| blob | 1370b34f466aaa23fd98240d145a9419840bc368 |
1 2018-04-04 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/85168
4 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5 propagating abnormals.
7 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12 Kito Cheng <kito.cheng@gmail.com>
14 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
17 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18 * config/nds32/nds32.md (sibcall_internal): New.
19 (sibcall_register): Remove.
20 (sibcall_immediate): Remove.
21 (sibcall_value_internal): New.
22 (sibcall_value_register): Remove.
23 (sibcall_value_immediate): Remove.
24 * config/nds32/predicates.md (nds32_general_register_operand): New.
25 (nds32_call_address_operand): New.
27 2018-04-03 Jakub Jelinek <jakub@redhat.com>
29 PR rtl-optimization/85167
30 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
31 bb_defs if *split_p, instead preinitialize it to NULL.
33 PR tree-optimization/85156
34 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
35 evaluating the argument multiple times.
37 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
39 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
40 than vector.
41 (_mm_cvtpd_ps): Likewise.
42 (_mm_cvttpd_epi32): Likewise.
43 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
44 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
45 vector, pixel, and bool following altivec.h include.
47 2018-04-03 Martin Sebor <msebor@redhat.com>
49 * doc/extend.texi (Common Function Attributes): Clarify.
50 (const attribute): Likewise.
51 (pure attribute): Likewise.
53 2018-04-03 Jakub Jelinek <jakub@redhat.com>
55 PR target/85169
56 * config/i386/i386.c (ix86_expand_vector_set): Use
57 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
59 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
61 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
62 instructions when changing rounding bits to preserve precision bits
63 in the x87 control word.
65 2018-04-03 Martin Liska <mliska@suse.cz>
67 PR tree-optimization/82491
68 * rtl.h (strip_offset_and_add): Replace += suboffset with
69 poly_uint64 () + suboffset.
71 2018-03-29 Martin Liska <mliska@suse.cz>
72 Martin Jambor <mjambor@suse.cz>
74 PR ipa/84947
75 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
76 param_type is not an integral or pointer type.
78 2018-04-03 Richard Biener <rguenther@suse.de>
80 * sese.h (recompute_all_dominators): Remove.
82 2018-04-02 Martin Sebor <msebor@redhat.com>
84 * doc/invoke.texi (-Wrestrict): Fix typos.
86 2018-04-02 Jim Wilson <jimw@sifive.com>
88 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
89 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
90 (<optab>di3, <optab>si3_extend): Likewise.
91 (<optab>si3_mask, <optab>si3_mask_1): New.
92 (<optab>di3_mask, <optab>di3_mask_1): New.
93 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
94 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
95 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
97 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
99 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
100 example.
102 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
104 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
105 (nds32_canonicalize_comparison): New function.
107 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
108 Kito Cheng <kito.cheng@gmail.com>
109 Kuan-Lin Chen <kuanlinchentw@gmail.com>
111 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
112 * config/nds32/constants.md (unspec_volatile_element): Add
113 UNSPEC_VOLATILE_RELAX_GROUP.
114 * config/nds32/nds32-relax-opt.c: New file.
115 * config/nds32/nds32-predicates.c
116 (nds32_symbol_load_store_p): New function.
117 * config/nds32/nds32-protos.h
118 (nds32_symbol_load_store_p): Declare function.
119 (make_pass_nds32_relax_opt): Declare new rtl pass function.
120 * config/nds32/nds32.c
121 (nds32_register_pass): New function to register pass.
122 (nds32_register_passes): New function to register passes.
123 * config/nds32/nds32.md (relax_group): New pattern.
124 * config/nds32/nds32.opt (mrelax-hint): New option.
125 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
127 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
129 * config/nds32/t-nds32: Modify files dependency.
131 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
133 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
134 (PROFILE_HOOK): Define its implementation.
136 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
138 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
139 type and 32-bit size.
141 2018-04-01 Jakub Jelinek <jakub@redhat.com>
143 PR middle-end/85090
144 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
145 (V_128_256): New mode iterator.
146 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
147 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
148 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
149 of V.
150 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
151 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
153 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
155 PR target/83315
156 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
157 NaN inputs correctly.
159 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
161 PR target/80546
162 * config/rs6000/vsx.md (??r): New mode attribute.
163 (*vsx_mov<mode>_64bit): Use it.
164 (*vsx_mov<mode>_32bit): Likewise.
166 2018-03-30 Martin Sebor <msebor@redhat.com>
168 PR tree-optimization/84818
169 * builtins.c (check_access): Use warning_n.
171 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
173 PR target/83822
174 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
175 condition.
176 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
177 condition.
179 2018-03-30 Julia Koval <julia.koval@intel.com>
181 PR target/84413
182 * x86-tune.def (movx, partial_reg_dependency): Enable for
183 m_SKYLAKE_AVX512.
185 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
187 PR inline-asm/84985
188 * lra-constraints.c (process_alt_operands): Move setting
189 this_alternative_matches below.
191 2018-03-29 Martin Liska <mliska@suse.cz>
193 PR lto/84995.
194 * doc/invoke.texi: Document how LTO works with debug info.
195 Describe auto-load support of binutils. Mention 'x86-64'
196 as valid option value of -march option.
198 2018-03-29 Jakub Jelinek <jakub@redhat.com>
200 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
202 PR c/85094
203 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
204 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
205 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
206 checking.
208 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
210 PR target/84912
211 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
212 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
213 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
214 for RS6000_BTM_POWERPC64.
215 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
216 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
217 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
218 definition.
219 (DIVDE): Use it.
220 (DIVDEU): Likewise.
222 2018-03-28 Carl Love <cel@us.ibm.com>
224 Revert
225 2017-09-27 Carl Love <cel@us.ibm.com>
227 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
228 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
229 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
230 fctiw instruction.
232 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
234 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
235 instead of __vector bool.
236 (_mm_max_pu8): Likewise.
237 (_mm_min_pi16): Likewise.
239 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
241 PR target/84912
242 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
243 (DIVWEUO): Likewise.
244 (DIVDEO): Likewise.
245 (DIVDEUO): Likewise.
246 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
247 DIVWEUO and DIVDEUO.
248 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
249 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
250 (div_extend): Likewise.
251 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
252 builtin function.
253 (__builtin_divweuo): Likewise.
254 (__builtin_divdeo): Likewise.
255 (__builtin_divdeuo): Likewise.
257 2018-03-28 Jakub Jelinek <jakub@redhat.com>
259 PR target/85095
260 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
261 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
263 PR tree-optimization/82004
264 * gimple-match-head.c (optimize_pow_to_exp): New function.
265 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
266 Don't fold to exp if optimize_pow_to_exp is false.
268 2018-03-28 Martin Liska <mliska@suse.cz>
270 PR other/84819
271 * calls.c (initialize_argument_information): Fix trailing space.
272 * common.opt: Fix typo and provide better explanation for
273 -fsanitize-coverage option.
274 * config/i386/i386.opt: Fix typo.
276 2018-03-28 Jakub Jelinek <jakub@redhat.com>
277 Martin Liska <mliska@suse.cz>
279 PR sanitizer/85081
280 * gimplify.c (asan_poison_variable): Don't do the check for
281 gimplify_omp_ctxp here.
282 (gimplify_decl_expr): Do it here.
283 (gimplify_target_expr): Likewise.
285 2018-03-28 Martin Liska <mliska@suse.cz>
287 PR target/84988
288 * config/i386/i386.c (ix86_function_arg_advance): Do not call
289 chkp_type_bounds_count if MPX is not enabled.
291 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
293 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
295 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
297 PR target/84914
298 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
299 function to create the function decl for complex long double
300 multiply and divide for -mabi=ieeelongdouble.
301 (init_float128_ieee): Call it.
303 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
305 PR target/85044
306 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
307 -fcf-protection=branch -mibt.
308 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
310 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
312 PR target/81863
313 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
315 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
317 PR target/85056
318 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
319 extern array declarations.
321 2018-03-27 Richard Biener <rguenther@suse.de>
323 PR middle-end/84067
324 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
325 explicit single_use checks.
327 2018-03-27 Richard Biener <rguenther@suse.de>
329 PR tree-optimization/85082
330 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
331 Valueize the VUSE.
333 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
335 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
336 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
337 Turn on fasynchronous-unwind-tables and funwind-tables.
339 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
341 PR target/85073
342 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
343 (*bmi_blsr_<mode>_ccz): Ditto.
345 2018-03-26 Tom de Vries <tom@codesourcery.com>
347 PR tree-optimization/85063
348 * omp-general.c (offloading_function_p): New function. Factor out
349 of ...
350 * omp-offload.c (pass_omp_target_link::gate): ... here.
351 * omp-general.h (offloading_function_p): Declare.
352 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
353 with attribute omp declare target for offloading functions.
355 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
357 PR tree-optimization/84005
358 * tree-data-ref.h (get_base_for_alignment): Declare.
359 * tree-data-ref.c (get_base_for_alignment_1): New function.
360 (get_base_for_alignment): Likewise.
361 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
362 get_base_for_alignment to find a suitable base object, instead
363 of always using drb->base_address.
365 2018-03-23 Jakub Jelinek <jakub@redhat.com>
367 PR inline-asm/85022
368 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
369 known size by default.
371 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
373 PR inline-asm/85030
374 * lra-constraints.c (process_alt_operands): Don't match BLKmode
375 and non BLKmode operands.
377 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
379 PR target/85026
380 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
381 Clean up attributes.
383 2018-03-23 Richard Biener <rguenther@suse.de>
385 PR debug/85020
386 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
387 we are going to emit early debug for LTO.
389 2018-03-23 Jakub Jelinek <jakub@redhat.com>
391 PR inline-asm/85034
392 * function.c (match_asm_constraints_1): Don't optimize if input
393 doesn't satisfy general_operand predicate for output's mode.
395 PR inline-asm/85022
396 * alias.c (write_dependence_p): Don't require for x_canonicalized
397 non-VOIDmode if x has VOIDmode.
399 PR sanitizer/85029
400 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
401 just don't try to optimize it rather than assert it never happens.
403 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
405 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
406 macro expansions for definition of ST_INTERNAL_<mode> and
407 LD_INTERNAL_<mode> builtins.
408 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
409 Remove prototype.
410 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
411 function.
412 (altivec_expand_st_builtin): Likewise.
413 (altivec_expand_builtin): Remove calls to deleted functions.
414 (rs6000_address_for_altivec): Delete this function.
415 * config/rs6000/vector.md: Remove expands for
416 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
418 2018-03-22 Sudakshina Das <sudi.das@arm.com>
420 PR target/84826
421 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
422 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
423 re-computing once computed.
424 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
425 (arm_init_machine_status): Initialize
426 machine->static_chain_stack_bytes.
428 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
430 PR target/84760
431 * doc/extend.texi: Add four new prototypes for vec_ld.
432 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
433 definitions for more logical presentation.
434 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
435 entries for V1TI variants of __builtin_altivec_ld builtin.
436 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
437 handling of V1TI variant of LVX icode pattern.
438 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
439 (rs6000_gimple_fold_builtin): Likewise.
440 (altivec_init_builtins): Add code to define
441 __builtin_altivec_lvx_v1ti function.
443 2018-03-22 Jakub Jelinek <jakub@redhat.com>
445 PR inline-asm/84941
446 * function.c (match_asm_constraints_1): Don't do the optimization
447 if input isn't a REG, SUBREG, MEM or constant.
449 2018-03-22 Tom de Vries <tom@codesourcery.com>
451 PR tree-optimization/84956
452 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
453 bb_has_abnormal_pred.
455 2018-03-22 Jakub Jelinek <jakub@redhat.com>
457 PR sanitizer/85018
458 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
459 DECL_INITIAL (decl) to decl at the end.
460 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
461 adjust the comment.
463 2018-03-21 Joseph Myers <joseph@codesourcery.com>
465 * doc/extend.texi (__builtin_tgmath): Document when complex
466 integer types are treated as _Complex _Float64.
468 2018-03-21 Tom de Vries <tom@codesourcery.com>
470 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
472 2018-03-21 Jakub Jelinek <jakub@redhat.com>
474 PR tree-optimization/84960
475 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
476 if it is ENTRY block, move them into single succ of ENTRY in that case.
478 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
480 PR tree-optimization/84811
481 * poly-int.h (poly_span_traits): Remove the T3 parameter and
482 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
483 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
484 (known_subrange_p): Update accordingly. Cast each value involved
485 in the size comparison, rather than casting the result of the
486 subtraction.
488 2018-03-21 Jakub Jelinek <jakub@redhat.com>
490 PR tree-optimization/84982
491 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
492 by flipping the least significant bit rather than all bits from
493 bitpos to bitpos + bitsize - 1.
495 2018-03-21 Nathan Sidwell <nathan@acm.org>
497 * doc/extend.texi (Deprecated Features): Remove mention of
498 long-deleted deprecations.
500 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
502 PR jit/84288
503 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
504 * configure: Regenerate.
506 2018-03-21 Tom de Vries <tom@codesourcery.com>
508 PR tree-optimization/83126
509 * tree-parloops.c (num_phis): New function.
510 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
512 2018-03-21 Nathan Sidwell <nathan@acm.org>
514 * doc/extend.texi (Deprecated Features): Update deprecated flags,
515 mention anon-struct/union members and trailing attributes.
517 2018-03-21 Bin Cheng <bin.cheng@arm.com>
519 PR tree-optimization/84969
520 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
521 builtin memset partitions if they set different rhs values.
523 2018-03-21 Jakub Jelinek <jakub@redhat.com>
525 PR rtl-optimization/84989
526 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
527 VEC_DUPLICATE with scalar result mode.
529 2018-03-21 Martin Liska <mliska@suse.cz>
531 PR ipa/84963
532 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
533 not intended return statement.
535 2018-03-21 Martin Liska <mliska@suse.cz>
537 PR target/84988
538 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
539 (chkp_find_bound_slots_1): Limit number of iterations.
541 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
543 PR target/84838
544 * Minor grammar fixes for x86 options.
546 2018-03-20 Jakub Jelinek <jakub@redhat.com>
548 PR debug/84875
549 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
550 holding REG_CFA_RESTORE notes, instead turn them into a USE.
552 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
554 PR target/83789
555 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
556 (altivec_lvx_<mode>_1op): Likewise.
557 (altivec_stvx_<mode>_2op): Likewise.
558 (altivec_stvx_<mode>_1op): Likewise.
559 (altivec_lvx_<VM2:mode>): New define_expand.
560 (altivec_stvx_<VM2:mode>): Likewise.
561 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
562 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
563 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
564 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
565 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
566 (rs6000_gen_lvx): Likewise.
567 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
568 (altivec_expand_stv_builtin): Likewise.
569 (altivec_expand_builtin): Likewise.
570 * config/rs6000/vector.md: Likewise.
572 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
574 PR target/82518
575 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
576 BYTES_BIG_ENDIAN.
578 2018-03-20 Richard Biener <rguenther@suse.de>
580 PR target/84986
581 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
582 sign-conversions as zero, fall back to standard scalar_stmt
583 cost for the rest.
585 2018-03-20 Martin Liska <mliska@suse.cz>
587 PR ipa/84825
588 * predict.c (rebuild_frequencies): Handle case when we have
589 PROFILE_ABSENT, but flag_guess_branch_prob is false.
591 2018-03-20 Jakub Jelinek <jakub@redhat.com>
593 PR target/84990
594 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
595 flag_section_anchors.
596 * varasm.c (use_blocks_for_decl_p): Remove hack for
597 dw2_force_const_mem.
599 PR target/84845
600 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
601 to ...
602 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
603 be created, use lowpart_subreg of operands[0] rather than operands[0]
604 itself.
605 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
606 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
607 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
608 and n constraint instead of aarch64_shift_imm_di and Usd.
609 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
610 (*aarch64_<optab>_reg_minus<mode>3): ... this.
612 2018-03-20 Sudakshina Das <sudi.das@arm.com>
614 PR target/82989
615 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
616 to favor GPR over NEON registers.
617 (<shift>di3_neon): Likewise.
619 2018-03-20 Tom de Vries <tom@codesourcery.com>
621 PR target/84952
622 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
623 (nvptx_process_pars): Emit bar.sync asap and alap.
625 2018-03-20 Tom de Vries <tom@codesourcery.com>
627 PR target/84954
628 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
629 seen_label if seen_label is already set.
631 2018-03-20 Jakub Jelinek <jakub@redhat.com>
633 PR target/84945
634 * config/i386/i386.c (fold_builtin_cpu): For features above 31
635 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
636 Use 1U instead of 1. Formatting fixes.
638 PR c/84953
639 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
640 instead of TREE_TYPE (s1) for the return value.
642 2018-03-19 Jakub Jelinek <jakub@redhat.com>
644 PR tree-optimization/84946
645 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
646 bitsize + bitsize in poly_uint64 rather than poly_int64.
648 PR sanitizer/78651
649 * dwarf2asm.c: Include fold-const.c.
650 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
651 of decl rather than decl itself.
653 PR rtl-optimization/84643
654 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
656 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
658 PR sanitizer/78651
659 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
660 calling assemble_variable.
662 2018-03-19 Sudakshina Das <sudi.das@arm.com>
664 PR target/81647
665 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
666 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
668 2018-03-19 Jim Wilson <jimw@sifive.com>
670 PR bootstrap/84856
671 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
672 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
673 (riscv_first_stack_step): Likewise.
674 (riscv_option_override): Use STACK_BOUNDARY instead of
675 MIN_STACK_BOUNDARY.
676 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
677 MIN_STACK_BOUNDARY.
678 (BIGGEST_ALIGNMENT): Set to 128.
679 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
680 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
681 STACK_BOUNDARY.
683 2018-03-19 Richard Biener <rguenther@suse.de>
685 PR tree-optimization/84933
686 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
687 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
689 2018-03-19 Richard Biener <rguenther@suse.de>
691 PR tree-optimization/84859
692 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
693 (cond_if_else_store_replacement): Perform sinking operation on
694 single-store BBs regardless of MAX_STORES_TO_SINK setting.
695 Generalize what a BB with a single eligible store is.
697 2018-03-19 Richard Biener <rguenther@suse.de>
699 PR tree-optimization/84929
700 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
701 chrec_is_positive against non-chrec arg.
703 2018-03-19 Tamar Christina <tamar.christina@arm.com>
705 PR target/84711
706 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
708 2018-03-18 Martin Liska <mliska@suse.cz>
710 PR rtl-optimization/84635
711 * regrename.c (build_def_use): Use matches_mode only when
712 matches >= 0.
714 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
716 PR tree-optimization/84913
717 * tree-vect-loop.c (vectorizable_reduction): Don't try to
718 vectorize chains of COND_EXPRs.
720 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
722 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
724 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
726 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
728 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
730 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
732 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
733 Kito Cheng <kito.cheng@gmail.com>
735 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
736 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
737 (nds32_adjust_reg_alloc_order): New function.
738 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
740 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
742 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
743 nds32_print_operand, nds32_print_operand_address): Use
744 HOST_WIDE_INT_PRINT_DEC instead.
746 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
748 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
750 2018-03-17 Jakub Jelinek <jakub@redhat.com>
752 PR target/84902
753 * config/i386/i386.c (initial_ix86_tune_features,
754 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
755 unsigned long long.
756 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
757 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
758 rather than 1u << ix86_tune. Formatting fix.
759 (ix86_option_override_internal): Change ix86_arch_mask from
760 unsigned int to unsigned HOST_WIDE_INT, initialize to
761 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
762 (ix86_function_specific_restore): Likewise.
764 2018-03-16 Jakub Jelinek <jakub@redhat.com>
766 PR target/84899
767 * postreload.c (reload_combine_recognize_pattern): Perform
768 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
769 truncate_int_for_mode the result for the destination's mode.
771 PR c/84909
772 * hsa-gen.c (mem_type_for_type): Fix comment typo.
773 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
774 Likewise.
775 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
776 Likewise.
778 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
780 PR target/84876
781 * lra-assigns.c (lra_split_hard_reg_for): Don't use
782 regno_allocno_class_array and sorted_pseudos.
783 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
784 insns where regno is used.
786 2018-03-16 Martin Liska <mliska@suse.cz>
788 PR ipa/84833
789 * multiple_target.c (create_dispatcher_calls): Redirect
790 reference in the symbol table.
792 2018-03-16 Martin Liska <mliska@suse.cz>
794 PR ipa/84722
795 * multiple_target.c (create_dispatcher_calls): Redirect also
796 an alias.
798 2018-03-16 Jakub Jelinek <jakub@redhat.com>
800 PR c++/79937
801 PR c++/82410
802 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
803 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
804 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
806 2018-03-16 Julia Koval <julia.koval@intel.com>
808 * doc/invoke.texi (Skylake Server): Add CLWB.
809 Cannonlake): Remove CLWB.
811 2018-03-16 Jakub Jelinek <jakub@redhat.com>
813 PR tree-optimization/84841
814 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
815 1 << 3.
816 (FLOAT_ONE_CONST_TYPE): Define.
817 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
818 (sort_by_operand_rank): Put entries with higher constant_type last
819 rather than first to match comments.
821 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
823 * config/nios2/nios2.md (movsi_internal): Fix thinko in
824 split predicate.
826 2018-03-15 Jakub Jelinek <jakub@redhat.com>
828 PR c++/79085
829 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
830 check and use address of target always.
832 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
834 PR target/84574
835 * config/i386/i386.c (indirect_thunk_needed): Update comments.
836 (indirect_thunk_bnd_needed): Likewise.
837 (indirect_thunks_used): Likewise.
838 (indirect_thunks_bnd_used): Likewise.
839 (indirect_return_needed): New.
840 (indirect_return_bnd_needed): Likewise.
841 (output_indirect_thunk_function): Add a bool argument for
842 function return.
843 (output_indirect_thunk_function): Don't generate alias for
844 function return thunk.
845 (ix86_code_end): Call output_indirect_thunk_function to generate
846 function return thunks.
847 (ix86_output_function_return): Set indirect_return_bnd_needed
848 and indirect_return_needed instead of indirect_thunk_bnd_needed
849 and indirect_thunk_needed.
851 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
853 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
854 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
855 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
857 2018-03-15 David Malcolm <dmalcolm@redhat.com>
858 Paul Hua <paul.hua.gm@gmail.com>
860 PR c/84852
861 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
863 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
865 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
866 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
867 resp. SFmode cases.
869 2018-03-15 Tamar Christina <tamar.christina@arm.com>
871 PR target/84711
872 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
873 instead of GET_MODE_SIZE when comparing Units.
875 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
877 PR target/68256
878 * varasm.c (hash_section): Return an unchangeble hash value
879 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
880 Return !aarch64_can_use_per_function_literal_pools_p ().
882 2018-03-15 Jakub Jelinek <jakub@redhat.com>
884 PR target/84860
885 * optabs.c (emit_conditional_move): Pass address of cmode's copy
886 rather than address of cmode as last argument to prepare_cmp_insn.
888 2018-03-15 Julia Koval <julia.koval@intel.com>
890 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
891 F_AVX512VNNI, F_AVX512BITALG): New.
893 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
895 PR target/83451
896 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
897 insn for floating-point loads and stores.
899 2018-03-14 Carl Love <cel@us.ibm.com>
901 * config/rs6000/rs6000-c.c: Add macro definitions for
902 ALTIVEC_BUILTIN_VEC_PERMXOR.
903 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
904 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
905 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
906 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
907 UNSPEC_VPERMXOR.
908 * config/doc/extend.texi: Add prototypes for vec_permxor.
910 2018-03-14 David Malcolm <dmalcolm@redhat.com>
912 PR c/84852
913 * diagnostic-show-locus.c (class layout_point): Convert m_line
914 from int to linenum_type.
915 (line_span::comparator): Use linenum "compare" function when
916 comparing line numbers.
917 (test_line_span): New function.
918 (layout_range::contains_point): Convert param "row" from int to
919 linenum_type.
920 (layout_range::intersects_line_p): Likewise.
921 (layout::will_show_line_p): Likewise.
922 (layout::print_source_line): Likewise.
923 (layout::should_print_annotation_line_p): Likewise.
924 (layout::print_annotation_line): Likewise.
925 (layout::print_leading_fixits): Likewise.
926 (layout::annotation_line_showed_range_p): Likewise.
927 (struct line_corrections): Likewise for field m_row.
928 (line_corrections::line_corrections): Likewise for param "row".
929 (layout::print_trailing_fixits): Likewise.
930 (layout::get_state_at_point): Likewise.
931 (layout::get_x_bound_for_row): Likewise.
932 (layout::print_line): Likewise.
933 (diagnostic_show_locus): Likewise for locals "last_line" and
934 "row".
935 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
936 * input.c (selftest::test_linenum_comparisons): New function.
937 (selftest::input_c_tests): Call it.
938 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
939 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
940 * selftest.h (ASSERT_GT): New macro.
941 (ASSERT_GT_AT): New macro.
942 (ASSERT_LT): New macro.
943 (ASSERT_LT_AT): New macro.
945 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
947 PR rtl-optimization/84780
948 * combine.c (distribute_links): Don't make a link based on pc_rtx.
950 2018-03-14 Martin Liska <mliska@suse.cz>
952 * tree.c (record_node_allocation_statistics): Use
953 get_stats_node_kind.
954 (get_stats_node_kind): New function extracted from
955 record_node_allocation_statistics.
956 (free_node): Use get_stats_node_kind.
958 2018-03-14 Richard Biener <rguenther@suse.de>
960 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
961 that the value-set of ANTIC_IN doesn't grow.
963 Revert
964 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
965 member.
966 (BB_VISITED_WITH_VISITED_SUCCS): New define.
967 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
969 2018-03-14 Julia Koval <julia.koval@intel.com>
971 * config.gcc (icelake-client, icelake-server): New.
972 (icelake): Remove.
973 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
974 (initial_ix86_arch_features): Ditto.
975 (PTA_SKYLAKE): Add SGX.
976 (PTA_ICELAKE): Remove.
977 (PTA_ICELAKE_CLIENT): New.
978 (PTA_ICELAKE_SERVER): New.
979 (ix86_option_override_internal): Split up icelake on icelake client and
980 icelake server.
981 (get_builtin_code_for_version): Ditto.
982 (fold_builtin_cpu): Ditto.
983 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
984 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
985 * config/i386/i386.h (processor_type): Ditto.
986 * doc/invoke.texi: Ditto.
988 2018-03-14 Jakub Jelinek <jakub@redhat.com>
990 PR sanitizer/83392
991 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
992 INTEGER_CST offset, add it together with bitpos / 8 and
993 sign extend based on POINTER_SIZE.
995 PR target/84844
996 Revert
997 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
999 PR target/78090
1000 * config/i386/constraints.md (Yc): New register constraint.
1001 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1002 Use Yc constraint for alternative 2 of operand 0. Remove
1003 preferred_for_speed attribute.
1005 2018-03-14 Richard Biener <rguenther@suse.de>
1007 PR tree-optimization/84830
1008 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1009 with the old one to avoid oscillations.
1011 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1013 PR target/83712
1014 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1015 pseudos.
1016 (assign_by_spills): Return a flag of reload assignment failure.
1017 Do not process the reload assignment failures. Do not spill other
1018 reload pseudos if they has the same reg class. Update n if
1019 necessary.
1020 (lra_assign): Add a return arg. Set up from the result of
1021 assign_by_spills call.
1022 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1023 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1024 usage_insns if it is not NULL.
1025 (spill_hard_reg_in_range): New function.
1026 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1027 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1028 function prototypes.
1029 (lra_assign): Change prototype.
1030 * lra.c (lra): Add code to deal with fails by splitting hard reg
1031 live ranges.
1033 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1035 * config/riscv/riscv.opt (mrelax): New option.
1036 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1037 "norelax" when riscv_mrelax is disabled.
1038 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1040 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1042 PR target/84743
1043 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1044 reassociation for int modes.
1046 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1048 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1049 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1050 for big-endian.
1051 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1052 * config/aarch64/aarch64-sve.md
1053 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1054 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1055 (*extend<mode><Vwide>2): Rename to...
1056 (aarch64_sve_extend<mode><Vwide>2): ...this.
1057 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1058 renaming the old pattern to...
1059 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1060 unsigned packs.
1061 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1062 define_expand, renaming the old pattern to...
1063 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1064 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1065 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1066 account when deciding which SVE instruction the optab should use.
1067 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1069 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1071 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1072 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1073 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1074 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1075 (tlsdesc_small_<mode>): Turn a define_expand and use
1076 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1077 (tlsdesc_small_advsimd_<mode>): ...this.
1078 (tlsdesc_small_sve_<mode>): New pattern.
1080 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1082 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1083 (UNSPEC_UMUL_HIGHPART): New constants.
1084 (MUL_HIGHPART): New int iteraor.
1085 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1086 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1087 define_expand.
1088 (*<su>mul<mode>3_highpart): New define_insn.
1090 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1092 PR lto/84805
1093 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1094 incomplete types.
1096 2018-03-13 Martin Liska <mliska@suse.cz>
1098 PR ipa/84658.
1099 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1100 vector.
1101 (sem_item_optimizer::~sem_item_optimizer): Release it.
1102 (sem_item_optimizer::merge_classes): Register variable aliases.
1103 (sem_item_optimizer::fixup_pt_set): New function.
1104 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1105 * ipa-icf.h: Declare new variables and functions.
1107 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1109 PR middle-end/84834
1110 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1111 integer_pow2p@2 and test integer_pow2p in condition.
1112 (A < 0 ? C : 0): Similarly for @1.
1114 PR middle-end/84831
1115 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1116 characters starting at p contain '\0' character, don't look beyond
1117 that.
1119 PR target/84827
1120 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1121 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1123 PR target/84828
1124 * reg-stack.c (change_stack): Change update_end var from int to
1125 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1126 also call set_block_for_insn on the newly added insns and rescan.
1128 PR target/84786
1129 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1130 on the last operand.
1132 PR c++/84704
1133 * tree.c (stabilize_reference_1): Return save_expr (e) for
1134 STATEMENT_LIST even if it doesn't have side-effects.
1136 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1138 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1140 2018-03-12 Renlin Li <renlin.li@arm.com>
1142 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1143 aarch64_output_scalar_simd_mov_immediate.
1145 2018-03-12 Martin Sebor <msebor@redhat.com>
1147 PR tree-optimization/83456
1148 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1149 for perfectly overlapping calls to memcpy.
1150 (gimple_fold_builtin_memory_chk): Same.
1151 (gimple_fold_builtin_strcpy): Handle no-warning.
1152 (gimple_fold_builtin_stxcpy_chk): Same.
1153 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1155 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1157 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1158 parameter. Use it for SFmode.
1159 (rs6000_function_arg_advance_1): Adjust.
1160 (rs6000_function_arg): Adjust.
1161 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1163 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1165 PR rtl-optimization/84169
1166 PR rtl-optimization/84780
1167 * combine.c (can_combine_p): Check for a 2-insn combination whether
1168 the destination register is used between the two insns, too.
1170 2018-03-12 Richard Biener <rguenther@suse.de>
1172 PR tree-optimization/84803
1173 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1174 for refs DR analysis didn't process.
1176 2018-03-12 Richard Biener <rguenther@suse.de>
1178 PR tree-optimization/84777
1179 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1180 force-vectorize loops ignore whether we are optimizing for size.
1182 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1184 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1185 (TARGET_MD_ASM_ADJUST): Define.
1187 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1188 Kito Cheng <kito.cheng@gmail.com>
1189 Chung-Ju Wu <jasonwucj@gmail.com>
1191 * config/nds32/nds32.c (nds32_compute_stack_frame,
1192 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1193 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1194 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1195 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1196 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1197 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1198 * config/nds32/nds32.md (prologue, epilogue): Use macro
1199 NDS32_V3PUSH_AVAILABLE_P to do checking.
1201 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1203 PR debug/58150
1204 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1205 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1206 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1207 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1208 addition of most attributes on !orig_type_die or the attribute not
1209 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1211 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1212 Chung-Ju Wu <jasonwucj@gmail.com>
1214 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1215 __NDS32_VH__ macro.
1216 * config/nds32/nds32.opt (mvh): New option.
1218 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1219 Chung-Ju Wu <jasonwucj@gmail.com>
1221 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1222 function.
1223 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1224 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1225 definition.
1227 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1228 Chung-Ju Wu <jasonwucj@gmail.com>
1230 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1231 function.
1232 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1233 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1235 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1236 Kito Cheng <kito.cheng@gmail.com>
1237 Chung-Ju Wu <jasonwucj@gmail.com>
1239 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1240 UNSPEC_FFMISM and UNSPEC_FLMISM.
1241 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1242 for ffb, ffmism and flmism.
1243 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1244 (unspec_ffmism): Ditto.
1245 (unspec_flmism): Ditto.
1246 (nds32_expand_builtin_impl): Check if string extension is available.
1247 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1248 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1250 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1252 Reverting patch:
1253 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1255 PR target/83712
1256 * lra-assigns.c (assign_by_spills): Return a flag of reload
1257 assignment failure. Do not process the reload assignment
1258 failures. Do not spill other reload pseudos if they has the same
1259 reg class.
1260 (lra_assign): Add a return arg. Set up from the result of
1261 assign_by_spills call.
1262 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1263 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1264 usage_insns if it is not NULL.
1265 (spill_hard_reg_in_range): New function.
1266 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1267 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1268 function prototypes.
1269 (lra_assign): Change prototype.
1270 * lra.c (lra): Add code to deal with fails by splitting hard reg
1271 live ranges.
1273 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1275 PR target/84807
1276 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1278 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1280 PR debug/84620
1281 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1282 (dw_val_node): Add val_symbolic_view.
1283 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1284 (symview_upper_bound): New.
1285 (new_line_info_table): Initialize symviews_since_reset.
1286 (dwarf2out_source_line): Count symviews_since_reset and set
1287 symview_upper_bound.
1288 (dw_val_equal_p): Handle symview.
1289 (add_AT_symview): New.
1290 (print_dw_val): Handle symview.
1291 (attr_checksum, attr_checksum_ordered): Likewise.
1292 (same_dw_val_p, size_of_die): Likewise.
1293 (value_format, output_die): Likewise.
1294 (add_high_low_attributes): Use add_AT_symview for entry_view.
1295 (dwarf2out_finish): Reset symview_upper_bound, clear
1296 zero_view_p.
1298 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1300 PR target/83969
1301 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1302 Add strict argument and use it.
1303 (rs6000_split_multireg_move): Update for new strict argument.
1304 (mem_operand_gpr): Disallow all non-offsettable addresses.
1305 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1307 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1309 PR target/84772
1310 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1311 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1312 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1314 PR c++/84767
1315 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1316 decl, use remap_type if we want to use the type.
1318 2018-03-09 Martin Sebor <msebor@redhat.com>
1320 PR tree-optimization/84526
1321 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1322 Remove dead code.
1323 (builtin_access::generic_overlap): Be prepared to handle non-array
1324 base objects.
1326 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1328 PR rtl-optimization/84682
1329 * lra-constraints.c (process_address_1): Check is_address flag
1330 for address constraints.
1331 (process_alt_operands): Likewise.
1332 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1333 preprocess_constraints.
1334 * recog.h (preprocess_constraints): Add oploc parameter.
1335 Adjust callers.
1336 * recog.c (preprocess_constraints): Test address_operand for
1337 CT_ADDRESS constraints.
1339 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1341 PR target/83712
1342 * lra-assigns.c (assign_by_spills): Return a flag of reload
1343 assignment failure. Do not process the reload assignment
1344 failures. Do not spill other reload pseudos if they has the same
1345 reg class.
1346 (lra_assign): Add a return arg. Set up from the result of
1347 assign_by_spills call.
1348 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1349 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1350 usage_insns if it is not NULL.
