1 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2 Kito Cheng <kito.cheng@gmail.com>
4 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10 Kito Cheng <kito.cheng@gmail.com>
12 * config/nds32/constraints.md (Ufe): New memory constraint.
13 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
14 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
17 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22 * config/nds32/nds32.md: Use optimize_size in the condition for
23 alu-shift instructions.
25 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
27 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
29 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
31 * config/nds32/nds32.md (negsi2): Refine pattern.
33 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
34 Chung-Ju Wu <jasonwucj@gmail.com>
36 * config/nds32/iterators.md (shift_rotate): New code iterator.
37 (shift): New code attribute.
38 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
39 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
40 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
41 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
43 (andsi3, *andsi3): Ditto.
44 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
45 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
46 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
47 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
48 nds32_ior_operand, nds32_xor_operand): New predicates.
50 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
52 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
53 (addsi3, subsi3): ... this.
55 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
57 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
59 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
61 * config/nds32/nds32.md: Adjust indention.
63 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
65 * config/nds32/nds32.md (feature): New attribute.
67 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
69 * config/nds32/nds32.md (subtype): New attribute.
71 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
74 * config/arm/arm-builtins.c (arm_expand_builtin): Change
75 expansion to perform a bitwise AND of the argument followed by a
76 boolean negation of the result.
78 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
80 PR rtl-optimization/84878
81 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
82 the basic block. Assert the use reference is not artificial and that
83 it has an associated insn.
85 2018-04-04 Michael Matz <matz@suse.de>
87 * builtins.c (compute_objsize): Pass correct operand
88 to array_at_struct_end_p.
90 2018-04-04 Richard Biener <rguenther@suse.de>
93 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
94 from contexts for DINFO_LEVEL_TERSE and below.
96 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
98 * config/nds32/nds32-doubleword.md (move_<mode>): Require
99 resiter_operand condition.
100 * config/nds32/nds32.md (*move<mode>): Ditto.
102 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
103 Monk Chiang <sh.chiang04@gmail.com>
105 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
107 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
109 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
111 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
112 Kito Cheng <kito.cheng@gmail.com>
114 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
115 nds32_cond_code_str, output_cond_branch,
116 output_cond_branch_compare_zero, nds32_expand_cbranch,
117 nds32_expand_cstore, nds32_expand_movcc,
118 nds32_output_cbranchsi4_equality_zero,
119 nds32_output_cbranchsi4_equality_reg,
120 nds32_output_cbranchsi4_equality_reg_or_const_int,
121 nds32_output_cbranchsi4_greater_less_zero: New functions.
122 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
123 nds32_expand_cstore, nds32_expand_movcc,
124 nds32_output_cbranchsi4_equality_zero,
125 nds32_output_cbranchsi4_equality_reg,
126 nds32_output_cbranchsi4_equality_reg_or_const_int,
127 nds32_output_cbranchsi4_greater_less_zero): Declare.
128 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
129 nds32_rimm11s_operand): New predicates.
130 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
131 * config/nds32/nds32.md: Rewrite all the branch and conditional move
134 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
136 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
137 * config/nds32/nds32.md: Ditto.
138 * config/nds32/pipelines.md: Ditto.
140 2018-04-04 Richard Biener <rguenther@suse.de>
142 PR tree-optimization/85168
143 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
144 propagating abnormals.
146 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
148 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
150 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
151 Kito Cheng <kito.cheng@gmail.com>
153 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
154 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
155 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
156 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
157 * config/nds32/nds32.md (sibcall_internal): New.
158 (sibcall_register): Remove.
159 (sibcall_immediate): Remove.
160 (sibcall_value_internal): New.
161 (sibcall_value_register): Remove.
162 (sibcall_value_immediate): Remove.
163 * config/nds32/predicates.md (nds32_general_register_operand): New.
164 (nds32_call_address_operand): New.
166 2018-04-03 Jakub Jelinek <jakub@redhat.com>
168 PR rtl-optimization/85167
169 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
170 bb_defs if *split_p, instead preinitialize it to NULL.
172 PR tree-optimization/85156
173 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
174 evaluating the argument multiple times.
176 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
178 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
180 (_mm_cvtpd_ps): Likewise.
181 (_mm_cvttpd_epi32): Likewise.
182 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
183 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
184 vector, pixel, and bool following altivec.h include.
186 2018-04-03 Martin Sebor <msebor@redhat.com>
188 * doc/extend.texi (Common Function Attributes): Clarify.
189 (const attribute): Likewise.
190 (pure attribute): Likewise.
192 2018-04-03 Jakub Jelinek <jakub@redhat.com>
195 * config/i386/i386.c (ix86_expand_vector_set): Use
196 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
198 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
200 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
201 instructions when changing rounding bits to preserve precision bits
202 in the x87 control word.
204 2018-04-03 Martin Liska <mliska@suse.cz>
206 PR tree-optimization/82491
207 * rtl.h (strip_offset_and_add): Replace += suboffset with
208 poly_uint64 () + suboffset.
210 2018-03-29 Martin Liska <mliska@suse.cz>
211 Martin Jambor <mjambor@suse.cz>
214 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
215 param_type is not an integral or pointer type.
217 2018-04-03 Richard Biener <rguenther@suse.de>
219 * sese.h (recompute_all_dominators): Remove.
221 2018-04-02 Martin Sebor <msebor@redhat.com>
223 * doc/invoke.texi (-Wrestrict): Fix typos.
225 2018-04-02 Jim Wilson <jimw@sifive.com>
227 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
228 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
229 (<optab>di3, <optab>si3_extend): Likewise.
230 (<optab>si3_mask, <optab>si3_mask_1): New.
231 (<optab>di3_mask, <optab>di3_mask_1): New.
232 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
233 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
234 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
236 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
238 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
241 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
243 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
244 (nds32_canonicalize_comparison): New function.
246 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
247 Kito Cheng <kito.cheng@gmail.com>
248 Kuan-Lin Chen <kuanlinchentw@gmail.com>
250 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
251 * config/nds32/constants.md (unspec_volatile_element): Add
252 UNSPEC_VOLATILE_RELAX_GROUP.
253 * config/nds32/nds32-relax-opt.c: New file.
254 * config/nds32/nds32-predicates.c
255 (nds32_symbol_load_store_p): New function.
256 * config/nds32/nds32-protos.h
257 (nds32_symbol_load_store_p): Declare function.
258 (make_pass_nds32_relax_opt): Declare new rtl pass function.
259 * config/nds32/nds32.c
260 (nds32_register_pass): New function to register pass.
261 (nds32_register_passes): New function to register passes.
262 * config/nds32/nds32.md (relax_group): New pattern.
263 * config/nds32/nds32.opt (mrelax-hint): New option.
264 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
266 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
268 * config/nds32/t-nds32: Modify files dependency.
270 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
273 (PROFILE_HOOK): Define its implementation.
275 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
277 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
278 type and 32-bit size.
280 2018-04-01 Jakub Jelinek <jakub@redhat.com>
283 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
284 (V_128_256): New mode iterator.
285 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
286 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
287 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
289 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
290 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
292 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
295 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
296 NaN inputs correctly.
298 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
301 * config/rs6000/vsx.md (??r): New mode attribute.
302 (*vsx_mov<mode>_64bit): Use it.
303 (*vsx_mov<mode>_32bit): Likewise.
305 2018-03-30 Martin Sebor <msebor@redhat.com>
307 PR tree-optimization/84818
308 * builtins.c (check_access): Use warning_n.
310 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
313 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
315 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
318 2018-03-30 Julia Koval <julia.koval@intel.com>
321 * x86-tune.def (movx, partial_reg_dependency): Enable for
324 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
327 * lra-constraints.c (process_alt_operands): Move setting
328 this_alternative_matches below.
330 2018-03-29 Martin Liska <mliska@suse.cz>
333 * doc/invoke.texi: Document how LTO works with debug info.
334 Describe auto-load support of binutils. Mention 'x86-64'
335 as valid option value of -march option.
337 2018-03-29 Jakub Jelinek <jakub@redhat.com>
339 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
342 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
343 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
344 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
347 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
350 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
351 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
352 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
353 for RS6000_BTM_POWERPC64.
354 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
355 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
356 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
361 2018-03-28 Carl Love <cel@us.ibm.com>
364 2017-09-27 Carl Love <cel@us.ibm.com>
366 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
367 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
368 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
371 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
373 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
374 instead of __vector bool.
375 (_mm_max_pu8): Likewise.
376 (_mm_min_pi16): Likewise.
378 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
381 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
385 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
387 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
388 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
389 (div_extend): Likewise.
390 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
392 (__builtin_divweuo): Likewise.
393 (__builtin_divdeo): Likewise.
394 (__builtin_divdeuo): Likewise.
396 2018-03-28 Jakub Jelinek <jakub@redhat.com>
399 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
400 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
402 PR tree-optimization/82004
403 * gimple-match-head.c (optimize_pow_to_exp): New function.
404 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
405 Don't fold to exp if optimize_pow_to_exp is false.
407 2018-03-28 Martin Liska <mliska@suse.cz>
410 * calls.c (initialize_argument_information): Fix trailing space.
411 * common.opt: Fix typo and provide better explanation for
412 -fsanitize-coverage option.
