mask_mode): Likewise.
9180 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9181 * doc/tm.texi: Regenerate.
9183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9184 Alan Hayward <alan.hayward@arm.com>
9185 David Sherwood <david.sherwood@arm.com>
9187 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9188 * omp-general.c (omp_max_vf): Likewise.
9189 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9190 (expand_omp_simd): Handle polynomial safelen.
9191 * omp-low.c (omplow_simd_context): Add a default constructor.
9192 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9193 (lower_rec_simd_input_clauses): Update accordingly.
9194 (lower_rec_input_clauses): Likewise.
9196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9197 Alan Hayward <alan.hayward@arm.com>
9198 David Sherwood <david.sherwood@arm.com>
9200 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9201 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9202 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9203 (vect_analyze_slp_cost): Likewise.
9204 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9205 (vect_model_load_cost): Likewise.
9207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9208 Alan Hayward <alan.hayward@arm.com>
9209 David Sherwood <david.sherwood@arm.com>
9211 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9212 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9213 from an unsigned int * to a poly_uint64_pod *.
9214 (calculate_unrolling_factor): New function.
9215 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9218 Alan Hayward <alan.hayward@arm.com>
9219 David Sherwood <david.sherwood@arm.com>
9221 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9222 from an unsigned int to a poly_uint64.
9223 (_loop_vec_info::slp_unrolling_factor): Likewise.
9224 (_loop_vec_info::vectorization_factor): Change from an int
9225 to a poly_uint64.
9226 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9227 (vect_get_num_vectors): New function.
9228 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9229 (vect_get_num_copies): Use vect_get_num_vectors.
9230 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9231 to an unsigned int *.
9232 (vect_analyze_data_refs): Change min_vf from an int * to a
9233 poly_uint64 *.
9234 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9235 than an unsigned HOST_WIDE_INT.
9236 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9237 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9238 to an unsigned int *.
9239 (vect_analyze_data_ref_dependences): Likewise.
9240 (vect_compute_data_ref_alignment): Handle polynomial vf.
9241 (vect_enhance_data_refs_alignment): Likewise.
9242 (vect_prune_runtime_alias_test_list): Likewise.
9243 (vect_shift_permute_load_chain): Likewise.
9244 (vect_supportable_dr_alignment): Likewise.
9245 (dependence_distance_ge_vf): Take the vectorization factor as a
9246 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9247 (vect_analyze_data_refs): Change min_vf from an int * to a
9248 poly_uint64 *.
9249 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9250 vfm1 as a poly_uint64 rather than an int. Make the same change
9251 for the returned bound_scalar.
9252 (vect_gen_vector_loop_niters): Handle polynomial vf.
9253 (vect_do_peeling): Likewise. Update call to
9254 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9255 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9256 be constant.
9257 * tree-vect-loop.c (vect_determine_vectorization_factor)
9258 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9259 (vect_get_known_peeling_cost): Likewise.
9260 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9261 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9262 (vect_transform_loop): Likewise. Use the lowest possible VF when
9263 updating the upper bounds of the loop.
9264 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9265 rather than an int.
9266 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9267 polynomial unroll factors.
9268 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9269 (vect_make_slp_decision): Likewise.
9270 (vect_supported_load_permutation_p): Likewise, and polynomial
9271 vf too.
9272 (vect_analyze_slp_cost): Handle polynomial vf.
9273 (vect_slp_analyze_node_operations): Likewise.
9274 (vect_slp_analyze_bb_1): Likewise.
9275 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9276 than an unsigned HOST_WIDE_INT.
9277 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9278 (vectorizable_load): Handle polynomial vf.
9279 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9280 a poly_uint64.
9281 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9284 Alan Hayward <alan.hayward@arm.com>
9285 David Sherwood <david.sherwood@arm.com>
9287 * match.pd: Handle bit operations involving three constants
9288 and try to fold one pair.