1351 (spill_hard_reg_in_range): New function.
1352 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1353 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1354 function prototypes.
1355 (lra_assign): Change prototype.
1356 * lra.c (lra): Add code to deal with fails by splitting hard reg
1357 live ranges.
1359 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1361 PR target/83193
1362 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1363 Accept complain bool parameter. Only emit errors if it is true.
1364 (arm_parse_cpu_option_name): Likewise.
1365 (arm_target_thumb_only): Adjust callers of the above.
1366 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1367 prototype to take a default true bool parameter.
1368 (arm_parse_arch_option_name): Likewise.
1370 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1371 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1373 PR jit/64089
1374 PR jit/84288
1375 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1376 * configure: Regenerate.
1377 * configure.ac ("linker --version-script option"): New.
1378 ("linker soname option"): New.
1380 2018-03-09 Richard Biener <rguenther@suse.de>
1382 PR tree-optimization/84775
1383 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1384 immediate uses of predicate stmts and mark them modified.
1386 Revert
1387 PR tree-optimization/84178
1388 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1389 to caller.
1390 (version_loop_for_if_conversion): Delay update_ssa call.
1391 (tree_if_conversion): Delay update_ssa until after predicate
1392 insertion.
1394 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1396 PR target/84763
1397 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1398 when the function accesses prior frames.
1400 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1402 PR debug/84456
1403 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1404 gen_llsym, otherwise call maybe_gen_llsym.
1406 PR inline-asm/84742
1407 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1408 has ',' character inside of it.
1410 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1412 PR target/84748
1413 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1414 as clobbering CC_REGNUM.
1416 2018-03-08 Richard Biener <rguenther@suse.de>
1418 PR middle-end/84552
1419 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1420 (follow_copies_to_constant): Do not follow SSA names registered
1421 for update.
1423 2018-03-08 Richard Biener <rguenther@suse.de>
1425 PR tree-optimization/84178
1426 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1427 to caller.
1428 (version_loop_for_if_conversion): Delay update_ssa call.
1429 (tree_if_conversion): Delay update_ssa until after predicate
1430 insertion.
1432 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1434 PR tree-optimization/84178
1435 * tree-if-conv.c (release_bb_predicate): Remove the
1436 the assertion that the stmts have NULL use_ops.
1437 Discard the statements, asserting that they haven't
1438 yet been added to a BB.
1440 2018-03-08 Richard Biener <rguenther@suse.de>
1442 PR tree-optimization/84746
1443 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1444 (phi_translate): Pass in destination ANTIC_OUT set.
1445 (phi_translate_1): Likewise. For a simplified result lookup
1446 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1447 (phi_translate_set): Adjust.
1448 (do_pre_regular_insertion): Likewise.
1449 (do_pre_partial_partial_insertion): Likewise.
1451 2018-03-08 Martin Liska <mliska@suse.cz>
1453 PR gcov-profile/84735
1454 * doc/gcov.texi: Document usage of profile files.
1455 * gcov-io.h: Document changes in the format.
1457 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1459 PR debug/84404
1460 PR debug/84408
1461 * dwarf2out.c (struct dw_line_info_table): Update comments for
1462 view == -1.
1463 (FORCE_RESET_NEXT_VIEW): New.
1464 (FORCE_RESETTING_VIEW_P): New.
1465 (RESETTING_VIEW_P): Check for -1 too.
1466 (ZERO_VIEW_P): Likewise.
1467 (new_line_info_table): Force-reset next view.
1468 (dwarf2out_begin_function): Likewise.
1469 (dwarf2out_source_line): Simplify zero_view_p initialization.
1470 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1471 view directly. Omit view when omitting .loc at line 0.
1473 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1475 PR tree-optimization/84740
1476 * tree-switch-conversion.c (process_switch): Call build_constructors
1477 only if info.phi_count is non-zero.
1479 PR tree-optimization/84739
1480 * tree-tailcall.c (find_tail_calls): Check call arguments against
1481 DECL_ARGUMENTS (current_function_decl) rather than
1482 DECL_ARGUMENTS (func) when checking for tail recursion.
1484 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1486 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1487 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1488 Volker Reichelt's entry and add entries for people that perform
1489 GCC fuzzy testing and report numerous bugs.
1491 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1493 PR target/82411
1494 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1495 readonly data in sdata, if that is disabled.
1496 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1497 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1498 -mreadonly-in-sdata option.
1500 2018-03-07 Martin Sebor <msebor@redhat.com>
1502 PR tree-optimization/84468
1503 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1504 basic block when looking for nul assignment.
1506 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1508 PR target/84277
1509 * except.h (output_function_exception_table): Adjust prototype.
1510 * except.c (output_function_exception_table): Remove FNNAME parameter
1511 and add SECTION parameter. Ouput one part of the table at a time.
1512 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1513 the first part of the exception table and emit unwind directives.
1514 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1515 (i386_pe_seh_cold_init): Likewise.
1516 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1517 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1518 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1519 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1520 * config/i386/winnt.c: Include except.h.
1521 (struct seh_frame_state): Add reg_offset, after_prologue and
1522 in_cold_section fields.
1523 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1524 (i386_pe_seh_cold_init): New function.
1525 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1526 to seh->in_cold_section.
1527 (seh_emit_push): Record the offset of the push.
1528 (seh_emit_save): Record the offet of the save.
1529 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1530 Test seh->after_prologue to disregard the epilogue.
1531 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1532 (i386_pe_end_cold_function): New function.
1534 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1536 PR fortran/84565
1537 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1538 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1540 PR c++/84704
1541 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1542 on tmp_var.
1543 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1544 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1546 PR middle-end/84723
1547 * multiple_target.c: Include tree-inline.h and intl.h.
1548 (expand_target_clones): Diagnose and fail if node->definition and
1549 !tree_versionable_function_p (node->decl).
1551 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1553 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1554 sprint_ul.
1555 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1556 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1557 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1559 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1561 PR target/84710
1562 * combine.c (try_combine): Use reg_or_subregno instead of handling
1563 just paradoxical SUBREGs and REGs.
1565 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1567 * config/arc/arc.c (arc_finalize_pic): Remove function.
1568 (arc_must_save_register): We use single base PIC register, remove
1569 checks to save/restore the PIC register.
1570 (arc_expand_prologue): Likewise.
1571 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1572 Remove.
1573 (arc_verify_short): Likewise.
1574 (arc_attr_type): Likewise.
1575 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1576 (walk_stores): Likewise.
1577 (arc_address_cost): Make it static.
1578 (arc_verify_short): Likewise.
1579 (branch_dest): Likewise.
1580 (arc_attr_type): Likewise.
1581 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1582 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1583 (arc_final_prescan_insn): Remove inserting the nops due to
1584 hardware hazards. It is done in reorg step.
1585 (insn_length_variant_t): Remove.
1586 (insn_length_parameters_t): Likewise.
1587 (arc_insn_length_parameters): Likewise.
1588 (arc_get_insn_variants): Likewise.
1589 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1591 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1593 PR inline-asm/84683
1594 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1595 assertion failure.
1597 PR tree-optimization/84687
1598 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1599 on new_node->decl.
1600 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1602 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1604 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1605 Rename to ppc_speculation_barrier.
1606 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1607 __builtin_ppc_speculation_barrier.
1609 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1611 PR target/84700
1612 * combine.c (combine_simplify_rtx): Don't try to simplify if
1613 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1614 are equal to x.
1616 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1618 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1619 to 32 bytes when compiling for POWER9.
1621 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1623 PR target/84564
1624 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1625 regparm >= 3 with no arg reg available also for calls with
1626 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1628 PR target/84524
1629 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1630 orig,vex.
1631 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1633 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1635 PR target/84264
1636 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1638 2018-03-05 Richard Biener <rguenther@suse.de>
1640 PR tree-optimization/84486
1641 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1642 When inserting a __builtin_assume_aligned call set the LHS
1643 SSA name alignment info accordingly.
1645 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1647 PR tree-optimization/84114
1648 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1649 Avoid reassociation of FLOAT_MODE addition.
1651 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1653 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1654 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1655 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1656 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1657 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1658 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1659 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1660 and -mwbnoinvd.
1661 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1662 __builtin_ia32_wbinvd): New builtins.
1663 (SPECIAL_ARGS2): New.
1664 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1665 (SPECIAL_ARGS2): New.
1666 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1667 (ix86_valid_target_attribute_inner_p): Ditto.
1668 (ix86_init_mmx_sse_builtins): Add special_args2.
1669 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1670 TARGET_WBNOINVD_P): New.
1671 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1672 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1673 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1674 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1675 * config/i386/pconfigintrin.h: New file.
1676 * config/i386/wbnoinvdintrin.h: Ditto.
1677 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1678 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1680 2018-03-05 Richard Biener <rguenther@suse.de>
1682 PR tree-optimization/84670
1683 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1684 member.
1685 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1686 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1687 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1688 doesn't grow if all successors (recursively) were visited at least
1689 once.
1691 2018-03-05 Richard Biener <rguenther@suse.de>
1693 PR tree-optimization/84650
1694 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1695 if executed in the loop pipeline.
1697 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1699 * doc/configfiles.texi (Configuration Files): Move info about
1700 conditionalizing $target-protos.h to...
1701 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1702 differs from $target-protos.h.
1704 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1705 Chung-Ju Wu <jasonwucj@gmail.com>
1707 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1708 * config/nds32/nds32-multiple.md (setmemsi): Define.
1709 * config/nds32/nds32-memory-manipulation.c
1710 (nds32_gen_dup_4_byte_to_word_value): New.
1711 (emit_setmem_word_loop): New.
1712 (emit_setmem_byte_loop): New.
1713 (nds32_expand_setmem_loop): New.
1714 (nds32_expand_setmem_loop_v3m): New.
1715 (nds32_expand_setmem_unroll): New.
1716 (nds32_expand_setmem): New.
1718 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1719 Chung-Ju Wu <jasonwucj@gmail.com>
1721 * config/nds32/nds32-memory-manipulation.c
1722 (nds32_emit_load_store): New.
1723 (nds32_emit_post_inc_load_store): New.
1724 (nds32_emit_mem_move): New.
1725 (nds32_emit_mem_move_block): New.
1726 (nds32_expand_movmemsi_loop_unknown_size): New.
1727 (nds32_expand_movmemsi_loop_known_size): New.
1728 (nds32_expand_movmemsi_loop): New.
1729 (nds32_expand_movmemsi_unroll): New.
1730 (nds32_expand_movmemqi): Rename ...
1731 (nds32_expand_movmemsi): ... to this.
1732 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1733 (movmemsi): ... to this.
1734 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1735 (nds32_expand_movmemsi): ... to this.
1737 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1738 Monk Chiang <sh.chiang04@gmail.com>
1739 Chung-Ju Wu <jasonwucj@gmail.com>
1741 * config/nds32/nds32-protos.h
1742 (nds32_expand_load_multiple): New arguments.
1743 (nds32_expand_store_multiple): Ditto.
1744 (nds32_valid_multiple_load_store): Rename ...
1745 (nds32_valid_multiple_load_store_p): ... to this.
1746 * config/nds32/nds32-memory-manipulation.c
1747 (nds32_expand_load_multiple): Refine implementation.
1748 (nds32_expand_store_multiple): Ditto.
1749 * config/nds32/nds32-multiple.md
1750 (load_multiple): Update nds32_expand_load_multiple interface.
1751 (store_multiple): Update nds32_expand_store_multiple interface.
1752 * config/nds32/nds32-predicates.c
1753 (nds32_valid_multiple_load_store): Rename ...
1754 (nds32_valid_multiple_load_store_p): ... to this and refine
1755 implementation.
1756 * config/nds32/predicates.md
1757 (nds32_load_multiple_and_update_address_operation): New predicate.
1758 (nds32_store_multiple_and_update_address_operation): New predicate.
1760 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1761 Chung-Ju Wu <jasonwucj@gmail.com>
1763 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1764 (combo): New attribute.
1765 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1767 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1769 * config/nds32/nds32.opt: Change -mcmodel= default value.
1771 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1772 Monk Chiang <sh.chiang04@gmail.com>
1773 Chung-Ju Wu <jasonwucj@gmail.com>
1775 * config/nds32/constants.md (unspec_element): New enum.
1776 * config/nds32/constraints.md (Umw): New constraint.
1777 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1778 * config/nds32/nds32-intrinsic.md: Likewise.
1779 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1780 (nds32_valid_smw_lwm_base_p): New.
1781 (nds32_output_smw_single_word): New.
1782 (nds32_output_lmw_single_word): New.
1783 (nds32_expand_unaligned_load): New.
1784 (nds32_expand_unaligned_store): New.
1785 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1786 (nds32_output_smw_single_word): Declare.
1787 (nds32_output_lmw_single_word): Declare.
1788 (nds32_expand_unaligned_load): Declare.
1789 (nds32_expand_unaligned_store): Declare.
1790 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1791 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1792 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1793 NDS32_BUILTIN_UASTORE_DW.
1794 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1795 predicate.
1797 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1798 Kito Cheng <kito.cheng@gmail.com>
1799 Chung-Ju Wu <jasonwucj@gmail.com>
1801 * config/nds32/nds32-intrinsic.c
1802 (nds32_expand_builtin_null_ftype_reg): Delete.
1803 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1804 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1805 (nds32_read_argument): New.
1806 (nds32_legitimize_target): Ditto.
1807 (nds32_legitimize_argument): Ditto.
1808 (nds32_check_constant_argument): Ditto.
1809 (nds32_expand_unop_builtin): Ditto.
1810 (nds32_expand_unopimm_builtin): Ditto.
1811 (nds32_expand_binop_builtin): Ditto.
1812 (nds32_builtin_decl_impl): Ditto.
1813 (builtin_description): Ditto.
1814 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1815 (nds32_init_builtins_impl): Ditto.
1816 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1817 (nds32_builtin_decl): New.
1818 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1819 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1821 2018-03-02 Jeff Law <law@redhat.com>
1823 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1824 (redundant_insn, fill_simple_delay_slots): Likewise.
1825 (fill_slots_from_thread): Likewise.
1826 * resource.c (mark_referenced_resources): Likewise.
1827 (mark_set_resources, find_dead_or_set_registers): Likewise.
1829 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1831 * substring-locations.h (format_warning_va): Formatting fix for
1832 ATTRIBUTE_GCC_DIAG.
1833 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1834 argument.
1835 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1836 * substring-locations.c: Include intl.h.
1837 (format_warning_va): Turned into small wrapper around
1838 format_warning_n_va, renamed to ...
1839 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1840 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1841 use ngettext.
1842 (format_warning_at_substring_n): New function.
1843 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1844 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1845 format_warning_at_substring with just a shorter name instead of
1846 const function pointer.
1847 (fmtwarn_n): New function.
1848 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1849 appropriate, get rid of all the fmtstr temporaries, move conditionals
1850 with G_() wrapped string literals directly into fmtwarn arguments,
1851 cast dir.len to (int), formatting fixes.
1853 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1855 * doc/invoke.texi: Remove "Cilk Plus" references.
1857 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1858 Richard Biener <rguenther@suse.de>
1860 PR ipa/84628
1861 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1862 for error or warning attributes if CALL_FROM_THUNK_P is set.
1863 Formatting fixes.
1865 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1867 PR target/56540
1868 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1869 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1871 PR target/56540
1872 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1873 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1875 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1876 instead of -1U in last predictors element's probability member.
1878 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1880 PR ipa/83983
1881 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1882 arguments if they are comparable.
1884 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1886 PR tree-optimization/84634
1887 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1888 masks and masked_loop_p with a single loop_masks, making sure it's
1889 null for bb vectorization.
1891 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1893 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1894 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1895 loop->force_vectorize to check whether there is no alias.
1897 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1899 PR target/84614
1900 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1901 prototypes.
1902 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1903 comments.
1904 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1905 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1906 instead of a loop around prev_real_insn.
1907 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1908 prev_real_insn.
1910 PR inline-asm/84625
1911 * config/i386/i386.c (ix86_print_operand): Use conditional
1912 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1913 zero vector.
1915 2018-03-02 Richard Biener <rguenther@suse.de>
1917 PR tree-optimization/84427
1918 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1919 (bitmap_set_subtract_values): Rewrite to handle multiple
1920 exprs per value.
1921 (clean): Likewise.
1922 (prune_clobbered_mems): Likewise.
1923 (phi_translate): Take edge instead of pred/phiblock.
1924 (phi_translate_1): Likewise.
1925 (phi_translate_set): Likewise. Insert all translated
1926 exprs for a value into the set, keeping possibly multiple
1927 expressions per value.
1928 (compute_antic_aux): Adjust for phi_translate changes.
1929 When intersecting union the expressions and prune those
1930 not in the final value set, keeping possibly multiple
1931 expressions per value. Do not use value-insertion
1932 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1933 all expressions. Add verification that the value-sets
1934 only shrink during iteration.
1935 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1936 (do_pre_regular_insertion): Likewise.
1937 (do_pre_partial_partial_insertion): Likewise.
1939 2018-03-02 Richard Biener <rguenther@suse.de>
1941 PR target/82005
1942 * config/darwin.c (saved_debug_info_level): New static global.
1943 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1944 (darwin_asm_lto_end): Restore debug info generation settings.
1946 2018-03-01 Martin Liska <mliska@suse.cz>
1948 PR sanitizer/82484
1949 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1950 volatile arguments.
1952 2018-03-01 Richard Biener <rguenther@suse.de>
1954 PR debug/84645
1955 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1956 type annotation with LTO when debug was disabled at compile-time.
1958 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1960 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1961 XINT with INTVAL.
1962 (mips_final_postscan_insn): Likewise.
1964 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1966 PR rtl-optimization/84528
1967 * alias.c (init_alias_target): Add commentary.
1968 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1969 a unique base value if the frame pointer is not eliminated
1970 to the stack pointer.
1972 2018-03-01 Tom de Vries <tom@codesourcery.com>
1974 PR rtl-optimization/83327
1975 * lra-int.h (hard_regs_spilled_into): Declare.
1976 * lra.c (hard_regs_spilled_into): Define.
1977 (init_reg_info): Init hard_regs_spilled_into.
1978 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1979 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1980 (process_bb_lives): Handle hard_regs_spilled_into.
1981 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1982 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1984 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1986 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1987 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1988 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1989 * config/rs6000/aix72.h: New file.
1991 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1993 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1994 instead of warning_at with conditional singular and plural messages
1995 where possible.
1997 PR target/52991
1998 * stor-layout.c (update_alignment_for_field): For
1999 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2000 && !DECL_PACKED (field), do the alignment update, just use
2001 only desired_align instead of MAX (type_align, desired_align)
2002 as the alignment.
2003 (place_field): Don't do known_align < desired_align handling
2004 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2005 is non-NULL, instead do it after rli->prev_field handling and
2006 only if not within a bitfield word. For DECL_PACKED (field)
2007 use type_align of BITS_PER_UNIT.
2009 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2011 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2012 superfluous parentheses and trailing spaces.
2014 2018-02-28 Richard Biener <rguenther@suse.de>
2016 PR tree-optimization/84584
2017 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2018 SCoPs with fake exit edge.
2020 2018-02-28 Martin Liska <mliska@suse.cz>
2022 PR testsuite/84597
2023 * timevar.c (timer::print): Fix format to properly print 100%
2024 values.
2026 2018-02-28 Richard Biener <rguenther@suse.de>
2028 PR middle-end/84607
2029 * genmatch.c (capture_info::walk_match): Do not mark
2030 captured expressions without operands as expr_p given
2031 they act more like predicates and should be subject to
2032 "lost tail" side-effect preserving.
2034 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2036 PR rtl-optimization/81611
2037 * auto-inc-dec.c (attempt_change): Move dead note from
2038 mem_insn if it's the next use of regno
2039 (find_address): Take address use of reg holding
2040 non-incremented value. Add parm to limit search to the named
2041 reg only.
2042 (merge_in_block): Attempt to use a mem insn that is the next
2043 use of the original regno.
2045 2018-02-27 Martin Sebor <msebor@redhat.com>
2047 PR c++/83871
2048 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2049 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2051 2018-02-27 Martin Sebor <msebor@redhat.com>
2053 PR translation/84207
2054 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2055 n argument to unsigned HOST_WIDE_INT.
2056 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2057 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2058 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2059 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2061 2018-02-27 Richard Biener <rguenther@suse.de>
2063 PR tree-optimization/84512
2064 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2065 Do not use the estimate returned from record_stmt_cost for
2066 the scalar iteration cost but sum properly using add_stmt_cost.
2068 2018-02-27 Richard Biener <rguenther@suse.de>
2070 PR tree-optimization/84466
2071 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2072 Adjust last change to less strictly validate use operands.
2074 2018-02-27 Martin Liska <mliska@suse.cz>
2076 PR gcov-profile/84548
2077 * gcov.c (process_file): Allow partial overlap and consider it
2078 also as group functions.
2079 (output_lines): Properly calculate range of lines for a group.
2081 2018-02-27 Martin Liska <mliska@suse.cz>
2083 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2084 'ggc' suffixes. Change first column width.
2085 (timer::print): Fix formatting of the column.
2087 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2089 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2090 preserve inline entry blocks for the sake of debug inline
2091 entry point markers alone.
2092 (remove_unused_locals): Suggest in comments a better place to
2093 force the preservation of inline entry blocks that are
2094 otherwise unused, but do not preserve them.
2096 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2098 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2100 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2102 PR target/84039
2103 * config/i386/constraints.md (Bs): Replace
2104 ix86_indirect_branch_register with
2105 TARGET_INDIRECT_BRANCH_REGISTER.
2106 (Bw): Likewise.
2107 * config/i386/i386.md (indirect_jump): Likewise.
2108 (tablejump): Likewise.
2109 (*sibcall_memory): Likewise.
2110 (*sibcall_value_memory): Likewise.
2111 Peepholes of indirect call and jump via memory: Likewise.
2112 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2113 (*sibcall_value_GOT_32): Likewise.
2114 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2115 (GOT_memory_operand): Likewise.
2116 (call_insn_operand): Likewise.
2117 (sibcall_insn_operand): Likewise.
2118 (GOT32_symbol_operand): Likewise.
2119 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2121 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2123 PR rtl-optimization/83496
2124 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2125 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2126 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2127 redundant insn, if any.
2128 (relax_delay_slots): Likewise.
2129 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2131 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2133 PR tree-optimization/83965
2134 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2135 that grouped statements are part of a reduction chain. Return
2136 true if the statement is not marked as a reduction itself but
2137 is part of a group.
2138 (vect_recog_dot_prod_pattern): Don't check whether the statement
2139 is part of a group here.
2140 (vect_recog_sad_pattern): Likewise.
2141 (vect_recog_widen_sum_pattern): Likewise.
2143 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2145 PR debug/84545
2146 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2148 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2150 PR target/84530
2151 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2152 the bool argument.
2153 (ix86_output_indirect_function_return): New prototype.
2154 (ix86_split_simple_return_pop_internal): Likewise.
2155 * config/i386/i386.c (indirect_return_via_cx): New.
2156 (indirect_return_via_cx_bnd): Likewise.
2157 (indirect_thunk_name): Handle return va CX_REG.
2158 (output_indirect_thunk_function): Create alias for
2159 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2160 (ix86_output_indirect_jmp): Remove the bool argument.
2161 (ix86_output_indirect_function_return): New function.
2162 (ix86_split_simple_return_pop_internal): Likewise.
2163 * config/i386/i386.md (*indirect_jump): Don't pass false
2164 to ix86_output_indirect_jmp.
2165 (*tablejump_1): Likewise.
2166 (simple_return_pop_internal): Change it to define_insn_and_split.
2167 Call ix86_split_simple_return_pop_internal to split it for
2168 -mfunction-return=.
2169 (simple_return_indirect_internal): Call
2170 ix86_output_indirect_function_return instead of
2171 ix86_output_indirect_jmp.
2173 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2175 PR bootstrap/84405
2176 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2177 memset and value initialization afterwards.
2179 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2181 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2183 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2185 PR target/84521
2186 * common/config/aarch64/aarch64-common.c
2187 (aarch_option_optimization_table[]): Switch
2188 off fomit-frame-pointer
2190 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2191 Chung-Ju Wu <jasonwucj@gmail.com>
2193 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2194 volatile memory.
2195 (store_multiple): Ditto.
2197 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2199 * config.gcc: Add --with-cpu support for nds32 target.
2200 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2201 * config/nds32/nds32.opt: Add -mcpu= option.
2203 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2205 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2206 isel=yes): Warn for these deprecated options.
2208 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2210 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2211 ISA_2_5_MASKS_EMBEDDED.
2213 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2215 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2216 p->max as pointers rather than using iterative_hash_expr.
2218 2018-02-23 Carl Love <cel@us.ibm.com>
2220 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2221 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2222 BU_P8V_OVERLOAD_2.
2223 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2224 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2225 P8V_BUILTIN_VEC_VUNSIGNED2.
2227 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2229 PR target/81572
2230 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2231 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2232 LRA_UNKNOWN_ALT.
2233 * lra-constraints.c (curr_insn_transform): Set up
2234 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2235 LRA_UNKNOWN_ALT.
2236 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2237 * lra-eliminations.c (spill_pseudos): Ditto.
2238 (process_insn_for_elimination): Ditto.
2239 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2240 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2241 LRA_NON_CLOBBERED_ALT.
2243 2018-02-22 Martin Sebor <msebor@redhat.com>
2245 PR tree-optimization/84480
2246 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2247 to maybe_diag_stxncpy_trunc. Call it.
2248 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2249 from gimple_fold_builtin_strcpy. Print inlining stack.
2250 (handle_builtin_stxncpy): Print inlining stack.
2251 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2253 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2255 PR target/84176
2256 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2257 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2258 and -fcheck-pointer-bounds are used together.
2259 (indirect_thunk_prefix): New enum.
2260 (indirect_thunk_need_prefix): New function.
2261 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2262 "_nt" instead of "_bnd" for NOTRACK prefix.
2263 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2264 (output_indirect_thunk_function): Likewise.
2265 (): Likewise.
2266 (ix86_code_end): Update output_indirect_thunk_function calls.
2267 (ix86_output_indirect_branch_via_reg): Replace
2268 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2269 (ix86_output_indirect_branch_via_push): Likewise.
2270 (ix86_output_function_return): Likewise.
2271 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2272 incompatible with -fcf-protection=branch and
2273 -fcheck-pointer-bounds.
2275 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2277 PR target/83335
2278 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2279 Change gcc_assert call to output_operand_lossage.
2281 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2283 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2285 2018-02-22 DJ Delorie <dj@redhat.com>
2286 Sebastian Perta <sebastian.perta@renesas.com>
2287 Oleg Endo <olegendo@gcc.gnu.org>
2289 * config/rx/rx.c (rx_rtx_costs): New function.
2290 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2292 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2294 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2296 2018-02-22 Martin Liska <mliska@suse.cz>
2298 PR driver/83193
2299 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2300 Add "native" as a possible value.
2302 2018-02-22 Martin Liska <mliska@suse.cz>
2304 PR driver/83193
2305 * config/i386/i386.c (ix86_option_override_internal):
2306 Add "native" as a possible value for -march and -mtune.
2308 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2310 PR target/84502
2311 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2312 to all type variants.
2314 PR tree-optimization/84503
2315 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2316 width as info->bitpos + info->bitsize - start.
2317 (merged_store_group::merge_overlapping): Simplify width computation.
2318 (check_no_overlap): New function.
2319 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2320 start + width and last_order of the group, fail if check_no_overlap
2321 fails.
2322 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2323 to group if check_no_overlap fails.
2325 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2327 * config/rs6000/altivec.md: Delete contraint arguments to
2328 define_expand, define_split, and define_peephole2, and in
2329 define_insn_and_split if always unused.
2330 * config/rs6000/darwin.md: Ditto.
2331 * config/rs6000/dfp.md: Ditto.
2332 * config/rs6000/rs6000.md: Ditto.
2333 * config/rs6000/sync.md: Ditto.
2334 * config/rs6000/vector.md: Ditto.
2335 * config/rs6000/vsx.md: Ditto.
2337 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2339 * config/rs6000/altivec.md: Write output control strings as braced
2340 blocks instead of double-quoted strings.
2341 * config/rs6000/darwin.md: Ditto.
2342 * config/rs6000/rs6000.md: Ditto.
2343 * config/rs6000/vector.md: Ditto.
2344 * config/rs6000/vsx.md: Ditto.
2346 2018-02-21 Jason Merrill <jason@redhat.com>
2348 PR c++/84314 - ICE with templates and fastcall attribute.
2349 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2351 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2353 * ipa-cp.c (determine_versionability): Fix comment typos.
2355 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2357 PR c/84229
2358 * ipa-cp.c (determine_versionability): Do not version functions caling
2359 va_arg_pack.
2361 2018-02-21 Martin Liska <mliska@suse.cz>
2363 PR driver/83193
2364 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2365 Add "native" as a possible value.
2366 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2367 the macro when native cpu detection is available.
2369 2018-02-21 Martin Liska <mliska@suse.cz>
2371 PR driver/83193
2372 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2373 Add "native" as a possible value.
2374 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2375 when native cpu detection is available.
2377 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2378 Martin Sebor <msebor@redhat.com>
2380 PR tree-optimization/84478
2381 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2382 false.
2383 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2384 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2385 support which is conservatively correct, for 2 only stay conservative
2386 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2387 argument to the 2 argument get_range_strlen, adjust 6 arg
2388 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2389 false.
2390 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2391 (gimple_fold_builtin_strlen): Pass true as last argument to
2392 get_range_strlen.
2394 2018-02-20 Martin Sebor <msebor@redhat.com>
2396 PR middle-end/84095
2397 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2398 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2399 (builtin_memref::builtin_memref): Factor out parts into
2400 set_base_and_offset and call it.
2402 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2404 PR middle-end/84406
2405 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2406 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2407 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2408 search at the associated MODE_INT.
2410 2018-02-20 Jeff Law <law@redhat.com>
2412 PR middle-end/82123
2413 PR tree-optimization/81592
2414 PR middle-end/79257
2415 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2416 for range data rather than using global data.
2417 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2418 range data rather than using global data.
2419 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2420 pass it to children as needed.
2421 (struct directive::fmtresult): Similarly.
2422 (struct directive::set_width): Similarly.
2423 (struct directive::set_precision): Similarly.
2424 (format_integer, format_directive, parse_directive): Similarly.
2425 (format_none): Accept unnamed vr_values parameter.
2426 (format_percent, format_floating, format_character): Similarly.
2427 (format_string, format_plain): Similarly.
2428 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2429 the EVRP range analyzer for range data rather than using global data.
2430 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2431 gimple-ssa-evrp-analyze.h
2432 (class sprintf_dom_walker): Add after_dom_children member function.
2433 Add evrp_range_analyzer member.
2434 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2435 range analyzer as needed.
2436 (sprintf_dom_walker::after_dom_children): New member function.
2437 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2438 if not optimizing.
2439 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2440 (evrp_range_analyzer::pop_to_marker): Likewise.
2442 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2444 PR tree-optimization/84419
2445 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2446 with the required type if its current type is compatible but
2447 different.
2449 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2451 PR middle-end/82004
2452 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2453 after vectorization.
2455 2018-02-20 Martin Liska <mliska@suse.cz>
2457 PR driver/83193
2458 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2459 possible values if we don't have a hint.
2461 2018-02-20 Martin Liska <mliska@suse.cz>
2463 PR c/84310
2464 PR target/79747
2465 * final.c (shorten_branches): Build align_tab array with one
2466 more element.
2467 * opts.c (finish_options): Add alignment option limit check.
2468 (MAX_CODE_ALIGN): Likewise.
2469 (MAX_CODE_ALIGN_VALUE): Likewise.
2470 * doc/invoke.texi: Document maximum allowed option value for
2471 all -falign-* options.
2473 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2475 PR target/84146
2476 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2477 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2478 * var-tracking.c (emit_note_insn_var_location): Remove all references
2479 to NOTE_INSN_CALL_ARG_LOCATION.
2480 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2481 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2482 Use copy_rtx_if_shared.
2483 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2484 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2485 (dwarf2out_var_location): Remove handling of
2486 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2487 on call_insn.
2488 * final.c (final_scan_insn): Remove all references to
2489 NOTE_INSN_CALL_ARG_LOCATION.
2490 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2491 before dumping final insns.
2492 * except.c (emit_note_eh_region_end): Remove all references to
2493 NOTE_INSN_CALL_ARG_LOCATION.
2494 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2495 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2496 * config/arc/arc.c (hwloop_optimize): Likewise.
2497 * config/arm/arm.c (create_fix_barrier): Likewise.
2498 * config/s390/s390.c (s390_chunkify_start): Likewise.
2499 * config/sh/sh.c (find_barrier): Likewise.
2500 * config/i386/i386.c (rest_of_insert_endbranch,
2501 ix86_seh_fixup_eh_fallthru): Likewise.
2502 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2503 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2504 * config/frv/frv.c (frv_function_prologue): Likewise.
2505 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2506 reg note.
2507 (note_outside_basic_block_p): Remove all references to
2508 NOTE_INSN_CALL_ARG_LOCATION.
2509 * gengtype.c (adjust_field_rtx_def): Likewise.
2510 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2511 Likewise.
2512 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2513 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2515 PR c++/84444
2516 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2517 is ADDR_EXPR.
2519 PR tree-optimization/84452
2520 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2521 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2522 is NULL.
2524 2018-02-19 Martin Liska <mliska@suse.cz>
2526 PR sanitizer/82183
2527 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2529 2018-02-19 Martin Liska <mliska@suse.cz>
2530 Richard Sandiford <richard.sandiford@linaro.org>
2532 PR tree-optimization/82491
2533 * gimple-fold.c (get_base_constructor): Make earlier bail out
2534 to prevent ubsan.
2536 2018-02-19 Carl Love <cel@us.ibm.com>
2538 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2539 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2540 BU_P8V_OVERLOAD_1.
2541 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2542 P8V_BUILTIN_VEC_NEG.
2544 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2546 * config/rl78/rl78.md (movdf): New define expand.
2548 2018-02-19 Martin Liska <mliska@suse.cz>
2550 PR other/80589
2551 * doc/invoke.texi: Fix typo.
2552 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2554 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2556 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2557 handle rs6000_single_float and rs6000_double_float specially for
2558 e500 family CPUs.
2560 2018-02-16 Jeff Law <law@redhat.com>
2562 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2563 (pop_regs): Use it.
2565 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2567 PR ipa/84425
2568 * ipa-inline.c (inline_small_functions): Fix a typo.
2570 2018-02-16 Nathan Sidwell <nathan@acm.org>
2572 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2574 2018-02-16 Carl Love <cel@us.ibm.com>
2576 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2577 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2578 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2579 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2580 expansion to P8V_BUILTIN_VEC_FLOAT2.
2582 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2584 PR rtl-optimization/70023
2585 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2586 src_regno into account.
2588 2018-02-16 Carl Love <cel@us.ibm.com>
2590 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2591 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2592 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2593 * config/rs6000/rs6000.c: Remove case statements for
2594 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2595 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2596 and P9V_BUILTIN_VEC_VINSERT4B.
2597 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2598 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2599 * config/rs6000/vsx.md:
2600 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2601 vec_insert4b.
2603 2018-02-16 Carl Love <cel@us.ibm.com>
2605 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2606 vec_insert4b.
2607 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2608 definitions.
2609 * config/rs6000/rs6000-c.c: Add the definitions for
2610 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2611 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2612 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2613 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2614 definition for insert4b and define insn *insert3b_internal.
2615 * doc/extend.texi: Add documentation for vec_extract4b.