413 * config/i386/i386.opt: Fix typo.
415 2018-03-28 Jakub Jelinek <jakub@redhat.com>
416 Martin Liska <mliska@suse.cz>
419 * gimplify.c (asan_poison_variable): Don't do the check for
420 gimplify_omp_ctxp here.
421 (gimplify_decl_expr): Do it here.
422 (gimplify_target_expr): Likewise.
424 2018-03-28 Martin Liska <mliska@suse.cz>
427 * config/i386/i386.c (ix86_function_arg_advance): Do not call
428 chkp_type_bounds_count if MPX is not enabled.
430 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
432 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
434 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
437 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
438 function to create the function decl for complex long double
439 multiply and divide for -mabi=ieeelongdouble.
440 (init_float128_ieee): Call it.
442 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
445 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
446 -fcf-protection=branch -mibt.
447 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
449 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
452 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
454 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
457 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
458 extern array declarations.
460 2018-03-27 Richard Biener <rguenther@suse.de>
463 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
464 explicit single_use checks.
466 2018-03-27 Richard Biener <rguenther@suse.de>
468 PR tree-optimization/85082
469 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
472 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
474 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
475 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
476 Turn on fasynchronous-unwind-tables and funwind-tables.
478 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
481 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
482 (*bmi_blsr_<mode>_ccz): Ditto.
484 2018-03-26 Tom de Vries <tom@codesourcery.com>
486 PR tree-optimization/85063
487 * omp-general.c (offloading_function_p): New function. Factor out
489 * omp-offload.c (pass_omp_target_link::gate): ... here.
490 * omp-general.h (offloading_function_p): Declare.
491 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
492 with attribute omp declare target for offloading functions.
494 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
496 PR tree-optimization/84005
497 * tree-data-ref.h (get_base_for_alignment): Declare.
498 * tree-data-ref.c (get_base_for_alignment_1): New function.
499 (get_base_for_alignment): Likewise.
500 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
501 get_base_for_alignment to find a suitable base object, instead
502 of always using drb->base_address.
504 2018-03-23 Jakub Jelinek <jakub@redhat.com>
507 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
508 known size by default.
510 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
513 * lra-constraints.c (process_alt_operands): Don't match BLKmode
514 and non BLKmode operands.
516 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
519 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
522 2018-03-23 Richard Biener <rguenther@suse.de>
525 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
526 we are going to emit early debug for LTO.
528 2018-03-23 Jakub Jelinek <jakub@redhat.com>
531 * function.c (match_asm_constraints_1): Don't optimize if input
532 doesn't satisfy general_operand predicate for output's mode.
535 * alias.c (write_dependence_p): Don't require for x_canonicalized
536 non-VOIDmode if x has VOIDmode.
539 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
540 just don't try to optimize it rather than assert it never happens.
542 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
544 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
545 macro expansions for definition of ST_INTERNAL_<mode> and
546 LD_INTERNAL_<mode> builtins.
547 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
549 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
551 (altivec_expand_st_builtin): Likewise.
552 (altivec_expand_builtin): Remove calls to deleted functions.
553 (rs6000_address_for_altivec): Delete this function.
554 * config/rs6000/vector.md: Remove expands for
555 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
557 2018-03-22 Sudakshina Das <sudi.das@arm.com>
560 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
561 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
562 re-computing once computed.
563 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
564 (arm_init_machine_status): Initialize
565 machine->static_chain_stack_bytes.
567 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
570 * doc/extend.texi: Add four new prototypes for vec_ld.
571 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
572 definitions for more logical presentation.
573 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
574 entries for V1TI variants of __builtin_altivec_ld builtin.
575 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
576 handling of V1TI variant of LVX icode pattern.
577 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
578 (rs6000_gimple_fold_builtin): Likewise.
579 (altivec_init_builtins): Add code to define
580 __builtin_altivec_lvx_v1ti function.
582 2018-03-22 Jakub Jelinek <jakub@redhat.com>
585 * function.c (match_asm_constraints_1): Don't do the optimization
586 if input isn't a REG, SUBREG, MEM or constant.
588 2018-03-22 Tom de Vries <tom@codesourcery.com>
590 PR tree-optimization/84956
591 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
592 bb_has_abnormal_pred.
594 2018-03-22 Jakub Jelinek <jakub@redhat.com>
597 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
598 DECL_INITIAL (decl) to decl at the end.
599 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
602 2018-03-21 Joseph Myers <joseph@codesourcery.com>
604 * doc/extend.texi (__builtin_tgmath): Document when complex
605 integer types are treated as _Complex _Float64.
607 2018-03-21 Tom de Vries <tom@codesourcery.com>
609 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
611 2018-03-21 Jakub Jelinek <jakub@redhat.com>
613 PR tree-optimization/84960
614 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
615 if it is ENTRY block, move them into single succ of ENTRY in that case.
617 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
619 PR tree-optimization/84811
620 * poly-int.h (poly_span_traits): Remove the T3 parameter and
621 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
622 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
623 (known_subrange_p): Update accordingly. Cast each value involved
624 in the size comparison, rather than casting the result of the
627 2018-03-21 Jakub Jelinek <jakub@redhat.com>
629 PR tree-optimization/84982
630 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
631 by flipping the least significant bit rather than all bits from
632 bitpos to bitpos + bitsize - 1.
634 2018-03-21 Nathan Sidwell <nathan@acm.org>
636 * doc/extend.texi (Deprecated Features): Remove mention of
637 long-deleted deprecations.
639 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
642 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
643 * configure: Regenerate.
645 2018-03-21 Tom de Vries <tom@codesourcery.com>
647 PR tree-optimization/83126
648 * tree-parloops.c (num_phis): New function.
649 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
651 2018-03-21 Nathan Sidwell <nathan@acm.org>
653 * doc/extend.texi (Deprecated Features): Update deprecated flags,
654 mention anon-struct/union members and trailing attributes.
656 2018-03-21 Bin Cheng <bin.cheng@arm.com>
658 PR tree-optimization/84969
659 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
660 builtin memset partitions if they set different rhs values.
662 2018-03-21 Jakub Jelinek <jakub@redhat.com>
664 PR rtl-optimization/84989
665 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
666 VEC_DUPLICATE with scalar result mode.
668 2018-03-21 Martin Liska <mliska@suse.cz>
671 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
672 not intended return statement.
674 2018-03-21 Martin Liska <mliska@suse.cz>
677 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
678 (chkp_find_bound_slots_1): Limit number of iterations.
680 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
683 * Minor grammar fixes for x86 options.
685 2018-03-20 Jakub Jelinek <jakub@redhat.com>
688 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
689 holding REG_CFA_RESTORE notes, instead turn them into a USE.
691 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
694 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
695 (altivec_lvx_<mode>_1op): Likewise.
696 (altivec_stvx_<mode>_2op): Likewise.
697 (altivec_stvx_<mode>_1op): Likewise.
698 (altivec_lvx_<VM2:mode>): New define_expand.
699 (altivec_stvx_<VM2:mode>): Likewise.
700 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
701 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
702 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
703 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
704 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
705 (rs6000_gen_lvx): Likewise.
706 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
707 (altivec_expand_stv_builtin): Likewise.
708 (altivec_expand_builtin): Likewise.
709 * config/rs6000/vector.md: Likewise.
711 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
714 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
717 2018-03-20 Richard Biener <rguenther@suse.de>
720 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
721 sign-conversions as zero, fall back to standard scalar_stmt
724 2018-03-20 Martin Liska <mliska@suse.cz>
727 * predict.c (rebuild_frequencies): Handle case when we have
728 PROFILE_ABSENT, but flag_guess_branch_prob is false.
730 2018-03-20 Jakub Jelinek <jakub@redhat.com>
733 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
734 flag_section_anchors.
735 * varasm.c (use_blocks_for_decl_p): Remove hack for
739 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
741 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
742 be created, use lowpart_subreg of operands[0] rather than operands[0]
744 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
745 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
746 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
747 and n constraint instead of aarch64_shift_imm_di and Usd.
748 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
749 (*aarch64_<optab>_reg_minus<mode>3): ... this.
751 2018-03-20 Sudakshina Das <sudi.das@arm.com>
754 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
755 to favor GPR over NEON registers.
756 (<shift>di3_neon): Likewise.
758 2018-03-20 Tom de Vries <tom@codesourcery.com>
761 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
762 (nvptx_process_pars): Emit bar.sync asap and alap.
764 2018-03-20 Tom de Vries <tom@codesourcery.com>
767 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
768 seen_label if seen_label is already set.
770 2018-03-20 Jakub Jelinek <jakub@redhat.com>
773 * config/i386/i386.c (fold_builtin_cpu): For features above 31
774 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
775 Use 1U instead of 1. Formatting fixes.
778 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
779 instead of TREE_TYPE (s1) for the return value.
781 2018-03-19 Jakub Jelinek <jakub@redhat.com>
783 PR tree-optimization/84946
784 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
785 bitsize + bitsize in poly_uint64 rather than poly_int64.
788 * dwarf2asm.c: Include fold-const.c.
789 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
790 of decl rather than decl itself.
792 PR rtl-optimization/84643
793 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
795 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
798 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
799 calling assemble_variable.