9290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9292 * tree-vect-loop-manip.c: Include gimple-fold.h.
9293 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9294 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9295 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9296 Add a path that uses a step of VF instead of 1, but disable it
9297 for now.
9298 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9299 and niters_no_overflow parameters. Update calls to
9300 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9301 Create a new SSA name if the latter choses to use a ste other
9302 than zero, and return it via niters_vector_mult_vf_var.
9303 * tree-vect-loop.c (vect_transform_loop): Update calls to
9304 vect_do_peeling, vect_gen_vector_loop_niters and
9305 slpeel_make_loop_iterate_ntimes.
9306 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9307 (vect_gen_vector_loop_niters): Update declarations after above changes.
9309 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9311 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9312 128-bit round to integer instructions.
9313 (ceil<mode>2): Likewise.
9314 (btrunc<mode>2): Likewise.
9315 (round<mode>2): Likewise.
9317 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9319 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9320 unaligned VSX load/store on P8/P9.
9321 (expand_block_clear): Allow the use of unaligned VSX
9322 load/store on P8/P9.
9324 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9326 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9327 New function.
9328 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9329 swap associated with both a load and a store.
9331 2018-01-02 Andrew Waterman <andrew@sifive.com>
9333 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9334 * config/riscv/riscv.md (clear_cache): Use it.
9336 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9338 * web.c: Remove out-of-date comment.
9340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9342 * expr.c (fixup_args_size_notes): Check that any existing
9343 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9344 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9345 (emit_single_push_insn): ...here.
9347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9349 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9350 (const_vector_encoded_nelts): New function.
9351 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9352 (const_vector_int_elt, const_vector_elt): Declare.
9353 * emit-rtl.c (const_vector_int_elt_1): New function.
9354 (const_vector_elt): Likewise.
9355 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9356 of CONST_VECTOR_ELT.
9358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9360 * expr.c: Include rtx-vector-builder.h.
9361 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9362 directly on the tree encoding.
9363 (const_vector_from_tree): Likewise.
9364 * optabs.c: Include rtx-vector-builder.h.
9365 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9366 sequence of "u" values.
9367 * vec-perm-indices.c: Include rtx-vector-builder.h.
9368 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9369 directly on the vec_perm_indices encoding.
9371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9373 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9374 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9375 * rtx-vector-builder.h: New file.
9376 * rtx-vector-builder.c: Likewise.
9377 * rtl.h (rtx_def::u2): Add a const_vector field.
9378 (CONST_VECTOR_NPATTERNS): New macro.
9379 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9380 (CONST_VECTOR_DUPLICATE_P): Likewise.
9381 (CONST_VECTOR_STEPPED_P): Likewise.
9382 (CONST_VECTOR_ENCODED_ELT): Likewise.
9383 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9384 (unwrap_const_vec_duplicate): Likewise.
9385 (const_vec_series_p): Check for a non-duplicated vector encoding.
9386 Say that the function only returns true for integer vectors.
9387 * emit-rtl.c: Include rtx-vector-builder.h.
9388 (gen_const_vec_duplicate_1): Delete.
9389 (gen_const_vector): Call gen_const_vec_duplicate instead of
9390 gen_const_vec_duplicate_1.
9391 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9392 (gen_const_vec_duplicate): Use rtx_vector_builder.
9393 (gen_const_vec_series): Likewise.
9394 (gen_rtx_CONST_VECTOR): Likewise.
9395 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9396 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9397 Build a new vector rather than modifying a CONST_VECTOR in-place.
9398 (handle_special_swappables): Update call accordingly.
9399 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9400 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9401 Build a new vector rather than modifying a CONST_VECTOR in-place.
9402 (handle_special_swappables): Update call accordingly.
9404 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9406 * simplify-rtx.c (simplify_const_binary_operation): Use
9407 CONST_VECTOR_ELT instead of XVECEXP.
9409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9411 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9412 the selector elements to be different from the data elements
9413 if the selector is a VECTOR_CST.