2617 2018-02-16 Nathan Sidwell <nathan@acm.org>
2619 * doc/extend.texi (Backwards Compatibility): Mention friend
2620 injection. Note for-scope is deprecated.
2621 * doc/invoke.texi (-ffriend-injection): Deprecate.
2623 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2625 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2626 that moved to I2, also allow destinations that are a paradoxical
2627 subreg (instead of a normal reg).
2629 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2631 PR target/83831
2632 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2633 to QImode.
2635 2018-02-16 Richard Biener <rguenther@suse.de>
2637 PR tree-optimization/84037
2638 PR tree-optimization/84016
2639 PR target/82862
2640 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2641 Adjust vec_construct for the fact we need additional higher latency
2642 128bit inserts for AVX256 and AVX512 vector builds.
2643 (ix86_add_stmt_cost): Scale vector construction cost for
2644 elementwise loads.
2646 2018-02-16 Richard Biener <rguenther@suse.de>
2648 PR tree-optimization/84417
2649 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2650 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2651 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2653 2018-02-16 Martin Liska <mliska@suse.cz>
2655 PR sanitizer/84307
2656 * internal-fn.def (ASAN_CHECK): Set proper flags.
2657 (ASAN_MARK): Likewise.
2659 2018-02-16 Julia Koval <julia.koval@intel.com>
2661 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2662 from PTA_CANNONLAKE.
2664 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2666 PR target/84272
2667 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2668 Use ++iter rather than iter++ for std::list iterators.
2669 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2670 defer deleting them until all nodes in the forest are processed. Do
2671 free even leaf nodes. Change to_process into auto_vec.
2673 PR bootstrap/84405
2674 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2675 * vec.h (vec_default_construct): Use memset instead of placement new
2676 if BROKEN_VALUE_INITIALIZATION is defined.
2677 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2678 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2679 is defined.
2681 PR rtl-optimization/83723
2682 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2683 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2684 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2685 recursive calls.
2686 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2687 callers.
2688 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2690 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2692 PR rtl-optimization/81443
2693 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2694 from inner REGs to paradoxical SUBREGs.
2696 2018-02-16 Richard Biener <rguenther@suse.de>
2698 PR tree-optimization/84399
2699 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2700 For operands we can analyze at their definition make sure we can
2701 analyze them at each use as well.
2703 2018-02-16 Richard Biener <rguenther@suse.de>
2705 PR tree-optimization/84190
2706 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2707 volatile accesses if the decl isn't volatile.
2709 2018-02-15 Jason Merrill <jason@redhat.com>
2711 PR c++/84314 - ICE with templates and fastcall attribute.
2712 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2713 TYPE_CANONICAL on an existing type.
2715 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2717 PR tree-optimization/84383
2718 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2719 dstoff nor call operand_equal_p if dstbase is NULL.
2721 PR tree-optimization/84334
2722 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2723 also a CONSTANT_CLASS_P, punt.
2725 2018-02-14 Jim Wilson <jimw@sifive.com>
2727 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2728 first SMALL_OPERAND check. New local min_second_step. Move assert
2729 to where locals are set. Add TARGET_RVC support.
2730 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2732 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2734 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2736 2018-02-14 Martin Sebor <msebor@redhat.com>
2738 PR tree-optimization/83698
2739 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2740 arrays constrain the offset range to their bounds.
2741 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2742 (builtin_access::overlap): Avoid setting the size of overlap if it's
2743 already been set.
2744 (maybe_diag_overlap): Also consider arrays when deciding what values
2745 of offsets to include in diagnostics.
2747 2018-02-14 Martin Sebor <msebor@redhat.com>
2749 PR c/84108
2750 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2751 that correspond to the kind of a declaration.
2753 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2755 PR target/83984
2756 * config/pa/pa.md: Load address of PIC label using the linkage table
2757 if the label is nonlocal.
2759 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2761 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2762 warning message if user requests -maltivec=be.
2763 * doc/invoke.texi: Document deprecation of -maltivec=be.
2765 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2767 PR target/84220
2768 * config/rs6000/rs6000-c.c: Update definitions for
2769 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2770 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2772 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2774 PR target/84239
2775 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2776 add _get_ssp intrinsics. Remove argument from
2777 __builtin_ia32_rdssp[d|q].
2778 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2779 * config/i386/i386-builtin.def: Remove argument from
2780 __builtin_ia32_rdssp[d|q].
2781 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2782 ix86_expand_special_args_builtin for _rdssp[d|q].
2783 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2784 Clear register before usage.
2785 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2786 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2788 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2790 PR tree-optimization/84357
2791 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2792 operand 1 of an ARRAY_REF too.
2794 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2796 PR target/83831
2797 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2798 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2799 declarations.
2800 (set_of_reg): New struct.
2801 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2802 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2803 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2804 functions.
2805 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2806 Split into bitclr, bitset, bitinvert patterns if appropriate.
2807 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2808 use rx_fuse_in_memory_bitop.
2809 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2810 to named insn, correct maximum insn length.
2812 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2814 PR target/79242
2815 * machmode.def: Define a complex mode for PARTIAL_INT.
2816 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2817 MODE_PARTIAL_INT.
2818 * doc/rtl.texi: Document CSPImode.
2819 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2820 handling.
2821 (msp430_hard_regno_nregs_with_padding): Likewise.
2823 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2825 PR target/84279
2826 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2828 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2830 PR rtl-optimization/84169
2831 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2832 we generated a parallel as new i3 and we split that to new i2 and i3
2833 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2834 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2835 those to i2, not i1. Partially rewrite this scan code.
2837 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2839 PR c/82210
2840 * stor-layout.c (place_field): For variable length fields, adjust
2841 offset_align afterwards not just based on the field's alignment,
2842 but also on the size.
2844 PR middle-end/84309
2845 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2846 of exps and logs in the use_exp2 case.
2848 2018-02-13 Jeff Law <law@redhat.com>
2850 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2851 entry for "vector".
2853 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2854 ARGS as unused.
2856 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2858 PR debug/84342
2859 PR debug/84319
2860 * common.opt (gas-loc-support, gas-locview-support): New.
2861 (ginline-points, ginternal-reset-location-views): New.
2862 * doc/invoke.texi: Document them. Use @itemx where intended.
2863 (gvariable-location-views): Adjust.
2864 * target.def (reset_location_view): New.
2865 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2866 (TARGET_RESET_LOCATION_VIEW): New.
2867 * doc/tm.texi: Rebuilt.
2868 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2869 (dwarf2out_default_as_locview_support): New.
2870 (output_asm_line_debug_info): Use option variables.
2871 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2872 (output_loc_list): Likewise.
2873 (add_high_low_attributes): Check option variables.
2874 Don't output entry view attribute in strict mode.
2875 (gen_inlined_subroutine_die): Check option variables.
2876 (dwarf2out_inline_entry): Likewise.
2877 (init_sections_and_labels): Likewise.
2878 (dwarf2out_early_finish): Likewise.
2879 (maybe_reset_location_view): New, from...
2880 (dwarf2out_var_location): ... here. Call it.
2881 * debug.h (dwarf2out_default_as_loc_support): Declare.
2882 (dwarf2out_default_as_locview_support): Declare.
2883 * hooks.c (hook_int_rtx_insn_0): New.
2884 * hooks.h (hook_int_rtx_insn_0): Declare.
2885 * toplev.c (process_options): Take -gas-loc-support and
2886 -gas-locview-support from dwarf2out. Enable
2887 -gvariable-location-views by default only with locview
2888 assembler support. Enable -ginternal-reset-location-views by
2889 default only if the target defines the corresponding hook.
2890 Enable -ginline-points by default if location views are
2891 enabled; force it disabled if statement frontiers are
2892 disabled.
2893 * tree-inline.c (expand_call_inline): Check option variables.
2894 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2896 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2898 PR tree-optimization/84321
2899 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2900 handling. Also check whether the anti-range contains any values
2901 that satisfy the mask; switch to a VR_RANGE if not.
2903 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2905 PR sanitizer/84340
2906 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2908 2018-02-13 Martin Jambor <mjambor@suse.cz>
2910 PR c++/83990
2911 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2912 of call statements, also set location of a load to a temporary.
2914 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2916 * config/rl78/rl78.c (add_vector_labels): New function.
2917 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2918 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2919 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2920 which checks that no arguments are passed.
2921 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2922 * doc/extend.texi: Documentation for the new attribute.
2924 2018-02-13 Andreas Schwab <schwab@suse.de>
2926 * config/riscv/linux.h (CPP_SPEC): Define.
2928 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2930 PR target/84335
2931 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2932 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2933 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2934 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2935 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2936 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2937 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2938 temporarily for AES and PCLMUL builtins.
2940 PR tree-optimization/84339
2941 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2942 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2943 Formatting fixes.
2945 PR middle-end/84309
2946 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2947 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2948 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2949 inline function.
2950 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2951 inline function.
2952 * omp-simd-clone.h: New file.
2953 * omp-simd-clone.c: Include omp-simd-clone.h.
2954 (expand_simd_clones): No longer static.
2955 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2956 cgraph.h and omp-simd-clone.h.
2957 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2958 (vect_recog_widen_shift_pattern): Formatting fix.
2959 (vect_pattern_recog_1): Don't check optab for calls.
2961 PR target/84336
2962 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2963 operands[2] into a REG before using gen_lowpart on it.
2965 2018-02-12 Jeff Law <law@redhat.com>
2967 PR target/83760
2968 * config/sh/sh.c (find_barrier): Consider a sibling call
2969 a barrier as well.
2971 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2972 successfully back substituting a reg.
2974 2018-02-12 Richard Biener <rguenther@suse.de>
2976 PR tree-optimization/84037
2977 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2978 parameter, move visited init to caller.
2979 (vect_slp_analyze_operations): Separate cost from validity
2980 check, initialize visited once for all instances.
2981 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2982 for all instances.
2983 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2984 out an assert.
2985 (vect_model_promotion_demotion_cost): Likewise.
2986 (vectorizable_bswap): Guard cost modeling with !slp_node
2987 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2988 SLP stmts.
2989 (vectorizable_call): Likewise.
2990 (vectorizable_conversion): Likewise.
2991 (vectorizable_assignment): Likewise.
2992 (vectorizable_shift): Likewise.
2993 (vectorizable_operation): Likewise.
2994 (vectorizable_store): Likewise.
2995 (vectorizable_load): Likewise.
2996 (vectorizable_condition): Likewise.
2997 (vectorizable_comparison): Likewise.
2999 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3001 PR sanitizer/84307
3002 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3003 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3004 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3006 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3008 PR middle-end/83665
3009 * params.def (inline-min-speedup): Increase from 8 to 15.
3010 (max-inline-insns-auto): Decrease from 40 to 30.
3011 * ipa-split.c (consider_split): Add some buffer for function to
3012 be considered inlining candidate.
3013 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3014 default values.
3016 2018-02-12 Richard Biener <rguenther@suse.de>
3018 PR tree-optimization/84037
3019 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3020 matched stmts if we cannot swap the non-matched ones.
3022 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3024 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3025 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3026 _mm_maskz_scalef_round_ss): New intrinsics.
3027 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3028 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3029 __builtin_ia32_scalefss_round): Remove.
3030 (__builtin_ia32_scalefsd_mask_round,
3031 __builtin_ia32_scalefss_mask_round): New intrinsics.
3032 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3033 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3034 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3035 "<round_constraint>")): Changed to ...
3036 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3037 "<round_scalar_constraint>")): ... this.
3038 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3039 %0, %1, %2<round_op3>}"): Changed to ...
3040 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3041 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3042 %2<round_scalar_mask_op3>}"): ... this.
3043 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3045 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3047 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3048 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3049 (_mm_maskz_sqrt_round_ss): New intrinsics.
3050 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3051 (__builtin_ia32_sqrtsd_mask_round)
3052 (__builtin_ia32_sqrtss_mask_round): New builtins.
3053 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3054 (__builtin_ia32_sqrtss_round): Remove.
3055 (__builtin_ia32_sqrtsd_mask_round)
3056 (__builtin_ia32_sqrtss_mask_round): New builtins.
3057 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3058 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3059 ((match_operand:VF_128 1 "vector_operand"
3060 "xBm,<round_constraint>")): Changed to ...
3061 ((match_operand:VF_128 1 "vector_operand"
3062 "xBm,<round_scalar_constraint>")): ... this.
3063 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3064 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3065 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3066 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3067 %<iptr>1<round_scalar_mask_op3>}): ... this.
3068 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3069 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3071 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3073 PR target/84266
3074 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3075 Cast vec_cmpeq result to correct type.
3076 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3077 Cast vec_cmpgt result to correct type.
3079 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3081 * final.c (final_scan_insn_1): Renamed from...
3082 (final_scan_insn): ... this. New wrapper, to recover
3083 seen from the outermost call in recursive ones.
3084 * config/sparc/sparc.c (output_return): Drop seen from call.
3085 (output_sibcall): Likewise.
3086 * config/visium/visium.c (output_branch): Likewise.
3088 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3090 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3091 function label.
3093 2018-02-10 Alan Modra <amodra@gmail.com>
3095 PR target/84300
3096 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3097 Specify LR as an input.
3099 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3101 PR sanitizer/83987
3102 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3103 remove_member_access_dummy_vars): New functions.
3104 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3105 lower_omp_1, execute_lower_omp): Use them.
3107 PR rtl-optimization/84308
3108 * shrink-wrap.c (spread_components): Release todo vector.
3110 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3112 PR rtl-optimization/57193
3113 * ira-color.c (struct allocno_color_data): Add member
3114 conflict_allocno_hard_prefs.
3115 (update_conflict_allocno_hard_prefs): New.
3116 (bucket_allocno_compare_func): Add a preference based on
3117 conflict_allocno_hard_prefs.
3118 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3119 (color_allocnos): Remove a dead code. Initiate
3120 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3122 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3124 PR target/84226
3125 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3126 constraint from =wa to wa. Avoid a subreg on the output operand,
3127 instead use a pseudo and subreg it in a move.
3128 (p9_xxbrd_<mode>): Changed to ...
3129 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3130 (p9_xxbrd_v2df): New expander.
3131 (p9_xxbrw_<mode>): Changed to ...
3132 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3133 (p9_xxbrw_v4sf): New expander.
3135 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3137 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3139 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3141 PR target/83926
3142 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3143 multiply in 32-bit mode.
3144 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3145 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3146 mode.
3148 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3150 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3151 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3152 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3153 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3155 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3157 PR lto/84213
3158 * dwarf2out.c (is_trivial_indirect_ref): New function.
3159 (dwarf2out_late_global_decl): Do not generate a location
3160 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3161 and that are not defined in the current unit.
3163 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3165 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3166 instead of a libcall for UNORDERED.
3168 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3170 PR target/82641
3171 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3172 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3174 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3176 PR target/PR84295
3177 * config/s390/s390.c (s390_set_current_function): Invoke
3178 s390_indirect_branch_settings also if fndecl didn't change.
3180 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3182 * config/rs6000/rs6000.md (blockage): Set length to zero.
3184 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3186 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3188 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3190 PR sanitizer/84285
3191 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3192 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3193 -static-lib*san.
3195 PR debug/84252
3196 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3197 PARALLEL incoming that failed vt_get_decl_and_offset check.
3199 PR middle-end/84237
3200 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3201 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3202 TREE_READONLY bit.
3203 (get_variable_section): For decls in named .bss* sections pass true as
3204 second argument to bss_initializer_p.
3206 2018-02-09 Marek Polacek <polacek@redhat.com>
3207 Jakub Jelinek <jakub@redhat.com>
3209 PR c++/83659
3210 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3211 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3212 Sync some changes from cxx_fold_indirect_ref.
3214 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3216 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3217 markers.
3218 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3219 (BLOCK_INLINE_ENTRY_LABEL): New.
3220 (dwarf2out_var_location): Disregard inline entry markers.
3221 (inline_entry_data): New struct.
3222 (inline_entry_data_hasher): New hashtable type.
3223 (inline_entry_data_hasher::hash): New.
3224 (inline_entry_data_hasher::equal): New.
3225 (inline_entry_data_table): New variable.
3226 (add_high_low_attributes): Add DW_AT_entry_pc and
3227 DW_AT_GNU_entry_view attributes if a pending entry is found
3228 in inline_entry_data_table. Add old entry_pc attribute only
3229 if debug nonbinding markers are disabled.
3230 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3231 markers are enabled.
3232 (block_within_block_p, dwarf2out_inline_entry): New.
3233 (dwarf2out_finish): Check that no entries remained in
3234 inline_entry_data_table.
3235 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3236 (final_scan_insn, notice_source_line): Likewise.
3237 (rest_of_clean_state): Skip inline entry markers.
3238 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3239 markers.
3240 * gimple.c (gimple_build_debug_inline_entry): New.
3241 * gimple.h (enum gimple_debug_subcode): Add
3242 GIMPLE_DEBUG_INLINE_ENTRY.
3243 (gimple_build_debug_inline_entry): Declare.
3244 (gimple_debug_inline_entry_p): New.
3245 (gimple_debug_nonbind_marker_p): Adjust.
3246 * insn-notes.def (INLINE_ENTRY): New.
3247 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3248 inline entry marker notes.
3249 (print_insn): Likewise.
3250 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3251 (INSN_DEBUG_MARKER_KIND): Likewise.
3252 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3253 * tree-inline.c (expand_call_inline): Build and insert
3254 debug_inline_entry stmt.
3255 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3256 inline entry blocks early, if nonbind markers are enabled.
3257 (dump_scope_block): Dump fragment info.
3258 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3259 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3260 (gimple_build_debug_inline_entry): New.
3261 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3262 Enable/disable inline entry points too.
3263 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3264 (DEBUG_INSN): Describe inline entry markers.
3266 * common.opt (gvariable-location-views): New.
3267 (gvariable-location-views=incompat5): New.
3268 * config.in: Rebuilt.
3269 * configure: Rebuilt.
3270 * configure.ac: Test assembler for view support.
3271 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3272 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3273 * dwarf2out.c (var_loc_view): New typedef.
3274 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3275 (dwarf2out_locviews_in_attribute): New.
3276 (dwarf2out_locviews_in_loclist): New.
3277 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3278 (enum dw_line_info_opcode): Add LI_adv_address.
3279 (struct dw_line_info_table): Add view.
3280 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3281 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3282 (zero_view_p): New variable.
3283 (ZERO_VIEW_P): New macro.
3284 (output_asm_line_debug_info): New.
3285 (struct var_loc_node): Add view.
3286 (add_AT_view_list, AT_loc_list): New.
3287 (add_var_loc_to_decl): Add view param. Test it against last.
3288 (new_loc_list): Add view params. Record them.
3289 (AT_loc_list_ptr): Handle loc and view lists.
3290 (view_list_to_loc_list_val_node): New.
3291 (print_dw_val): Handle dw_val_class_view_list.
3292 (size_of_die): Likewise.
3293 (value_format): Likewise.
3294 (loc_list_has_views): New.
3295 (gen_llsym): Set vl_symbol too.
3296 (maybe_gen_llsym, skip_loc_list_entry): New.
3297 (dwarf2out_maybe_output_loclist_view_pair): New.
3298 (output_loc_list): Output view list or entries too.
3299 (output_view_list_offset): New.
3300 (output_die): Handle dw_val_class_view_list.
3301 (output_dwarf_version): New.
3302 (output_compilation_unit_header): Use it.
3303 (output_skeleton_debug_sections): Likewise.
3304 (output_rnglists, output_line_info): Likewise.
3305 (output_pubnames, output_aranges): Update version comments.
3306 (output_one_line_info_table): Output view numbers in asm comments.
3307 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3308 Call maybe_gen_llsym.
3309 (loc_list_from_tree_1): Adjust.
3310 (add_AT_location_description): Create view list attribute if
3311 needed, check it's absent otherwise.
3312 (convert_cfa_to_fb_loc_list): Adjust.
3313 (maybe_emit_file): Call output_asm_line_debug_info for test.
3314 (dwarf2out_var_location): Reset views as needed. Precompute
3315 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3316 attribute. Set view.
3317 (new_line_info_table): Reset next view.
3318 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3319 (dwarf2out_source_line): Likewise. Output view resets and labels to
3320 the assembler, or select appropriate line info opcodes.
3321 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3322 (optimize_string_length): Catch it. Adjust.
3323 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3324 dw_val_class_view_list, and remove it if no longer needed.
3325 (hash_loc_list): Hash view numbers.
3326 (loc_list_hasher::equal): Compare them.
3327 (optimize_location_lists): Check whether a view list symbol is
3328 needed, and whether the locview attribute is present, and
3329 whether they match. Remove the locview attribute if no longer
3330 needed.
3331 (index_location_lists): Call skip_loc_list_entry for test.
3332 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3333 Use output_dwarf_version.
3334 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3335 (struct dw_val_node): Add val_view_list.
3336 * final.c (SEEN_NEXT_VIEW): New.
3337 (set_next_view_needed): New.
3338 (clear_next_view_needed): New.
3339 (maybe_output_next_view): New.
3340 (final_start_function): Rename to...
3341 (final_start_function_1): ... this. Take pointer to FIRST,
3342 add SEEN parameter. Emit param bindings in the initial view.
3343 (final_start_function): Reintroduce SEEN-less interface.
3344 (final): Rename to...
3345 (final_1): ... this. Take SEEN parameter. Output final pending
3346 next view at the end.
3347 (final): Reintroduce seen-less interface.
3348 (final_scan_insn): Output pending next view before switching
3349 sections or ending a block. Mark the next view as needed when
3350 outputting variable locations. Notify debug backend of section
3351 changes, and of location view changes.
3352 (rest_of_handle_final): Adjust.
3353 * toplev.c (process_options): Autodetect value for debug variable
3354 location views option. Warn on incompat5 without -gdwarf-5.
3355 * doc/invoke.texi (gvariable-location-views): New.
3356 (gvariable-location-views=incompat5): New.
3357 (gno-variable-location-views): New.
3359 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3361 PR tree-optimization/84136
3362 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3363 that the result of find_edge is non-NULL.
3365 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3367 PR target/83008
3368 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3369 storing integer register in SImode. Fix cost of 256 and 512
3370 byte aligned SSE register store.
3372 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3374 * config/i386/i386.c (ix86_multiplication_cost): Fix
3375 multiplication cost for TARGET_AVX512DQ.
3377 2018-02-08 Marek Polacek <polacek@redhat.com>
3379 PR tree-optimization/84238
3380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3381 get_range_strlen.
3383 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3385 PR tree-optimization/84265
3386 * tree-vect-stmts.c (vectorizable_store): Don't treat
3387 VMAT_CONTIGUOUS accesses as grouped.
3388 (vectorizable_load): Likewise.
3390 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3392 PR tree-optimization/81635
3393 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3394 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3395 (test_round_for_mask): New functions.
3396 (wide_int_cc_tests): Call test_round_for_mask.
3397 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3398 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3399 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3400 range returned by get_range_info.
3402 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3404 PR ipa/81360
3405 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3406 * symtab.c: Include builtins.h
3407 (symtab_node::output_to_lto_symbol_table_p): Move here
3408 from lto-streamer-out.c:output_symbol_p.
3409 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3410 (output_symbol_p): Move all logic to symtab.c
3411 (produce_symtab): Update.
3413 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3415 * config/s390/s390-opts.h (enum indirect_branch): Define.
3416 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3417 (s390_indirect_branch_via_thunk)
3418 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3419 (enum s390_indirect_branch_type): Define.
3420 * config/s390/s390.c (struct s390_frame_layout, struct
3421 machine_function): Remove.
3422 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3423 (indirect_branch_table_label_no, indirect_branch_table_name):
3424 Define variables.
3425 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3426 (enum s390_indirect_branch_option): Define.
3427 (s390_return_addr_from_memory): New function.
3428 (s390_handle_string_attribute): New function.
3429 (s390_attribute_table): Add new attribute handler.
3430 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3431 (s390_indirect_branch_via_thunk): New function.
3432 (s390_indirect_branch_via_inline_thunk): New function.
3433 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3434 sibling call optimization for non z10 compiles.
3435 (s390_emit_call): Force indirect branch target to be a single
3436 register. Add r1 clobber for non-z10 compiles.
3437 (s390_emit_epilogue): Emit return jump via return_use expander.
3438 (s390_reorg): Handle JUMP_INSNs as execute targets.
3439 (s390_option_override_internal): Perform validity checks for the
3440 new command line options.
3441 (s390_indirect_branch_attrvalue): New function.
3442 (s390_indirect_branch_settings): New function.
3443 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3444 (s390_output_indirect_thunk_function): New function.
3445 (s390_code_end): Implement target hook.
3446 (s390_case_values_threshold): Implement target hook.
3447 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3448 macros.
3449 * config/s390/s390.h (struct s390_frame_layout)
3450 (struct machine_function): Move here from s390.c.
3451 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3452 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3453 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3454 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3455 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3456 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3457 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3458 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3459 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3460 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3461 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3462 (mnemonic attribute): Add values which aren't recognized
3463 automatically.
3464 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3465 pattern for branch conversion. Fix mnemonic attribute.
3466 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3467 indirect branch via thunk if requested.
3468 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3469 ("*indirect_jump"): Disable for branch conversion using out of
3470 line thunks.
3471 ("indirect_jump_via_thunk<mode>_z10")
3472 ("indirect_jump_via_thunk<mode>")
3473 ("indirect_jump_via_inlinethunk<mode>_z10")
3474 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3475 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3476 ("casesi_jump_via_inlinethunk<mode>_z10")
3477 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3478 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3479 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3480 ("*indirect2_jump"): Disable for branch conversion.
3481 ("casesi_jump"): Turn into expander and expand patterns for branch
3482 conversion.
3483 ("return_use"): New expander.
3484 ("*return"): Emit return via thunk and rename it to ...
3485 ("*return<mode>"): ... this one.
3486 * config/s390/s390.opt: Add new options and and enum for the
3487 option values.
3489 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3491 * lra-constraints.c (match_reload): Unconditionally use
3492 gen_lowpart_SUBREG, rather than selecting between that
3493 and equivalent gen_rtx_SUBREG code.
3495 2018-02-08 Richard Biener <rguenther@suse.de>
3497 PR tree-optimization/84233
3498 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3499 changed flag instead of boguously re-using phi_inserted.
3501 2018-02-08 Martin Jambor <mjambor@suse.cz>
3503 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3504 static local variables.
3506 2018-02-08 Richard Biener <rguenther@suse.de>
3508 PR tree-optimization/84278
3509 * tree-vect-stmts.c (vectorizable_store): When looking for
3510 smaller vector types to perform grouped strided loads/stores
3511 make sure the mode is supported by the target.
3512 (vectorizable_load): Likewise.
3514 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3516 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3517 Increase LDP/STP opportunities by adding adjacent callee-saves.
3519 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3521 PR rtl-optimization/84068
3522 PR rtl-optimization/83459
3523 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3525 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3527 PR tree-optimization/84224
3528 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3529 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3530 non-zero arguments.
3532 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3534 PR target/84113
3535 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3536 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3537 count, remove one USE.
3539 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3541 * doc/install.texi (Configuration): Document the
3542 --with-long-double-format={ibm,ieee} PowerPC configuration
3543 options.
3545 PR target/84154
3546 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3547 Convert from define_expand to be define_insn_and_split. Rework
3548 float/double/_Float128 conversions to QI/HI/SImode to work with
3549 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3550 conversions to QI/HImode types did a store and then a load to
3551 truncate the value. For conversions to VSX registers, don't split
3552 the insn, instead emit the code directly. Use the code iterator
3553 any_fix to combine signed and unsigned conversions.
3554 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3555 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3556 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3557 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3558 (fix_<mode>di2_hw): Likewise.
3559 (fixuns_<mode>di2_hw): Likewise.
3560 (fix_<mode>si2_hw): Likewise.
3561 (fixuns_<mode>si2_hw): Likewise.
3562 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3563 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3564 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3565 fix<uns>_trunc<SFDF:mode>si2_p8.
3566 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3567 used.
3568 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3569 (fix<uns>_<mode>_mem): Likewise.
3570 (fctiw<u>z_<mode>_mem): Likewise.
3571 (fix<uns>_<mode>_mem): Likewise.
3572 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3573 the register allocator from doing a direct move to the GPRs to do
3574 a store, and instead use the ISA 3.0 store byte/half-word from
3575 vector register instruction. For IEEE 128-bit floating point,
3576 also optimize stores of 32-bit ints.
3577 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3579 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3581 * genextract.c (push_pathstr_operand): New function to support
3582 [a-zA-Z].
3583 (walk_rtx): Call push_pathstr_operand.
3584 (print_path): Support [a-zA-Z].
3586 2018-02-07 Richard Biener <rguenther@suse.de>
3588 PR tree-optimization/84037
3589 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3590 (cse_and_gimplify_to_preheader): Declare.
3591 (vect_get_place_in_interleaving_chain): Likewise.
3592 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3593 ivexpr_map.
3594 (_loop_vec_info::~_loop_vec_info): Delete it.
3595 (cse_and_gimplify_to_preheader): New function.
3596 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3597 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3598 (vectorizable_load): Likewise. For grouped stores always base
3599 the IV on the first element.
3600 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3601 condition before gimplifying.
3603 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3605 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3606 *DIV_EXPR and *MOD_EXPR.
3608 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3610 PR target/84248
3611 * config/i386/i386.c (ix86_option_override_internal): Mask out
3612 the CF_SET bit when checking -fcf-protection.
3614 2018-02-07 Tom de Vries <tom@codesourcery.com>
3616 PR libgomp/84217
3617 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3618 enough.
3620 2018-02-07 Richard Biener <rguenther@suse.de>
3622 PR tree-optimization/84204
3623 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3624 this place.
3626 PR tree-optimization/84205
3627 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3628 special-case isl_ast_op_zdiv_r.
3630 PR tree-optimization/84223
3631 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3632 Only add conditions from within the region.
3633 (gather_bbs::after_dom_children): Adjust.
3635 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3637 PR target/84209
3638 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3639 * config/avr/avr.md: Only post-reload split REG-REG moves if
3640 either register is GENERAL_REG_P.
3642 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3644 PR tree-optimization/84235
3645 * tree-ssa-scopedtables.c
3646 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3647 if the subtraction is performed in floating point type where NaNs are
3648 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3649 build 1. Formatting fix.
3651 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3653 PR target/84146
3654 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3655 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3656 and skip it regardless of bb boundaries. Use CALL_P macro,
3657 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3658 unnecessarily, formatting fix.
3660 2018-02-06 Michael Collison <michael.collison@arm.com>
3662 * config/arm/thumb2.md:
3663 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3664 (*thumb_mov_notscc): Ditto.
3666 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3668 PR target/84154
3669 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3670 unsigned_fix, not "s".
3672 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3674 * configure.ac (gcc_fn_eh_frame_ro): New function.
3675 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3676 correct .eh_frame permissions.
3677 * configure: Regenerate.
3679 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3681 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3682 irrelevant options.
3684 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3686 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3687 Display warning message for -mno-speculate-indirect-jumps.
3689 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3691 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3692 Undocumented.
3693 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3695 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3697 PR tree-optimization/84225
3698 * tree-eh.c (find_trapping_overflow): Only call
3699 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3701 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3703 PR target/84145
3704 * config/i386/i386.c: Reimplement the check of possible options
3705 -mibt/-mshstk conbination. Change error messages.
3706 * doc/invoke.texi: Fix a typo: remove extra '='.
3708 2018-02-06 Marek Polacek <polacek@redhat.com>
3710 PR tree-optimization/84228
3711 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3713 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3715 PR target/82641
3716 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3717 emitted arch directives.
3718 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3719 __ARM_FEATURE_COPROC before changing architectures.
3721 2018-02-06 Richard Biener <rguenther@suse.de>
3723 * config/i386/i386.c (print_reg): Fix typo.
3724 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3726 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3728 * configure: Regenerate.
3730 2018-02-05 Martin Sebor <msebor@redhat.com>
3732 PR tree-optimization/83369
3733 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3734 inlining context.
3736 2018-02-05 Martin Liska <mliska@suse.cz>
3738 * doc/invoke.texi: Cherry-pick upstream r323995.
3740 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3742 * ira.c (ira_init_register_move_cost): Adjust comment.
3744 2018-02-05 Martin Liska <mliska@suse.cz>
3746 PR gcov-profile/84137
3747 * doc/gcov.texi: Fix typo in documentation.
3749 2018-02-05 Martin Liska <mliska@suse.cz>
3751 PR gcov-profile/83879
3752 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3753 using dlopen functionality.
3755 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3757 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3758 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3759 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3760 _mm_maskz_range_round_ss): New intrinsics.
3761 (__builtin_ia32_rangesd128_round)
3762 (__builtin_ia32_rangess128_round): Remove.
3763 (__builtin_ia32_rangesd128_mask_round,
3764 __builtin_ia32_rangess128_mask_round): New builtins.
3765 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3766 __builtin_ia32_rangess128_round): Remove.
3767 (__builtin_ia32_rangesd128_mask_round,
3768 __builtin_ia32_rangess128_mask_round): New builtins.
3769 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3770 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3771 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3772 "<round_saeonly_constraint>")): Changed to ...
3773 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3774 "<round_saeonly_scalar_constraint>")): ... this.
3775 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3776 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3777 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3778 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3779 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3781 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3783 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3784 options.
3785 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3786 Remove all values except native, 8540 and 8548.
3788 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3790 * config/i386/i386.c (ix86_output_function_return): Pass
3791 INVALID_REGNUM, instead of -1, as invalid register number to
3792 indirect_thunk_name and output_indirect_thunk.
3794 2018-02-02 Julia Koval <julia.koval@intel.com>
3796 * config.gcc: Add -march=icelake.
3797 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3798 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3799 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3800 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3801 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3802 (processor_target_table): Add icelake.
3803 (ix86_option_override_internal): Handle new PTAs.
3804 (get_builtin_code_for_version): Handle icelake.
3805 (M_INTEL_COREI7_ICELAKE): New.
3806 (fold_builtin_cpu): Handle icelake.
3807 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3808 * doc/invoke.texi: Add -march=icelake.
3810 2018-02-02 Julia Koval <julia.koval@intel.com>
3812 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3813 to wide_int_bitmask.
3814 * wide-int-bitmask.h: New.
3816 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3818 PR target/84066
3819 * config/i386/i386.md: Replace Pmode with word_mode in
3820 builtin_setjmp_setup and builtin_longjmp to support x32.
3822 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3824 PR target/56010
3825 PR target/83743
3826 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3827 #include "opts.h".
3828 (rs6000_supported_cpu_names): New static variable.
3829 (linux_cpu_translation_table): Likewise.
3830 (elf_platform) <cpu>: Define new static variable and use it.
3831 Translate kernel AT_PLATFORM name to canonical name if needed.
3832 Error if platform name is unknown.
3834 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3836 PR target/84089
3837 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3839 2018-02-01 Jeff Law <law@redhat.com>
3841 PR target/84128
3842 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3843 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3844 the scratch if RELEASE_VIA_POP is false.
3845 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3846 If we have to save a temporary register, decrement SIZE appropriately.
3847 Pass new arguments to release_scratch_register_on_entry.
3848 (ix86_adjust_stack_and_probe): Likewise.
3849 (ix86_emit_probe_stack_range): Pass new arguments to
3850 release_scratch_register_on_entry.
3852 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3854 PR rtl-optimization/84157
3855 * combine.c (change_zero_ext): Use REG_P predicate in
3856 front of HARD_REGISTER_P predicate.
3858 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3860 * config/avr/avr.c (avr_option_override): Move disabling of
3861 -fdelete-null-pointer-checks to...
3862 * common/config/avr/avr-common.c (avr_option_optimization_table):
3863 ...here.