801 2018-03-19 Sudakshina Das <sudi.das@arm.com>
804 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
805 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
807 2018-03-19 Jim Wilson <jimw@sifive.com>
810 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
811 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
812 (riscv_first_stack_step): Likewise.
813 (riscv_option_override): Use STACK_BOUNDARY instead of
815 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
817 (BIGGEST_ALIGNMENT): Set to 128.
818 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
819 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
822 2018-03-19 Richard Biener <rguenther@suse.de>
824 PR tree-optimization/84933
825 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
826 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
828 2018-03-19 Richard Biener <rguenther@suse.de>
830 PR tree-optimization/84859
831 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
832 (cond_if_else_store_replacement): Perform sinking operation on
833 single-store BBs regardless of MAX_STORES_TO_SINK setting.
834 Generalize what a BB with a single eligible store is.
836 2018-03-19 Richard Biener <rguenther@suse.de>
838 PR tree-optimization/84929
839 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
840 chrec_is_positive against non-chrec arg.
842 2018-03-19 Tamar Christina <tamar.christina@arm.com>
845 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
847 2018-03-18 Martin Liska <mliska@suse.cz>
849 PR rtl-optimization/84635
850 * regrename.c (build_def_use): Use matches_mode only when
853 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
855 PR tree-optimization/84913
856 * tree-vect-loop.c (vectorizable_reduction): Don't try to
857 vectorize chains of COND_EXPRs.
859 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
861 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
863 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
865 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
867 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
869 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
871 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
872 Kito Cheng <kito.cheng@gmail.com>
874 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
875 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
876 (nds32_adjust_reg_alloc_order): New function.
877 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
879 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
881 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
882 nds32_print_operand, nds32_print_operand_address): Use
883 HOST_WIDE_INT_PRINT_DEC instead.
885 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
887 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
889 2018-03-17 Jakub Jelinek <jakub@redhat.com>
892 * config/i386/i386.c (initial_ix86_tune_features,
893 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
895 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
896 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
897 rather than 1u << ix86_tune. Formatting fix.
898 (ix86_option_override_internal): Change ix86_arch_mask from
899 unsigned int to unsigned HOST_WIDE_INT, initialize to
900 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
901 (ix86_function_specific_restore): Likewise.
903 2018-03-16 Jakub Jelinek <jakub@redhat.com>
906 * postreload.c (reload_combine_recognize_pattern): Perform
907 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
908 truncate_int_for_mode the result for the destination's mode.
911 * hsa-gen.c (mem_type_for_type): Fix comment typo.
912 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
914 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
917 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
920 * lra-assigns.c (lra_split_hard_reg_for): Don't use
921 regno_allocno_class_array and sorted_pseudos.
922 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
923 insns where regno is used.
925 2018-03-16 Martin Liska <mliska@suse.cz>
928 * multiple_target.c (create_dispatcher_calls): Redirect
929 reference in the symbol table.
931 2018-03-16 Martin Liska <mliska@suse.cz>
934 * multiple_target.c (create_dispatcher_calls): Redirect also
937 2018-03-16 Jakub Jelinek <jakub@redhat.com>
941 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
942 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
943 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
945 2018-03-16 Julia Koval <julia.koval@intel.com>
947 * doc/invoke.texi (Skylake Server): Add CLWB.
948 Cannonlake): Remove CLWB.
950 2018-03-16 Jakub Jelinek <jakub@redhat.com>
952 PR tree-optimization/84841
953 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
955 (FLOAT_ONE_CONST_TYPE): Define.
956 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
957 (sort_by_operand_rank): Put entries with higher constant_type last
958 rather than first to match comments.
960 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
962 * config/nios2/nios2.md (movsi_internal): Fix thinko in
965 2018-03-15 Jakub Jelinek <jakub@redhat.com>
968 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
969 check and use address of target always.
971 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
974 * config/i386/i386.c (indirect_thunk_needed): Update comments.
975 (indirect_thunk_bnd_needed): Likewise.
976 (indirect_thunks_used): Likewise.
977 (indirect_thunks_bnd_used): Likewise.
978 (indirect_return_needed): New.
979 (indirect_return_bnd_needed): Likewise.
980 (output_indirect_thunk_function): Add a bool argument for
982 (output_indirect_thunk_function): Don't generate alias for
983 function return thunk.
984 (ix86_code_end): Call output_indirect_thunk_function to generate
985 function return thunks.
986 (ix86_output_function_return): Set indirect_return_bnd_needed
987 and indirect_return_needed instead of indirect_thunk_bnd_needed
988 and indirect_thunk_needed.
990 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
992 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
993 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
994 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
996 2018-03-15 David Malcolm <dmalcolm@redhat.com>
997 Paul Hua <paul.hua.gm@gmail.com>
1000 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1002 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1004 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1005 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1008 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1011 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1012 instead of GET_MODE_SIZE when comparing Units.
1014 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1017 * varasm.c (hash_section): Return an unchangeble hash value
1018 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1019 Return !aarch64_can_use_per_function_literal_pools_p ().
1021 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1024 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1025 rather than address of cmode as last argument to prepare_cmp_insn.
1027 2018-03-15 Julia Koval <julia.koval@intel.com>
1029 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1030 F_AVX512VNNI, F_AVX512BITALG): New.
1032 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1035 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1036 insn for floating-point loads and stores.
1038 2018-03-14 Carl Love <cel@us.ibm.com>
1040 * config/rs6000/rs6000-c.c: Add macro definitions for
1041 ALTIVEC_BUILTIN_VEC_PERMXOR.
1042 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1043 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1044 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1045 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1047 * config/doc/extend.texi: Add prototypes for vec_permxor.
1049 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1052 * diagnostic-show-locus.c (class layout_point): Convert m_line
1053 from int to linenum_type.
1054 (line_span::comparator): Use linenum "compare" function when
1055 comparing line numbers.
1056 (test_line_span): New function.
1057 (layout_range::contains_point): Convert param "row" from int to
1059 (layout_range::intersects_line_p): Likewise.
1060 (layout::will_show_line_p): Likewise.
1061 (layout::print_source_line): Likewise.
1062 (layout::should_print_annotation_line_p): Likewise.
1063 (layout::print_annotation_line): Likewise.
1064 (layout::print_leading_fixits): Likewise.
1065 (layout::annotation_line_showed_range_p): Likewise.
1066 (struct line_corrections): Likewise for field m_row.
1067 (line_corrections::line_corrections): Likewise for param "row".
1068 (layout::print_trailing_fixits): Likewise.
1069 (layout::get_state_at_point): Likewise.
1070 (layout::get_x_bound_for_row): Likewise.
1071 (layout::print_line): Likewise.
1072 (diagnostic_show_locus): Likewise for locals "last_line" and
1074 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1075 * input.c (selftest::test_linenum_comparisons): New function.
1076 (selftest::input_c_tests): Call it.
1077 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1078 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1079 * selftest.h (ASSERT_GT): New macro.
1080 (ASSERT_GT_AT): New macro.
1081 (ASSERT_LT): New macro.
1082 (ASSERT_LT_AT): New macro.
1084 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1086 PR rtl-optimization/84780
1087 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1089 2018-03-14 Martin Liska <mliska@suse.cz>
1091 * tree.c (record_node_allocation_statistics): Use
1092 get_stats_node_kind.
1093 (get_stats_node_kind): New function extracted from
1094 record_node_allocation_statistics.
1095 (free_node): Use get_stats_node_kind.
1097 2018-03-14 Richard Biener <rguenther@suse.de>
1099 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1100 that the value-set of ANTIC_IN doesn't grow.
1103 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1105 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1106 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1108 2018-03-14 Julia Koval <julia.koval@intel.com>
1110 * config.gcc (icelake-client, icelake-server): New.
1112 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1113 (initial_ix86_arch_features): Ditto.
1114 (PTA_SKYLAKE): Add SGX.
1115 (PTA_ICELAKE): Remove.
1116 (PTA_ICELAKE_CLIENT): New.
1117 (PTA_ICELAKE_SERVER): New.
1118 (ix86_option_override_internal): Split up icelake on icelake client and
1120 (get_builtin_code_for_version): Ditto.
1121 (fold_builtin_cpu): Ditto.
1122 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1123 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1124 * config/i386/i386.h (processor_type): Ditto.
1125 * doc/invoke.texi: Ditto.
1127 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1130 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1131 INTEGER_CST offset, add it together with bitpos / 8 and
1132 sign extend based on POINTER_SIZE.
1136 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1139 * config/i386/constraints.md (Yc): New register constraint.
1140 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1141 Use Yc constraint for alternative 2 of operand 0. Remove
1142 preferred_for_speed attribute.
1144 2018-03-14 Richard Biener <rguenther@suse.de>
1146 PR tree-optimization/84830
1147 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1148 with the old one to avoid oscillations.
1150 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1153 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1155 (assign_by_spills): Return a flag of reload assignment failure.
1156 Do not process the reload assignment failures. Do not spill other
1157 reload pseudos if they has the same reg class. Update n if
1159 (lra_assign): Add a return arg. Set up from the result of
1160 assign_by_spills call.
1161 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1162 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1163 usage_insns if it is not NULL.
1164 (spill_hard_reg_in_range): New function.