9414 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9415 ssizetype for the selector.
9417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9419 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9420 before testing each element individually.
9421 * tree-vect-generic.c (lower_vec_perm): Likewise.
9423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9425 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9426 * selftest-run-tests.c (selftest::run_tests): Call it.
9427 * vector-builder.h (vector_builder::operator ==): New function.
9428 (vector_builder::operator !=): Likewise.
9429 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9430 (vec_perm_indices::all_from_input_p): New function.
9431 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9432 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9433 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9434 instead of reading the VECTOR_CST directly. Detect whether both
9435 vector inputs are the same before constructing the vec_perm_indices,
9436 and update the number of inputs argument accordingly. Use the
9437 utility functions added above. Only construct sel2 if we need to.
9439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9441 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9442 the broadcast of the low byte.
9443 (expand_mult_highpart): Use an explicit encoding for the permutes.
9444 * optabs-query.c (can_mult_highpart_p): Likewise.
9445 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9446 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9447 (vectorizable_bswap): Likewise.
9448 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9449 explicit encoding for the power-of-2 permutes.
9450 (vect_permute_store_chain): Likewise.
9451 (vect_grouped_load_supported): Likewise.
9452 (vect_permute_load_chain): Likewise.
9454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9456 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9457 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9458 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9459 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9460 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9461 (vect_gen_perm_mask_any): Likewise.
9463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9465 * int-vector-builder.h: New file.
9466 * vec-perm-indices.h: Include int-vector-builder.h.
9467 (vec_perm_indices): Redefine as an int_vector_builder.
9468 (auto_vec_perm_indices): Delete.
9469 (vec_perm_builder): Redefine as a stand-alone class.
9470 (vec_perm_indices::vec_perm_indices): New function.
9471 (vec_perm_indices::clamp): Likewise.
9472 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9473 (vec_perm_indices::new_vector): New function.
9474 (vec_perm_indices::new_expanded_vector): Update for new
9475 vec_perm_indices class.
9476 (vec_perm_indices::rotate_inputs): New function.
9477 (vec_perm_indices::all_in_range_p): Operate directly on the
9478 encoded form, without computing elided elements.
9479 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9480 encoding. Update for new vec_perm_indices class.
9481 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9482 the given vec_perm_builder.
9483 (expand_vec_perm_var): Update vec_perm_builder constructor.
9484 (expand_mult_highpart): Use vec_perm_builder instead of
9485 auto_vec_perm_indices.
9486 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9487 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9488 or double series encoding as appropriate.
9489 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9490 vec_perm_indices instead of auto_vec_perm_indices.
9491 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9492 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9493 (vect_permute_store_chain): Likewise.
9494 (vect_grouped_load_supported): Likewise.
9495 (vect_permute_load_chain): Likewise.
9496 (vect_shift_permute_load_chain): Likewise.
9497 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9498 (vect_transform_slp_perm_load): Likewise.
9499 (vect_schedule_slp_instance): Likewise.
9500 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9501 (vectorizable_mask_load_store): Likewise.
9502 (vectorizable_bswap): Likewise.
9503 (vectorizable_store): Likewise.
9504 (vectorizable_load): Likewise.
9505 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9506 vec_perm_indices instead of auto_vec_perm_indices. Use
9507 tree_to_vec_perm_builder to read the vector from a tree.
9508 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9509 vec_perm_builder instead of a vec_perm_indices.
9510 (have_whole_vector_shift): Use vec_perm_builder and
9511 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9512 truncation to calc_vec_perm_mask_for_shift.
9513 (vect_create_epilog_for_reduction): Likewise.
9514 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9515 from auto_vec_perm_indices to vec_perm_indices.
9516 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9517 instead of changing individual elements.
9518 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9519 the vector in d.perm.
9520 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9521 from auto_vec_perm_indices to vec_perm_indices.