3865 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3867 PR tree-optimization/81635
3868 * tree-data-ref.c (split_constant_offset_1): For types that
3869 wrap on overflow, try to use range info to prove that wrapping
3870 cannot occur.
3872 2018-02-01 Renlin Li <renlin.li@arm.com>
3874 PR target/83370
3875 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3876 TAILCALL_ADDR_REGS.
3877 (aarch64_register_move_cost): Likewise.
3878 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3879 TAILCALL_ADDR_REGS.
3880 (REG_CLASS_NAMES): Likewise.
3881 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3882 TAILCALL_ADDR_REGS. Remove IP registers.
3883 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3885 2018-02-01 Richard Biener <rguenther@suse.de>
3887 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3888 for specifying RPO order and allow NULL for that.
3889 * domwalk.c (dom_walker::dom_walker): Likewise.
3890 (dom_walker::walk): Handle NULL RPO order.
3891 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3892 in RPO order.
3893 (rewrite_update_dom_walker): Likewise.
3894 (mark_def_dom_walker): Likewise.
3896 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3898 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3899 (aarch64_maybe_expand_sve_subreg_move): Declare.
3900 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3901 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3902 predicate.
3903 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3904 that are semantically a reverse operation.
3905 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3906 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3907 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3908 functions.
3909 (aarch64_can_change_mode_class): For big-endian, forbid changes
3910 between two SVE modes if they have different element sizes.
3912 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3914 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3915 the TImode handling for big-endian targets.
3917 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3919 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3920 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3921 not just bytes.
3922 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3923 Remove BSWAP handing for big-endian targets and use the form of
3924 LD1RQ appropariate for the mode.
3926 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3928 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3929 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3930 duplicated element.
3932 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3934 PR tearget/83845
3935 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3936 check for operands that need to go through aarch64_sve_reload_be.
3938 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3940 PR tree-optimization/81661
3941 PR tree-optimization/84117
3942 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3943 * tree-eh.c: Include gimplify.h.
3944 (find_trapping_overflow, replace_trapping_overflow,
3945 rewrite_to_non_trapping_overflow): New functions.
3946 * tree-vect-loop.c: Include tree-eh.h.
3947 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3948 * tree-data-ref.c: Include tree-eh.h.
3949 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3951 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3953 PR rtl-optimization/84123
3954 * combine.c (change_zero_ext): Check if hard register satisfies
3955 can_change_dest_mode before calling gen_lowpart_SUBREG.
3957 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3959 PR target/82444
3960 * ira.c (ira_init_register_move_cost): Remove assert.
3962 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3964 PR rtl-optimization/84071
3965 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3966 * doc/tm.texi: Regenerate.
3968 2018-01-31 Richard Biener <rguenther@suse.de>
3970 PR tree-optimization/84132
3971 * tree-data-ref.c (analyze_miv_subscript): Properly
3972 check whether evolution_function_is_affine_multivariate_p
3973 before calling gcd_of_steps_may_divide_p.
3975 2018-01-31 Julia Koval <julia.koval@intel.com>
3977 PR target/83618
3978 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3979 * config/i386/i386.md (rdpid_rex64) New.
3980 (rdpid): Make 32bit only.
3982 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3984 PR lto/84105
3985 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3986 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3988 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3990 Revert
3991 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3993 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3995 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3997 PR rtl-optimization/84071
3998 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3999 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4001 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4003 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4004 (arc_attribute_table): Add 'aux' attribute.
4005 (arc_in_small_data_p): Consider aux like variables.
4006 (arc_is_aux_reg_p): New function.
4007 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4008 (arc_get_aux_arg): New function.
4009 (prepare_move_operands): Handle aux-register access.
4010 (arc_handle_aux_attribute): New function.
4011 * doc/extend.texi (ARC Variable attributes): Add subsection.
4013 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4015 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4016 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4017 (arc_attribute_table): Add 'uncached' attribute.
4018 (arc_print_operand): Print '.di' flag for uncached memory
4019 accesses.
4020 (arc_in_small_data_p): Do not consider for small data the uncached
4021 types.
4022 (arc_is_uncached_mem_p): New function.
4023 * config/arc/predicates.md (compact_store_memory_operand): Check
4024 for uncached memory accesses.
4025 (nonvol_nonimm_operand): Likewise.
4026 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4028 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4030 PR c/84100
4031 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4032 falign-loops=): Add Optimization flag.
4034 2018-01-30 Jeff Law <law@redhat.com>
4036 PR target/84064
4037 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4038 INT_REGISTERS_SAVED. Check it prior to calling
4039 get_scratch_register_on_entry.
4040 (ix86_adjust_stack_and_probe): Similarly.
4041 (ix86_emit_probe_stack_range): Similarly.
4042 (ix86_expand_prologue): Corresponding changes.
4044 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4046 PR target/40411
4047 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4048 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4050 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4052 PR target/84112
4053 * lra-constraints.c (curr_insn_transform): Process AND in the
4054 address.
4056 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4058 PR rtl-optimization/83986
4059 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4060 dependence against last_pending_memory_flush in addition to
4061 pending_jump_insns.
4063 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4065 PR tree-optimization/81611
4066 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4067 copies.
4069 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4071 PR target/83758
4072 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4073 a reg rtx.
4075 2018-01-30 Richard Biener <rguenther@suse.de>
4076 Jakub Jelinek <jakub@redhat.com>
4078 PR tree-optimization/84111
4079 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4080 inner loops added during recursion, as they don't have up-to-date
4081 SSA form.
4083 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4085 PR ipa/81360
4086 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4087 (can_inline_edge_by_limits_p): ... here.
4088 (can_early_inline_edge_p, check_callers,
4089 update_caller_keys, update_callee_keys, recursive_inlining,
4090 add_new_edges_to_heap, speculation_useful_p,
4091 inline_small_functions,
4092 inline_small_functions, flatten_function,
4093 inline_to_all_callers_1): Update.
4095 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4097 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4098 zeros correctly.
4100 2018-01-30 Richard Biener <rguenther@suse.de>
4102 PR tree-optimization/83008
4103 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4104 invariant and constant vector uses in stmts when they need
4105 more than one stmt.
4107 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4109 PR bootstrap/84017
4110 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4111 * configure: Regenerate.
4113 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4115 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4116 pattern.
4117 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4118 Use gen_rtx_REG rather than gen_lowpart.
4120 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4122 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4123 rather than 0 when creating partial subregs.
4125 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4127 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4128 of usage.
4130 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4132 PR target/81550
4133 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4134 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4135 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4136 flags. This restores the settings used before the 2017-07-24.
4137 Turning off pre increment/decrement/modify allows IVOPTS to
4138 optimize DF/SF loops where the index is an int.
4140 2018-01-29 Richard Biener <rguenther@suse.de>
4141 Kelvin Nilsen <kelvin@gcc.gnu.org>
4143 PR bootstrap/80867
4144 * tree-vect-stmts.c (vectorizable_call): Don't call
4145 targetm.vectorize_builtin_md_vectorized_function if callee is
4146 NULL.
4148 2018-01-22 Carl Love <cel@us.ibm.com>
4150 * doc/extend.tex: Fix typo in second arg in
4151 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4153 2018-01-29 Richard Biener <rguenther@suse.de>
4155 PR tree-optimization/84086
4156 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4157 (flush_ssaname_freelist): When SSA names were released reset
4158 the SCEV hash table.
4160 2018-01-29 Richard Biener <rguenther@suse.de>
4162 PR tree-optimization/84057
4163 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4164 removed paths when removing edges.
4166 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4168 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4169 -mfunction-return=@var{choice}.
4171 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4173 PR diagnostic/84034
4174 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4175 Handle CR like TAB.
4176 (layout::print_source_line): Likewise.
4177 (test_get_line_width_without_trailing_whitespace): Add test cases.
4179 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4181 PR middle-end/84040
4182 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4183 debug insns.
4185 2018-01-26 Jim Wilson <jimw@sifive.com>
4187 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4189 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4190 specified.
4192 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4194 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4195 and CMP + SUB-immediate -> SUBS.
4197 2018-01-26 Martin Sebor <msebor@redhat.com>
4199 PR tree-optimization/83896
4200 * tree-ssa-strlen.c (get_string_len): Rename...
4201 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4202 Avoid assuming length is constant.
4203 (handle_char_store): Use HOST_WIDE_INT for string length.
4205 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4207 PR target/81763
4208 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4209 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4211 2018-01-26 Richard Biener <rguenther@suse.de>
4213 PR rtl-optimization/84003
4214 * dse.c (record_store): Only record redundant stores when
4215 the earlier store aliases at least all accesses the later one does.
4217 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4219 PR rtl-optimization/83985
4220 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4221 REG_CFA_RESTORE insns.
4222 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4223 REG_CFA_RESTORE insns here.
4225 PR c/83989
4226 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4227 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4229 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4231 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4232 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4233 (arc_init): Likewise.
4234 (arc_override_options): Likewise.
4235 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4236 value.
4237 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4238 support.
4239 * config/arc/arc.h (TARGET_DBNZ): Define.
4240 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4241 properly set the tune attribute.
4242 (dbnz): Use TARGET_DBNZ guard.
4243 * config/arc/arc.opt (mtune): Add core3 option.
4245 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4247 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4248 recognize new pic like addresses.
4249 (arc_delegitimize_address): Clean up.
4251 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4253 * config/arc/arc-arches.def: Option mrf16 valid for all
4254 architectures.
4255 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4256 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4257 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4258 * config/arc/arc-tables.opt: Regenerate.
4259 * config/arc/arc.c (arc_conditional_register_usage): Handle
4260 reduced register file case.
4261 (arc_file_start): Set must have build attributes.
4262 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4263 mrf16 option value.
4264 * config/arc/arc.opt (mrf16): Add new option.
4265 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4266 * config/arc/genmultilib.awk: Handle new mrf16 option.
4267 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4268 * config/arc/t-multilib: Regenerate.
4269 * doc/invoke.texi (ARC Options): Document mrf16 option.
4271 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4273 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4274 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4275 (arc_attribute_table): Add 'secure_call' attribute.
4276 (arc_print_operand): Print secure call operand.
4277 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4278 secure.
4279 (arc_is_secure_call_p): New function. * config/arc/arc.md
4280 (call_i): Add support for sjli instruction.
4281 (call_value_i): Likewise.
4282 * config/arc/constraints.md (Csc): New constraint.
4284 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4285 John Eric Martin <John.Martin@emmicro-us.com>
4287 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4288 * config/arc/arc.c (_arc_jli_section): New struct.
4289 (arc_jli_section): New type.
4290 (rc_jli_sections): New static variable.
4291 (arc_handle_jli_attribute): New function.
4292 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4293 (arc_file_end): New function.
4294 (TARGET_ASM_FILE_END): Define.
4295 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4296 (arc_add_jli_section): New function.
4297 (jli_call_scan): Likewise.
4298 (arc_reorg): Call jli_call_scan.
4299 (arc_output_addsi): Remove 'S' from printing asm operand.
4300 (arc_is_jli_call_p): New function.
4301 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4302 operand.
4303 (movhi_insn): Likewise.
4304 (movsi_insn): Likewise.
4305 (movsi_set_cc_insn): Likewise.
4306 (loadqi_update): Likewise.
4307 (load_zeroextendqisi_update): Likewise.
4308 (load_signextendqisi_update): Likewise.
4309 (loadhi_update): Likewise.
4310 (load_zeroextendhisi_update): Likewise.
4311 (load_signextendhisi_update): Likewise.
4312 (loadsi_update): Likewise.
4313 (loadsf_update): Likewise.
4314 (movsicc_insn): Likewise.
4315 (bset_insn): Likewise.
4316 (bxor_insn): Likewise.
4317 (bclr_insn): Likewise.
4318 (bmsk_insn): Likewise.
4319 (bicsi3_insn): Likewise.
4320 (cmpsi_cc_c_insn): Likewise.
4321 (movsi_ne): Likewise.
4322 (movsi_cond_exec): Likewise.
4323 (clrsbsi2): Likewise.
4324 (norm_f): Likewise.
4325 (normw): Likewise.
4326 (swap): Likewise.
4327 (divaw): Likewise.
4328 (flag): Likewise.
4329 (sr): Likewise.
4330 (kflag): Likewise.
4331 (ffs): Likewise.
4332 (ffs_f): Likewise.
4333 (fls): Likewise.
4334 (call_i): Remove 'S' asm letter, add jli instruction.
4335 (call_value_i): Likewise.
4336 * config/arc/arc.op (mjli-always): New option.
4337 * config/arc/constraints.md (Cji): New constraint.
4338 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4339 operand.
4340 (subsf3_fpx): Likewise.
4341 (mulsf3_fpx): Likewise.
4342 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4343 asm operand.
4344 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4345 function attrbutes.
4346 * doc/invoke.texi (ARC): Document mjli-always option.
4348 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4350 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4351 avoid addition with 0 and use incw and decw where possible.
4353 2018-01-26 Richard Biener <rguenther@suse.de>
4355 PR tree-optimization/81082
4356 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4357 association if it requires casting to unsigned.
4358 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4359 from fold_plusminus_mult_expr to catch important cases late when
4360 range info is available.
4362 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4364 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4365 * configure.ac (hidden_linkonce): New test.
4366 * configure: Regenerate.
4367 * config.in: Regenerate.
4369 2018-01-26 Julia Koval <julia.koval@intel.com>
4371 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4372 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4373 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4374 _mm_mask_bitshuffle_epi64_mask): Fix type.
4375 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4376 USI_FTYPE_V4DI_V4DI_USI): Remove.
4377 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4378 __builtin_ia32_vpshufbitqmb256_mask,
4379 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4380 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4381 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4383 2018-01-26 Alan Modra <amodra@gmail.com>
4385 PR target/84033
4386 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4387 UNSPEC_VBPERMQ. Sort other unspecs.
4389 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4391 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4393 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4395 PR middle-end/83055
4396 * predict.c (drop_profile): Do not push/pop cfun; update also
4397 node->count.
4398 (handle_missing_profiles): Fix logic looking for zero profiles.
4400 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4402 PR middle-end/83977
4403 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4404 on functions with #pragma omp declare simd or functions with simd
4405 attribute.
4406 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4407 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4408 Remove trailing \n from warning_at calls.
4410 2018-01-25 Tom de Vries <tom@codesourcery.com>
4412 PR target/84028
4413 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4414 for neutered workers.
4416 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4418 PR target/68467
4419 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4420 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4422 2018-01-24 Jeff Law <law@redhat.com>
4424 PR target/83994
4425 * i386.c (get_probe_interval): Move to earlier point.
4426 (ix86_compute_frame_layout): If -fstack-clash-protection and
4427 the frame is larger than the probe interval, then use pushes
4428 to save registers rather than reg->mem moves.
4429 (ix86_expand_prologue): Remove conditional for int_registers_saved
4430 assertion.
4432 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4434 PR target/84014
4435 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4436 min/max for never referenced object.
4438 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4440 PR middle-end/83977
4441 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4442 here.
4443 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4444 attributes from DECL_ATTRIBUTES (decl) without affecting
4445 DECL_ATTRIBUTES (current_function_decl).
4446 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4447 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4449 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4451 PR tree-optimization/83979
4452 * fold-const.c (fold_comparison): Use constant_boolean_node
4453 instead of boolean_{true,false}_node.
4455 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4457 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4458 with zero counts.
4460 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4462 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4463 Simplify the clause that sets the length attribute.
4464 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4465 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4466 clause that sets the length attribute.
4467 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4469 2018-01-24 Tom de Vries <tom@codesourcery.com>
4471 PR target/83589
4472 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4473 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4474 Add strict parameter.
4475 (prevent_branch_around_nothing): Insert dummy insn between branch to
4476 label and label with no ptx insn inbetween.
4477 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4479 2018-01-24 Tom de Vries <tom@codesourcery.com>
4481 PR target/81352
4482 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4483 for neutered threads in warp.
4484 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4486 2018-01-24 Richard Biener <rguenther@suse.de>
4488 PR tree-optimization/83176
4489 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4490 operands.
4492 2018-01-24 Richard Biener <rguenther@suse.de>
4494 PR tree-optimization/82819
4495 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4496 code generating pluses that are no-ops in the target precision.
4498 2018-01-24 Richard Biener <rguenther@suse.de>
4500 PR middle-end/84000
4501 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4503 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4505 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4506 to merge probabilities.
4507 * predict.c (probably_never_executed): Also mark as cold functions
4508 with global 0 profile and guessed local profile.
4509 * profile-count.c (profile_probability::combine_with_count): New
4510 member function.
4511 * profile-count.h (profile_probability::operator*,
4512 profile_probability::operator*=, profile_probability::operator/,
4513 profile_probability::operator/=): Reduce precision to adjusted
4514 and set value to guessed on contradictory divisions.
4515 (profile_probability::combine_with_freq): Remove.
4516 (profile_probability::combine_wiht_count): Declare.
4517 (profile_count::force_nonzero):: Set to adjusted.
4518 (profile_count::probability_in):: Set quality to adjusted.
4519 * tree-ssa-tail-merge.c (replace_block_by): Use
4520 combine_with_count.
4522 2018-01-23 Andrew Waterman <andrew@sifive.com>
4523 Jim Wilson <jimw@sifive.com>
4525 * config/riscv/riscv.c (riscv_stack_boundary): New.
4526 (riscv_option_override): Set riscv_stack_boundary. Handle
4527 riscv_preferred_stack_boundary_arg.
4528 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4529 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4530 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4531 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4532 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4533 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4535 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4537 PR target/83905
4538 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4539 of struct ix86_frame.
4540 (ix86_expand_epilogue): Likewise. Add a local variable for
4541 the reg_save_offset field in struct ix86_frame.
4543 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4545 PR tree-optimization/82604
4546 * tree-loop-distribution.c (enum partition_kind): New enum item
4547 PKIND_PARTIAL_MEMSET.
4548 (partition_builtin_p): Support above new enum item.
4549 (generate_code_for_partition): Ditto.
4550 (compute_access_range): Differentiate cases that equality can be
4551 proven at all loops, the innermost loops or no loops.
4552 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4553 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4554 (finalize_partitions, distribute_loop): Don't fuse partition of
4555 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4556 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4557 parloop is enabled.
4559 2018-01-23 Martin Liska <mliska@suse.cz>
4561 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4562 order to ignore the predictor.
4563 (PRED_POLYMORPHIC_CALL): Likewise.
4564 (PRED_RECURSIVE_CALL): Likewise.
4566 2018-01-23 Martin Liska <mliska@suse.cz>
4568 * tree-profile.c (tree_profiling): Print function header to
4569 aware reader which function we are working on.
4570 * value-prof.c (gimple_find_values_to_profile): Do not print
4571 not interesting value histograms.
4573 2018-01-23 Martin Liska <mliska@suse.cz>
4575 * profile-count.h (enum profile_quality): Add
4576 profile_uninitialized as the first value. Do not number values
4577 as they are zero based.
4578 (profile_count::verify): Update sanity check.
4579 (profile_probability::verify): Likewise.
4581 2018-01-23 Nathan Sidwell <nathan@acm.org>
4583 * doc/invoke.texi (ffor-scope): Deprecate.
4585 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4587 PR tree-optimization/83510
4588 * domwalk.c (set_all_edges_as_executable): New function.
4589 (dom_walker::dom_walker): Convert bool param
4590 "skip_unreachable_blocks" to enum reachability. Move setup of
4591 edge flags to set_all_edges_as_executable and only do it when
4592 reachability is REACHABLE_BLOCKS.
4593 * domwalk.h (enum dom_walker::reachability): New enum.
4594 (dom_walker::dom_walker): Convert bool param
4595 "skip_unreachable_blocks" to enum reachability.
4596 (set_all_edges_as_executable): New decl.
4597 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4598 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4599 "reachability".
4600 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4601 but converting true to REACHABLE_BLOCKS.
4602 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4603 * tree-vrp.c
4604 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4605 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4606 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4607 REACHABLE_BLOCKS.
4608 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4609 if check_all_array_refs will be called.
4611 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4613 * tree.c (selftest::test_location_wrappers): Add more test
4614 coverage.
4616 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4618 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4619 (selftest::test_bit_in_range): Likewise.
4621 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4623 PR testsuite/83888
4624 * doc/sourcebuild.texi (vect_float): Say that the selector
4625 only describes the situation when -funsafe-math-optimizations is on.
4626 (vect_float_strict): Document.
4628 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4630 PR tree-optimization/83965
4631 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4632 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4633 instead of checking only for a reduction.
4634 (vect_recog_widen_sum_pattern): Likewise.
4636 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4638 * predict.c (probably_never_executed): Only use precise profile info.
4639 (compute_function_frequency): Skip after inlining hack since we now
4640 have quality checking.
4642 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4644 * profile-count.h (profile_probability::very_unlikely,
4645 profile_probability::unlikely, profile_probability::even): Set
4646 precision to guessed.
4648 2018-01-23 Richard Biener <rguenther@suse.de>
4650 PR tree-optimization/83963
4651 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4652 Properly terminate dominator walk when crossing the exit edge not
4653 when visiting its source block.
4655 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4657 PR c++/83918
4658 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4659 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4661 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4663 PR tree-optimization/83957
4664 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4665 semicolon after for body surrounded by braces.
4667 PR tree-optimization/83081
4668 * profile-count.h (profile_probability::split): New method.
4669 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4670 Use profile_probability::split.
4671 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4672 when splitting a single conditional jump into 2.
4674 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4676 PR tree-optimization/69452
4677 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4678 decl.
4680 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4682 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4683 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4684 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4686 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4688 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4689 * config/rl78/rl78.md (movdi): New define_expand.
4690 * config/rl78/rl78.c (rl78_split_movdi): New function.
4692 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4694 PR target/83862
4695 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4696 no longer used.
4697 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4698 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4699 128-bit to produce an UNSPEC move to get the double word with the
4700 signbit and then a shift directly to do signbit.
4701 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4702 implementation with a new version that just does either a direct
4703 move or a regular move. Move memory interface to separate insns.
4704 Move insns so they are next to the expander.
4705 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4706 with signbit move. Split big and little endian case.
4707 (signbit<mode>2_dm_mem_le): Likewise.
4708 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4709 (signbit<mode>2_dm2): Likewise.
4711 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4713 * config/rl78/rl78.md (anddi3): New define_expand.
4715 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4717 * config/rl78/rl78.md (umindi3): New define_expand.
4719 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4721 * config/rl78/rl78.md (smindi3): New define_expand.
4723 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4725 * config/rl78/rl78.md (smaxdi3): New define_expand.
4727 2018-01-22 Carl Love <cel@us.ibm.com>
4729 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4730 LVX_V1TI): Add macro expansion.
4731 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4732 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4733 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4734 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4735 Change check to determine if the instruction is a byte reversing
4736 entry. Fix typo in comment.
4737 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4738 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4739 Add def_builtin calls for new builtins.
4740 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4741 Add define_insn expansion.
4743 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4745 * config/rl78/rl78.md (umaxdi3): New define_expand.
4747 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4749 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4750 for non-QImode registers.
4752 2018-01-22 Richard Biener <rguenther@suse.de>
4754 PR tree-optimization/83963
4755 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4756 including the loop exit block.
4757 (scop_detection::merge_sese): Likewise.
4758 (scop_detection::add_scop): Do it here instead.
4760 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4762 * doc/sourcebuild.texi (arm_softfloat): Document.
4764 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4766 PR gcc/77734
4767 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4768 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4769 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4771 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4772 David Edelsohn <dje.gcc@gmail.com>
4774 PR target/83946
4775 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4776 Change "crset eq" to "crset 2".
4777 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4778 (*call_indirect_aix<mode>_nospec): Likewise.
4779 (*call_value_indirect_aix<mode>_nospec): Likewise.
4780 (*call_indirect_elfv2<mode>_nospec): Likewise.
4781 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4782 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4783 change assembly output from . to $.
4784 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4785 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4786 (*tablejump<mode>_internal1_nospec): Likewise.
4788 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4790 PR target/80870
4791 * config/sh/sh_optimize_sett_clrt.cc:
4792 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4794 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4796 PR tree-optimization/83940
4797 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4798 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4799 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4800 use it to pass back the definition type.
4801 (vect_check_store_rhs): Likewise rhs_dt_out.
4802 (vect_build_gather_load_calls): Add a mask_dt argument and use
4803 it instead of a call to vect_is_simple_use.
4804 (vectorizable_store): Update calls to vect_check_load_store_mask
4805 and vect_check_store_rhs. Use the dt returned by the latter instead
4806 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4807 instead of calls to vect_is_simple_use. Pass the scalar rather
4808 than the vector operand to vect_is_simple_use when handling
4809 second and subsequent copies of an rhs value.
4810 (vectorizable_load): Update calls to vect_check_load_store_mask
4811 and vect_build_gather_load_calls. Use the cached mask_dt and
4812 gs_info.offset_dt instead of calls to vect_is_simple_use.
4814 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4816 PR middle-end/83945
4817 * tree-emutls.c: Include gimplify.h.
4818 (lower_emutls_2): New function.
4819 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4820 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4821 it before further processing.
4823 PR target/83930
4824 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4825 UINTVAL (trueop1) instead of INTVAL (op1).
4827 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4829 PR debug/81570
4830 PR debug/83728
4831 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4832 INCOMING_FRAME_SP_OFFSET if not defined.
4833 (scan_trace): Add ENTRY argument. If true and
4834 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4835 emit a note to adjust the CFA offset.
4836 (create_cfi_notes): Adjust scan_trace callers.
4837 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4838 INCOMING_FRAME_SP_OFFSET in the CIE.
4839 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4840 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4841 Likewise.
4842 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4843 * doc/tm.texi: Regenerated.
4845 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4847 PR rtl-optimization/83147
4848 * lra-constraints.c (remove_inheritance_pseudos): Use
4849 lra_substitute_pseudo_within_insn.
4851 2018-01-19 Tom de Vries <tom@codesourcery.com>
4852 Cesar Philippidis <cesar@codesourcery.com>
4854 PR target/83920
4855 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4857 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4859 PR target/83790
4860 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4861 spaces for function labels.
4863 2018-01-19 Martin Liska <mliska@suse.cz>
4865 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4866 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4867 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4868 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4869 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4870 (PRED_CONST_RETURN): Change from 69 to 65.
4871 (PRED_NULL_RETURN): Change from 91 to 71.
4872 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4873 (PRED_LOOP_GUARD): Change from 66 to 73.
4875 2018-01-19 Martin Liska <mliska@suse.cz>
4877 * predict.c (predict_insn_def): Add new assert.
4878 (struct branch_predictor): Change type to signed integer.
4879 (test_prediction_value_range): Amend test to cover
4880 PROB_UNINITIALIZED.
4881 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4882 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4883 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4884 (PRED_LOOP_IV_COMPARE): Likewise.
4885 * predict.h (PROB_UNINITIALIZED): Define new constant.
4887 2018-01-19 Martin Liska <mliska@suse.cz>
4889 * predict.c (dump_prediction): Add new format for
4890 analyze_brprob.py script which is enabled with -details
4891 suboption.
4892 * profile-count.h (precise_p): New function.
4894 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4896 PR tree-optimization/83922
4897 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4898 there are no statements that need masking.
4899 (vect_active_double_reduction_p): New function.
4900 (vect_analyze_loop_operations): Use it when handling phis that
4901 are not in the loop header.
4903 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4905 PR tree-optimization/83914
4906 * tree-vect-loop.c (vectorizable_induction): Don't convert
4907 init_expr or apply the peeling adjustment for inductions
4908 that are nested within the vectorized loop.
4910 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4912 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4913 instead of NEG.
4915 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4917 PR sanitizer/81715
4918 PR testsuite/83882
4919 * function.h (gimplify_parameters): Add gimple_seq * argument.
4920 * function.c: Include gimple.h and options.h.
4921 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4922 for the added local temporaries if needed.
4923 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4924 if there are any parameter cleanups, wrap whole body into a
4925 try/finally with the cleanups.
4927 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4929 PR target/82964
4930 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4931 Use GET_MODE_CLASS for scalar floating point.
4933 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4935 PR ipa/82256
4936 patch by PaX Team
4937 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4938 Fix call of call_cgraph_insertion_hooks.
4940 2018-01-18 Martin Sebor <msebor@redhat.com>
4942 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4944 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4946 PR ipa/83619
4947 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4948 frequencies.
4950 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4952 PR other/70268
4953 * common.opt: (-ffile-prefix-map): New option.
4954 * opts.c (common_handle_option): Defer it.
4955 * opts-global.c (handle_common_deferred_options): Handle it.
4956 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4957 * file-prefix-map.h: New file.
4958 (remap_debug_filename, add_debug_prefix_map): ...here.
4959 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4960 * final.c (debug_prefix_map, add_debug_prefix_map
4961 remap_debug_filename): Move to...
4962 * file-prefix-map.c: New file.
4963 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4964 generalize, get rid of alloca(), use strrchr() instead of strchr().
4965 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4966 Implement in terms of add_prefix_map().
4967 (remap_macro_filename, remap_debug_filename): Implement in term of
4968 remap_filename().
4969 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4970 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4971 * dbxout.c: Include file-prefix-map.h.
4972 * varasm.c: Likewise.
4973 * vmsdbgout.c: Likewise.
4974 * xcoffout.c: Likewise.
4975 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4976 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4977 * doc/invoke.texi (-ffile-prefix-map): Document.
4978 (-fdebug-prefix-map): Update description.
4980 2018-01-18 Martin Liska <mliska@suse.cz>
4982 * config/i386/i386.c (indirect_thunk_name): Document that also
4983 lfence is emitted.
4984 (output_indirect_thunk): Document why both instructions
4985 (pause and lfence) are generated.
4987 2018-01-18 Richard Biener <rguenther@suse.de>
4989 PR tree-optimization/83887
4990 * graphite-scop-detection.c
4991 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4992 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4993 (scop_detection::merge_sese): Re-implement with a flood-fill
4994 algorithm that properly finds a SESE region if it exists.
4996 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4998 PR c/61240
4999 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5000 pointer_diff optimizations use view_convert instead of convert.
5002 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5004 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5005 Generate different code for -mno-speculate-indirect-jumps.
5006 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5007 (*call_indirect_aix<mode>): Disable for
5008 -mno-speculate-indirect-jumps.
5009 (*call_indirect_aix<mode>_nospec): New define_insn.
5010 (*call_value_indirect_aix<mode>): Disable for
5011 -mno-speculate-indirect-jumps.
5012 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5013 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5014 -mno-speculate-indirect-jumps.
5015 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5017 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5019 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5020 long double type, set the flags for noting the default long double
5021 type, even if we don't pass or return a long double type.
5023 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5025 PR ipa/83051
5026 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5027 failure.
5029 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5031 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5032 support for merge[hl].
5033 (fold_mergehl_helper): New helper function.
5034 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5035 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5036 (altivec_vmrglw_direct): Add xxmrglw insn.
5038 2018-01-17 Andrew Waterman <andrew@sifive.com>
5040 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5041 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5043 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5045 PR lto/83121
5046 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5047 call the lto_location_cache before reading the
5048 DECL_SOURCE_LOCATION of the types.
5050 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5051 Richard Sandiford <richard.sandiford@linaro.org>
5053 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5054 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5055 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5056 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5057 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5058 Add declaration.
5059 * config/aarch64/constraints.md (aarch64_movti_operand):
5060 Limit immediates.
5061 * config/aarch64/predicates.md (Uti): Add new constraint.
5063 2018-01-17 Carl Love <cel@us.ibm.com>
5065 * config/rs6000/vsx.md (define_expand xl_len_r,
5066 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5067 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5068 lxvll.
5069 (define_expand, define_insn): Move the shift left from the
5070 define_insn to the define_expand for lxvl and stxvl instructions.
5071 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5072 and XL_LEN_R definitions to PURE.
5074 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5076 * config/i386/i386.c (indirect_thunk_name): Declare regno
5077 as unsigned int. Compare regno with INVALID_REGNUM.
5078 (output_indirect_thunk): Ditto.
5079 (output_indirect_thunk_function): Ditto.
5080 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5081 in the call to output_indirect_thunk_function.
5083 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5085 PR middle-end/83884
5086 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5087 rather than the size of inner_type to determine the stack slot size
5088 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5090 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5092 PR target/83546
5093 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5094 to PTA_SILVERMONT.
5096 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5098 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5099 endian Linux systems to optionally enable multilibs for selecting
5100 the long double type if the user configured an explicit type.
5101 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5102 have no long double multilibs if not defined.
5103 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5104 warn if the user used -mabi={ieee,ibm}longdouble and we built
5105 multilibs for long double.
5106 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5107 appropriate multilib option.
5108 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5109 multilib options.
5110 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5111 for building long double multilibs.
5112 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5114 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5116 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5117 copies.
5119 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5120 64 bits.
5121 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5122 128 bits.
5124 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5125 variables.
5127 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5128 return value.
5130 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5132 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5133 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5135 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5137 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5138 different rtl trees depending on TARGET_64BIT.
5139 (rs6000_gen_lvx): Likewise.
5141 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5143 * config/visium/visium.md (nop): Tweak comment.
5144 (hazard_nop): Likewise.
5146 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5148 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5149 -mspeculate-indirect-jumps.
5150 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5151 for -mno-speculate-indirect-jumps.
5152 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5153 (*call_value_indirect_elfv2<mode>): Disable for
5154 -mno-speculate-indirect-jumps.
5155 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5156 (indirect_jump): Emit different RTL for
5157 -mno-speculate-indirect-jumps.
5158 (*indirect_jump<mode>): Disable for
5159 -mno-speculate-indirect-jumps.
5160 (*indirect_jump<mode>_nospec): New define_insn.
5161 (tablejump): Emit different RTL for
5162 -mno-speculate-indirect-jumps.
5163 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5164 (tablejumpsi_nospec): New define_expand.
5165 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5166 (tablejumpdi_nospec): New define_expand.
5167 (*tablejump<mode>_internal1): Disable for
5168 -mno-speculate-indirect-jumps.
5169 (*tablejump<mode>_internal1_nospec): New define_insn.
5170 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5171 option.
5173 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5175 * caller-save.c (insert_save): Drop unnecessary parameter. All
5176 callers updated.
5178 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5179 Richard Biener <rguenth@suse.de>
5181 PR libgomp/83590
5182 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5183 return early, inline manually is_gimple_sizepos. Make sure if we
5184 call gimplify_expr we don't end up with a gimple constant.
5185 * tree.c (variably_modified_type_p): Don't return true for
5186 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5187 * gimplify.h (is_gimple_sizepos): Remove.
5189 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5191 PR tree-optimization/83857
5192 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5193 vectorizable_live_operation for pure SLP statements.
5194 (vectorizable_live_operation): Handle PHIs.
5196 2018-01-16 Richard Biener <rguenther@suse.de>
5198 PR tree-optimization/83867
5199 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5200 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5202 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5204 PR c/83844
5205 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5206 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5207 If off is not INTEGER_CST, issue a may not be aligned warning
5208 rather than isn't aligned. Use isn%'t rather than isn't.
5209 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5210 into MULT_EXPR.
5211 <case MULT_EXPR>: Improve the case when bottom and one of the
5212 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5213 operand, in that case check if the other operand is multiple of
5214 bottom divided by the INTEGER_CST operand.
5216 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5218 PR target/83858
5219 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5220 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5221 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5222 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5223 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5224 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5225 (pa_function_arg_size): New function.
5227 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5229 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5230 in a separate statement.
5232 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5234 PR tree-optimization/83847
5235 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5236 group gathers and scatters.
5238 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5240 PR rtl-optimization/86620
5241 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5243 PR rtl-optimization/83213
5244 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5245 to last if both are JUMP_INSNs.