1165 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1166 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1167 function prototypes.
1168 (lra_assign): Change prototype.
1169 * lra.c (lra): Add code to deal with fails by splitting hard reg
1172 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1174 * config/riscv/riscv.opt (mrelax): New option.
1175 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1176 "norelax" when riscv_mrelax is disabled.
1177 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1179 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1182 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1183 reassociation for int modes.
1185 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1187 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1188 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1190 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1191 * config/aarch64/aarch64-sve.md
1192 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1193 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1194 (*extend<mode><Vwide>2): Rename to...
1195 (aarch64_sve_extend<mode><Vwide>2): ...this.
1196 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1197 renaming the old pattern to...
1198 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1200 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1201 define_expand, renaming the old pattern to...
1202 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1203 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1204 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1205 account when deciding which SVE instruction the optab should use.
1206 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1208 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1210 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1211 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1212 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1213 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1214 (tlsdesc_small_<mode>): Turn a define_expand and use
1215 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1216 (tlsdesc_small_advsimd_<mode>): ...this.
1217 (tlsdesc_small_sve_<mode>): New pattern.
1219 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1221 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1222 (UNSPEC_UMUL_HIGHPART): New constants.
1223 (MUL_HIGHPART): New int iteraor.
1224 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1225 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1227 (*<su>mul<mode>3_highpart): New define_insn.
1229 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1232 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1235 2018-03-13 Martin Liska <mliska@suse.cz>
1238 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1240 (sem_item_optimizer::~sem_item_optimizer): Release it.
1241 (sem_item_optimizer::merge_classes): Register variable aliases.
1242 (sem_item_optimizer::fixup_pt_set): New function.
1243 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1244 * ipa-icf.h: Declare new variables and functions.
1246 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1249 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1250 integer_pow2p@2 and test integer_pow2p in condition.
1251 (A < 0 ? C : 0): Similarly for @1.
1254 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1255 characters starting at p contain '\0' character, don't look beyond
1259 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1260 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1263 * reg-stack.c (change_stack): Change update_end var from int to
1264 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1265 also call set_block_for_insn on the newly added insns and rescan.
1268 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1269 on the last operand.
1272 * tree.c (stabilize_reference_1): Return save_expr (e) for
1273 STATEMENT_LIST even if it doesn't have side-effects.
1275 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1277 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1279 2018-03-12 Renlin Li <renlin.li@arm.com>
1281 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1282 aarch64_output_scalar_simd_mov_immediate.
1284 2018-03-12 Martin Sebor <msebor@redhat.com>
1286 PR tree-optimization/83456
1287 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1288 for perfectly overlapping calls to memcpy.
1289 (gimple_fold_builtin_memory_chk): Same.
1290 (gimple_fold_builtin_strcpy): Handle no-warning.
1291 (gimple_fold_builtin_stxcpy_chk): Same.
1292 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1294 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1296 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1297 parameter. Use it for SFmode.
1298 (rs6000_function_arg_advance_1): Adjust.
1299 (rs6000_function_arg): Adjust.
1300 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1302 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1304 PR rtl-optimization/84169
1305 PR rtl-optimization/84780
1306 * combine.c (can_combine_p): Check for a 2-insn combination whether
1307 the destination register is used between the two insns, too.
1309 2018-03-12 Richard Biener <rguenther@suse.de>
1311 PR tree-optimization/84803
1312 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1313 for refs DR analysis didn't process.
1315 2018-03-12 Richard Biener <rguenther@suse.de>
1317 PR tree-optimization/84777
1318 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1319 force-vectorize loops ignore whether we are optimizing for size.
1321 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1323 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1324 (TARGET_MD_ASM_ADJUST): Define.
1326 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1327 Kito Cheng <kito.cheng@gmail.com>
1328 Chung-Ju Wu <jasonwucj@gmail.com>
1330 * config/nds32/nds32.c (nds32_compute_stack_frame,
1331 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1332 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1333 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1334 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1335 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1336 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1337 * config/nds32/nds32.md (prologue, epilogue): Use macro
1338 NDS32_V3PUSH_AVAILABLE_P to do checking.
1340 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1343 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1344 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1345 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1346 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1347 addition of most attributes on !orig_type_die or the attribute not
1348 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1350 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1351 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1355 * config/nds32/nds32.opt (mvh): New option.
1357 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1358 Chung-Ju Wu <jasonwucj@gmail.com>
1360 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1362 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1363 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1366 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1367 Chung-Ju Wu <jasonwucj@gmail.com>
1369 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1371 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1372 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1374 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1375 Kito Cheng <kito.cheng@gmail.com>
1376 Chung-Ju Wu <jasonwucj@gmail.com>
1378 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1379 UNSPEC_FFMISM and UNSPEC_FLMISM.
1380 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1381 for ffb, ffmism and flmism.
1382 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1383 (unspec_ffmism): Ditto.
1384 (unspec_flmism): Ditto.
1385 (nds32_expand_builtin_impl): Check if string extension is available.
1386 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1387 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1389 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1392 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1395 * lra-assigns.c (assign_by_spills): Return a flag of reload
1396 assignment failure. Do not process the reload assignment
1397 failures. Do not spill other reload pseudos if they has the same
1399 (lra_assign): Add a return arg. Set up from the result of
1400 assign_by_spills call.
1401 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1402 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1403 usage_insns if it is not NULL.
1404 (spill_hard_reg_in_range): New function.
1405 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1406 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1407 function prototypes.
1408 (lra_assign): Change prototype.
1409 * lra.c (lra): Add code to deal with fails by splitting hard reg
1412 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1415 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1417 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1420 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1421 (dw_val_node): Add val_symbolic_view.
1422 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1423 (symview_upper_bound): New.
1424 (new_line_info_table): Initialize symviews_since_reset.
1425 (dwarf2out_source_line): Count symviews_since_reset and set
1426 symview_upper_bound.
1427 (dw_val_equal_p): Handle symview.
1428 (add_AT_symview): New.
1429 (print_dw_val): Handle symview.
1430 (attr_checksum, attr_checksum_ordered): Likewise.
1431 (same_dw_val_p, size_of_die): Likewise.
1432 (value_format, output_die): Likewise.
1433 (add_high_low_attributes): Use add_AT_symview for entry_view.
1434 (dwarf2out_finish): Reset symview_upper_bound, clear
1437 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1440 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1441 Add strict argument and use it.
1442 (rs6000_split_multireg_move): Update for new strict argument.
1443 (mem_operand_gpr): Disallow all non-offsettable addresses.
1444 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1446 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1449 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1450 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1451 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1454 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1455 decl, use remap_type if we want to use the type.
1457 2018-03-09 Martin Sebor <msebor@redhat.com>
1459 PR tree-optimization/84526
1460 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1462 (builtin_access::generic_overlap): Be prepared to handle non-array
1465 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1467 PR rtl-optimization/84682
1468 * lra-constraints.c (process_address_1): Check is_address flag
1469 for address constraints.
1470 (process_alt_operands): Likewise.
1471 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1472 preprocess_constraints.
1473 * recog.h (preprocess_constraints): Add oploc parameter.
1475 * recog.c (preprocess_constraints): Test address_operand for
1476 CT_ADDRESS constraints.
1478 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1481 * lra-assigns.c (assign_by_spills): Return a flag of reload
1482 assignment failure. Do not process the reload assignment
1483 failures. Do not spill other reload pseudos if they has the same
1485 (lra_assign): Add a return arg. Set up from the result of
1486 assign_by_spills call.
1487 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1488 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1489 usage_insns if it is not NULL.
1490 (spill_hard_reg_in_range): New function.
1491 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1492 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1493 function prototypes.
1494 (lra_assign): Change prototype.
1495 * lra.c (lra): Add code to deal with fails by splitting hard reg
1498 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1501 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1502 Accept complain bool parameter. Only emit errors if it is true.
1503 (arm_parse_cpu_option_name): Likewise.
1504 (arm_target_thumb_only): Adjust callers of the above.
1505 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1506 prototype to take a default true bool parameter.
1507 (arm_parse_arch_option_name): Likewise.
1509 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1510 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1514 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1515 * configure: Regenerate.
1516 * configure.ac ("linker --version-script option"): New.
1517 ("linker soname option"): New.
1519 2018-03-09 Richard Biener <rguenther@suse.de>
1521 PR tree-optimization/84775
1522 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1523 immediate uses of predicate stmts and mark them modified.
1526 PR tree-optimization/84178
1527 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1529 (version_loop_for_if_conversion): Delay update_ssa call.
1530 (tree_if_conversion): Delay update_ssa until after predicate
1533 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1536 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1537 when the function accesses prior frames.
1539 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1542 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1543 gen_llsym, otherwise call maybe_gen_llsym.
1546 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1547 has ',' character inside of it.
1549 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1552 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1553 as clobbering CC_REGNUM.
1555 2018-03-08 Richard Biener <rguenther@suse.de>
1558 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1559 (follow_copies_to_constant): Do not follow SSA names registered
1562 2018-03-08 Richard Biener <rguenther@suse.de>
1564 PR tree-optimization/84178
1565 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1567 (version_loop_for_if_conversion): Delay update_ssa call.