9522 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9523 instead of changing individual elements.
9524 (arm_vectorize_vec_perm_const): Use new_vector to install
9525 the vector in d.perm.
9526 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9527 Update vec_perm_builder constructor.
9528 (rs6000_expand_interleave): Likewise.
9529 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9530 (rs6000_expand_interleave): Likewise.
9532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9534 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9535 to qimode could truncate the indices.
9536 * optabs.c (expand_vec_perm_var): Likewise.
9538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9540 * Makefile.in (OBJS): Add vec-perm-indices.o.
9541 * vec-perm-indices.h: New file.
9542 * vec-perm-indices.c: Likewise.
9543 * target.h (vec_perm_indices): Replace with a forward class
9544 declaration.
9545 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9546 * optabs.h: Include vec-perm-indices.h.
9547 (expand_vec_perm): Delete.
9548 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9549 (expand_vec_perm_const): Declare.
9550 * target.def (vec_perm_const_ok): Replace with...
9551 (vec_perm_const): ...this new hook.
9552 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9553 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9554 * doc/tm.texi: Regenerate.
9555 * optabs.def (vec_perm_const): Delete.
9556 * doc/md.texi (vec_perm_const): Likewise.
9557 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9558 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9559 expand_vec_perm for constant permutation vectors. Assert that
9560 the mode of variable permutation vectors is the integer equivalent
9561 of the mode that is being permuted.
9562 * optabs-query.h (selector_fits_mode_p): Declare.
9563 * optabs-query.c: Include vec-perm-indices.h.
9564 (selector_fits_mode_p): New function.
9565 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9566 is defined, instead of checking whether the vec_perm_const_optab
9567 exists. Use targetm.vectorize.vec_perm_const instead of
9568 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9569 fit in the vector mode before using a variable permute.
9570 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9571 vec_perm_indices instead of an rtx.
9572 (expand_vec_perm): Replace with...
9573 (expand_vec_perm_const): ...this new function. Take the selector
9574 as a vec_perm_indices rather than an rtx. Also take the mode of
9575 the selector. Update call to shift_amt_for_vec_perm_mask.
9576 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9577 Use vec_perm_indices::new_expanded_vector to expand the original
9578 selector into bytes. Check whether the indices fit in the vector
9579 mode before using a variable permute.
9580 (expand_vec_perm_var): Make global.
9581 (expand_mult_highpart): Use expand_vec_perm_const.
9582 * fold-const.c: Includes vec-perm-indices.h.
9583 * tree-ssa-forwprop.c: Likewise.
9584 * tree-vect-data-refs.c: Likewise.
9585 * tree-vect-generic.c: Likewise.
9586 * tree-vect-loop.c: Likewise.
9587 * tree-vect-slp.c: Likewise.
9588 * tree-vect-stmts.c: Likewise.
9589 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9590 Delete.
9591 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9592 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9593 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9594 (aarch64_vectorize_vec_perm_const): ...this new function.
9595 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9596 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9597 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9598 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9599 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9601 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9602 into...
9603 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9604 check for NEON modes.
9605 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9606 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9607 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9608 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9609 into...
9610 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9611 the old VEC_PERM_CONST conditions.
9612 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9613 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9614 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9615 (ia64_vectorize_vec_perm_const_ok): Merge into...
9616 (ia64_vectorize_vec_perm_const): ...this new function.
9617 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9618 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9619 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9620 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9621 * config/mips/mips.c (mips_expand_vec_perm_const)
9622 (mips_vectorize_vec_perm_const_ok): Merge into...
9623 (mips_vectorize_vec_perm_const): ...this new function.
9624 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9625 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9626 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9627 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9628 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9629 (rs6000_expand_vec_perm_const): Delete.
9630 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9631 Delete.
9632 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9633 (altivec_expand_vec_perm_const_le): Take each operand individually.