5247 PR tree-optimization/83843
5248 * gimple-ssa-store-merging.c
5249 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5250 store_immediate_info for bswap/nop orig_stores.
5252 2018-01-15 Andrew Waterman <andrew@sifive.com>
5254 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5255 !TARGET_MUL.
5256 <UDIV>: Increase cost if !TARGET_DIV.
5258 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5260 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5261 (define_attr "cr_logical_3op"): New.
5262 (cceq_ior_compare): Adjust.
5263 (cceq_ior_compare_complement): Adjust.
5264 (*cceq_rev_compare): Adjust.
5265 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5266 (is_cracked_insn): Adjust.
5267 (insn_must_be_first_in_group): Adjust.
5268 * config/rs6000/40x.md: Adjust.
5269 * config/rs6000/440.md: Adjust.
5270 * config/rs6000/476.md: Adjust.
5271 * config/rs6000/601.md: Adjust.
5272 * config/rs6000/603.md: Adjust.
5273 * config/rs6000/6xx.md: Adjust.
5274 * config/rs6000/7450.md: Adjust.
5275 * config/rs6000/7xx.md: Adjust.
5276 * config/rs6000/8540.md: Adjust.
5277 * config/rs6000/cell.md: Adjust.
5278 * config/rs6000/e300c2c3.md: Adjust.
5279 * config/rs6000/e500mc.md: Adjust.
5280 * config/rs6000/e500mc64.md: Adjust.
5281 * config/rs6000/e5500.md: Adjust.
5282 * config/rs6000/e6500.md: Adjust.
5283 * config/rs6000/mpc.md: Adjust.
5284 * config/rs6000/power4.md: Adjust.
5285 * config/rs6000/power5.md: Adjust.
5286 * config/rs6000/power6.md: Adjust.
5287 * config/rs6000/power7.md: Adjust.
5288 * config/rs6000/power8.md: Adjust.
5289 * config/rs6000/power9.md: Adjust.
5290 * config/rs6000/rs64.md: Adjust.
5291 * config/rs6000/titan.md: Adjust.
5293 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5295 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5296 ix86_indirect_branch_register logic.
5298 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5300 * config/i386/constraints.md (Bs): Update
5301 ix86_indirect_branch_register check. Don't check
5302 ix86_indirect_branch_register with GOT_memory_operand.
5303 (Bw): Likewise.
5304 * config/i386/predicates.md (GOT_memory_operand): Don't check
5305 ix86_indirect_branch_register here.
5306 (GOT32_symbol_operand): Likewise.
5308 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5310 * config/i386/predicates.md (constant_call_address_operand):
5311 Rewrite ix86_indirect_branch_register logic.
5312 (sibcall_insn_operand): Likewise.
5314 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5316 * config/i386/constraints.md (Bs): Replace
5317 ix86_indirect_branch_thunk_register with
5318 ix86_indirect_branch_register.
5319 (Bw): Likewise.
5320 * config/i386/i386.md (indirect_jump): Likewise.
5321 (tablejump): Likewise.
5322 (*sibcall_memory): Likewise.
5323 (*sibcall_value_memory): Likewise.
5324 Peepholes of indirect call and jump via memory: Likewise.
5325 * config/i386/i386.opt: Likewise.
5326 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5327 (GOT_memory_operand): Likewise.
5328 (call_insn_operand): Likewise.
5329 (sibcall_insn_operand): Likewise.
5330 (GOT32_symbol_operand): Likewise.
5332 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5334 PR middle-end/83837
5335 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5336 type rather than type addr's type points to.
5337 (expand_omp_atomic_mutex): Likewise.
5338 (expand_omp_atomic): Likewise.
5340 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5342 PR target/83839
5343 * config/i386/i386.c (output_indirect_thunk_function): Use
5344 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5345 for __x86_return_thunk.
5347 2018-01-15 Richard Biener <rguenther@suse.de>
5349 PR middle-end/83850
5350 * expmed.c (extract_bit_field_1): Fix typo.
5352 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5354 PR target/83687
5355 * config/arm/iterators.md (VF): New mode iterator.
5356 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5357 Remove integer-related logic from pattern.
5358 (neon_vabd<mode>_3): Likewise.
5360 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5362 PR middle-end/82694
5363 * common.opt (fstrict-overflow): No longer an alias.
5364 (fwrapv-pointer): New option.
5365 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5366 also for pointer types based on flag_wrapv_pointer.
5367 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5368 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5369 opts->x_flag_wrapv got set.
5370 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5371 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5372 POINTER_TYPE_OVERFLOW_UNDEFINED.
5373 * match.pd: Likewise in address comparison pattern.
5374 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5376 2018-01-15 Richard Biener <rguenther@suse.de>
5378 PR lto/83804
5379 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5380 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5381 Reset type names to their identifier if their TYPE_DECL doesn't
5382 have linkage (and thus is used for ODR and devirt).
5383 (save_debug_info_for_decl): Remove.
5384 (save_debug_info_for_type): Likewise.
5385 (add_tree_to_fld_list): Adjust.
5386 * tree-pretty-print.c (dump_generic_node): Make dumping of
5387 type names more robust.
5389 2018-01-15 Richard Biener <rguenther@suse.de>
5391 * BASE-VER: Bump to 8.0.1.
5393 2018-01-14 Martin Sebor <msebor@redhat.com>
5395 PR other/83508
5396 * builtins.c (check_access): Avoid warning when the no-warning bit
5397 is set.
5399 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5401 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5402 * ira-color (allocno_hard_regs_compare): Likewise.
5404 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5406 PR target/83013
5407 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5408 Use .pushsection/.popsection.
5410 2018-01-14 Martin Sebor <msebor@redhat.com>
5412 PR c++/81327
5413 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5415 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5417 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5418 entry from extra_headers.
5419 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5420 extra_headers, make the list bitwise identical to the i?86-*-* one.
5422 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5424 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5425 -mcmodel=large with -mindirect-branch=thunk,
5426 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5427 -mfunction-return=thunk-extern.
5428 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5429 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5430 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5432 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5434 * config/i386/i386.c (print_reg): Print the name of the full
5435 integer register without '%'.
5436 (ix86_print_operand): Handle 'V'.
5437 * doc/extend.texi: Document 'V' modifier.
5439 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5441 * config/i386/constraints.md (Bs): Disallow memory operand for
5442 -mindirect-branch-register.
5443 (Bw): Likewise.
5444 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5445 (GOT_memory_operand): Likewise.
5446 (call_insn_operand): Likewise.
5447 (sibcall_insn_operand): Likewise.
5448 (GOT32_symbol_operand): Likewise.
5449 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5450 for -mindirect-branch-register.
5451 (tablejump): Likewise.
5452 (*sibcall_memory): Likewise.
5453 (*sibcall_value_memory): Likewise.
5454 Disallow peepholes of indirect call and jump via memory for
5455 -mindirect-branch-register.
5456 (*call_pop): Replace m with Bw.
5457 (*call_value_pop): Likewise.
5458 (*sibcall_pop_memory): Replace m with Bs.
5459 * config/i386/i386.opt (mindirect-branch-register): New option.
5460 * doc/invoke.texi: Document -mindirect-branch-register option.
5462 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5464 * config/i386/i386-protos.h (ix86_output_function_return): New.
5465 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5466 set function_return_type.
5467 (indirect_thunk_name): Add ret_p to indicate thunk for function
5468 return.
5469 (output_indirect_thunk_function): Pass false to
5470 indirect_thunk_name.
5471 (ix86_output_indirect_branch_via_reg): Likewise.
5472 (ix86_output_indirect_branch_via_push): Likewise.
5473 (output_indirect_thunk_function): Create alias for function
5474 return thunk if regno < 0.
5475 (ix86_output_function_return): New function.
5476 (ix86_handle_fndecl_attribute): Handle function_return.
5477 (ix86_attribute_table): Add function_return.
5478 * config/i386/i386.h (machine_function): Add
5479 function_return_type.
5480 * config/i386/i386.md (simple_return_internal): Use
5481 ix86_output_function_return.
5482 (simple_return_internal_long): Likewise.
5483 * config/i386/i386.opt (mfunction-return=): New option.
5484 (indirect_branch): Mention -mfunction-return=.
5485 * doc/extend.texi: Document function_return function attribute.
5486 * doc/invoke.texi: Document -mfunction-return= option.
5488 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5490 * config/i386/i386-opts.h (indirect_branch): New.
5491 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5492 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5493 with local indirect jump when converting indirect call and jump.
5494 (ix86_set_indirect_branch_type): New.
5495 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5496 (indirectlabelno): New.
5497 (indirect_thunk_needed): Likewise.
5498 (indirect_thunk_bnd_needed): Likewise.
5499 (indirect_thunks_used): Likewise.
5500 (indirect_thunks_bnd_used): Likewise.
5501 (INDIRECT_LABEL): Likewise.
5502 (indirect_thunk_name): Likewise.
5503 (output_indirect_thunk): Likewise.
5504 (output_indirect_thunk_function): Likewise.
5505 (ix86_output_indirect_branch_via_reg): Likewise.
5506 (ix86_output_indirect_branch_via_push): Likewise.
5507 (ix86_output_indirect_branch): Likewise.
5508 (ix86_output_indirect_jmp): Likewise.
5509 (ix86_code_end): Call output_indirect_thunk_function if needed.
5510 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5511 needed.
5512 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5513 (ix86_attribute_table): Add indirect_branch.
5514 * config/i386/i386.h (machine_function): Add indirect_branch_type
5515 and has_local_indirect_jump.
5516 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5517 to true.
5518 (tablejump): Likewise.
5519 (*indirect_jump): Use ix86_output_indirect_jmp.
5520 (*tablejump_1): Likewise.
5521 (simple_return_indirect_internal): Likewise.
5522 * config/i386/i386.opt (mindirect-branch=): New option.
5523 (indirect_branch): New.
5524 (keep): Likewise.
5525 (thunk): Likewise.
5526 (thunk-inline): Likewise.
5527 (thunk-extern): Likewise.
5528 * doc/extend.texi: Document indirect_branch function attribute.
5529 * doc/invoke.texi: Document -mindirect-branch= option.
5531 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5533 PR ipa/83051
5534 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5536 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5538 * ipa-inline.c (want_inline_small_function_p): Return false if
5539 inlining has already failed with CIF_FINAL_ERROR.
5540 (update_caller_keys): Call want_inline_small_function_p before
5541 can_inline_edge_p.
5542 (update_callee_keys): Likewise.
5544 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5546 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5547 New function.
5548 (rs6000_quadword_masked_address_p): Likewise.
5549 (quad_aligned_load_p): Likewise.
5550 (quad_aligned_store_p): Likewise.
5551 (const_load_sequence_p): Add comment to describe the outer-most loop.
5552 (mimic_memory_attributes_and_flags): New function.
5553 (rs6000_gen_stvx): Likewise.
5554 (replace_swapped_aligned_store): Likewise.
5555 (rs6000_gen_lvx): Likewise.
5556 (replace_swapped_aligned_load): Likewise.
5557 (replace_swapped_load_constant): Capitalize argument name in
5558 comment describing this function.
5559 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5560 and stores that access quad-word aligned addresses and replace
5561 with stvx or lvx instructions when appropriate.
5562 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5563 New function prototype.
5564 (rs6000_quadword_masked_address_p): Likewise.
5565 (rs6000_gen_lvx): Likewise.
5566 (rs6000_gen_stvx): Likewise.
5567 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5568 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5569 when memory address is aligned.
5570 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5571 this split to select lvx instruction when memory address is aligned.
5572 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5573 instruction when memory address is aligned.
5574 (*vsx_le_perm_load_v16qi): Likewise.
5575 (four unnamed splitters): Modify to select the stvx instruction
5576 when memory is aligned.
5578 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5580 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5581 which appears in the queue multiple times.
5583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5584 Alan Hayward <alan.hayward@arm.com>
5585 David Sherwood <david.sherwood@arm.com>
5587 * tree-vectorizer.h (vec_lower_bound): New structure.
5588 (_loop_vec_info): Add check_nonzero and lower_bounds.
5589 (LOOP_VINFO_CHECK_NONZERO): New macro.
5590 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5591 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5592 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5593 fields. Make seg_len the distance travelled, not including the
5594 access size.
5595 (dr_direction_indicator): Declare.
5596 (dr_zero_step_indicator): Likewise.
5597 (dr_known_forward_stride_p): Likewise.
5598 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5599 tree-ssanames.h.
5600 (runtime_alias_check_p): Allow runtime alias checks with
5601 variable strides.
5602 (operator ==): Compare access_size and align.
5603 (prune_runtime_alias_test_list): Rework for new distinction between
5604 the access_size and seg_len.
5605 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5606 segment lengths.
5607 (get_segment_min_max): New function.
5608 (create_intersect_range_checks): Use it.
5609 (dr_step_indicator): New function.
5610 (dr_direction_indicator): Likewise.
5611 (dr_zero_step_indicator): Likewise.
5612 (dr_known_forward_stride_p): Likewise.
5613 * tree-loop-distribution.c (data_ref_segment_size): Return
5614 DR_STEP * (niters - 1).
5615 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5616 constructor.
5617 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5618 (vect_preserves_scalar_order_p): New function, split out from...
5619 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5620 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5621 (vect_vfa_access_size): New function.
5622 (vect_vfa_align): Likewise.
5623 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5624 (dump_lower_bound): New function.
5625 (vect_check_lower_bound): Likewise.
5626 (vect_small_gap_p): Likewise.
5627 (vectorizable_with_step_bound_p): Likewise.
5628 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5629 depencies if the vectorization factor is 1. Convert the checks
5630 for nonzero steps into checks on the bounds of DR_STEP. Try using
5631 a bunds check for variable steps if the minimum required step is
5632 relatively small. Update calls to the dr_with_seg_len
5633 constructor and to vect_compile_time_alias.
5634 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5635 function.
5636 (vect_loop_versioning): Call it.
5637 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5638 when retrying.
5639 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5642 Alan Hayward <alan.hayward@arm.com>
5643 David Sherwood <david.sherwood@arm.com>
5645 * doc/sourcebuild.texi (vect_scatter_store): Document.
5646 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5647 optabs.
5648 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5649 Document.
5650 * genopinit.c (main): Add supports_vec_scatter_store and
5651 supports_vec_scatter_store_cached to target_optabs.
5652 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5653 IFN_MASK_SCATTER_STORE.
5654 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5655 functions.
5656 * internal-fn.h (internal_store_fn_p): Declare.
5657 (internal_fn_stored_value_index): Likewise.
5658 * internal-fn.c (scatter_store_direct): New macro.
5659 (expand_scatter_store_optab_fn): New function.
5660 (direct_scatter_store_optab_supported_p): New macro.
5661 (internal_store_fn_p): New function.
5662 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5663 IFN_MASK_SCATTER_STORE.
5664 (internal_fn_mask_index): Likewise.
5665 (internal_fn_stored_value_index): New function.
5666 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5667 for scatter stores.
5668 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5669 * optabs-query.c (supports_vec_scatter_store_p): New function.
5670 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5671 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5672 true for scatter stores.
5673 (vect_gather_scatter_fn_p): Handle scatter stores too.
5674 (vect_check_gather_scatter): Consider using scatter stores if
5675 supports_vec_scatter_store_p.
5676 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5677 scatter stores too.
5678 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5679 internal_fn_stored_value_index.
5680 (check_load_store_masking): Handle scatter stores too.
5681 (vect_get_store_rhs): Make public.
5682 (vectorizable_call): Use internal_store_fn_p.
5683 (vectorizable_store): Handle scatter store internal functions.
5684 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5685 when deciding whether the end of the group has been reached.
5686 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5687 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5688 (mask_scatter_store<mode>): New insns.
5690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5691 Alan Hayward <alan.hayward@arm.com>
5692 David Sherwood <david.sherwood@arm.com>
5694 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5695 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5696 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5697 function.
5698 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5699 Use vect_truncate_gather_scatter_offset if we can't treat the
5700 operation as a normal gather load or scatter store.
5701 (get_group_load_store_type): Take the gather_scatter_info
5702 as argument. Try using a gather load or scatter store for
5703 single-element groups.
5704 (get_load_store_type): Update calls to get_group_load_store_type
5705 and vect_use_strided_gather_scatters_p.
5707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5708 Alan Hayward <alan.hayward@arm.com>
5709 David Sherwood <david.sherwood@arm.com>
5711 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5712 optional tree argument.
5713 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5714 null target hooks.
5715 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5716 but continue to use the current value as a fallback.
5717 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5718 to compare the updates.
5719 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5720 (get_load_store_type): Use it when handling a strided access.
5721 (vect_get_strided_load_store_ops): New function.
5722 (vect_get_data_ptr_increment): Likewise.
5723 (vectorizable_load): Handle strided gather loads. Always pass
5724 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5727 Alan Hayward <alan.hayward@arm.com>
5728 David Sherwood <david.sherwood@arm.com>
5730 * doc/md.texi (gather_load@var{m}): Document.
5731 (mask_gather_load@var{m}): Likewise.
5732 * genopinit.c (main): Add supports_vec_gather_load and
5733 supports_vec_gather_load_cached to target_optabs.
5734 * optabs-tree.c (init_tree_optimization_optabs): Use
5735 ggc_cleared_alloc to allocate target_optabs.
5736 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5737 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5738 functions.
5739 * internal-fn.h (internal_load_fn_p): Declare.
5740 (internal_gather_scatter_fn_p): Likewise.
5741 (internal_fn_mask_index): Likewise.
5742 (internal_gather_scatter_fn_supported_p): Likewise.
5743 * internal-fn.c (gather_load_direct): New macro.
5744 (expand_gather_load_optab_fn): New function.
5745 (direct_gather_load_optab_supported_p): New macro.
5746 (direct_internal_fn_optab): New function.
5747 (internal_load_fn_p): Likewise.
5748 (internal_gather_scatter_fn_p): Likewise.
5749 (internal_fn_mask_index): Likewise.
5750 (internal_gather_scatter_fn_supported_p): Likewise.
5751 * optabs-query.c (supports_at_least_one_mode_p): New function.
5752 (supports_vec_gather_load_p): Likewise.
5753 * optabs-query.h (supports_vec_gather_load_p): Declare.
5754 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5755 and memory_type field.
5756 (NUM_PATTERNS): Bump to 15.
5757 * tree-vect-data-refs.c: Include internal-fn.h.
5758 (vect_gather_scatter_fn_p): New function.
5759 (vect_describe_gather_scatter_call): Likewise.
5760 (vect_check_gather_scatter): Try using internal functions for
5761 gather loads. Recognize existing calls to a gather load function.
5762 (vect_analyze_data_refs): Consider using gather loads if
5763 supports_vec_gather_load_p.
5764 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5765 (vect_get_gather_scatter_offset_type): Likewise.
5766 (vect_convert_mask_for_vectype): Likewise.
5767 (vect_add_conversion_to_patterm): Likewise.
5768 (vect_try_gather_scatter_pattern): Likewise.
5769 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5770 (vect_vect_recog_func_ptrs): Add it.
5771 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5772 internal_fn_mask_index and internal_gather_scatter_fn_p.
5773 (check_load_store_masking): Take the gather_scatter_info as an
5774 argument and handle gather loads.
5775 (vect_get_gather_scatter_ops): New function.
5776 (vectorizable_call): Check internal_load_fn_p.
5777 (vectorizable_load): Likewise. Handle gather load internal
5778 functions.
5779 (vectorizable_store): Update call to check_load_store_masking.
5780 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5781 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5782 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5783 (aarch64_gather_scale_operand_d): New predicates.
5784 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5785 (mask_gather_load<mode>): New insns.
5787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5788 Alan Hayward <alan.hayward@arm.com>
5789 David Sherwood <david.sherwood@arm.com>
5791 * optabs.def (fold_left_plus_optab): New optab.
5792 * doc/md.texi (fold_left_plus_@var{m}): Document.
5793 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5794 * internal-fn.c (fold_left_direct): Define.
5795 (expand_fold_left_optab_fn): Likewise.
5796 (direct_fold_left_optab_supported_p): Likewise.
5797 * fold-const-call.c (fold_const_fold_left): New function.
5798 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5799 * tree-parloops.c (valid_reduction_p): New function.
5800 (gather_scalar_reductions): Use it.
5801 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5802 (vect_finish_replace_stmt): Declare.
5803 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5804 (needs_fold_left_reduction_p): New function, split out from...
5805 (vect_is_simple_reduction): ...here. Accept reductions that
5806 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5807 (vect_force_simple_reduction): Also store the reduction type in
5808 the assignment's STMT_VINFO_REDUC_TYPE.
5809 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5810 (merge_with_identity): New function.
5811 (vect_expand_fold_left): Likewise.
5812 (vectorize_fold_left_reduction): Likewise.
5813 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5814 scalar phi in place for it. Check for target support and reject
5815 cases that would reassociate the operation. Defer the transform
5816 phase to vectorize_fold_left_reduction.
5817 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5818 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5819 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5823 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5824 call to ifc_temp_var.
5826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5827 Alan Hayward <alan.hayward@arm.com>
5828 David Sherwood <david.sherwood@arm.com>
5830 * target.def (legitimize_address_displacement): Take the original
5831 offset as a poly_int.
5832 * targhooks.h (default_legitimize_address_displacement): Update
5833 accordingly.
5834 * targhooks.c (default_legitimize_address_displacement): Likewise.
5835 * doc/tm.texi: Regenerate.
5836 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5837 as an argument, moving assert of ad->disp == ad->disp_term to...
5838 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5839 Try calling targetm.legitimize_address_displacement before expanding
5840 the address rather than afterwards, and adjust for the new interface.
5841 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5842 Match the new hook interface. Handle SVE addresses.
5843 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5844 new hook interface.
5846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5848 * Makefile.in (OBJS): Add early-remat.o.
5849 * target.def (select_early_remat_modes): New hook.
5850 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5851 * doc/tm.texi: Regenerate.
5852 * targhooks.h (default_select_early_remat_modes): Declare.
5853 * targhooks.c (default_select_early_remat_modes): New function.
5854 * timevar.def (TV_EARLY_REMAT): New timevar.
5855 * passes.def (pass_early_remat): New pass.
5856 * tree-pass.h (make_pass_early_remat): Declare.
5857 * early-remat.c: New file.
5858 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5859 function.
5860 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5863 Alan Hayward <alan.hayward@arm.com>
5864 David Sherwood <david.sherwood@arm.com>
5866 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5867 vfm1 with a bound_epilog parameter.
5868 (vect_do_peeling): Update calls accordingly, and move the prologue
5869 call earlier in the function. Treat the base bound_epilog as 0 for
5870 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5871 this base when peeling for gaps.
5872 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5873 with fully-masked loops.
5874 (vect_estimate_min_profitable_iters): Handle the single peeled
5875 iteration in that case.
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 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5882 single-element interleaving even if the size is not a power of 2.
5883 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5884 accesses for single-element interleaving if the group size is
5885 not a power of 2.
5887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5888 Alan Hayward <alan.hayward@arm.com>
5889 David Sherwood <david.sherwood@arm.com>
5891 * doc/md.texi (fold_extract_last_@var{m}): Document.
5892 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5893 * optabs.def (fold_extract_last_optab): New optab.
5894 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5895 * internal-fn.c (fold_extract_direct): New macro.
5896 (expand_fold_extract_optab_fn): Likewise.
5897 (direct_fold_extract_optab_supported_p): Likewise.
5898 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5899 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5900 EXTRACT_LAST_REDUCTION.
5901 (get_initial_def_for_reduction): Do not create an initial vector
5902 for EXTRACT_LAST_REDUCTION reductions.
5903 (vectorizable_reduction): Leave the scalar phi in place for
5904 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5905 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5906 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5907 transform phase to vectorizable_condition.
5908 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5909 split out from...
5910 (vect_finish_stmt_generation): ...here.
5911 (vect_finish_replace_stmt): New function.
5912 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5913 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5914 pattern.
5915 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5918 Alan Hayward <alan.hayward@arm.com>
5919 David Sherwood <david.sherwood@arm.com>
5921 * doc/md.texi (extract_last_@var{m}): Document.
5922 * optabs.def (extract_last_optab): New optab.
5923 * internal-fn.def (EXTRACT_LAST): New internal function.
5924 * internal-fn.c (cond_unary_direct): New macro.
5925 (expand_cond_unary_optab_fn): Likewise.
5926 (direct_cond_unary_optab_supported_p): Likewise.
5927 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5928 loops using EXTRACT_LAST.
5929 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5930 (extract_last_<mode>): ...this optab.
5931 (vec_extract<mode><Vel>): Update accordingly.
5933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5934 Alan Hayward <alan.hayward@arm.com>
5935 David Sherwood <david.sherwood@arm.com>
5937 * target.def (empty_mask_is_expensive): New hook.
5938 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5939 * doc/tm.texi: Regenerate.
5940 * targhooks.h (default_empty_mask_is_expensive): Declare.
5941 * targhooks.c (default_empty_mask_is_expensive): New function.
5942 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5943 if the target says that empty masks are expensive.
5944 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5945 New function.
5946 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5949 Alan Hayward <alan.hayward@arm.com>
5950 David Sherwood <david.sherwood@arm.com>
5952 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5953 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5954 (vect_use_loop_mask_for_alignment_p): New function.
5955 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5956 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5957 niters_skip argument. Make sure that the first niters_skip elements
5958 of the first iteration are inactive.
5959 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5960 Update call to vect_set_loop_masks_directly.
5961 (get_misalign_in_elems): New function, split out from...
5962 (vect_gen_prolog_loop_niters): ...here.
5963 (vect_update_init_of_dr): Take a code argument that specifies whether
5964 the adjustment should be added or subtracted.
5965 (vect_update_init_of_drs): Likewise.
5966 (vect_prepare_for_masked_peels): New function.
5967 (vect_do_peeling): Skip prologue peeling if we're using a mask
5968 instead. Update call to vect_update_inits_of_drs.
5969 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5970 mask_skip_niters.
5971 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5972 alignment. Do not include the number of peeled iterations in
5973 the minimum threshold in that case.
5974 (vectorizable_induction): Adjust the start value down by
5975 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5976 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5977 Take the number of skipped iterations into account when calculating
5978 the loop bounds.
5979 * tree-vect-stmts.c (vect_gen_while_not): New function.
5981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5982 Alan Hayward <alan.hayward@arm.com>
5983 David Sherwood <david.sherwood@arm.com>
5985 * doc/sourcebuild.texi (vect_fully_masked): Document.
5986 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5987 default value to 0.
5988 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5989 split out from...
5990 (vect_analyze_loop_2): ...here. Don't check the vectorization
5991 factor against the number of loop iterations if the loop is
5992 fully-masked.
5994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5995 Alan Hayward <alan.hayward@arm.com>
5996 David Sherwood <david.sherwood@arm.com>
5998 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5999 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6000 (dump_groups): Update accordingly.
6001 (iv_use::mem_type): New member variable.
6002 (address_p): New function.
6003 (record_use): Add a mem_type argument and initialize the new
6004 mem_type field.
6005 (record_group_use): Add a mem_type argument. Use address_p.
6006 Remove obsolete null checks of base_object. Update call to record_use.
6007 (find_interesting_uses_op): Update call to record_group_use.
6008 (find_interesting_uses_cond): Likewise.
6009 (find_interesting_uses_address): Likewise.
6010 (get_mem_type_for_internal_fn): New function.
6011 (find_address_like_use): Likewise.
6012 (find_interesting_uses_stmt): Try find_address_like_use before
6013 calling find_interesting_uses_op.
6014 (addr_offset_valid_p): Use the iv mem_type field as the type
6015 of the addressed memory.
6016 (add_autoinc_candidates): Likewise.
6017 (get_address_cost): Likewise.
6018 (split_small_address_groups_p): Use address_p.
6019 (split_address_groups): Likewise.
6020 (add_iv_candidate_for_use): Likewise.
6021 (autoinc_possible_for_pair): Likewise.
6022 (rewrite_groups): Likewise.
6023 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6024 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6025 (get_alias_ptr_type_for_ptr_address): New function.
6026 (rewrite_use_address): Rewrite address uses in calls that were
6027 identified by find_address_like_use.
6029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6030 Alan Hayward <alan.hayward@arm.com>
6031 David Sherwood <david.sherwood@arm.com>
6033 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6034 TARGET_MEM_REFs.
6035 * gimple-expr.h (is_gimple_addressable: Likewise.
6036 * gimple-expr.c (is_gimple_address): Likewise.
6037 * internal-fn.c (expand_call_mem_ref): New function.
6038 (expand_mask_load_optab_fn): Use it.
6039 (expand_mask_store_optab_fn): Likewise.
6041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6042 Alan Hayward <alan.hayward@arm.com>
6043 David Sherwood <david.sherwood@arm.com>
6045 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6046 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6047 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6048 (cond_umax@var{mode}): Document.
6049 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6050 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6051 (cond_umin_optab, cond_umax_optab): New optabs.
6052 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6053 (COND_IOR, COND_XOR): New internal functions.
6054 * internal-fn.h (get_conditional_internal_fn): Declare.
6055 * internal-fn.c (cond_binary_direct): New macro.
6056 (expand_cond_binary_optab_fn): Likewise.
6057 (direct_cond_binary_optab_supported_p): Likewise.
6058 (get_conditional_internal_fn): New function.
6059 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6060 Cope with reduction statements that are vectorized as calls rather
6061 than assignments.
6062 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6063 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6064 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6065 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6066 (UNSPEC_COND_EOR): New unspecs.
6067 (optab): Add mappings for them.
6068 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6069 (sve_int_op, sve_fp_op): New int attributes.
6071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6072 Alan Hayward <alan.hayward@arm.com>
6073 David Sherwood <david.sherwood@arm.com>
6075 * optabs.def (while_ult_optab): New optab.
6076 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6077 * internal-fn.def (WHILE_ULT): New internal function.
6078 * internal-fn.h (direct_internal_fn_supported_p): New override
6079 that takes two types as argument.
6080 * internal-fn.c (while_direct): New macro.
6081 (expand_while_optab_fn): New function.
6082 (convert_optab_supported_p): Likewise.
6083 (direct_while_optab_supported_p): New macro.
6084 * wide-int.h (wi::udiv_ceil): New function.
6085 * tree-vectorizer.h (rgroup_masks): New structure.
6086 (vec_loop_masks): New typedef.
6087 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6088 and fully_masked_p.
6089 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6090 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6091 (vect_max_vf): New function.
6092 (slpeel_make_loop_iterate_ntimes): Delete.
6093 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6094 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6095 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6096 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6097 internal-fn.h, stor-layout.h and optabs-query.h.
6098 (vect_set_loop_mask): New function.
6099 (add_preheader_seq): Likewise.
6100 (add_header_seq): Likewise.
6101 (interleave_supported_p): Likewise.
6102 (vect_maybe_permute_loop_masks): Likewise.
6103 (vect_set_loop_masks_directly): Likewise.
6104 (vect_set_loop_condition_masked): Likewise.
6105 (vect_set_loop_condition_unmasked): New function, split out from
6106 slpeel_make_loop_iterate_ntimes.
6107 (slpeel_make_loop_iterate_ntimes): Rename to..
6108 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6109 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6110 (vect_do_peeling): Update call accordingly.
6111 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6112 loops.
6113 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6114 mask_compare_type, can_fully_mask_p and fully_masked_p.
6115 (release_vec_loop_masks): New function.
6116 (_loop_vec_info): Use it to free the loop masks.
6117 (can_produce_all_loop_masks_p): New function.
6118 (vect_get_max_nscalars_per_iter): Likewise.
6119 (vect_verify_full_masking): Likewise.
6120 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6121 retries, and free the mask rgroups before retrying. Check loop-wide
6122 reasons for disallowing fully-masked loops. Make the final decision
6123 about whether use a fully-masked loop or not.
6124 (vect_estimate_min_profitable_iters): Do not assume that peeling
6125 for the number of iterations will be needed for fully-masked loops.
6126 (vectorizable_reduction): Disable fully-masked loops.
6127 (vectorizable_live_operation): Likewise.
6128 (vect_halve_mask_nunits): New function.
6129 (vect_double_mask_nunits): Likewise.
6130 (vect_record_loop_mask): Likewise.
6131 (vect_get_loop_mask): Likewise.
6132 (vect_transform_loop): Handle the case in which the final loop
6133 iteration might handle a partial vector. Call vect_set_loop_condition
6134 instead of slpeel_make_loop_iterate_ntimes.
6135 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6136 (check_load_store_masking): New function.
6137 (prepare_load_store_mask): Likewise.
6138 (vectorizable_store): Handle fully-masked loops.
6139 (vectorizable_load): Likewise.
6140 (supportable_widening_operation): Use vect_halve_mask_nunits for
6141 booleans.
6142 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6143 (vect_gen_while): New function.
6144 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6145 (aarch64_uqdec<mode>): New insn.
6147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6148 Alan Hayward <alan.hayward@arm.com>
6149 David Sherwood <david.sherwood@arm.com>
6151 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6152 (reduc_xor_scal_optab): New optabs.
6153 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6154 (reduc_xor_scal_@var{m}): Document.
6155 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6156 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6157 internal functions.
6158 * fold-const-call.c (fold_const_call): Handle them.
6159 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6160 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6161 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6162 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6163 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6164 (UNSPEC_XORV): New unspecs.
6165 (optab): Add entries for them.
6166 (BITWISEV): New int iterator.
6167 (bit_reduc_op): New int attributes.
6169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6170 Alan Hayward <alan.hayward@arm.com>
6171 David Sherwood <david.sherwood@arm.com>
6173 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6174 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6175 * optabs.def (vec_shl_insert_optab): New optab.
6176 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6177 (duplicate_and_interleave): Likewise.
6178 * tree-vect-loop.c: Include internal-fn.h.
6179 (neutral_op_for_slp_reduction): New function, split out from
6180 get_initial_defs_for_reduction.
6181 (get_initial_def_for_reduction): Handle option 2 for variable-length
6182 vectors by loading the neutral value into a vector and then shifting
6183 the initial value into element 0.
6184 (get_initial_defs_for_reduction): Replace the code argument with
6185 the neutral value calculated by neutral_op_for_slp_reduction.
6186 Use gimple_build_vector for constant-length vectors.
6187 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6188 but the first group_size elements have a neutral value.
6189 Use duplicate_and_interleave otherwise.
6190 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6191 Update call to get_initial_defs_for_reduction. Handle SLP
6192 reductions for variable-length vectors by creating one vector
6193 result for each scalar result, with the elements associated
6194 with other scalar results stubbed out with the neutral value.
6195 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6196 Require IFN_VEC_SHL_INSERT for double reductions on
6197 variable-length vectors, or SLP reductions that have
6198 a neutral value. Require can_duplicate_and_interleave_p
6199 support for variable-length unchained SLP reductions if there
6200 is no neutral value, such as for MIN/MAX reductions. Also require
6201 the number of vector elements to be a multiple of the number of
6202 SLP statements when doing variable-length unchained SLP reductions.
6203 Update call to vect_create_epilog_for_reduction.
6204 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6205 and remove initial values.
6206 (duplicate_and_interleave): Make public.
6207 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6208 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6211 Alan Hayward <alan.hayward@arm.com>
6212 David Sherwood <david.sherwood@arm.com>
6214 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6215 (can_duplicate_and_interleave_p): New function.
6216 (vect_get_and_check_slp_defs): Take the vector of statements
6217 rather than just the current one. Remove excess parentheses.
6218 Restriction rejectinon of vect_constant_def and vect_external_def
6219 for variable-length vectors to boolean types, or types for which
6220 can_duplicate_and_interleave_p is false.
6221 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6222 (duplicate_and_interleave): New function.