1568 (tree_if_conversion): Delay update_ssa until after predicate
1571 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1573 PR tree-optimization/84178
1574 * tree-if-conv.c (release_bb_predicate): Remove the
1575 the assertion that the stmts have NULL use_ops.
1576 Discard the statements, asserting that they haven't
1577 yet been added to a BB.
1579 2018-03-08 Richard Biener <rguenther@suse.de>
1581 PR tree-optimization/84746
1582 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1583 (phi_translate): Pass in destination ANTIC_OUT set.
1584 (phi_translate_1): Likewise. For a simplified result lookup
1585 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1586 (phi_translate_set): Adjust.
1587 (do_pre_regular_insertion): Likewise.
1588 (do_pre_partial_partial_insertion): Likewise.
1590 2018-03-08 Martin Liska <mliska@suse.cz>
1592 PR gcov-profile/84735
1593 * doc/gcov.texi: Document usage of profile files.
1594 * gcov-io.h: Document changes in the format.
1596 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1600 * dwarf2out.c (struct dw_line_info_table): Update comments for
1602 (FORCE_RESET_NEXT_VIEW): New.
1603 (FORCE_RESETTING_VIEW_P): New.
1604 (RESETTING_VIEW_P): Check for -1 too.
1605 (ZERO_VIEW_P): Likewise.
1606 (new_line_info_table): Force-reset next view.
1607 (dwarf2out_begin_function): Likewise.
1608 (dwarf2out_source_line): Simplify zero_view_p initialization.
1609 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1610 view directly. Omit view when omitting .loc at line 0.
1612 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1614 PR tree-optimization/84740
1615 * tree-switch-conversion.c (process_switch): Call build_constructors
1616 only if info.phi_count is non-zero.
1618 PR tree-optimization/84739
1619 * tree-tailcall.c (find_tail_calls): Check call arguments against
1620 DECL_ARGUMENTS (current_function_decl) rather than
1621 DECL_ARGUMENTS (func) when checking for tail recursion.
1623 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1625 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1626 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1627 Volker Reichelt's entry and add entries for people that perform
1628 GCC fuzzy testing and report numerous bugs.
1630 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1633 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1634 readonly data in sdata, if that is disabled.
1635 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1636 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1637 -mreadonly-in-sdata option.
1639 2018-03-07 Martin Sebor <msebor@redhat.com>
1641 PR tree-optimization/84468
1642 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1643 basic block when looking for nul assignment.
1645 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1648 * except.h (output_function_exception_table): Adjust prototype.
1649 * except.c (output_function_exception_table): Remove FNNAME parameter
1650 and add SECTION parameter. Ouput one part of the table at a time.
1651 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1652 the first part of the exception table and emit unwind directives.
1653 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1654 (i386_pe_seh_cold_init): Likewise.
1655 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1656 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1657 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1658 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1659 * config/i386/winnt.c: Include except.h.
1660 (struct seh_frame_state): Add reg_offset, after_prologue and
1661 in_cold_section fields.
1662 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1663 (i386_pe_seh_cold_init): New function.
1664 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1665 to seh->in_cold_section.
1666 (seh_emit_push): Record the offset of the push.
1667 (seh_emit_save): Record the offet of the save.
1668 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1669 Test seh->after_prologue to disregard the epilogue.
1670 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1671 (i386_pe_end_cold_function): New function.
1673 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1676 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1677 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1680 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1682 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1683 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1686 * multiple_target.c: Include tree-inline.h and intl.h.
1687 (expand_target_clones): Diagnose and fail if node->definition and
1688 !tree_versionable_function_p (node->decl).
1690 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1692 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1694 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1695 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1696 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1698 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1701 * combine.c (try_combine): Use reg_or_subregno instead of handling
1702 just paradoxical SUBREGs and REGs.
1704 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1706 * config/arc/arc.c (arc_finalize_pic): Remove function.
1707 (arc_must_save_register): We use single base PIC register, remove
1708 checks to save/restore the PIC register.
1709 (arc_expand_prologue): Likewise.
1710 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1712 (arc_verify_short): Likewise.
1713 (arc_attr_type): Likewise.
1714 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1715 (walk_stores): Likewise.
1716 (arc_address_cost): Make it static.
1717 (arc_verify_short): Likewise.
1718 (branch_dest): Likewise.
1719 (arc_attr_type): Likewise.
1720 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1721 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1722 (arc_final_prescan_insn): Remove inserting the nops due to
1723 hardware hazards. It is done in reorg step.
1724 (insn_length_variant_t): Remove.
1725 (insn_length_parameters_t): Likewise.
1726 (arc_insn_length_parameters): Likewise.
1727 (arc_get_insn_variants): Likewise.
1728 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1730 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1733 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1736 PR tree-optimization/84687
1737 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1739 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1741 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1743 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1744 Rename to ppc_speculation_barrier.
1745 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1746 __builtin_ppc_speculation_barrier.
1748 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1751 * combine.c (combine_simplify_rtx): Don't try to simplify if
1752 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1755 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1757 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1758 to 32 bytes when compiling for POWER9.
1760 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1763 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1764 regparm >= 3 with no arg reg available also for calls with
1765 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1768 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1770 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1772 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1775 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1777 2018-03-05 Richard Biener <rguenther@suse.de>
1779 PR tree-optimization/84486
1780 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1781 When inserting a __builtin_assume_aligned call set the LHS
1782 SSA name alignment info accordingly.
1784 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1786 PR tree-optimization/84114
1787 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1788 Avoid reassociation of FLOAT_MODE addition.
1790 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1792 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1793 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1794 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1795 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1796 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1797 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1798 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1800 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1801 __builtin_ia32_wbinvd): New builtins.
1802 (SPECIAL_ARGS2): New.
1803 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1804 (SPECIAL_ARGS2): New.
1805 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1806 (ix86_valid_target_attribute_inner_p): Ditto.
1807 (ix86_init_mmx_sse_builtins): Add special_args2.
1808 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1809 TARGET_WBNOINVD_P): New.
1810 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1811 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1812 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1813 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1814 * config/i386/pconfigintrin.h: New file.
1815 * config/i386/wbnoinvdintrin.h: Ditto.
1816 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1817 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1819 2018-03-05 Richard Biener <rguenther@suse.de>
1821 PR tree-optimization/84670
1822 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1824 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1825 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1826 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1827 doesn't grow if all successors (recursively) were visited at least
1830 2018-03-05 Richard Biener <rguenther@suse.de>
1832 PR tree-optimization/84650
1833 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1834 if executed in the loop pipeline.
1836 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1838 * doc/configfiles.texi (Configuration Files): Move info about
1839 conditionalizing $target-protos.h to...
1840 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1841 differs from $target-protos.h.
1843 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1844 Chung-Ju Wu <jasonwucj@gmail.com>
1846 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1847 * config/nds32/nds32-multiple.md (setmemsi): Define.
1848 * config/nds32/nds32-memory-manipulation.c
1849 (nds32_gen_dup_4_byte_to_word_value): New.
1850 (emit_setmem_word_loop): New.
1851 (emit_setmem_byte_loop): New.
1852 (nds32_expand_setmem_loop): New.
1853 (nds32_expand_setmem_loop_v3m): New.
1854 (nds32_expand_setmem_unroll): New.
1855 (nds32_expand_setmem): New.
1857 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1858 Chung-Ju Wu <jasonwucj@gmail.com>
1860 * config/nds32/nds32-memory-manipulation.c
1861 (nds32_emit_load_store): New.
1862 (nds32_emit_post_inc_load_store): New.
1863 (nds32_emit_mem_move): New.
1864 (nds32_emit_mem_move_block): New.
1865 (nds32_expand_movmemsi_loop_unknown_size): New.
1866 (nds32_expand_movmemsi_loop_known_size): New.
1867 (nds32_expand_movmemsi_loop): New.
1868 (nds32_expand_movmemsi_unroll): New.
1869 (nds32_expand_movmemqi): Rename ...
1870 (nds32_expand_movmemsi): ... to this.
1871 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1872 (movmemsi): ... to this.
1873 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1874 (nds32_expand_movmemsi): ... to this.
1876 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1877 Monk Chiang <sh.chiang04@gmail.com>
1878 Chung-Ju Wu <jasonwucj@gmail.com>
1880 * config/nds32/nds32-protos.h
1881 (nds32_expand_load_multiple): New arguments.
1882 (nds32_expand_store_multiple): Ditto.
1883 (nds32_valid_multiple_load_store): Rename ...
1884 (nds32_valid_multiple_load_store_p): ... to this.
1885 * config/nds32/nds32-memory-manipulation.c
1886 (nds32_expand_load_multiple): Refine implementation.
1887 (nds32_expand_store_multiple): Ditto.
1888 * config/nds32/nds32-multiple.md
1889 (load_multiple): Update nds32_expand_load_multiple interface.
1890 (store_multiple): Update nds32_expand_store_multiple interface.
1891 * config/nds32/nds32-predicates.c
1892 (nds32_valid_multiple_load_store): Rename ...
1893 (nds32_valid_multiple_load_store_p): ... to this and refine
1895 * config/nds32/predicates.md
1896 (nds32_load_multiple_and_update_address_operation): New predicate.
1897 (nds32_store_multiple_and_update_address_operation): New predicate.