9634 Operate on constant selectors rather than rtxes.
9635 (altivec_expand_vec_perm_const): Likewise. Update call to
9636 altivec_expand_vec_perm_const_le.
9637 (rs6000_expand_vec_perm_const): Delete.
9638 (rs6000_vectorize_vec_perm_const_ok): Delete.
9639 (rs6000_vectorize_vec_perm_const): New function.
9640 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9641 an element count and rtx array.
9642 (rs6000_expand_extract_even): Update call accordingly.
9643 (rs6000_expand_interleave): Likewise.
9644 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9645 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9646 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9647 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9648 (rs6000_expand_vec_perm_const): Delete.
9649 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9650 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9651 (altivec_expand_vec_perm_const_le): Take each operand individually.
9652 Operate on constant selectors rather than rtxes.
9653 (altivec_expand_vec_perm_const): Likewise. Update call to
9654 altivec_expand_vec_perm_const_le.
9655 (rs6000_expand_vec_perm_const): Delete.
9656 (rs6000_vectorize_vec_perm_const_ok): Delete.
9657 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9658 reference to the SPE evmerge intructions.
9659 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9660 an element count and rtx array.
9661 (rs6000_expand_extract_even): Update call accordingly.
9662 (rs6000_expand_interleave): Likewise.
9663 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9664 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9665 new function.
9666 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9668 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9670 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9671 vector mode and that that mode matches the mode of the data
9672 being permuted.
9673 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9674 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9675 directly using expand_vec_perm_1 when forcing selectors into
9676 registers.
9677 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9681 * optabs-query.h (can_vec_perm_p): Delete.
9682 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9683 * optabs-query.c (can_vec_perm_p): Split into...
9684 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9685 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9686 particular selector is valid.
9687 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9688 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9689 (vect_grouped_load_supported): Likewise.
9690 (vect_shift_permute_load_chain): Likewise.
9691 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9692 (vect_transform_slp_perm_load): Likewise.
9693 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9694 (vectorizable_bswap): Likewise.
9695 (vect_gen_perm_mask_checked): Likewise.
9696 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9697 implementations of variable permutation vectors into account
9698 when deciding which selector to use.
9699 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9700 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9701 with a false third argument.
9702 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9703 to test whether the constant selector is valid and can_vec_perm_var_p
9704 to test whether a variable selector is valid.
9706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9708 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9709 * optabs-query.c (can_vec_perm_p): Likewise.
9710 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9711 instead of vec_perm_indices.
9712 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9713 (vect_gen_perm_mask_checked): Likewise,
9714 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9715 (vect_gen_perm_mask_checked): Likewise,
9717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9719 * optabs-query.h (qimode_for_vec_perm): Declare.
9720 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9721 (qimode_for_vec_perm): ...this new function.
9722 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9724 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9726 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9727 does not have a conditional at the top.
9729 2018-01-02 Richard Biener <rguenther@suse.de>
9731 * ipa-inline.c (big_speedup_p): Fix expression.
9733 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9735 PR target/81616
9736 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9737 for generic 4->6.
9739 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9741 PR target/81616
9742 Generic tuning.
9743 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9744 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9745 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9746 cond_taken_branch_cost 3->4.
9748 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9750 PR tree-optimization/83581
9751 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9752 TODO_cleanup_cfg if any changes have been made.
9754 PR middle-end/83608
9755 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9756 convert_modes if target mode has the right side, but different mode
9757 class.
9759 PR middle-end/83609
9760 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9761 last argument when extracting from CONCAT. If either from_real or
9762 from_imag is NULL, use expansion through memory. If result is not
9763 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9764 the parts directly to inner mode, if even that fails, use expansion
9765 through memory.
9767 PR middle-end/83623
9768 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9769 check for bswap in mode rather than HImode and use that in expand_unop
9770 too.
9771 \f
9772 Copyright (C) 2018 Free Software Foundation, Inc.
9774 Copying and distribu