6223 (vect_get_constant_vectors): Use gimple_build_vector for
6224 constant-length vectors and suitable variable-length constant
6225 vectors. Use duplicate_and_interleave for other variable-length
6226 vectors. Don't defer the update when inserting new statements.
6228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6229 Alan Hayward <alan.hayward@arm.com>
6230 David Sherwood <david.sherwood@arm.com>
6232 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6233 min_profitable_iters doesn't go negative.
6235 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6236 Alan Hayward <alan.hayward@arm.com>
6237 David Sherwood <david.sherwood@arm.com>
6239 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6240 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6241 * optabs.def (vec_mask_load_lanes_optab): New optab.
6242 (vec_mask_store_lanes_optab): Likewise.
6243 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6244 (MASK_STORE_LANES): Likewise.
6245 * internal-fn.c (mask_load_lanes_direct): New macro.
6246 (mask_store_lanes_direct): Likewise.
6247 (expand_mask_load_optab_fn): Handle masked operations.
6248 (expand_mask_load_lanes_optab_fn): New macro.
6249 (expand_mask_store_optab_fn): Handle masked operations.
6250 (expand_mask_store_lanes_optab_fn): New macro.
6251 (direct_mask_load_lanes_optab_supported_p): Likewise.
6252 (direct_mask_store_lanes_optab_supported_p): Likewise.
6253 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6254 parameter.
6255 (vect_load_lanes_supported): Likewise.
6256 * tree-vect-data-refs.c (strip_conversion): New function.
6257 (can_group_stmts_p): Likewise.
6258 (vect_analyze_data_ref_accesses): Use it instead of checking
6259 for a pair of assignments.
6260 (vect_store_lanes_supported): Take a masked_p parameter.
6261 (vect_load_lanes_supported): Likewise.
6262 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6263 vect_store_lanes_supported and vect_load_lanes_supported.
6264 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6265 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6266 parameter. Don't allow gaps for masked accesses.
6267 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6268 and vect_load_lanes_supported.
6269 (get_load_store_type): Take a masked_p parameter and update
6270 call to get_group_load_store_type.
6271 (vectorizable_store): Update call to get_load_store_type.
6272 Handle IFN_MASK_STORE_LANES.
6273 (vectorizable_load): Update call to get_load_store_type.
6274 Handle IFN_MASK_LOAD_LANES.
6276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6277 Alan Hayward <alan.hayward@arm.com>
6278 David Sherwood <david.sherwood@arm.com>
6280 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6281 modes for SVE.
6282 * config/aarch64/aarch64-protos.h
6283 (aarch64_sve_struct_memory_operand_p): Declare.
6284 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6285 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6286 (VPRED, vpred): Handle SVE structure modes.
6287 * config/aarch64/constraints.md (Utx): New constraint.
6288 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6289 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6290 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6291 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6292 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6293 structure modes. Split into pieces after RA.
6294 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6295 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6296 New patterns.
6297 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6298 SVE structure modes.
6299 (aarch64_classify_address): Likewise.
6300 (sizetochar): Move earlier in file.
6301 (aarch64_print_operand): Handle SVE register lists.
6302 (aarch64_array_mode): New function.
6303 (aarch64_sve_struct_memory_operand_p): Likewise.
6304 (TARGET_ARRAY_MODE): Redefine.
6306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6307 Alan Hayward <alan.hayward@arm.com>
6308 David Sherwood <david.sherwood@arm.com>
6310 * target.def (array_mode): New target hook.
6311 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6312 * doc/tm.texi: Regenerate.
6313 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6314 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6315 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6316 targetm.array_mode.
6317 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6318 type sizes.
6320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6321 Alan Hayward <alan.hayward@arm.com>
6322 David Sherwood <david.sherwood@arm.com>
6324 * fold-const.c (fold_binary_loc): Check the argument types
6325 rather than the result type when testing for a vector operation.
6327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6329 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6330 * doc/tm.texi: Regenerate.
6332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6333 Alan Hayward <alan.hayward@arm.com>
6334 David Sherwood <david.sherwood@arm.com>
6336 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6337 (sve): Document new AArch64 extension.
6338 * doc/md.texi (w): Extend the description of the AArch64
6339 constraint to include SVE vectors.
6340 (Upl, Upa): Document new AArch64 predicate constraints.
6341 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6342 enum.
6343 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6344 (msve-vector-bits=): New option.
6345 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6346 SVE when these are disabled.
6347 (sve): New extension.
6348 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6349 modes. Adjust their number of units based on aarch64_sve_vg.
6350 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6351 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6352 aarch64_addr_query_type.
6353 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6354 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6355 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6356 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6357 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6358 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6359 (aarch64_simd_imm_zero_p): Delete.
6360 (aarch64_check_zero_based_sve_index_immediate): Declare.
6361 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6362 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6363 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6364 (aarch64_sve_float_mul_immediate_p): Likewise.
6365 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6366 rather than an rtx.
6367 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6368 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6369 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6370 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6371 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6372 (aarch64_regmode_natural_size): Likewise.
6373 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6374 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6375 left one place.
6376 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6377 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6378 for VG and the SVE predicate registers.
6379 (V_ALIASES): Add a "z"-prefixed alias.
6380 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6381 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6382 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6383 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6384 (REG_CLASS_NAMES): Add entries for them.
6385 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6386 and the predicate registers.
6387 (aarch64_sve_vg): Declare.
6388 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6389 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6390 (REGMODE_NATURAL_SIZE): Define.
6391 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6392 SVE macros.
6393 * config/aarch64/aarch64.c: Include cfgrtl.h.
6394 (simd_immediate_info): Add a constructor for series vectors,
6395 and an associated step field.
6396 (aarch64_sve_vg): New variable.
6397 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6398 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6399 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6400 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6401 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6402 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6403 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6404 (aarch64_get_mask_mode): New functions.
6405 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6406 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6407 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6408 predicate modes and predicate registers. Explicitly restrict
6409 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6410 to store a vector mode if it is recognized by
6411 aarch64_classify_vector_mode.
6412 (aarch64_regmode_natural_size): New function.
6413 (aarch64_hard_regno_caller_save_mode): Return the original mode
6414 for predicates.
6415 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6416 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6417 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6418 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6419 functions.
6420 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6421 does not overlap dest if the function is frame-related. Handle
6422 SVE constants.
6423 (aarch64_split_add_offset): New function.
6424 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6425 them aarch64_add_offset.
6426 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6427 and update call to aarch64_sub_sp.
6428 (aarch64_add_cfa_expression): New function.
6429 (aarch64_expand_prologue): Pass extra temporary registers to the
6430 functions above. Handle the case in which we need to emit new
6431 DW_CFA_expressions for registers that were originally saved
6432 relative to the stack pointer, but now have to be expressed
6433 relative to the frame pointer.
6434 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6435 functions above.
6436 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6437 IP0 and IP1 values for SVE frames.
6438 (aarch64_expand_vec_series): New function.
6439 (aarch64_expand_sve_widened_duplicate): Likewise.
6440 (aarch64_expand_sve_const_vector): Likewise.
6441 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6442 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6443 into the register, rather than emitting a SET directly.
6444 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6445 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6446 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6447 (offset_9bit_signed_scaled_p): New functions.
6448 (aarch64_replicate_bitmask_imm): New function.
6449 (aarch64_bitmask_imm): Use it.
6450 (aarch64_cannot_force_const_mem): Reject expressions involving
6451 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6452 (aarch64_classify_index): Handle SVE indices, by requiring
6453 a plain register index with a scale that matches the element size.
6454 (aarch64_classify_address): Handle SVE addresses. Assert that
6455 the mode of the address is VOIDmode or an integer mode.
6456 Update call to aarch64_classify_symbol.
6457 (aarch64_classify_symbolic_expression): Update call to
6458 aarch64_classify_symbol.
6459 (aarch64_const_vec_all_in_range_p): New function.
6460 (aarch64_print_vector_float_operand): Likewise.
6461 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6462 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6463 and the FP immediates 1.0 and 0.5.
6464 (aarch64_print_address_internal): Handle SVE addresses.
6465 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6466 (aarch64_regno_regclass): Handle predicate registers.
6467 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6468 data modes.
6469 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6470 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6471 (aarch64_convert_sve_vector_bits): New function.
6472 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6473 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6474 rather than an rtx.
6475 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6476 Handle SVE vector and predicate modes. Accept VL-based constants
6477 that need only one temporary register, and VL offsets that require
6478 no temporary registers.
6479 (aarch64_conditional_register_usage): Mark the predicate registers
6480 as fixed if SVE isn't available.
6481 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6482 Return true for SVE vector and predicate modes.
6483 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6484 rather than an unsigned int. Handle SVE modes.
6485 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6486 SVE modes.
6487 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6488 if SVE is enabled.
6489 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6490 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6491 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6492 (aarch64_sve_float_mul_immediate_p): New functions.
6493 (aarch64_sve_valid_immediate): New function.
6494 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6495 Explicitly reject structure modes. Check for INDEX constants.
6496 Handle PTRUE and PFALSE constants.
6497 (aarch64_check_zero_based_sve_index_immediate): New function.
6498 (aarch64_simd_imm_zero_p): Delete.
6499 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6500 vector modes. Accept constants in the range of CNT[BHWD].
6501 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6502 ask for an Advanced SIMD mode.
6503 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6504 (aarch64_simd_vector_alignment): Handle SVE predicates.
6505 (aarch64_vectorize_preferred_vector_alignment): New function.
6506 (aarch64_simd_vector_alignment_reachable): Use it instead of
6507 the vector size.
6508 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6509 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6510 functions.
6511 (MAX_VECT_LEN): Delete.
6512 (expand_vec_perm_d): Add a vec_flags field.
6513 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6514 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6515 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6516 for SVE modes.
6517 (aarch64_evpc_rev): Rename to...
6518 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6519 (aarch64_evpc_rev_global): New function.
6520 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6521 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6522 MAX_VECT_LEN.
6523 (aarch64_evpc_sve_tbl): New function.
6524 (aarch64_expand_vec_perm_const_1): Update after rename of
6525 aarch64_evpc_rev. Handle SVE permutes too, trying
6526 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6527 than aarch64_evpc_tbl.
6528 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6529 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6530 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6531 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6532 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6533 (aarch64_expand_sve_vcond): New functions.
6534 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6535 of aarch64_vector_mode_p.
6536 (aarch64_dwarf_poly_indeterminate_value): New function.
6537 (aarch64_compute_pressure_classes): Likewise.
6538 (aarch64_can_change_mode_class): Likewise.
6539 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6540 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6541 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6542 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6543 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6544 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6545 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6546 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6547 constraints.
6548 (Dn, Dl, Dr): Accept const as well as const_vector.
6549 (Dz): Likewise. Compare against CONST0_RTX.
6550 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6551 of "vector" where appropriate.
6552 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6553 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6554 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6555 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6556 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6557 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6558 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6559 (v_int_equiv): Extend to SVE modes.
6560 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6561 mode attributes.
6562 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6563 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6564 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6565 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6566 (SVE_COND_FP_CMP): New int iterators.
6567 (perm_hilo): Handle the new unpack unspecs.
6568 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6569 attributes.
6570 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6571 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6572 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6573 (aarch64_equality_operator, aarch64_constant_vector_operand)
6574 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6575 (aarch64_sve_nonimmediate_operand): Likewise.
6576 (aarch64_sve_general_operand): Likewise.
6577 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6578 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6579 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6580 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6581 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6582 (aarch64_sve_float_arith_immediate): Likewise.
6583 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6584 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6585 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6586 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6587 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6588 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6589 (aarch64_sve_float_arith_operand): Likewise.
6590 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6591 (aarch64_sve_float_mul_operand): Likewise.
6592 (aarch64_sve_vec_perm_operand): Likewise.
6593 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6594 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6595 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6596 as well as const_vector.
6597 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6598 in file. Use CONST0_RTX and CONSTM1_RTX.
6599 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6600 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6601 Use aarch64_simd_imm_zero.
6602 * config/aarch64/aarch64-sve.md: New file.
6603 * config/aarch64/aarch64.md: Include it.
6604 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6605 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6606 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6607 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6608 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6609 (sve): New attribute.
6610 (enabled): Disable instructions with the sve attribute unless
6611 TARGET_SVE.
6612 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6613 aarch64_expand_mov_immediate.
6614 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6615 CNT[BHSD] immediates.
6616 (movti): Split CONST_POLY_INT moves into two halves.
6617 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6618 Split additions that need a temporary here if the destination
6619 is the stack pointer.
6620 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6621 (*add<mode>3_poly_1): New instruction.
6622 (set_clobber_cc): New expander.
6624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6626 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6627 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6628 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6629 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6630 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6631 Change innermode from fixed_mode_size to machine_mode.
6632 (simplify_subreg): Update call accordingly. Handle a constant-sized
6633 subreg of a variable-length CONST_VECTOR.
6635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6636 Alan Hayward <alan.hayward@arm.com>
6637 David Sherwood <david.sherwood@arm.com>
6639 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6640 (add_offset_to_base): New function, split out from...
6641 (create_mem_ref): ...here. When handling a scale other than 1,
6642 check first whether the address is valid without the offset.
6643 Add it into the base if so, leaving the index and scale as-is.
6645 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6647 PR c++/83778
6648 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6649 fold_for_warn before checking if arg2 is INTEGER_CST.
6651 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6653 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6654 (store_multiple_operation): Delete.
6655 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6656 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6657 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6658 guarded by TARGET_STRING.
6659 (rs6000_output_load_multiple): Delete.
6660 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6661 OPTION_MASK_STRING / TARGET_STRING handling.
6662 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6663 (const rs6000_opt_masks) <"string">: Change mask to 0.
6664 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6665 (MASK_STRING): Delete.
6666 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6667 parts. Simplify.
6668 (load_multiple): Delete.
6669 (*ldmsi8): Delete.
6670 (*ldmsi7): Delete.
6671 (*ldmsi6): Delete.
6672 (*ldmsi5): Delete.
6673 (*ldmsi4): Delete.
6674 (*ldmsi3): Delete.
6675 (store_multiple): Delete.
6676 (*stmsi8): Delete.
6677 (*stmsi7): Delete.
6678 (*stmsi6): Delete.
6679 (*stmsi5): Delete.
6680 (*stmsi4): Delete.
6681 (*stmsi3): Delete.
6682 (movmemsi_8reg): Delete.
6683 (corresponding unnamed define_insn): Delete.
6684 (movmemsi_6reg): Delete.
6685 (corresponding unnamed define_insn): Delete.
6686 (movmemsi_4reg): Delete.
6687 (corresponding unnamed define_insn): Delete.
6688 (movmemsi_2reg): Delete.
6689 (corresponding unnamed define_insn): Delete.
6690 (movmemsi_1reg): Delete.
6691 (corresponding unnamed define_insn): Delete.
6692 * config/rs6000/rs6000.opt (mno-string): New.
6693 (mstring): Replace by deprecation warning stub.
6694 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6696 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6698 * regrename.c (regrename_do_replace): If replacing the same
6699 reg multiple times, try to reuse last created gen_raw_REG.
6701 PR debug/81155
6702 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6703 main to workaround a bug in GDB.
6705 2018-01-12 Tom de Vries <tom@codesourcery.com>
6707 PR target/83737
6708 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6710 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6712 PR rtl-optimization/80481
6713 * ira-color.c (get_cap_member): New function.
6714 (allocnos_conflict_by_live_ranges_p): Use it.
6715 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6716 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6718 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6720 PR target/83628
6721 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6722 (*saddl_se_1): Ditto.
6723 (*ssubsi_1): Ditto.
6724 (*ssubl_se_1): Ditto.
6726 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6728 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6729 rather than wi::to_widest for DR_INITs.
6730 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6731 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6732 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6733 INTEGER_CSTs.
6734 (vect_analyze_group_access_1): Note that here.
6736 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6738 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6739 polynomial type sizes.
6741 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6743 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6744 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6745 (gimple_add_tmp_var): Likewise.
6747 2018-01-12 Martin Liska <mliska@suse.cz>
6749 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6750 (gimple_alloc_sizes): Likewise.
6751 (dump_gimple_statistics): Use PRIu64 in printf format.
6752 * gimple.h: Change uint64_t to int.
6754 2018-01-12 Martin Liska <mliska@suse.cz>
6756 * tree-core.h: Use uint64_t instead of int.
6757 * tree.c (tree_node_counts): Likewise.
6758 (tree_node_sizes): Likewise.
6759 (dump_tree_statistics): Use PRIu64 in printf format.
6761 2018-01-12 Martin Liska <mliska@suse.cz>
6763 * Makefile.in: As qsort_chk is implemented in vec.c, add
6764 vec.o to linkage of gencfn-macros.
6765 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6766 passing the info to record_node_allocation_statistics.
6767 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6768 and pass the info.
6769 * ggc-common.c (struct ggc_usage): Add operator== and use
6770 it in operator< and compare function.
6771 * mem-stats.h (struct mem_usage): Likewise.
6772 * vec.c (struct vec_usage): Remove operator< and compare
6773 function. Can be simply inherited.
6775 2018-01-12 Martin Jambor <mjambor@suse.cz>
6777 PR target/81616
6778 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6779 * tree-ssa-math-opts.c: Include domwalk.h.
6780 (convert_mult_to_fma_1): New function.
6781 (fma_transformation_info): New type.
6782 (fma_deferring_state): Likewise.
6783 (cancel_fma_deferring): New function.
6784 (result_of_phi): Likewise.
6785 (last_fma_candidate_feeds_initial_phi): Likewise.
6786 (convert_mult_to_fma): Added deferring logic, split actual
6787 transformation to convert_mult_to_fma_1.
6788 (math_opts_dom_walker): New type.
6789 (math_opts_dom_walker::after_dom_children): New method, body moved
6790 here from pass_optimize_widening_mul::execute, added deferring logic
6791 bits.
6792 (pass_optimize_widening_mul::execute): Moved most of code to
6793 math_opts_dom_walker::after_dom_children.
6794 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6795 * config/i386/i386.c (ix86_option_override_internal): Added
6796 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6798 2018-01-12 Richard Biener <rguenther@suse.de>
6800 PR debug/83157
6801 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6802 inline instance vars.
6804 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6806 PR target/81819
6807 * config/rx/rx.c (rx_is_restricted_memory_address):
6808 Handle SUBREG case.
6810 2018-01-12 Richard Biener <rguenther@suse.de>
6812 PR tree-optimization/80846
6813 * target.def (split_reduction): New target hook.
6814 * targhooks.c (default_split_reduction): New function.
6815 * targhooks.h (default_split_reduction): Declare.
6816 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6817 target requests first reduce vectors by combining low and high
6818 parts.
6819 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6820 (get_vectype_for_scalar_type_and_size): Export.
6821 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6822 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6823 * doc/tm.texi: Regenerate.
6824 * config/i386/i386.c (ix86_split_reduction): Implement
6825 TARGET_VECTORIZE_SPLIT_REDUCTION.
6827 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6829 PR target/83368
6830 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6831 in PIC mode except for TARGET_VXWORKS_RTP.
6832 * config/sparc/sparc.c: Include cfgrtl.h.
6833 (TARGET_INIT_PIC_REG): Define.
6834 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6835 (sparc_pic_register_p): New predicate.
6836 (sparc_legitimate_address_p): Use it.
6837 (sparc_legitimize_pic_address): Likewise.
6838 (sparc_delegitimize_address): Likewise.
6839 (sparc_mode_dependent_address_p): Likewise.
6840 (gen_load_pcrel_sym): Remove 4th parameter.
6841 (load_got_register): Adjust call to above. Remove obsolete stuff.
6842 (sparc_expand_prologue): Do not call load_got_register here.
6843 (sparc_flat_expand_prologue): Likewise.
6844 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6845 (sparc_use_pseudo_pic_reg): New function.
6846 (sparc_init_pic_reg): Likewise.
6847 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6848 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6850 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6852 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6853 Add item for branch_cost.
6855 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6857 PR rtl-optimization/83565
6858 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6859 not extend the result to a larger mode for rotate operations.
6860 (num_sign_bit_copies1): Likewise.
6862 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6864 PR target/40411
6865 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6866 -symbolic.
6867 Use values-Xc.o for -pedantic.
6868 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6870 2018-01-12 Martin Liska <mliska@suse.cz>
6872 PR ipa/83054
6873 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6874 New function.
6875 (possible_polymorphic_call_targets): Use it.
6876 (ipa_devirt): Likewise.
6878 2018-01-12 Martin Liska <mliska@suse.cz>
6880 * profile-count.h (enum profile_quality): Use 0 as invalid
6881 enum value of profile_quality.
6883 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6885 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6886 -mext-string options.
6888 2018-01-12 Richard Biener <rguenther@suse.de>
6890 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6891 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6892 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6893 Likewise.
6894 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6896 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6898 * configure.ac (--with-long-double-format): Add support for the
6899 configuration option to change the default long double format on
6900 PowerPC systems.
6901 * config.gcc (powerpc*-linux*-*): Likewise.
6902 * configure: Regenerate.
6903 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6904 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6905 used without modification.
6907 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6909 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6910 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6911 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6912 MISC_BUILTIN_SPEC_BARRIER.
6913 (rs6000_init_builtins): Likewise.
6914 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6915 enum value.
6916 (speculation_barrier): New define_insn.
6917 * doc/extend.texi: Document __builtin_speculation_barrier.
6919 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6921 PR target/83203
6922 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6923 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6924 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6925 iterators.
6926 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6927 integral modes instead of "ss" and "sd".
6928 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6929 vectors with 32-bit and 64-bit elements.
6930 (vecdupssescalarmodesuffix): New mode attribute.
6931 (vec_dup<mode>): Use it.
6933 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6935 PR target/83330
6936 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6937 frame if argument is passed on stack.
6939 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6941 PR target/82682
6942 * ree.c (combine_reaching_defs): Optimize also
6943 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6944 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6946 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6948 PR middle-end/83189
6949 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6951 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6953 PR middle-end/83718
6954 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6955 after they are computed.
6957 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6959 PR tree-optimization/83695
6960 * gimple-loop-linterchange.cc
6961 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6962 reset cached scev information after interchange.
6963 (pass_linterchange::execute): Remove call to scev_reset_htab.
6965 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6967 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6968 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6969 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6970 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6971 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6972 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6973 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6974 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6975 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6976 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6977 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6978 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6979 (V_lane_reg): Likewise.
6980 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6981 New define_expand.
6982 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6983 (vfmal_lane_low<mode>_intrinsic,
6984 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6985 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6986 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6987 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6988 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6989 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6991 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6993 * config/arm/arm-cpus.in (fp16fml): New feature.
6994 (ALL_SIMD): Add fp16fml.
6995 (armv8.2-a): Add fp16fml as an option.
6996 (armv8.3-a): Likewise.
6997 (armv8.4-a): Add fp16fml as part of fp16.
6998 * config/arm/arm.h (TARGET_FP16FML): Define.
6999 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7000 when appropriate.
7001 * config/arm/arm-modes.def (V2HF): Define.
7002 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7003 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7004 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7005 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7006 vfmsl_low, vfmsl_high): New set of builtins.
7007 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7008 (vfml_op): New code attribute.
7009 (VFMLHALVES): New int iterator.
7010 (VFML, VFMLSEL): New mode attributes.
7011 (V_reg): Define mapping for V2HF.
7012 (V_hi, V_lo): New mode attributes.
7013 (VF_constraint): Likewise.
7014 (vfml_half, vfml_half_selector): New int attributes.
7015 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7016 define_expand.
7017 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7018 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7019 New define_insn.
7020 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7021 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7022 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7023 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7024 documentation.
7025 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7026 Document new effective target and option set.
7028 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7030 * config/arm/arm-cpus.in (armv8_4): New feature.
7031 (ARMv8_4a): New fgroup.
7032 (armv8.4-a): New arch.
7033 * config/arm/arm-tables.opt: Regenerate.
7034 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7035 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7036 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7037 Add matching rules for -march=armv8.4-a and extensions.
7038 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7040 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7042 PR target/81821
7043 * config/rx/rx.md (BW): New mode attribute.
7044 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7046 2018-01-11 Richard Biener <rguenther@suse.de>
7048 PR tree-optimization/83435
7049 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7050 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7051 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7053 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7054 Alan Hayward <alan.hayward@arm.com>
7055 David Sherwood <david.sherwood@arm.com>
7057 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7058 field.
7059 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7060 (aarch64_print_address_internal): Use it to check for a zero offset.
7062 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7063 Alan Hayward <alan.hayward@arm.com>
7064 David Sherwood <david.sherwood@arm.com>
7066 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7067 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7068 Return a poly_int64 rather than a HOST_WIDE_INT.
7069 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7070 rather than a HOST_WIDE_INT.
7071 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7072 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7073 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7074 final_offset from HOST_WIDE_INT to poly_int64.
7075 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7076 to_constant when getting the number of units in an Advanced SIMD
7077 mode.
7078 (aarch64_builtin_vectorized_function): Check for a constant number
7079 of units.
7080 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7081 GET_MODE_SIZE.
7082 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7083 attribute instead of GET_MODE_NUNITS.
7084 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7085 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7086 GET_MODE_SIZE for fixed-size registers.
7087 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7088 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7089 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7090 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7091 (aarch64_print_operand, aarch64_print_address_internal)
7092 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7093 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7094 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7095 Handle polynomial GET_MODE_SIZE.
7096 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7097 wider than SImode without modification.
7098 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7099 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7100 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7101 passing and returning SVE modes.
7102 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7103 rather than GEN_INT.
7104 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7105 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7106 (aarch64_allocate_and_probe_stack_space): Likewise.
7107 (aarch64_layout_frame): Cope with polynomial offsets.
7108 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7109 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7110 polynomial offsets.
7111 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7112 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7113 poly_int64 rather than a HOST_WIDE_INT.
7114 (aarch64_get_separate_components, aarch64_process_components)
7115 (aarch64_expand_prologue, aarch64_expand_epilogue)
7116 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7117 (aarch64_anchor_offset): New function, split out from...
7118 (aarch64_legitimize_address): ...here.
7119 (aarch64_builtin_vectorization_cost): Handle polynomial
7120 TYPE_VECTOR_SUBPARTS.
7121 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7122 GET_MODE_NUNITS.
7123 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7124 number of elements from the PARALLEL rather than the mode.
7125 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7126 rather than GET_MODE_BITSIZE.
7127 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7128 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7129 (aarch64_expand_vec_perm_const_1): Handle polynomial
7130 d->perm.length () and d->perm elements.
7131 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7132 Apply to_constant to d->perm elements.
7133 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7134 polynomial CONST_VECTOR_NUNITS.
7135 (aarch64_move_pointer): Take amount as a poly_int64 rather
7136 than an int.
7137 (aarch64_progress_pointer): Avoid temporary variable.
7138 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7139 the mode attribute instead of GET_MODE.
7141 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7142 Alan Hayward <alan.hayward@arm.com>
7143 David Sherwood <david.sherwood@arm.com>
7145 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7146 x exists before using it.
7147 (aarch64_add_constant_internal): Rename to...
7148 (aarch64_add_offset_1): ...this. Replace regnum with separate
7149 src and dest rtxes. Handle the case in which they're different,
7150 including when the offset is zero. Replace scratchreg with an rtx.
7151 Use 2 additions if there is no spare register into which we can
7152 move a 16-bit constant.
7153 (aarch64_add_constant): Delete.
7154 (aarch64_add_offset): Replace reg with separate src and dest
7155 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7156 Use aarch64_add_offset_1.
7157 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7158 an rtx rather than an int. Take the delta as a poly_int64
7159 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7160 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7161 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7162 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7163 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7164 and aarch64_add_sp.
7165 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7166 aarch64_add_constant.
7168 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7170 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7171 Use scalar_float_mode.
7173 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7175 * config/aarch64/aarch64-simd.md
7176 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7177 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7178 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7179 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7180 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7181 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7182 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7183 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7184 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7185 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7187 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7189 PR target/83514
7190 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7191 targ_options->x_arm_arch_string is non NULL.
7193 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7195 * config/aarch64/aarch64.h
7196 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7198 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7200 PR target/82096
7201 * expmed.c (emit_store_flag_force): Swap if const op0
7202 and change VOIDmode to mode of op0.
7204 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7206 PR rtl-optimization/83761
7207 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7208 than bytes to mode_for_size.
7210 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7212 PR middle-end/83189
7213 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7214 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7215 profile.
7217 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7219 PR middle-end/83575
7220 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7221 when in layout mode.
7222 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7223 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7224 partition fixup.
7226 2018-01-10 Michael Collison <michael.collison@arm.com>
7228 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7229 * config/aarch64/aarch64-option-extension.def: Add
7230 AARCH64_OPT_EXTENSION of 'fp16fml'.
7231 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7232 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7233 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7234 * config/aarch64/constraints.md (Ui7): New constraint.
7235 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7236 (VFMLA_SEL_W): Ditto.
7237 (f16quad): Ditto.
7238 (f16mac1): Ditto.
7239 (VFMLA16_LOW): New int iterator.
7240 (VFMLA16_HIGH): Ditto.
7241 (UNSPEC_FMLAL): New unspec.
7242 (UNSPEC_FMLSL): Ditto.
7243 (UNSPEC_FMLAL2): Ditto.
7244 (UNSPEC_FMLSL2): Ditto.
7245 (f16mac): New code attribute.
7246 * config/aarch64/aarch64-simd-builtins.def
7247 (aarch64_fmlal_lowv2sf): Ditto.
7248 (aarch64_fmlsl_lowv2sf): Ditto.
7249 (aarch64_fmlalq_lowv4sf): Ditto.
7250 (aarch64_fmlslq_lowv4sf): Ditto.
7251 (aarch64_fmlal_highv2sf): Ditto.
7252 (aarch64_fmlsl_highv2sf): Ditto.
7253 (aarch64_fmlalq_highv4sf): Ditto.
7254 (aarch64_fmlslq_highv4sf): Ditto.
7255 (aarch64_fmlal_lane_lowv2sf): Ditto.
7256 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7257 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7258 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7259 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7260 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7261 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7262 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7263 (aarch64_fmlal_lane_highv2sf): Ditto.
7264 (aarch64_fmlsl_lane_highv2sf): Ditto.
7265 (aarch64_fmlal_laneq_highv2sf): Ditto.
7266 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7267 (aarch64_fmlalq_lane_highv4sf): Ditto.
7268 (aarch64_fmlsl_lane_highv4sf): Ditto.
7269 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7270 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7271 * config/aarch64/aarch64-simd.md:
7272 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7273 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7274 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7275 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7276 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7277 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7278 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7279 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7280 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7281 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7282 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7283 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7284 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7285 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7286 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7287 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7288 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7289 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7290 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7291 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7292 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7293 (vfmlsl_low_u32): Ditto.
7294 (vfmlalq_low_u32): Ditto.
7295 (vfmlslq_low_u32): Ditto.
7296 (vfmlal_high_u32): Ditto.
7297 (vfmlsl_high_u32): Ditto.
7298 (vfmlalq_high_u32): Ditto.
7299 (vfmlslq_high_u32): Ditto.
7300 (vfmlal_lane_low_u32): Ditto.
7301 (vfmlsl_lane_low_u32): Ditto.
7302 (vfmlal_laneq_low_u32): Ditto.
7303 (vfmlsl_laneq_low_u32): Ditto.
7304 (vfmlalq_lane_low_u32): Ditto.
7305 (vfmlslq_lane_low_u32): Ditto.
7306 (vfmlalq_laneq_low_u32): Ditto.
7307 (vfmlslq_laneq_low_u32): Ditto.
7308 (vfmlal_lane_high_u32): Ditto.
7309 (vfmlsl_lane_high_u32): Ditto.
7310 (vfmlal_laneq_high_u32): Ditto.
7311 (vfmlsl_laneq_high_u32): Ditto.
7312 (vfmlalq_lane_high_u32): Ditto.
7313 (vfmlslq_lane_high_u32): Ditto.
7314 (vfmlalq_laneq_high_u32): Ditto.
7315 (vfmlslq_laneq_high_u32): Ditto.
7316 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7317 (AARCH64_FL_FOR_ARCH8_4): New.
7318 (AARCH64_ISA_F16FML): New ISA flag.
7319 (TARGET_F16FML): New feature flag for fp16fml.
7320 (doc/invoke.texi): Document new fp16fml option.
7322 2018-01-10 Michael Collison <michael.collison@arm.com>
7324 * config/aarch64/aarch64-builtins.c:
7325 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7326 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7327 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7328 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7329 (AARCH64_ISA_SHA3): New ISA flag.
7330 (TARGET_SHA3): New feature flag for sha3.
7331 * config/aarch64/iterators.md (sha512_op): New int attribute.
7332 (CRYPTO_SHA512): New int iterator.
7333 (UNSPEC_SHA512H): New unspec.
7334 (UNSPEC_SHA512H2): Ditto.
7335 (UNSPEC_SHA512SU0): Ditto.
7336 (UNSPEC_SHA512SU1): Ditto.
7337 * config/aarch64/aarch64-simd-builtins.def
7338 (aarch64_crypto_sha512hqv2di): New builtin.
7339 (aarch64_crypto_sha512h2qv2di): Ditto.
7340 (aarch64_crypto_sha512su0qv2di): Ditto.
7341 (aarch64_crypto_sha512su1qv2di): Ditto.
7342 (aarch64_eor3qv8hi): Ditto.
7343 (aarch64_rax1qv2di): Ditto.
7344 (aarch64_xarqv2di): Ditto.
7345 (aarch64_bcaxqv8hi): Ditto.
7346 * config/aarch64/aarch64-simd.md:
7347 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7348 (aarch64_crypto_sha512su0qv2di): Ditto.
7349 (aarch64_crypto_sha512su1qv2di): Ditto.
7350 (aarch64_eor3qv8hi): Ditto.
7351 (aarch64_rax1qv2di): Ditto.
7352 (aarch64_xarqv2di): Ditto.
7353 (aarch64_bcaxqv8hi): Ditto.
7354 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7355 (vsha512h2q_u64): Ditto.
7356 (vsha512su0q_u64): Ditto.
7357 (vsha512su1q_u64): Ditto.
7358 (veor3q_u16): Ditto.
7359 (vrax1q_u64): Ditto.
7360 (vxarq_u64): Ditto.
7361 (vbcaxq_u16): Ditto.
7362 * config/arm/types.md (crypto_sha512): New type attribute.
7363 (crypto_sha3): Ditto.
7364 (doc/invoke.texi): Document new sha3 option.
7366 2018-01-10 Michael Collison <michael.collison@arm.com>
7368 * config/aarch64/aarch64-builtins.c:
7369 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7370 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7371 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7372 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7373 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7374 (AARCH64_ISA_SM4): New ISA flag.
7375 (TARGET_SM4): New feature flag for sm4.
7376 * config/aarch64/aarch64-simd-builtins.def
7377 (aarch64_sm3ss1qv4si): Ditto.
7378 (aarch64_sm3tt1aq4si): Ditto.
7379 (aarch64_sm3tt1bq4si): Ditto.
7380 (aarch64_sm3tt2aq4si): Ditto.
7381 (aarch64_sm3tt2bq4si): Ditto.
7382 (aarch64_sm3partw1qv4si): Ditto.
7383 (aarch64_sm3partw2qv4si): Ditto.
7384 (aarch64_sm4eqv4si): Ditto.
7385 (aarch64_sm4ekeyqv4si): Ditto.
7386 * config/aarch64/aarch64-simd.md:
7387 (aarch64_sm3ss1qv4si): Ditto.
7388 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7389 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7390 (aarch64_sm4eqv4si): Ditto.
7391 (aarch64_sm4ekeyqv4si): Ditto.
7392 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7393 (sm3part_op): Ditto.
7394 (CRYPTO_SM3TT): Ditto.