1899 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1900 Chung-Ju Wu <jasonwucj@gmail.com>
1902 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1903 (combo): New attribute.
1904 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1906 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1908 * config/nds32/nds32.opt: Change -mcmodel= default value.
1910 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1911 Monk Chiang <sh.chiang04@gmail.com>
1912 Chung-Ju Wu <jasonwucj@gmail.com>
1914 * config/nds32/constants.md (unspec_element): New enum.
1915 * config/nds32/constraints.md (Umw): New constraint.
1916 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1917 * config/nds32/nds32-intrinsic.md: Likewise.
1918 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1919 (nds32_valid_smw_lwm_base_p): New.
1920 (nds32_output_smw_single_word): New.
1921 (nds32_output_lmw_single_word): New.
1922 (nds32_expand_unaligned_load): New.
1923 (nds32_expand_unaligned_store): New.
1924 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1925 (nds32_output_smw_single_word): Declare.
1926 (nds32_output_lmw_single_word): Declare.
1927 (nds32_expand_unaligned_load): Declare.
1928 (nds32_expand_unaligned_store): Declare.
1929 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1930 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1931 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1932 NDS32_BUILTIN_UASTORE_DW.
1933 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1936 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1937 Kito Cheng <kito.cheng@gmail.com>
1938 Chung-Ju Wu <jasonwucj@gmail.com>
1940 * config/nds32/nds32-intrinsic.c
1941 (nds32_expand_builtin_null_ftype_reg): Delete.
1942 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1943 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1944 (nds32_read_argument): New.
1945 (nds32_legitimize_target): Ditto.
1946 (nds32_legitimize_argument): Ditto.
1947 (nds32_check_constant_argument): Ditto.
1948 (nds32_expand_unop_builtin): Ditto.
1949 (nds32_expand_unopimm_builtin): Ditto.
1950 (nds32_expand_binop_builtin): Ditto.
1951 (nds32_builtin_decl_impl): Ditto.
1952 (builtin_description): Ditto.
1953 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1954 (nds32_init_builtins_impl): Ditto.
1955 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1956 (nds32_builtin_decl): New.
1957 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1958 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1960 2018-03-02 Jeff Law <law@redhat.com>
1962 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1963 (redundant_insn, fill_simple_delay_slots): Likewise.
1964 (fill_slots_from_thread): Likewise.
1965 * resource.c (mark_referenced_resources): Likewise.
1966 (mark_set_resources, find_dead_or_set_registers): Likewise.
1968 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1970 * substring-locations.h (format_warning_va): Formatting fix for
1972 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1974 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1975 * substring-locations.c: Include intl.h.
1976 (format_warning_va): Turned into small wrapper around
1977 format_warning_n_va, renamed to ...
1978 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1979 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1981 (format_warning_at_substring_n): New function.
1982 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1983 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1984 format_warning_at_substring with just a shorter name instead of
1985 const function pointer.
1986 (fmtwarn_n): New function.
1987 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1988 appropriate, get rid of all the fmtstr temporaries, move conditionals
1989 with G_() wrapped string literals directly into fmtwarn arguments,
1990 cast dir.len to (int), formatting fixes.
1992 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1994 * doc/invoke.texi: Remove "Cilk Plus" references.
1996 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1997 Richard Biener <rguenther@suse.de>
2000 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2001 for error or warning attributes if CALL_FROM_THUNK_P is set.
2004 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2007 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2008 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2011 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2012 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2014 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2015 instead of -1U in last predictors element's probability member.
2017 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2020 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2021 arguments if they are comparable.
2023 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2025 PR tree-optimization/84634
2026 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2027 masks and masked_loop_p with a single loop_masks, making sure it's
2028 null for bb vectorization.
2030 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2032 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2033 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2034 loop->force_vectorize to check whether there is no alias.
2036 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2039 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2041 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2043 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2044 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2045 instead of a loop around prev_real_insn.
2046 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2050 * config/i386/i386.c (ix86_print_operand): Use conditional
2051 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2054 2018-03-02 Richard Biener <rguenther@suse.de>
2056 PR tree-optimization/84427
2057 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2058 (bitmap_set_subtract_values): Rewrite to handle multiple
2061 (prune_clobbered_mems): Likewise.
2062 (phi_translate): Take edge instead of pred/phiblock.
2063 (phi_translate_1): Likewise.
2064 (phi_translate_set): Likewise. Insert all translated
2065 exprs for a value into the set, keeping possibly multiple
2066 expressions per value.
2067 (compute_antic_aux): Adjust for phi_translate changes.
2068 When intersecting union the expressions and prune those
2069 not in the final value set, keeping possibly multiple
2070 expressions per value. Do not use value-insertion
2071 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2072 all expressions. Add verification that the value-sets
2073 only shrink during iteration.
2074 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2075 (do_pre_regular_insertion): Likewise.
2076 (do_pre_partial_partial_insertion): Likewise.
2078 2018-03-02 Richard Biener <rguenther@suse.de>
2081 * config/darwin.c (saved_debug_info_level): New static global.
2082 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2083 (darwin_asm_lto_end): Restore debug info generation settings.
2085 2018-03-01 Martin Liska <mliska@suse.cz>
2088 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2091 2018-03-01 Richard Biener <rguenther@suse.de>
2094 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2095 type annotation with LTO when debug was disabled at compile-time.
2097 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2099 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2101 (mips_final_postscan_insn): Likewise.
2103 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2105 PR rtl-optimization/84528
2106 * alias.c (init_alias_target): Add commentary.
2107 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2108 a unique base value if the frame pointer is not eliminated
2109 to the stack pointer.
2111 2018-03-01 Tom de Vries <tom@codesourcery.com>
2113 PR rtl-optimization/83327
2114 * lra-int.h (hard_regs_spilled_into): Declare.
2115 * lra.c (hard_regs_spilled_into): Define.
2116 (init_reg_info): Init hard_regs_spilled_into.
2117 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2118 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2119 (process_bb_lives): Handle hard_regs_spilled_into.
2120 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2121 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2123 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2125 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2126 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2127 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2128 * config/rs6000/aix72.h: New file.
2130 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2132 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2133 instead of warning_at with conditional singular and plural messages
2137 * stor-layout.c (update_alignment_for_field): For
2138 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2139 && !DECL_PACKED (field), do the alignment update, just use
2140 only desired_align instead of MAX (type_align, desired_align)
2142 (place_field): Don't do known_align < desired_align handling
2143 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2144 is non-NULL, instead do it after rli->prev_field handling and
2145 only if not within a bitfield word. For DECL_PACKED (field)
2146 use type_align of BITS_PER_UNIT.
2148 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2150 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2151 superfluous parentheses and trailing spaces.
2153 2018-02-28 Richard Biener <rguenther@suse.de>
2155 PR tree-optimization/84584
2156 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2157 SCoPs with fake exit edge.
2159 2018-02-28 Martin Liska <mliska@suse.cz>
2162 * timevar.c (timer::print): Fix format to properly print 100%
2165 2018-02-28 Richard Biener <rguenther@suse.de>
2168 * genmatch.c (capture_info::walk_match): Do not mark
2169 captured expressions without operands as expr_p given
2170 they act more like predicates and should be subject to
2171 "lost tail" side-effect preserving.
2173 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2175 PR rtl-optimization/81611
2176 * auto-inc-dec.c (attempt_change): Move dead note from
2177 mem_insn if it's the next use of regno
2178 (find_address): Take address use of reg holding
2179 non-incremented value. Add parm to limit search to the named
2181 (merge_in_block): Attempt to use a mem insn that is the next
2182 use of the original regno.
2184 2018-02-27 Martin Sebor <msebor@redhat.com>
2187 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2188 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2190 2018-02-27 Martin Sebor <msebor@redhat.com>
2192 PR translation/84207
2193 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2194 n argument to unsigned HOST_WIDE_INT.
2195 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2196 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2197 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2198 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2200 2018-02-27 Richard Biener <rguenther@suse.de>
2202 PR tree-optimization/84512
2203 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2204 Do not use the estimate returned from record_stmt_cost for
2205 the scalar iteration cost but sum properly using add_stmt_cost.
2207 2018-02-27 Richard Biener <rguenther@suse.de>
2209 PR tree-optimization/84466
2210 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2211 Adjust last change to less strictly validate use operands.
2213 2018-02-27 Martin Liska <mliska@suse.cz>
2215 PR gcov-profile/84548
2216 * gcov.c (process_file): Allow partial overlap and consider it
2217 also as group functions.
2218 (output_lines): Properly calculate range of lines for a group.
2220 2018-02-27 Martin Liska <mliska@suse.cz>
2222 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2223 'ggc' suffixes. Change first column width.
2224 (timer::print): Fix formatting of the column.
2226 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2228 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2229 preserve inline entry blocks for the sake of debug inline
2230 entry point markers alone.
2231 (remove_unused_locals): Suggest in comments a better place to
2232 force the preservation of inline entry blocks that are
2233 otherwise unused, but do not preserve them.
2235 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2237 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2239 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2242 * config/i386/constraints.md (Bs): Replace
2243 ix86_indirect_branch_register with
2244 TARGET_INDIRECT_BRANCH_REGISTER.
2246 * config/i386/i386.md (indirect_jump): Likewise.