7395 (CRYPTO_SM3PART): Ditto.
7396 (UNSPEC_SM3SS1): New unspec.
7397 (UNSPEC_SM3TT1A): Ditto.
7398 (UNSPEC_SM3TT1B): Ditto.
7399 (UNSPEC_SM3TT2A): Ditto.
7400 (UNSPEC_SM3TT2B): Ditto.
7401 (UNSPEC_SM3PARTW1): Ditto.
7402 (UNSPEC_SM3PARTW2): Ditto.
7403 (UNSPEC_SM4E): Ditto.
7404 (UNSPEC_SM4EKEY): Ditto.
7405 * config/aarch64/constraints.md (Ui2): New constraint.
7406 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7407 * config/arm/types.md (crypto_sm3): New type attribute.
7408 (crypto_sm4): Ditto.
7409 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7410 (vsm3tt1aq_u32): Ditto.
7411 (vsm3tt1bq_u32): Ditto.
7412 (vsm3tt2aq_u32): Ditto.
7413 (vsm3tt2bq_u32): Ditto.
7414 (vsm3partw1q_u32): Ditto.
7415 (vsm3partw2q_u32): Ditto.
7416 (vsm4eq_u32): Ditto.
7417 (vsm4ekeyq_u32): Ditto.
7418 (doc/invoke.texi): Document new sm4 option.
7420 2018-01-10 Michael Collison <michael.collison@arm.com>
7422 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7423 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7424 (AARCH64_FL_FOR_ARCH8_4): New.
7425 (AARCH64_FL_V8_4): New flag.
7426 (doc/invoke.texi): Document new armv8.4-a option.
7428 2018-01-10 Michael Collison <michael.collison@arm.com>
7430 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7431 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7432 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7433 * config/aarch64/aarch64-option-extension.def: Add
7434 AARCH64_OPT_EXTENSION of 'sha2'.
7435 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7436 (crypto): Disable sha2 and aes if crypto disabled.
7437 (crypto): Enable aes and sha2 if enabled.
7438 (simd): Disable sha2 and aes if simd disabled.
7439 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7440 New flags.
7441 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7442 (TARGET_SHA2): New feature flag for sha2.
7443 (TARGET_AES): New feature flag for aes.
7444 * config/aarch64/aarch64-simd.md:
7445 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7446 conditional on TARGET_AES.
7447 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7448 (aarch64_crypto_sha1hsi): Make pattern conditional
7449 on TARGET_SHA2.
7450 (aarch64_crypto_sha1hv4si): Ditto.
7451 (aarch64_be_crypto_sha1hv4si): Ditto.
7452 (aarch64_crypto_sha1su1v4si): Ditto.
7453 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7454 (aarch64_crypto_sha1su0v4si): Ditto.
7455 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7456 (aarch64_crypto_sha256su0v4si): Ditto.
7457 (aarch64_crypto_sha256su1v4si): Ditto.
7458 (doc/invoke.texi): Document new aes and sha2 options.
7460 2018-01-10 Martin Sebor <msebor@redhat.com>
7462 PR tree-optimization/83781
7463 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7464 as string arrays.
7466 2018-01-11 Martin Sebor <msebor@gmail.com>
7467 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7469 PR tree-optimization/83501
7470 PR tree-optimization/81703
7472 * tree-ssa-strlen.c (get_string_cst): Rename...
7473 (get_string_len): ...to this. Handle global constants.
7474 (handle_char_store): Adjust.
7476 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7477 Jim Wilson <jimw@sifive.com>
7479 * config/riscv/riscv-protos.h (riscv_output_return): New.
7480 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7481 (riscv_attribute_table, riscv_output_return),
7482 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7483 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7484 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7485 function.
7486 (riscv_expand_prologue): Add early return for naked function.
7487 (riscv_expand_epilogue): Likewise.
7488 (riscv_function_ok_for_sibcall): Return false for naked function.
7489 (riscv_set_current_function): New.
7490 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7491 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7492 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7493 * doc/extend.texi (RISC-V Function Attributes): New.
7495 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7497 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7498 check for 128-bit long double before checking TCmode.
7499 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7500 128-bit long doubles before checking TFmode or TCmode.
7501 (FLOAT128_IBM_P): Likewise.
7503 2018-01-10 Martin Sebor <msebor@redhat.com>
7505 PR tree-optimization/83671
7506 * builtins.c (c_strlen): Unconditionally return zero for the empty
7507 string.
7508 Use -Warray-bounds for warnings.
7509 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7510 for non-constant array indices with COMPONENT_REF, arrays of
7511 arrays, and pointers to arrays.
7512 (gimple_fold_builtin_strlen): Determine and set length range for
7513 non-constant character arrays.
7515 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7517 PR middle-end/81897
7518 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7519 empty blocks.
7521 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7523 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7525 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7527 PR target/83399
7528 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7529 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7530 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7531 indexed_or_indirect_operand predicate.
7532 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7533 (*vsx_le_perm_load_v8hi): Likewise.
7534 (*vsx_le_perm_load_v16qi): Likewise.
7535 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7536 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7537 (*vsx_le_perm_store_v8hi): Likewise.
7538 (*vsx_le_perm_store_v16qi): Likewise.
7539 (eight unnamed splitters): Likewise.
7541 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7543 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7544 * config/rs6000/emmintrin.h: Likewise.
7545 * config/rs6000/mmintrin.h: Likewise.
7546 * config/rs6000/xmmintrin.h: Likewise.
7548 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7550 PR c++/43486
7551 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7552 "public_flag".
7553 * tree.c (tree_nop_conversion): Return true for location wrapper
7554 nodes.
7555 (maybe_wrap_with_location): New function.
7556 (selftest::check_strip_nops): New function.
7557 (selftest::test_location_wrappers): New function.
7558 (selftest::tree_c_tests): Call it.
7559 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7560 (maybe_wrap_with_location): New decl.
7561 (EXPR_LOCATION_WRAPPER_P): New macro.
7562 (location_wrapper_p): New inline function.
7563 (tree_strip_any_location_wrapper): New inline function.
7565 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7567 PR target/83735
7568 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7569 stack_realign_offset for the largest alignment of stack slot
7570 actually used.
7571 (ix86_find_max_used_stack_alignment): New function.
7572 (ix86_finalize_stack_frame_flags): Use it. Set
7573 max_used_stack_alignment if we don't realign stack.
7574 * config/i386/i386.h (machine_function): Add
7575 max_used_stack_alignment.
7577 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7579 * config/arm/arm.opt (-mbranch-cost): New option.
7580 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7581 account.
7583 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7585 PR target/83629
7586 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7587 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7589 2018-01-10 Richard Biener <rguenther@suse.de>
7591 PR debug/83765
7592 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7593 early out so it also covers the case where we have a non-NULL
7594 origin.
7596 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7598 PR tree-optimization/83753
7599 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7600 for non-strided grouped accesses if the number of elements is 1.
7602 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7604 PR target/81616
7605 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7606 * i386.h (TARGET_USE_GATHER): Define.
7607 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7609 2018-01-10 Martin Liska <mliska@suse.cz>
7611 PR bootstrap/82831
7612 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7613 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7614 partitioning.
7615 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7616 CLEANUP_NO_PARTITIONING is not set.
7618 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7620 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7621 for vectors, as a partial revert of r254296.
7622 * rtl.h (const_vec_p): Delete.
7623 (const_vec_duplicate_p): Don't test for vector CONSTs.
7624 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7625 * expmed.c (make_tree): Likewise.
7627 Revert:
7628 * common.md (E, F): Use CONSTANT_P instead of checking for
7629 CONST_VECTOR.
7630 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7631 checking for CONST_VECTOR.
7633 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7635 PR middle-end/83575
7636 * predict.c (force_edge_cold): Handle in more sane way edges
7637 with no prediction.
7639 2018-01-09 Carl Love <cel@us.ibm.com>
7641 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7642 V4SI, V4SF types.
7643 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7644 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7645 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7646 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7647 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7648 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7649 * config/rs6000/rs6000-protos.h: Add extern defition for
7650 rs6000_generate_float2_double_code.
7651 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7652 function.
7653 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7654 (float2_v2df): Add define_expand.
7656 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7658 PR target/83628
7659 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7660 op_mode in the force_to_mode call.
7662 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7664 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7665 instead of checking each element individually.
7666 (aarch64_evpc_uzp): Likewise.
7667 (aarch64_evpc_zip): Likewise.
7668 (aarch64_evpc_ext): Likewise.
7669 (aarch64_evpc_rev): Likewise.
7670 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7671 instead of checking each element individually. Return true without
7672 generating rtl if
7673 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7674 whether all selected elements come from the same input, instead of
7675 checking each element individually. Remove calls to gen_rtx_REG,
7676 start_sequence and end_sequence and instead assert that no rtl is
7677 generated.
7679 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7681 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7682 order of HIGH and CONST checks.
7684 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7686 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7687 if the destination isn't an SSA_NAME.
7689 2018-01-09 Richard Biener <rguenther@suse.de>
7691 PR tree-optimization/83668
7692 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7693 move prologue...
7694 (canonicalize_loop_form): ... here, renamed from ...
7695 (canonicalize_loop_closed_ssa_form): ... this and amended to
7696 swap successor edges for loop exit blocks to make us use
7697 the RPO order we need for initial schedule generation.
7699 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7701 PR tree-optimization/64811
7702 * match.pd: When optimizing comparisons with Inf, avoid
7703 introducing or losing exceptions from comparisons with NaN.
7705 2018-01-09 Martin Liska <mliska@suse.cz>
7707 PR sanitizer/82517
7708 * asan.c (shadow_mem_size): Add gcc_assert.
7710 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7712 Don't save registers in main().
7714 PR target/83738
7715 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7716 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7717 * config/avr/avr.c (avr_set_current_function): Don't error if
7718 naked, OS_task or OS_main are specified at the same time.
7719 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7720 OS_main.
7721 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7722 attribute.
7723 * common/config/avr/avr-common.c (avr_option_optimization_table):
7724 Switch on -mmain-is-OS_task for optimizing compilations.
7726 2018-01-09 Richard Biener <rguenther@suse.de>
7728 PR tree-optimization/83572
7729 * graphite.c: Include cfganal.h.
7730 (graphite_transform_loops): Connect infinite loops to exit
7731 and remove fake edges at the end.
7733 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7735 * ipa-inline.c (edge_badness): Revert accidental checkin.
7737 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7739 PR ipa/80763
7740 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7741 symbols; not inline clones.
7743 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7745 PR target/83507
7746 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7747 hard registers. Formatting fixes.
7749 PR preprocessor/83722
7750 * gcc.c (try_generate_repro): Pass
7751 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7752 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7753 do_report_bug.
7755 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7756 Kito Cheng <kito.cheng@gmail.com>
7758 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7759 (riscv_leaf_function_p): Delete.
7760 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7762 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7764 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7765 function.
7766 (do_ifelse): New function.
7767 (do_isel): New function.
7768 (do_sub3): New function.
7769 (do_add3): New function.
7770 (do_load_mask_compare): New function.
7771 (do_overlap_load_compare): New function.
7772 (expand_compare_loop): New function.
7773 (expand_block_compare): Call expand_compare_loop() when appropriate.
7774 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7775 option description.
7776 (-mblock-compare-inline-loop-limit): New option.
7778 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7780 PR target/83677
7781 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7782 Reverse order of second and third operands in first alternative.
7783 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7784 of first and second elements in UNSPEC_VPERMR vector.
7785 (altivec_expand_vec_perm_le): Likewise.
7787 2018-01-08 Jeff Law <law@redhat.com>
7789 PR rtl-optimizatin/81308
7790 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7791 (process_switch): If group_case_labels makes a change, then set
7792 cfg_altered.
7793 (pass_convert_switch::execute): If a switch is converted, then
7794 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7796 PR rtl-optimization/81308
7797 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7798 splitting insns.
7800 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7802 PR target/83663 - Revert r255946
7803 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7804 generation for cases where splatting a value is not useful.
7805 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7806 across a vec_duplicate and a paradoxical subreg forming a vector
7807 mode to a vec_concat.
7809 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7811 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7812 -march=armv8.3-a variants.
7813 * config/arm/t-multilib: Likewise.
7814 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7816 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7818 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7819 to generate rtl.
7820 (cceq_ior_compare_complement): Give it a name so I can use it, and
7821 change boolean_or_operator predicate to boolean_operator so it can
7822 be used to generate a crand.
7823 (eqne): New code iterator.
7824 (bd/bd_neg): New code_attrs.
7825 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7826 a single define_insn.
7827 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7828 decrement (bdnzt/bdnzf/bdzt/bdzf).
7829 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7830 with the new names of the branch decrement patterns, and added the
7831 names of the branch decrement conditional patterns.
7833 2018-01-08 Richard Biener <rguenther@suse.de>
7835 PR tree-optimization/83563
7836 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7837 cache.
7839 2018-01-08 Richard Biener <rguenther@suse.de>
7841 PR middle-end/83713
7842 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7844 2018-01-08 Richard Biener <rguenther@suse.de>
7846 PR tree-optimization/83685
7847 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7848 references to abnormals.
7850 2018-01-08 Richard Biener <rguenther@suse.de>
7852 PR lto/83719
7853 * dwarf2out.c (output_indirect_strings): Handle empty
7854 skeleton_debug_str_hash.
7855 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7857 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7859 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7860 (emit_store_direct): Likewise.
7861 (arc_trampoline_adjust_address): Likewise.
7862 (arc_asm_trampoline_template): New function.
7863 (arc_initialize_trampoline): Use asm_trampoline_template.
7864 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7865 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7866 * config/arc/arc.md (flush_icache): Delete pattern.
7868 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7870 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7871 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7872 munaligned-access.
7874 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7876 PR target/83681
7877 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7878 by not USED_FOR_TARGET.
7879 (make_pass_resolve_sw_modes): Likewise.
7881 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7883 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7884 USED_FOR_TARGET.
7886 2018-01-08 Richard Biener <rguenther@suse.de>
7888 PR middle-end/83580
7889 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7891 2018-01-08 Richard Biener <rguenther@suse.de>
7893 PR middle-end/83517
7894 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7896 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7898 PR middle-end/81897
7899 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7900 basic blocks with a small number of successors.
7901 (convert_control_dep_chain_into_preds): Improve handling of
7902 forwarder blocks.
7903 (dump_predicates): Split apart into...
7904 (dump_pred_chain): ...here...
7905 (dump_pred_info): ...and here.
7906 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7907 (can_chain_union_be_invalidated_p): Improve check for invalidation
7908 of paths.
7909 (uninit_uses_cannot_happen): Avoid unnecessary if
7910 convert_control_dep_chain_into_preds yielded nothing.
7912 2018-01-06 Martin Sebor <msebor@redhat.com>
7914 PR tree-optimization/83640
7915 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7916 subtracting negative offset from size.
7917 (builtin_access::overlap): Adjust offset bounds of the access to fall
7918 within the size of the object if possible.
7920 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7922 PR rtl-optimization/83699
7923 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7924 extract_bit_field_as_subreg to cases in which the extracted
7925 value is also a vector.
7927 * lra-constraints.c (process_alt_operands): Test for the equivalence
7928 substitutions when detecting a possible reload cycle.
7930 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7932 PR debug/83480
7933 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7934 by default if flag_selective_schedling{,2}. Formatting fixes.
7936 PR rtl-optimization/83682
7937 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7938 if it has non-VECTOR_MODE element mode.
7939 (vec_duplicate_p): Likewise.
7941 PR middle-end/83694
7942 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7943 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7945 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7947 PR target/83604
7948 * config/i386/i386-builtin.def
7949 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7950 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7951 Require also OPTION_MASK_ISA_AVX512F in addition to
7952 OPTION_MASK_ISA_GFNI.
7953 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7954 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7955 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7956 to OPTION_MASK_ISA_GFNI.
7957 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7958 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7959 OPTION_MASK_ISA_AVX512BW.
7960 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7961 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7962 addition to OPTION_MASK_ISA_GFNI.
7963 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7964 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7965 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7966 to OPTION_MASK_ISA_GFNI.
7967 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7968 a requirement for all ISAs rather than any of them with a few
7969 exceptions.
7970 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7971 processing.
7972 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7973 bitmasks to be enabled with 3 exceptions, instead of requiring any
7974 enabled ISA with lots of exceptions.
7975 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7976 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7977 Change avx512bw in isa attribute to avx512f.
7978 * config/i386/sgxintrin.h: Add license boilerplate.
7979 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7980 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7981 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7982 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7983 defined.
7984 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7985 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7986 temporarily sse2 rather than sse if not enabled already.
7988 PR target/83604
7989 * config/i386/sse.md (VI248_VLBW): Rename to ...
7990 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7991 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7992 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7993 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7994 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7995 mode iterator instead of VI248_VLBW.
7997 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7999 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8000 (record_modified): Skip clobbers; add debug output.
8001 (param_change_prob): Use sreal frequencies.
8003 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8005 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8006 punt for user-aligned variables.
8008 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8010 * tree-chrec.c (chrec_contains_symbols): Return true for
8011 POLY_INT_CST.
8013 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8015 PR target/82439
8016 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8017 of (x|y) == x for BICS pattern.
8019 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8021 PR tree-optimization/83605
8022 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8023 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8024 can throw.
8026 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8028 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8029 * config/epiphany/rtems.h: New file.
8031 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8032 Uros Bizjak <ubizjak@gmail.com>
8034 PR target/83554
8035 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8036 QIreg_operand instead of register_operand predicate.
8037 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8038 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8039 comments instead of -fmitigate[-_]rop.
8041 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8043 PR bootstrap/81926
8044 * cgraphunit.c (symbol_table::compile): Switch to text_section
8045 before calling assembly_start debug hook.
8046 * run-rtl-passes.c (run_rtl_passes): Likewise.
8047 Include output.h.
8049 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8051 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8052 range_int_cst_p rather than !symbolic_range_p before calling
8053 extract_range_from_multiplicative_op_1.
8055 2018-01-04 Jeff Law <law@redhat.com>
8057 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8058 redundant test in assertion.
8060 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8062 * doc/rtl.texi: Document machine_mode wrapper classes.
8064 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8066 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8067 using tree_to_uhwi.
8069 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8071 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8072 the VEC_PERM_EXPR fold to fail.
8074 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8076 PR debug/83585
8077 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8078 to switched_sections.
8080 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8082 PR target/83680
8083 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8084 test for d.testing.
8086 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8088 PR target/83387
8089 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8090 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8092 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8094 PR debug/83666
8095 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8096 is BLKmode and bitpos not zero or mode change is needed.
8098 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8100 PR target/83675
8101 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8102 TARGET_VIS2.
8104 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8106 PR target/83628
8107 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8108 instead of MULT rtx. Update all corresponding splitters.
8109 (*saddl_se): Ditto.
8110 (*ssub<modesuffix>): Ditto.
8111 (*ssubl_se): Ditto.
8112 (*cmp_sadd_di): Update split patterns.
8113 (*cmp_sadd_si): Ditto.
8114 (*cmp_sadd_sidi): Ditto.
8115 (*cmp_ssub_di): Ditto.
8116 (*cmp_ssub_si): Ditto.
8117 (*cmp_ssub_sidi): Ditto.
8118 * config/alpha/predicates.md (const23_operand): New predicate.
8119 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8120 Look for ASHIFT, not MULT inner operand.
8121 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8123 2018-01-04 Martin Liska <mliska@suse.cz>
8125 PR gcov-profile/83669
8126 * gcov.c (output_intermediate_file): Add version to intermediate
8127 gcov file.
8128 * doc/gcov.texi: Document new field 'version' in intermediate
8129 file format. Fix location of '-k' option of gcov command.
8131 2018-01-04 Martin Liska <mliska@suse.cz>
8133 PR ipa/82352
8134 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8136 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8138 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8140 2018-01-03 Martin Sebor <msebor@redhat.com>
8142 PR tree-optimization/83655
8143 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8144 checking calls with invalid arguments.
8146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8148 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8149 (vectorizable_mask_load_store): Delete.
8150 (vectorizable_call): Return false for masked loads and stores.
8151 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8152 instead of gimple_assign_rhs1.
8153 (vectorizable_load): Handle IFN_MASK_LOAD.
8154 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8158 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8159 split out from..,
8160 (vectorizable_mask_load_store): ...here.
8161 (vectorizable_load): ...and here.
8163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8165 * tree-vect-stmts.c (vect_build_all_ones_mask)
8166 (vect_build_zero_merge_argument): New functions, split out from...
8167 (vectorizable_load): ...here.
8169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8171 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8172 split out from...
8173 (vectorizable_mask_load_store): ...here.
8174 (vectorizable_store): ...and here.
8176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8178 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8179 split out from...
8180 (vectorizable_mask_load_store): ...here.
8182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8184 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8185 (vect_model_store_cost): Take a vec_load_store_type instead of a
8186 vect_def_type.
8187 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8188 (vect_model_store_cost): Take a vec_load_store_type instead of a
8189 vect_def_type.
8190 (vectorizable_mask_load_store): Update accordingly.
8191 (vectorizable_store): Likewise.
8192 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8196 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8197 IFN_MASK_LOAD calls here rather than...
8198 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8201 Alan Hayward <alan.hayward@arm.com>
8202 David Sherwood <david.sherwood@arm.com>
8204 * expmed.c (extract_bit_field_1): For vector extracts,
8205 fall back to extract_bit_field_as_subreg if vec_extract
8206 isn't available.
8208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8209 Alan Hayward <alan.hayward@arm.com>
8210 David Sherwood <david.sherwood@arm.com>
8212 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8213 they are variable or constant sized.
8214 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8215 slots for constant-sized data.
8217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8218 Alan Hayward <alan.hayward@arm.com>
8219 David Sherwood <david.sherwood@arm.com>
8221 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8222 handling COND_EXPRs with boolean comparisons, try to find a better
8223 basis for the mask type than the boolean itself.
8225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8227 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8228 is calculated and how it can be overridden.
8229 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8230 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8231 if defined.
8232 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8233 if nonzero.
8235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8236 Alan Hayward <alan.hayward@arm.com>
8237 David Sherwood <david.sherwood@arm.com>
8239 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8240 Remove the mode argument.
8241 (aarch64_simd_valid_immediate): Remove the mode and inverse
8242 arguments.
8243 * config/aarch64/iterators.md (bitsize): New iterator.
8244 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8245 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8246 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8247 aarch64_simd_valid_immediate.
8248 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8249 (aarch64_reg_or_bic_imm): Likewise.
8250 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8251 with an insn_type enum and msl with a modifier_type enum.
8252 Replace element_width with a scalar_mode. Change the shift
8253 to unsigned int. Add constructors for scalar_float_mode and
8254 scalar_int_mode elements.
8255 (aarch64_vect_float_const_representable_p): Delete.
8256 (aarch64_can_const_movi_rtx_p)
8257 (aarch64_simd_scalar_immediate_valid_for_move)
8258 (aarch64_simd_make_constant): Update call to
8259 aarch64_simd_valid_immediate.
8260 (aarch64_advsimd_valid_immediate_hs): New function.
8261 (aarch64_advsimd_valid_immediate): Likewise.
8262 (aarch64_simd_valid_immediate): Remove mode and inverse
8263 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8264 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8265 and aarch64_float_const_representable_p on the result.
8266 (aarch64_output_simd_mov_immediate): Remove mode argument.
8267 Update call to aarch64_simd_valid_immediate and use of
8268 simd_immediate_info.
8269 (aarch64_output_scalar_simd_mov_immediate): Update call
8270 accordingly.
8272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8273 Alan Hayward <alan.hayward@arm.com>
8274 David Sherwood <david.sherwood@arm.com>
8276 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8277 (mode_nunits): Likewise CONST_MODE_NUNITS.
8278 * machmode.def (ADJUST_NUNITS): Document.
8279 * genmodes.c (mode_data::need_nunits_adj): New field.
8280 (blank_mode): Update accordingly.
8281 (adj_nunits): New variable.
8282 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8283 parameter.
8284 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8285 listed in adj_nunits.
8286 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8287 listed in adj_nunits. Don't emit case statements for such modes.
8288 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8289 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8290 nothing if adj_nunits is nonnull.
8291 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8292 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8293 (emit_mode_fbit): Update use of print_maybe_const_decl.
8294 (emit_move_size): Likewise. Treat the array as non-const
8295 if adj_nunits.
8296 (emit_mode_adjustments): Handle adj_nunits.
8298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8300 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8301 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8302 (VECTOR_MODES): Use it.
8303 (make_vector_modes): Take the prefix as an argument.
8305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8306 Alan Hayward <alan.hayward@arm.com>
8307 David Sherwood <david.sherwood@arm.com>
8309 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8310 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8311 for MODE_VECTOR_BOOL.
8312 * machmode.def (VECTOR_BOOL_MODE): Document.
8313 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8314 (make_vector_bool_mode): New function.
8315 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8316 MODE_VECTOR_BOOL.
8317 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8318 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8319 Likewise.
8320 * stor-layout.c (int_mode_for_mode): Likewise.
8321 * tree.c (build_vector_type_for_mode): Likewise.
8322 * varasm.c (output_constant_pool_2): Likewise.
8323 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8324 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8325 for MODE_VECTOR_BOOL.
8326 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8327 of mode class checks.
8328 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8329 instead of a list of mode class checks.
8330 (expand_vector_scalar_condition): Likewise.
8331 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8334 Alan Hayward <alan.hayward@arm.com>
8335 David Sherwood <david.sherwood@arm.com>
8337 * machmode.h (mode_size): Change from unsigned short to
8338 poly_uint16_pod.
8339 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8340 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8341 or if measurement_type is not polynomial.
8342 (fixed_size_mode::includes_p): Check for constant-sized modes.
8343 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8344 return a poly_uint16 rather than an unsigned short.
8345 (emit_mode_size): Change the type of mode_size from unsigned short
8346 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8347 (emit_mode_adjustments): Cope with polynomial vector sizes.
8348 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8349 for GET_MODE_SIZE.
8350 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8351 for GET_MODE_SIZE.
8352 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8353 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8354 * caller-save.c (setup_save_areas): Likewise.
8355 (replace_reg_with_saved_mem): Likewise.
8356 * calls.c (emit_library_call_value_1): Likewise.
8357 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8358 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8359 (gen_lowpart_for_combine): Likewise.
8360 * convert.c (convert_to_integer_1): Likewise.
8361 * cse.c (equiv_constant, cse_insn): Likewise.
8362 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8363 (cselib_subst_to_values): Likewise.
8364 * dce.c (word_dce_process_block): Likewise.
8365 * df-problems.c (df_word_lr_mark_ref): Likewise.
8366 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8367 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8368 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8369 (rtl_for_decl_location): Likewise.
8370 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8371 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8372 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8373 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8374 (expand_expr_real_1): Likewise.
8375 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8376 (pad_below): Likewise.
8377 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8378 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8379 * ira.c (get_subreg_tracking_sizes): Likewise.
8380 * ira-build.c (ira_create_allocno_objects): Likewise.
8381 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8382 (ira_sort_regnos_for_alter_reg): Likewise.
8383 * ira-costs.c (record_operand_costs): Likewise.
8384 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8385 (resolve_simple_move): Likewise.
8386 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8387 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8388 (lra_constraints): Likewise.
8389 (CONST_POOL_OK_P): Reject variable-sized modes.
8390 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8391 (add_pseudo_to_slot, lra_spill): Likewise.
8392 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8393 * optabs-query.c (get_best_extraction_insn): Likewise.
8394 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8395 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8396 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8397 * recog.c (offsettable_address_addr_space_p): Likewise.
8398 * regcprop.c (maybe_mode_change): Likewise.
8399 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8400 * regrename.c (build_def_use): Likewise.
8401 * regstat.c (dump_reg_info): Likewise.
8402 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8403 (find_reloads, find_reloads_subreg_address): Likewise.
8404 * reload1.c (eliminate_regs_1): Likewise.
8405 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8406 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8407 (simplify_binary_operation_1, simplify_subreg): Likewise.
8408 * targhooks.c (default_function_arg_padding): Likewise.
8409 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8410 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8411 (verify_gimple_assign_ternary): Likewise.
8412 * tree-inline.c (estimate_move_cost): Likewise.
8413 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8414 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8415 (get_address_cost_ainc): Likewise.
8416 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8417 (vect_supportable_dr_alignment): Likewise.
8418 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8419 (vectorizable_reduction): Likewise.
8420 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8421 (vectorizable_operation, vectorizable_load): Likewise.
8422 * tree.c (build_same_sized_truth_vector_type): Likewise.
8423 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8424 * var-tracking.c (emit_note_insn_var_location): Likewise.
8425 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8426 (ADDR_VEC_ALIGN): Likewise.
8428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8429 Alan Hayward <alan.hayward@arm.com>
8430 David Sherwood <david.sherwood@arm.com>
8432 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8433 unsigned short.
8434 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8435 or if measurement_type is polynomial.
8436 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8437 * combine.c (make_extraction): Likewise.
8438 * dse.c (find_shift_sequence): Likewise.
8439 * dwarf2out.c (mem_loc_descriptor): Likewise.
8440 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8441 (extract_bit_field, extract_low_bits): Likewise.
8442 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8443 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8444 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8445 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8446 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8447 * reload.c (find_reloads): Likewise.
8448 * reload1.c (alter_reg): Likewise.
8449 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8450 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8451 * tree-if-conv.c (predicate_mem_writes): Likewise.
8452 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8453 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8454 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8455 * valtrack.c (dead_debug_insert_temp): Likewise.
8456 * varasm.c (mergeable_constant_section): Likewise.
8457 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8460 Alan Hayward <alan.hayward@arm.com>
8461 David Sherwood <david.sherwood@arm.com>
8463 * expr.c (expand_assignment): Cope with polynomial mode sizes
8464 when assigning to a CONCAT.
8466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8467 Alan Hayward <alan.hayward@arm.com>
8468 David Sherwood <david.sherwood@arm.com>
8470 * machmode.h (mode_precision): Change from unsigned short to
8471 poly_uint16_pod.
8472 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8473 short.
8474 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8475 or if measurement_type is not polynomial.
8476 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8477 in which the mode is already known to be a scalar_int_mode.
8478 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8479 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8480 initializer.
8481 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8482 for GET_MODE_PRECISION.
8483 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8484 for GET_MODE_PRECISION.
8485 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8486 as polynomial.
8487 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8488 (expand_field_assignment, make_extraction): Likewise.
8489 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8490 (get_last_value): Likewise.
8491 * convert.c (convert_to_integer_1): Likewise.
8492 * cse.c (cse_insn): Likewise.
8493 * expr.c (expand_expr_real_1): Likewise.
8494 * lra-constraints.c (simplify_operand_subreg): Likewise.
8495 * optabs-query.c (can_atomic_load_p): Likewise.
8496 * optabs.c (expand_atomic_load): Likewise.
8497 (expand_atomic_store): Likewise.
8498 * ree.c (combine_reaching_defs): Likewise.
8499 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8500 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8501 * tree.h (type_has_mode_precision_p): Likewise.
8502 * ubsan.c (instrument_si_overflow): Likewise.
8504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8505 Alan Hayward <alan.hayward@arm.com>
8506 David Sherwood <david.sherwood@arm.com>
8508 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8509 polynomial numbers of units.
8510 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8511 (valid_vector_subparts_p): New function.
8512 (build_vector_type): Remove temporary shim and take the number
8513 of units as a poly_uint64 rather than an int.
8514 (build_opaque_vector_type): Take the number of units as a
8515 poly_uint64 rather than an int.
8516 * tree.c (build_vector_from_ctor): Handle polynomial
8517 TYPE_VECTOR_SUBPARTS.
8518 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8519 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8520 (build_vector_from_val): If the number of units is variable,
8521 use build_vec_duplicate_cst for constant operands and
8522 VEC_DUPLICATE_EXPR otherwise.
8523 (make_vector_type): Remove temporary is_constant ().
8524 (build_vector_type, build_opaque_vector_type): Take the number of
8525 units as a poly_uint64 rather than an int.
8526 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8527 VECTOR_CST_NELTS.
8528 * cfgexpand.c (expand_debug_expr): Likewise.
8529 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8530 (store_constructor, expand_expr_real_1): Likewise.
8531 (const_scalar_mask_from_tree): Likewise.
8532 * fold-const-call.c (fold_const_reduction): Likewise.
8533 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8534 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8535 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8536 (fold_relational_const): Likewise.
8537 (native_interpret_vector): Likewise. Change the size from an
8538 int to an unsigned int.
8539 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8540 TYPE_VECTOR_SUBPARTS.
8541 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8542 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8543 duplicating a non-constant operand into a variable-length vector.
8544 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8545 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8546 * ipa-icf.c (sem_variable::equals): Likewise.
8547 * match.pd: Likewise.
8548 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8549 * print-tree.c (print_node): Likewise.
8550 * stor-layout.c (layout_type): Likewise.
8551 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8552 * tree-cfg.c (verify_gimple_comparison): Likewise.
8553 (verify_gimple_assign_binary): Likewise.
8554 (verify_gimple_assign_ternary): Likewise.
8555 (verify_gimple_assign_single): Likewise.
8556 * tree-pretty-print.c (dump_generic_node): Likewise.
8557 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8558 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8559 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8560 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8561 (vect_shift_permute_load_chain): Likewise.
8562 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8563 (expand_vector_condition, optimize_vector_constructor): Likewise.
8564 (lower_vec_perm, get_compute_type): Likewise.
8565 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8566 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8567 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8568 (vect_recog_mask_conversion_pattern): Likewise.
8569 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8570 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8571 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8572 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8573 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8574 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8575 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8576 (supportable_widening_operation): Likewise.
8577 (supportable_narrowing_operation): Likewise.
8578 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8579 Likewise.
8580 * varasm.c (output_constant): 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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8587 so that both the length == 3 and length != 3 cases set up their
8588 own permute vectors. Add comments explaining why we know the
8589 number of elements is constant.
8590 (vect_permute_load_chain): Likewise.
8592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8593 Alan Hayward <alan.hayward@arm.com>
8594 David Sherwood <david.sherwood@arm.com>
8596 * machmode.h (mode_nunits): Change from unsigned char to
8597 poly_uint16_pod.
8598 (ONLY_FIXED_SIZE_MODES): New macro.
8599 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8600 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8601 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8602 New typedefs.
8603 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8604 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8605 or if measurement_type is not polynomial.
8606 * genmodes.c (ZERO_COEFFS): New macro.
8607 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8608 poly_uint16.
8609 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8610 Use ZERO_COEFFS when emitting initializers.
8611 * data-streamer.h (bp_pack_poly_value): New function.
8612 (bp_unpack_poly_value): Likewise.
8613 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8614 for GET_MODE_NUNITS.
8615 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8616 for GET_MODE_NUNITS.
8617 * tree.c (make_vector_type): Remove temporary shim and make
8618 the real function take the number of units as a poly_uint64
8619 rather than an int.
8620 (build_vector_type_for_mode): Handle polynomial nunits.
8621 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8622 * emit-rtl.c (const_vec_series_p_1): Likewise.
8623 (gen_rtx_CONST_VECTOR): Likewise.
8624 * fold-const.c (test_vec_duplicate_folding): Likewise.
8625 * genrecog.c (validate_pattern): Likewise.
8626 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8627 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8628 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8629 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8630 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8631 * rtlanal.c (subreg_get_info): Likewise.
8632 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8633 (vect_grouped_load_supported): Likewise.
8634 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8635 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8636 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8637 (simplify_const_unary_operation, simplify_binary_operation_1)
8638 (simplify_const_binary_operation, simplify_ternary_operation)
8639 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8640 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8641 instead of CONST_VECTOR_NUNITS.