2247 (tablejump): Likewise.
2248 (*sibcall_memory): Likewise.
2249 (*sibcall_value_memory): Likewise.
2250 Peepholes of indirect call and jump via memory: Likewise.
2251 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2252 (*sibcall_value_GOT_32): Likewise.
2253 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2254 (GOT_memory_operand): Likewise.
2255 (call_insn_operand): Likewise.
2256 (sibcall_insn_operand): Likewise.
2257 (GOT32_symbol_operand): Likewise.
2258 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2260 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2262 PR rtl-optimization/83496
2263 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2264 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2265 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2266 redundant insn, if any.
2267 (relax_delay_slots): Likewise.
2268 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2270 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2272 PR tree-optimization/83965
2273 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2274 that grouped statements are part of a reduction chain. Return
2275 true if the statement is not marked as a reduction itself but
2277 (vect_recog_dot_prod_pattern): Don't check whether the statement
2278 is part of a group here.
2279 (vect_recog_sad_pattern): Likewise.
2280 (vect_recog_widen_sum_pattern): Likewise.
2282 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2285 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2287 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2290 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2292 (ix86_output_indirect_function_return): New prototype.
2293 (ix86_split_simple_return_pop_internal): Likewise.
2294 * config/i386/i386.c (indirect_return_via_cx): New.
2295 (indirect_return_via_cx_bnd): Likewise.
2296 (indirect_thunk_name): Handle return va CX_REG.
2297 (output_indirect_thunk_function): Create alias for
2298 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2299 (ix86_output_indirect_jmp): Remove the bool argument.
2300 (ix86_output_indirect_function_return): New function.
2301 (ix86_split_simple_return_pop_internal): Likewise.
2302 * config/i386/i386.md (*indirect_jump): Don't pass false
2303 to ix86_output_indirect_jmp.
2304 (*tablejump_1): Likewise.
2305 (simple_return_pop_internal): Change it to define_insn_and_split.
2306 Call ix86_split_simple_return_pop_internal to split it for
2308 (simple_return_indirect_internal): Call
2309 ix86_output_indirect_function_return instead of
2310 ix86_output_indirect_jmp.
2312 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2315 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2316 memset and value initialization afterwards.
2318 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2320 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2322 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2325 * common/config/aarch64/aarch64-common.c
2326 (aarch_option_optimization_table[]): Switch
2327 off fomit-frame-pointer
2329 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2330 Chung-Ju Wu <jasonwucj@gmail.com>
2332 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2334 (store_multiple): Ditto.
2336 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2338 * config.gcc: Add --with-cpu support for nds32 target.
2339 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2340 * config/nds32/nds32.opt: Add -mcpu= option.
2342 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2344 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2345 isel=yes): Warn for these deprecated options.
2347 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2349 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2350 ISA_2_5_MASKS_EMBEDDED.
2352 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2354 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2355 p->max as pointers rather than using iterative_hash_expr.
2357 2018-02-23 Carl Love <cel@us.ibm.com>
2359 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2360 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2362 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2363 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2364 P8V_BUILTIN_VEC_VUNSIGNED2.
2366 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2369 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2370 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2372 * lra-constraints.c (curr_insn_transform): Set up
2373 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2375 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2376 * lra-eliminations.c (spill_pseudos): Ditto.
2377 (process_insn_for_elimination): Ditto.
2378 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2379 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2380 LRA_NON_CLOBBERED_ALT.
2382 2018-02-22 Martin Sebor <msebor@redhat.com>
2384 PR tree-optimization/84480
2385 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2386 to maybe_diag_stxncpy_trunc. Call it.
2387 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2388 from gimple_fold_builtin_strcpy. Print inlining stack.
2389 (handle_builtin_stxncpy): Print inlining stack.
2390 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2392 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2395 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2396 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2397 and -fcheck-pointer-bounds are used together.
2398 (indirect_thunk_prefix): New enum.
2399 (indirect_thunk_need_prefix): New function.
2400 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2401 "_nt" instead of "_bnd" for NOTRACK prefix.
2402 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2403 (output_indirect_thunk_function): Likewise.
2405 (ix86_code_end): Update output_indirect_thunk_function calls.
2406 (ix86_output_indirect_branch_via_reg): Replace
2407 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2408 (ix86_output_indirect_branch_via_push): Likewise.
2409 (ix86_output_function_return): Likewise.
2410 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2411 incompatible with -fcf-protection=branch and
2412 -fcheck-pointer-bounds.
2414 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2417 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2418 Change gcc_assert call to output_operand_lossage.
2420 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2422 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2424 2018-02-22 DJ Delorie <dj@redhat.com>
2425 Sebastian Perta <sebastian.perta@renesas.com>
2426 Oleg Endo <olegendo@gcc.gnu.org>
2428 * config/rx/rx.c (rx_rtx_costs): New function.
2429 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2431 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2433 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2435 2018-02-22 Martin Liska <mliska@suse.cz>
2438 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2439 Add "native" as a possible value.
2441 2018-02-22 Martin Liska <mliska@suse.cz>
2444 * config/i386/i386.c (ix86_option_override_internal):
2445 Add "native" as a possible value for -march and -mtune.
2447 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2450 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2451 to all type variants.
2453 PR tree-optimization/84503
2454 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2455 width as info->bitpos + info->bitsize - start.
2456 (merged_store_group::merge_overlapping): Simplify width computation.
2457 (check_no_overlap): New function.
2458 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2459 start + width and last_order of the group, fail if check_no_overlap
2461 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2462 to group if check_no_overlap fails.
2464 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2466 * config/rs6000/altivec.md: Delete contraint arguments to
2467 define_expand, define_split, and define_peephole2, and in
2468 define_insn_and_split if always unused.
2469 * config/rs6000/darwin.md: Ditto.
2470 * config/rs6000/dfp.md: Ditto.
2471 * config/rs6000/rs6000.md: Ditto.
2472 * config/rs6000/sync.md: Ditto.
2473 * config/rs6000/vector.md: Ditto.
2474 * config/rs6000/vsx.md: Ditto.
2476 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2478 * config/rs6000/altivec.md: Write output control strings as braced
2479 blocks instead of double-quoted strings.
2480 * config/rs6000/darwin.md: Ditto.
2481 * config/rs6000/rs6000.md: Ditto.
2482 * config/rs6000/vector.md: Ditto.
2483 * config/rs6000/vsx.md: Ditto.
2485 2018-02-21 Jason Merrill <jason@redhat.com>
2487 PR c++/84314 - ICE with templates and fastcall attribute.
2488 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2490 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2492 * ipa-cp.c (determine_versionability): Fix comment typos.
2494 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2497 * ipa-cp.c (determine_versionability): Do not version functions caling
2500 2018-02-21 Martin Liska <mliska@suse.cz>
2503 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2504 Add "native" as a possible value.
2505 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2506 the macro when native cpu detection is available.
2508 2018-02-21 Martin Liska <mliska@suse.cz>
2511 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2512 Add "native" as a possible value.
2513 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2514 when native cpu detection is available.
2516 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2517 Martin Sebor <msebor@redhat.com>
2519 PR tree-optimization/84478
2520 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2522 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2523 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2524 support which is conservatively correct, for 2 only stay conservative
2525 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2526 argument to the 2 argument get_range_strlen, adjust 6 arg
2527 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2529 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2530 (gimple_fold_builtin_strlen): Pass true as last argument to
2533 2018-02-20 Martin Sebor <msebor@redhat.com>
2536 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2537 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2538 (builtin_memref::builtin_memref): Factor out parts into
2539 set_base_and_offset and call it.
2541 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2544 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2545 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2546 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2547 search at the associated MODE_INT.
2549 2018-02-20 Jeff Law <law@redhat.com>
2552 PR tree-optimization/81592
2554 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2555 for range data rather than using global data.
2556 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2557 range data rather than using global data.
2558 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2559 pass it to children as needed.
2560 (struct directive::fmtresult): Similarly.
2561 (struct directive::set_width): Similarly.
2562 (struct directive::set_precision): Similarly.
2563 (format_integer, format_directive, parse_directive): Similarly.
2564 (format_none): Accept unnamed vr_values parameter.
2565 (format_percent, format_floating, format_character): Similarly.
2566 (format_string, format_plain): Similarly.
2567 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2568 the EVRP range analyzer for range data rather than using global data.
2569 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2570 gimple-ssa-evrp-analyze.h
2571 (class sprintf_dom_walker): Add after_dom_children member function.
2572 Add evrp_range_analyzer member.
2573 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2574 range analyzer as needed.
2575 (sprintf_dom_walker::after_dom_children): New member function.
2576 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2578 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2579 (evrp_range_analyzer::pop_to_marker): Likewise.
2581 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2583 PR tree-optimization/84419
2584 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2585 with the required type if its current type is compatible but
2588 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2591 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2592 after vectorization.
2594 2018-02-20 Martin Liska <mliska@suse.cz>
2597 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2598 possible values if we don't have a hint.
2600 2018-02-20 Martin Liska <mliska@suse.cz>
2604 * final.c (shorten_branches): Build align_tab array with one
2606 * opts.c (finish_options): Add alignment option limit check.