8642 * varasm.c (output_constant_pool_2): Likewise.
8643 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8644 explicit-encoded elements in the XVEC for variable-length vectors.
8646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8648 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8651 Alan Hayward <alan.hayward@arm.com>
8652 David Sherwood <david.sherwood@arm.com>
8654 * coretypes.h (fixed_size_mode): Declare.
8655 (fixed_size_mode_pod): New typedef.
8656 * builtins.h (target_builtins::x_apply_args_mode)
8657 (target_builtins::x_apply_result_mode): Change type to
8658 fixed_size_mode_pod.
8659 * builtins.c (apply_args_size, apply_result_size, result_vector)
8660 (expand_builtin_apply_args_1, expand_builtin_apply)
8661 (expand_builtin_return): Update accordingly.
8663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8665 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8666 * cselib.c (cselib_hash_rtx): Likewise.
8667 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8668 CONST_VECTOR encoding.
8670 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8671 Jeff Law <law@redhat.com>
8673 PR target/83641
8674 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8675 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8676 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8677 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8679 PR target/83641
8680 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8681 explicitly probe *sp in a noreturn function if there were any callee
8682 register saves or frame pointer is needed.
8684 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8686 PR debug/83621
8687 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8688 BLKmode for ternary, binary or unary expressions.
8690 PR debug/83645
8691 * var-tracking.c (delete_vta_debug_insn): New inline function.
8692 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8693 insns from get_insns () to NULL instead of each bb separately.
8694 Use delete_vta_debug_insn. No longer static.
8695 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8696 delete_vta_debug_insns callers.
8697 * rtl.h (delete_vta_debug_insns): Declare.
8698 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8699 instead of variable_tracking_main.
8701 2018-01-03 Martin Sebor <msebor@redhat.com>
8703 PR tree-optimization/83603
8704 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8705 arguments past the endof the argument list in functions declared
8706 without a prototype.
8707 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8708 Avoid checking when arguments are null.
8710 2018-01-03 Martin Sebor <msebor@redhat.com>
8712 PR c/83559
8713 * doc/extend.texi (attribute const): Fix a typo.
8714 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8715 issuing -Wsuggest-attribute for void functions.
8717 2018-01-03 Martin Sebor <msebor@redhat.com>
8719 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8720 offset_int::from instead of wide_int::to_shwi.
8721 (maybe_diag_overlap): Remove assertion.
8722 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8723 * gimple-ssa-sprintf.c (format_directive): Same.
8724 (parse_directive): Same.
8725 (sprintf_dom_walker::compute_format_length): Same.
8726 (try_substitute_return_value): Same.
8728 2018-01-03 Jeff Law <law@redhat.com>
8730 PR middle-end/83654
8731 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8732 non-constant residual for zero at runtime and avoid probing in
8733 that case. Reorganize code for trailing problem to mirror handling
8734 of the residual.
8736 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8738 PR tree-optimization/83501
8739 * tree-ssa-strlen.c (get_string_cst): New.
8740 (handle_char_store): Call get_string_cst.
8742 2018-01-03 Martin Liska <mliska@suse.cz>
8744 PR tree-optimization/83593
8745 * tree-ssa-strlen.c: Include tree-cfg.h.
8746 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8747 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8748 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8749 to false.
8750 (strlen_dom_walker::before_dom_children): Call
8751 gimple_purge_dead_eh_edges. Dump tranformation with details
8752 dump flags.
8753 (strlen_dom_walker::before_dom_children): Update call by adding
8754 new argument cleanup_eh.
8755 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8757 2018-01-03 Martin Liska <mliska@suse.cz>
8759 PR ipa/83549
8760 * cif-code.def (VARIADIC_THUNK): New enum value.
8761 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8762 thunks.
8764 2018-01-03 Jan Beulich <jbeulich@suse.com>
8766 * sse.md (mov<mode>_internal): Tighten condition for when to use
8767 vmovdqu<ssescalarsize> for TI and OI modes.
8769 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8771 Update copyright years.
8773 2018-01-03 Martin Liska <mliska@suse.cz>
8775 PR ipa/83594
8776 * ipa-visibility.c (function_and_variable_visibility): Skip
8777 functions with noipa attribure.
8779 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8781 * gcc.c (process_command): Update copyright notice dates.
8782 * gcov-dump.c (print_version): Ditto.
8783 * gcov.c (print_version): Ditto.
8784 * gcov-tool.c (print_version): Ditto.
8785 * gengtype.c (create_file): Ditto.
8786 * doc/cpp.texi: Bump @copying's copyright year.
8787 * doc/cppinternals.texi: Ditto.
8788 * doc/gcc.texi: Ditto.
8789 * doc/gccint.texi: Ditto.
8790 * doc/gcov.texi: Ditto.
8791 * doc/install.texi: Ditto.
8792 * doc/invoke.texi: Ditto.
8794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8796 * vector-builder.h (vector_builder::m_full_nelts): Change from
8797 unsigned int to poly_uint64.
8798 (vector_builder::full_nelts): Update prototype accordingly.
8799 (vector_builder::new_vector): Likewise.
8800 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8801 (vector_builder::operator ==): Likewise.
8802 (vector_builder::finalize): Likewise.
8803 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8804 Take the number of elements as a poly_uint64 rather than an
8805 unsigned int.
8806 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8807 from unsigned int to poly_uint64.
8808 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8809 (vec_perm_indices::new_vector): Likewise.
8810 (vec_perm_indices::length): Likewise.
8811 (vec_perm_indices::nelts_per_input): Likewise.
8812 (vec_perm_indices::input_nelts): Likewise.
8813 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8814 number of elements per input as a poly_uint64 rather than an
8815 unsigned int. Use the original encoding for variable-length
8816 vectors, rather than clamping each individual element.
8817 For the second and subsequent elements in each pattern,
8818 clamp the step and base before clamping their sum.
8819 (vec_perm_indices::series_p): Handle polynomial element counts.
8820 (vec_perm_indices::all_in_range_p): Likewise.
8821 (vec_perm_indices_to_tree): Likewise.
8822 (vec_perm_indices_to_rtx): Likewise.
8823 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8824 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8825 (tree_vector_builder::new_binary_operation): Handle polynomial
8826 element counts. Return false if we need to know the number
8827 of elements at compile time.
8828 * fold-const.c (fold_vec_perm): Punt if the number of elements
8829 isn't known at compile time.
8831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8833 * vec-perm-indices.h (vec_perm_builder): Change element type
8834 from HOST_WIDE_INT to poly_int64.
8835 (vec_perm_indices::element_type): Update accordingly.
8836 (vec_perm_indices::clamp): Handle polynomial element_types.
8837 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8838 (vec_perm_indices::all_in_range_p): Likewise.
8839 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8840 than shwi trees.
8841 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8842 polynomial vec_perm_indices element types.
8843 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8844 * fold-const.c (fold_vec_perm): Likewise.
8845 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8846 * tree-vect-generic.c (lower_vec_perm): Likewise.
8847 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8848 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8849 element type to HOST_WIDE_INT.
8851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8852 Alan Hayward <alan.hayward@arm.com>
8853 David Sherwood <david.sherwood@arm.com>
8855 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8856 rather than an int. Use plus_constant.
8857 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8858 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8861 Alan Hayward <alan.hayward@arm.com>
8862 David Sherwood <david.sherwood@arm.com>
8864 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8865 a HOST_WIDE_INT to a poly_int64.
8867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8868 Alan Hayward <alan.hayward@arm.com>
8869 David Sherwood <david.sherwood@arm.com>
8871 * calls.c (load_register_parameters): Cope with polynomial
8872 mode sizes. Require a constant size for BLKmode parameters
8873 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8874 forces a parameter to be padded at the lsb end in order to
8875 fill a complete number of words, require the parameter size
8876 to be ordered wrt UNITS_PER_WORD.
8878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8879 Alan Hayward <alan.hayward@arm.com>
8880 David Sherwood <david.sherwood@arm.com>
8882 * reload1.c (spill_stack_slot_width): Change element type
8883 from unsigned int to poly_uint64_pod.
8884 (alter_reg): Treat mode sizes as polynomial.
8886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8887 Alan Hayward <alan.hayward@arm.com>
8888 David Sherwood <david.sherwood@arm.com>
8890 * reload.c (complex_word_subreg_p): New function.
8891 (reload_inner_reg_of_subreg, push_reload): Use it.
8893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8894 Alan Hayward <alan.hayward@arm.com>
8895 David Sherwood <david.sherwood@arm.com>
8897 * lra-constraints.c (process_alt_operands): Reject matched
8898 operands whose sizes aren't ordered.
8899 (match_reload): Refer to this check here.
8901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8902 Alan Hayward <alan.hayward@arm.com>
8903 David Sherwood <david.sherwood@arm.com>
8905 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8906 that the mode size is in the set {1, 2, 4, 8, 16}.
8908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8909 Alan Hayward <alan.hayward@arm.com>
8910 David Sherwood <david.sherwood@arm.com>
8912 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8913 Use plus_constant instead of gen_rtx_PLUS.
8915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8916 Alan Hayward <alan.hayward@arm.com>
8917 David Sherwood <david.sherwood@arm.com>
8919 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8920 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8921 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8922 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8923 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8924 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8925 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8926 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8927 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8928 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8929 a poly_int64.
8930 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8931 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8932 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8933 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8934 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8935 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8936 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8937 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8938 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8939 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8940 function.
8941 * expr.c (emit_move_resolve_push): Treat the input and result
8942 of PUSH_ROUNDING as a poly_int64.
8943 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8944 (emit_push_insn): Likewise.
8945 * lra-eliminations.c (mark_not_eliminable): Likewise.
8946 * recog.c (push_operand): Likewise.
8947 * reload1.c (elimination_effects): Likewise.
8948 * rtlanal.c (nonzero_bits1): Likewise.
8949 * calls.c (store_one_arg): Likewise. Require the padding to be
8950 known at compile time.
8952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8953 Alan Hayward <alan.hayward@arm.com>
8954 David Sherwood <david.sherwood@arm.com>
8956 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8957 Use plus_constant instead of gen_rtx_PLUS.
8959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8960 Alan Hayward <alan.hayward@arm.com>
8961 David Sherwood <david.sherwood@arm.com>
8963 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8964 rather than an int.
8966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8967 Alan Hayward <alan.hayward@arm.com>
8968 David Sherwood <david.sherwood@arm.com>
8970 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8971 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8972 via stack temporaries. Treat the mode size as polynomial too.
8974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8975 Alan Hayward <alan.hayward@arm.com>
8976 David Sherwood <david.sherwood@arm.com>
8978 * expr.c (expand_expr_real_2): When handling conversions involving
8979 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8980 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8981 as a poly_uint64 too.
8983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8984 Alan Hayward <alan.hayward@arm.com>
8985 David Sherwood <david.sherwood@arm.com>
8987 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8990 Alan Hayward <alan.hayward@arm.com>
8991 David Sherwood <david.sherwood@arm.com>
8993 * combine.c (can_change_dest_mode): Handle polynomial
8994 REGMODE_NATURAL_SIZE.
8995 * expmed.c (store_bit_field_1): Likewise.
8996 * expr.c (store_constructor): Likewise.
8997 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8998 and polynomial REGMODE_NATURAL_SIZE.
8999 (gen_lowpart_common): Likewise.
9000 * reginfo.c (record_subregs_of_mode): Likewise.
9001 * rtlanal.c (read_modify_subreg_p): Likewise.
9003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9004 Alan Hayward <alan.hayward@arm.com>
9005 David Sherwood <david.sherwood@arm.com>
9007 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9008 numbers of elements.
9010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9011 Alan Hayward <alan.hayward@arm.com>
9012 David Sherwood <david.sherwood@arm.com>
9014 * match.pd: Cope with polynomial numbers of vector elements.
9016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9017 Alan Hayward <alan.hayward@arm.com>
9018 David Sherwood <david.sherwood@arm.com>
9020 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9021 in a POINTER_PLUS_EXPR.
9023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9024 Alan Hayward <alan.hayward@arm.com>
9025 David Sherwood <david.sherwood@arm.com>
9027 * omp-simd-clone.c (simd_clone_subparts): New function.
9028 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9029 (ipa_simd_modify_function_body): Likewise.
9031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9032 Alan Hayward <alan.hayward@arm.com>
9033 David Sherwood <david.sherwood@arm.com>
9035 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9036 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9037 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9038 (expand_vector_condition, vector_element): Likewise.
9039 (subparts_gt): New function.
9040 (get_compute_type): Use subparts_gt.
9041 (count_type_subparts): Delete.
9042 (expand_vector_operations_1): Use subparts_gt instead of
9043 count_type_subparts.
9045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9046 Alan Hayward <alan.hayward@arm.com>
9047 David Sherwood <david.sherwood@arm.com>
9049 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9050 (vect_compile_time_alias): ...this new function. Do the calculation
9051 on poly_ints rather than trees.
9052 (vect_prune_runtime_alias_test_list): Update call accordingly.
9054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9055 Alan Hayward <alan.hayward@arm.com>
9056 David Sherwood <david.sherwood@arm.com>
9058 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9059 numbers of units.
9060 (vect_schedule_slp_instance): Likewise.
9062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9063 Alan Hayward <alan.hayward@arm.com>
9064 David Sherwood <david.sherwood@arm.com>
9066 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9067 constant and extern definitions for variable-length vectors.
9068 (vect_get_constant_vectors): Note that the number of units
9069 is known to be constant.
9071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9072 Alan Hayward <alan.hayward@arm.com>
9073 David Sherwood <david.sherwood@arm.com>
9075 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9076 of units as polynomial. Choose between WIDE and NARROW based
9077 on multiple_p.
9079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9080 Alan Hayward <alan.hayward@arm.com>
9081 David Sherwood <david.sherwood@arm.com>
9083 * tree-vect-stmts.c (simd_clone_subparts): New function.
9084 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9087 Alan Hayward <alan.hayward@arm.com>
9088 David Sherwood <david.sherwood@arm.com>
9090 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9091 vectors as polynomial. Use build_index_vector for
9092 IFN_GOMP_SIMD_LANE.
9094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9095 Alan Hayward <alan.hayward@arm.com>
9096 David Sherwood <david.sherwood@arm.com>
9098 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9099 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9100 for variable-length vectors.
9101 (vectorizable_mask_load_store): Treat the number of units as
9102 polynomial, asserting that it is constant if the condition has
9103 already been enforced.
9104 (vectorizable_store, vectorizable_load): Likewise.
9106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9107 Alan Hayward <alan.hayward@arm.com>
9108 David Sherwood <david.sherwood@arm.com>
9110 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9111 of units as polynomial. Punt if we can't tell at compile time
9112 which vector contains the final result.
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-vect-loop.c (vectorizable_induction): Treat the number
9119 of units as polynomial. Punt on SLP inductions. Use an integer
9120 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9121 cast of such a series for variable-length floating-point
9122 reductions.
9124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9125 Alan Hayward <alan.hayward@arm.com>
9126 David Sherwood <david.sherwood@arm.com>
9128 * tree.h (build_index_vector): Declare.
9129 * tree.c (build_index_vector): New function.
9130 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9131 of units as polynomial, forcibly converting it to a constant if
9132 vectorizable_reduction has already enforced the condition.
9133 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9134 to create a {1,2,3,...} vector.
9135 (vectorizable_reduction): Treat the number of units as polynomial.
9136 Choose vectype_in based on the largest scalar element size rather
9137 than the smallest number of units. Enforce the restrictions
9138 relied on above.
9140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9141 Alan Hayward <alan.hayward@arm.com>
9142 David Sherwood <david.sherwood@arm.com>
9144 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9145 number of units as polynomial.
9147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9148 Alan Hayward <alan.hayward@arm.com>
9149 David Sherwood <david.sherwood@arm.com>
9151 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9152 * target.def (autovectorize_vector_sizes): Return the vector sizes
9153 by pointer, using vector_sizes rather than a bitmask.
9154 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9155 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9156 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9157 Likewise.
9158 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9159 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9160 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9161 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9162 * omp-general.c (omp_max_vf): Likewise.
9163 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9164 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9165 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9166 * tree-vect-slp.c (vect_slp_bb): Likewise.
9167 * doc/tm.texi: Regenerate.
9168 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9169 to a poly_uint64.
9170 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9171 the vector size as a poly_uint64 rather than an unsigned int.
9172 (current_vector_size): Change from an unsigned int to a poly_uint64.
9173 (get_vectype_for_scalar_type): Update accordingly.
9174 * tree.h (build_truth_vector_type): Take the size and number of
9175 units as a poly_uint64 rather than an unsigned int.
9176 (build_vector_type): Add a temporary overload that takes
9177 the number of units as a poly_uint64 rather than an unsigned int.
9178 * tree.c (make_vector_type): Likewise.
9179 (build_truth_vector_type): Take the number of units as a poly_uint64
9180 rather than an unsigned int.
9182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9183 Alan Hayward <alan.hayward@arm.com>
9184 David Sherwood <david.sherwood@arm.com>
9186 * target.def (get_mask_mode): Take the number of units and length
9187 as poly_uint64s rather than unsigned ints.
9188 * targhooks.h (default_get_mask_mode): Update accordingly.
9189 * targhooks.c (default_get_mask_mode): Likewise.
9190 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9191 * doc/tm.texi: Regenerate.
9193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9194 Alan Hayward <alan.hayward@arm.com>
9195 David Sherwood <david.sherwood@arm.com>
9197 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9198 * omp-general.c (omp_max_vf): Likewise.
9199 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9200 (expand_omp_simd): Handle polynomial safelen.
9201 * omp-low.c (omplow_simd_context): Add a default constructor.
9202 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9203 (lower_rec_simd_input_clauses): Update accordingly.
9204 (lower_rec_input_clauses): Likewise.
9206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9207 Alan Hayward <alan.hayward@arm.com>
9208 David Sherwood <david.sherwood@arm.com>
9210 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9211 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9212 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9213 (vect_analyze_slp_cost): Likewise.
9214 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9215 (vect_model_load_cost): Likewise.
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-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9222 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9223 from an unsigned int * to a poly_uint64_pod *.
9224 (calculate_unrolling_factor): New function.
9225 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9228 Alan Hayward <alan.hayward@arm.com>
9229 David Sherwood <david.sherwood@arm.com>
9231 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9232 from an unsigned int to a poly_uint64.
9233 (_loop_vec_info::slp_unrolling_factor): Likewise.
9234 (_loop_vec_info::vectorization_factor): Change from an int
9235 to a poly_uint64.
9236 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9237 (vect_get_num_vectors): New function.
9238 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9239 (vect_get_num_copies): Use vect_get_num_vectors.
9240 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9241 to an unsigned int *.
9242 (vect_analyze_data_refs): Change min_vf from an int * to a
9243 poly_uint64 *.
9244 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9245 than an unsigned HOST_WIDE_INT.
9246 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9247 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9248 to an unsigned int *.
9249 (vect_analyze_data_ref_dependences): Likewise.
9250 (vect_compute_data_ref_alignment): Handle polynomial vf.
9251 (vect_enhance_data_refs_alignment): Likewise.
9252 (vect_prune_runtime_alias_test_list): Likewise.
9253 (vect_shift_permute_load_chain): Likewise.
9254 (vect_supportable_dr_alignment): Likewise.
9255 (dependence_distance_ge_vf): Take the vectorization factor as a
9256 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9257 (vect_analyze_data_refs): Change min_vf from an int * to a
9258 poly_uint64 *.
9259 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9260 vfm1 as a poly_uint64 rather than an int. Make the same change
9261 for the returned bound_scalar.
9262 (vect_gen_vector_loop_niters): Handle polynomial vf.
9263 (vect_do_peeling): Likewise. Update call to
9264 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9265 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9266 be constant.
9267 * tree-vect-loop.c (vect_determine_vectorization_factor)
9268 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9269 (vect_get_known_peeling_cost): Likewise.
9270 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9271 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9272 (vect_transform_loop): Likewise. Use the lowest possible VF when
9273 updating the upper bounds of the loop.
9274 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9275 rather than an int.
9276 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9277 polynomial unroll factors.
9278 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9279 (vect_make_slp_decision): Likewise.
9280 (vect_supported_load_permutation_p): Likewise, and polynomial
9281 vf too.
9282 (vect_analyze_slp_cost): Handle polynomial vf.
9283 (vect_slp_analyze_node_operations): Likewise.
9284 (vect_slp_analyze_bb_1): Likewise.
9285 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9286 than an unsigned HOST_WIDE_INT.
9287 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9288 (vectorizable_load): Handle polynomial vf.
9289 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9290 a poly_uint64.
9291 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9294 Alan Hayward <alan.hayward@arm.com>
9295 David Sherwood <david.sherwood@arm.com>
9297 * match.pd: Handle bit operations involving three constants
9298 and try to fold one pair.
9300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9302 * tree-vect-loop-manip.c: Include gimple-fold.h.
9303 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9304 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9305 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9306 Add a path that uses a step of VF instead of 1, but disable it
9307 for now.
9308 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9309 and niters_no_overflow parameters. Update calls to
9310 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9311 Create a new SSA name if the latter choses to use a ste other
9312 than zero, and return it via niters_vector_mult_vf_var.
9313 * tree-vect-loop.c (vect_transform_loop): Update calls to
9314 vect_do_peeling, vect_gen_vector_loop_niters and
9315 slpeel_make_loop_iterate_ntimes.
9316 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9317 (vect_gen_vector_loop_niters): Update declarations after above changes.
9319 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9321 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9322 128-bit round to integer instructions.
9323 (ceil<mode>2): Likewise.
9324 (btrunc<mode>2): Likewise.
9325 (round<mode>2): Likewise.
9327 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9329 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9330 unaligned VSX load/store on P8/P9.
9331 (expand_block_clear): Allow the use of unaligned VSX
9332 load/store on P8/P9.
9334 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9336 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9337 New function.
9338 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9339 swap associated with both a load and a store.
9341 2018-01-02 Andrew Waterman <andrew@sifive.com>
9343 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9344 * config/riscv/riscv.md (clear_cache): Use it.
9346 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9348 * web.c: Remove out-of-date comment.
9350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9352 * expr.c (fixup_args_size_notes): Check that any existing
9353 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9354 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9355 (emit_single_push_insn): ...here.
9357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9359 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9360 (const_vector_encoded_nelts): New function.
9361 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9362 (const_vector_int_elt, const_vector_elt): Declare.
9363 * emit-rtl.c (const_vector_int_elt_1): New function.
9364 (const_vector_elt): Likewise.
9365 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9366 of CONST_VECTOR_ELT.
9368 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9370 * expr.c: Include rtx-vector-builder.h.
9371 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9372 directly on the tree encoding.
9373 (const_vector_from_tree): Likewise.
9374 * optabs.c: Include rtx-vector-builder.h.
9375 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9376 sequence of "u" values.
9377 * vec-perm-indices.c: Include rtx-vector-builder.h.
9378 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9379 directly on the vec_perm_indices encoding.
9381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9383 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9384 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9385 * rtx-vector-builder.h: New file.
9386 * rtx-vector-builder.c: Likewise.
9387 * rtl.h (rtx_def::u2): Add a const_vector field.
9388 (CONST_VECTOR_NPATTERNS): New macro.
9389 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9390 (CONST_VECTOR_DUPLICATE_P): Likewise.
9391 (CONST_VECTOR_STEPPED_P): Likewise.
9392 (CONST_VECTOR_ENCODED_ELT): Likewise.
9393 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9394 (unwrap_const_vec_duplicate): Likewise.
9395 (const_vec_series_p): Check for a non-duplicated vector encoding.
9396 Say that the function only returns true for integer vectors.
9397 * emit-rtl.c: Include rtx-vector-builder.h.
9398 (gen_const_vec_duplicate_1): Delete.
9399 (gen_const_vector): Call gen_const_vec_duplicate instead of
9400 gen_const_vec_duplicate_1.
9401 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9402 (gen_const_vec_duplicate): Use rtx_vector_builder.
9403 (gen_const_vec_series): Likewise.
9404 (gen_rtx_CONST_VECTOR): Likewise.
9405 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9406 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9407 Build a new vector rather than modifying a CONST_VECTOR in-place.
9408 (handle_special_swappables): Update call accordingly.
9409 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9410 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9411 Build a new vector rather than modifying a CONST_VECTOR in-place.
9412 (handle_special_swappables): Update call accordingly.
9414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9416 * simplify-rtx.c (simplify_const_binary_operation): Use
9417 CONST_VECTOR_ELT instead of XVECEXP.
9419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9421 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9422 the selector elements to be different from the data elements
9423 if the selector is a VECTOR_CST.
9424 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9425 ssizetype for the selector.
9427 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9429 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9430 before testing each element individually.
9431 * tree-vect-generic.c (lower_vec_perm): Likewise.
9433 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9435 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9436 * selftest-run-tests.c (selftest::run_tests): Call it.
9437 * vector-builder.h (vector_builder::operator ==): New function.
9438 (vector_builder::operator !=): Likewise.
9439 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9440 (vec_perm_indices::all_from_input_p): New function.
9441 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9442 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9443 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9444 instead of reading the VECTOR_CST directly. Detect whether both
9445 vector inputs are the same before constructing the vec_perm_indices,
9446 and update the number of inputs argument accordingly. Use the
9447 utility functions added above. Only construct sel2 if we need to.
9449 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9451 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9452 the broadcast of the low byte.
9453 (expand_mult_highpart): Use an explicit encoding for the permutes.
9454 * optabs-query.c (can_mult_highpart_p): Likewise.
9455 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9456 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9457 (vectorizable_bswap): Likewise.
9458 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9459 explicit encoding for the power-of-2 permutes.
9460 (vect_permute_store_chain): Likewise.
9461 (vect_grouped_load_supported): Likewise.
9462 (vect_permute_load_chain): Likewise.
9464 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9466 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9467 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9468 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9469 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9470 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9471 (vect_gen_perm_mask_any): Likewise.
9473 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9475 * int-vector-builder.h: New file.
9476 * vec-perm-indices.h: Include int-vector-builder.h.
9477 (vec_perm_indices): Redefine as an int_vector_builder.
9478 (auto_vec_perm_indices): Delete.
9479 (vec_perm_builder): Redefine as a stand-alone class.
9480 (vec_perm_indices::vec_perm_indices): New function.
9481 (vec_perm_indices::clamp): Likewise.
9482 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9483 (vec_perm_indices::new_vector): New function.
9484 (vec_perm_indices::new_expanded_vector): Update for new
9485 vec_perm_indices class.
9486 (vec_perm_indices::rotate_inputs): New function.
9487 (vec_perm_indices::all_in_range_p): Operate directly on the
9488 encoded form, without computing elided elements.
9489 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9490 encoding. Update for new vec_perm_indices class.
9491 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9492 the given vec_perm_builder.
9493 (expand_vec_perm_var): Update vec_perm_builder constructor.
9494 (expand_mult_highpart): Use vec_perm_builder instead of
9495 auto_vec_perm_indices.
9496 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9497 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9498 or double series encoding as appropriate.
9499 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9500 vec_perm_indices instead of auto_vec_perm_indices.
9501 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9502 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9503 (vect_permute_store_chain): Likewise.
9504 (vect_grouped_load_supported): Likewise.
9505 (vect_permute_load_chain): Likewise.
9506 (vect_shift_permute_load_chain): Likewise.
9507 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9508 (vect_transform_slp_perm_load): Likewise.
9509 (vect_schedule_slp_instance): Likewise.
9510 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9511 (vectorizable_mask_load_store): Likewise.
9512 (vectorizable_bswap): Likewise.
9513 (vectorizable_store): Likewise.
9514 (vectorizable_load): Likewise.
9515 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9516 vec_perm_indices instead of auto_vec_perm_indices. Use
9517 tree_to_vec_perm_builder to read the vector from a tree.
9518 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9519 vec_perm_builder instead of a vec_perm_indices.
9520 (have_whole_vector_shift): Use vec_perm_builder and
9521 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9522 truncation to calc_vec_perm_mask_for_shift.
9523 (vect_create_epilog_for_reduction): Likewise.
9524 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9525 from auto_vec_perm_indices to vec_perm_indices.
9526 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9527 instead of changing individual elements.
9528 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9529 the vector in d.perm.
9530 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9531 from auto_vec_perm_indices to vec_perm_indices.
9532 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9533 instead of changing individual elements.
9534 (arm_vectorize_vec_perm_const): Use new_vector to install
9535 the vector in d.perm.
9536 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9537 Update vec_perm_builder constructor.
9538 (rs6000_expand_interleave): Likewise.
9539 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9540 (rs6000_expand_interleave): Likewise.
9542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9544 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9545 to qimode could truncate the indices.
9546 * optabs.c (expand_vec_perm_var): Likewise.
9548 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9550 * Makefile.in (OBJS): Add vec-perm-indices.o.
9551 * vec-perm-indices.h: New file.
9552 * vec-perm-indices.c: Likewise.
9553 * target.h (vec_perm_indices): Replace with a forward class
9554 declaration.
9555 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9556 * optabs.h: Include vec-perm-indices.h.
9557 (expand_vec_perm): Delete.
9558 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9559 (expand_vec_perm_const): Declare.
9560 * target.def (vec_perm_const_ok): Replace with...
9561 (vec_perm_const): ...this new hook.
9562 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9563 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9564 * doc/tm.texi: Regenerate.
9565 * optabs.def (vec_perm_const): Delete.
9566 * doc/md.texi (vec_perm_const): Likewise.
9567 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9568 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9569 expand_vec_perm for constant permutation vectors. Assert that
9570 the mode of variable permutation vectors is the integer equivalent
9571 of the mode that is being permuted.
9572 * optabs-query.h (selector_fits_mode_p): Declare.
9573 * optabs-query.c: Include vec-perm-indices.h.
9574 (selector_fits_mode_p): New function.
9575 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9576 is defined, instead of checking whether the vec_perm_const_optab
9577 exists. Use targetm.vectorize.vec_perm_const instead of
9578 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9579 fit in the vector mode before using a variable permute.
9580 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9581 vec_perm_indices instead of an rtx.
9582 (expand_vec_perm): Replace with...
9583 (expand_vec_perm_const): ...this new function. Take the selector
9584 as a vec_perm_indices rather than an rtx. Also take the mode of
9585 the selector. Update call to shift_amt_for_vec_perm_mask.
9586 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9587 Use vec_perm_indices::new_expanded_vector to expand the original
9588 selector into bytes. Check whether the indices fit in the vector
9589 mode before using a variable permute.
9590 (expand_vec_perm_var): Make global.
9591 (expand_mult_highpart): Use expand_vec_perm_const.
9592 * fold-const.c: Includes vec-perm-indices.h.
9593 * tree-ssa-forwprop.c: Likewise.
9594 * tree-vect-data-refs.c: Likewise.
9595 * tree-vect-generic.c: Likewise.
9596 * tree-vect-loop.c: Likewise.
9597 * tree-vect-slp.c: Likewise.
9598 * tree-vect-stmts.c: Likewise.
9599 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9600 Delete.
9601 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9602 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9603 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9604 (aarch64_vectorize_vec_perm_const): ...this new function.
9605 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9606 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9607 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9608 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9609 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9610 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9611 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9612 into...
9613 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9614 check for NEON modes.
9615 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9616 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9617 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9618 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9619 into...
9620 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9621 the old VEC_PERM_CONST conditions.
9622 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9623 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9624 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9625 (ia64_vectorize_vec_perm_const_ok): Merge into...
9626 (ia64_vectorize_vec_perm_const): ...this new function.
9627 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9628 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9629 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9630 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9631 * config/mips/mips.c (mips_expand_vec_perm_const)
9632 (mips_vectorize_vec_perm_const_ok): Merge into...
9633 (mips_vectorize_vec_perm_const): ...this new function.
9634 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9635 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9636 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9637 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9638 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9639 (rs6000_expand_vec_perm_const): Delete.
9640 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9641 Delete.
9642 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9643 (altivec_expand_vec_perm_const_le): Take each operand individually.
9644 Operate on constant selectors rather than rtxes.
9645 (altivec_expand_vec_perm_const): Likewise. Update call to
9646 altivec_expand_vec_perm_const_le.
9647 (rs6000_expand_vec_perm_const): Delete.
9648 (rs6000_vectorize_vec_perm_const_ok): Delete.
9649 (rs6000_vectorize_vec_perm_const): New function.
9650 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9651 an element count and rtx array.
9652 (rs6000_expand_extract_even): Update call accordingly.
9653 (rs6000_expand_interleave): Likewise.
9654 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9655 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9656 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9657 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9658 (rs6000_expand_vec_perm_const): Delete.
9659 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9660 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9661 (altivec_expand_vec_perm_const_le): Take each operand individually.
9662 Operate on constant selectors rather than rtxes.
9663 (altivec_expand_vec_perm_const): Likewise. Update call to
9664 altivec_expand_vec_perm_const_le.
9665 (rs6000_expand_vec_perm_const): Delete.
9666 (rs6000_vectorize_vec_perm_const_ok): Delete.
9667 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9668 reference to the SPE evmerge intructions.
9669 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9670 an element count and rtx array.
9671 (rs6000_expand_extract_even): Update call accordingly.
9672 (rs6000_expand_interleave): Likewise.
9673 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9674 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9675 new function.
9676 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9680 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9681 vector mode and that that mode matches the mode of the data
9682 being permuted.
9683 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9684 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9685 directly using expand_vec_perm_1 when forcing selectors into
9686 registers.
9687 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9691 * optabs-query.h (can_vec_perm_p): Delete.
9692 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9693 * optabs-query.c (can_vec_perm_p): Split into...
9694 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9695 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9696 particular selector is valid.
9697 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9698 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9699 (vect_grouped_load_supported): Likewise.
9700 (vect_shift_permute_load_chain): Likewise.
9701 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9702 (vect_transform_slp_perm_load): Likewise.
9703 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9704 (vectorizable_bswap): Likewise.
9705 (vect_gen_perm_mask_checked): Likewise.
9706 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9707 implementations of variable permutation vectors into account
9708 when deciding which selector to use.
9709 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9710 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9711 with a false third argument.
9712 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9713 to test whether the constant selector is valid and can_vec_perm_var_p
9714 to test whether a variable selector is valid.
9716 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9718 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9719 * optabs-query.c (can_vec_perm_p): Likewise.
9720 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9721 instead of vec_perm_indices.
9722 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9723 (vect_gen_perm_mask_checked): Likewise,
9724 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9725 (vect_gen_perm_mask_checked): Likewise,
9727 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9729 * optabs-query.h (qimode_for_vec_perm): Declare.
9730 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9731 (qimode_for_vec_perm): ...this new function.
9732 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9734 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9736 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9737 does not have a conditional at the top.
9739 2018-01-02 Richard Biener <rguenther@suse.de>
9741 * ipa-inline.c (big_speedup_p): Fix expression.
9743 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9745 PR target/81616
9746 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9747 for generic 4->6.
9749 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9751 PR target/81616
9752 Generic tuning.
9753 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9754 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9755 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9756 cond_taken_branch_cost 3->4.
9758 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9760 PR tree-optimization/83581
9761 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9762 TODO_cleanup_cfg if any changes have been made.
9764 PR middle-end/83608
9765 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9766 convert_modes if target mode has the right side, but different mode
9767 class.
9769 PR middle-end/83609
9770 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9771 last argument when extracting from CONCAT. If either from_real or
9772 from_imag is NULL, use expansion through memory. If result is not
9773 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9774 the parts directly to inner mode, if even that fails, use expansion
9775 through memory.
9777 PR middle-end/83623
9778 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9779 check for bswap in mode rather than HImode and use that in expand_unop
9780 too.
9781 \f
9782 Copyright (C) 2018 Free Software Foundation, Inc.
9784 Copying and distribu