2607 (MAX_CODE_ALIGN): Likewise.
2608 (MAX_CODE_ALIGN_VALUE): Likewise.
2609 * doc/invoke.texi: Document maximum allowed option value for
2610 all -falign-* options.
2612 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2615 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2616 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2617 * var-tracking.c (emit_note_insn_var_location): Remove all references
2618 to NOTE_INSN_CALL_ARG_LOCATION.
2619 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2620 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2621 Use copy_rtx_if_shared.
2622 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2623 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2624 (dwarf2out_var_location): Remove handling of
2625 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2627 * final.c (final_scan_insn): Remove all references to
2628 NOTE_INSN_CALL_ARG_LOCATION.
2629 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2630 before dumping final insns.
2631 * except.c (emit_note_eh_region_end): Remove all references to
2632 NOTE_INSN_CALL_ARG_LOCATION.
2633 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2634 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2635 * config/arc/arc.c (hwloop_optimize): Likewise.
2636 * config/arm/arm.c (create_fix_barrier): Likewise.
2637 * config/s390/s390.c (s390_chunkify_start): Likewise.
2638 * config/sh/sh.c (find_barrier): Likewise.
2639 * config/i386/i386.c (rest_of_insert_endbranch,
2640 ix86_seh_fixup_eh_fallthru): Likewise.
2641 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2642 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2643 * config/frv/frv.c (frv_function_prologue): Likewise.
2644 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2646 (note_outside_basic_block_p): Remove all references to
2647 NOTE_INSN_CALL_ARG_LOCATION.
2648 * gengtype.c (adjust_field_rtx_def): Likewise.
2649 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2651 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2652 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2655 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2658 PR tree-optimization/84452
2659 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2660 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2663 2018-02-19 Martin Liska <mliska@suse.cz>
2666 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2668 2018-02-19 Martin Liska <mliska@suse.cz>
2669 Richard Sandiford <richard.sandiford@linaro.org>
2671 PR tree-optimization/82491
2672 * gimple-fold.c (get_base_constructor): Make earlier bail out
2675 2018-02-19 Carl Love <cel@us.ibm.com>
2677 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2678 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2680 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2681 P8V_BUILTIN_VEC_NEG.
2683 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2685 * config/rl78/rl78.md (movdf): New define expand.
2687 2018-02-19 Martin Liska <mliska@suse.cz>
2690 * doc/invoke.texi: Fix typo.
2691 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2693 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2695 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2696 handle rs6000_single_float and rs6000_double_float specially for
2699 2018-02-16 Jeff Law <law@redhat.com>
2701 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2704 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2707 * ipa-inline.c (inline_small_functions): Fix a typo.
2709 2018-02-16 Nathan Sidwell <nathan@acm.org>
2711 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2713 2018-02-16 Carl Love <cel@us.ibm.com>
2715 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2716 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2717 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2718 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2719 expansion to P8V_BUILTIN_VEC_FLOAT2.
2721 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2723 PR rtl-optimization/70023
2724 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2725 src_regno into account.
2727 2018-02-16 Carl Love <cel@us.ibm.com>
2729 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2730 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2731 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2732 * config/rs6000/rs6000.c: Remove case statements for
2733 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2734 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2735 and P9V_BUILTIN_VEC_VINSERT4B.
2736 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2737 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2738 * config/rs6000/vsx.md:
2739 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2742 2018-02-16 Carl Love <cel@us.ibm.com>
2744 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2746 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2748 * config/rs6000/rs6000-c.c: Add the definitions for
2749 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2750 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2751 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2752 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2753 definition for insert4b and define insn *insert3b_internal.
2754 * doc/extend.texi: Add documentation for vec_extract4b.
2756 2018-02-16 Nathan Sidwell <nathan@acm.org>
2758 * doc/extend.texi (Backwards Compatibility): Mention friend
2759 injection. Note for-scope is deprecated.
2760 * doc/invoke.texi (-ffriend-injection): Deprecate.
2762 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2764 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2765 that moved to I2, also allow destinations that are a paradoxical
2766 subreg (instead of a normal reg).
2768 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2771 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2774 2018-02-16 Richard Biener <rguenther@suse.de>
2776 PR tree-optimization/84037
2777 PR tree-optimization/84016
2779 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2780 Adjust vec_construct for the fact we need additional higher latency
2781 128bit inserts for AVX256 and AVX512 vector builds.
2782 (ix86_add_stmt_cost): Scale vector construction cost for
2785 2018-02-16 Richard Biener <rguenther@suse.de>
2787 PR tree-optimization/84417
2788 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2789 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2790 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2792 2018-02-16 Martin Liska <mliska@suse.cz>
2795 * internal-fn.def (ASAN_CHECK): Set proper flags.
2796 (ASAN_MARK): Likewise.
2798 2018-02-16 Julia Koval <julia.koval@intel.com>
2800 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2801 from PTA_CANNONLAKE.
2803 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2806 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2807 Use ++iter rather than iter++ for std::list iterators.
2808 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2809 defer deleting them until all nodes in the forest are processed. Do
2810 free even leaf nodes. Change to_process into auto_vec.
2813 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2814 * vec.h (vec_default_construct): Use memset instead of placement new
2815 if BROKEN_VALUE_INITIALIZATION is defined.
2816 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2817 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2820 PR rtl-optimization/83723
2821 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2822 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2823 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2825 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2827 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2829 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2831 PR rtl-optimization/81443
2832 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2833 from inner REGs to paradoxical SUBREGs.
2835 2018-02-16 Richard Biener <rguenther@suse.de>
2837 PR tree-optimization/84399
2838 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2839 For operands we can analyze at their definition make sure we can
2840 analyze them at each use as well.
2842 2018-02-16 Richard Biener <rguenther@suse.de>
2844 PR tree-optimization/84190
2845 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2846 volatile accesses if the decl isn't volatile.
2848 2018-02-15 Jason Merrill <jason@redhat.com>
2850 PR c++/84314 - ICE with templates and fastcall attribute.
2851 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2852 TYPE_CANONICAL on an existing type.
2854 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2856 PR tree-optimization/84383
2857 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2858 dstoff nor call operand_equal_p if dstbase is NULL.
2860 PR tree-optimization/84334
2861 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2862 also a CONSTANT_CLASS_P, punt.
2864 2018-02-14 Jim Wilson <jimw@sifive.com>
2866 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2867 first SMALL_OPERAND check. New local min_second_step. Move assert
2868 to where locals are set. Add TARGET_RVC support.
2869 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2871 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2873 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2875 2018-02-14 Martin Sebor <msebor@redhat.com>
2877 PR tree-optimization/83698
2878 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2879 arrays constrain the offset range to their bounds.
2880 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2881 (builtin_access::overlap): Avoid setting the size of overlap if it's
2883 (maybe_diag_overlap): Also consider arrays when deciding what values
2884 of offsets to include in diagnostics.
2886 2018-02-14 Martin Sebor <msebor@redhat.com>
2889 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2890 that correspond to the kind of a declaration.
2892 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2895 * config/pa/pa.md: Load address of PIC label using the linkage table
2896 if the label is nonlocal.
2898 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2900 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2901 warning message if user requests -maltivec=be.
2902 * doc/invoke.texi: Document deprecation of -maltivec=be.
2904 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2907 * config/rs6000/rs6000-c.c: Update definitions for
2908 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2909 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2911 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2914 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2915 add _get_ssp intrinsics. Remove argument from
2916 __builtin_ia32_rdssp[d|q].
2917 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2918 * config/i386/i386-builtin.def: Remove argument from
2919 __builtin_ia32_rdssp[d|q].
2920 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2921 ix86_expand_special_args_builtin for _rdssp[d|q].
2922 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2923 Clear register before usage.
2924 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2925 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2927 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2929 PR tree-optimization/84357
2930 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2931 operand 1 of an ARRAY_REF too.
2933 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2936 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2937 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2939 (set_of_reg): New struct.
2940 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2941 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2942 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2944 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2945 Split into bitclr, bitset, bitinvert patterns if appropriate.
2946 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2947 use rx_fuse_in_memory_bitop.
2948 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2949 to named insn, correct maximum insn length.
2951 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2954 * machmode.def: Define a complex mode for PARTIAL_INT.
2955 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2957 * doc/rtl.texi: Document CSPImode.
2958 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2960 (msp430_hard_regno_nregs_with_padding): Likewise.
2962 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2965 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2967 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2969 PR rtl-optimization/84169
2970 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2971 we generated a parallel as new i3 and we split that to new i2 and i3
2972 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2973 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2974 those to i2, not i1. Partially rewrite this scan code.
2976 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2979 * stor-layout.c (place_field): For variable length fields, adjust
2980 offset_align afterwards not just based on the field's alignment,
2981 but also on the size.
2984 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2985 of exps and logs in the use_exp2 case.
2987 2018-02-13 Jeff Law <law@redhat.com>
2989 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2992 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2995 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2999 * common.opt (gas-loc-support, gas-locview-support): New.
3000 (ginline-points, ginternal-reset-location-views): New.
3001 * doc/invoke.texi: Document them. Use @itemx where intended.
3002 (gvariable-location-views): Adjust.
3003 * target.def (reset_location_view): New.
3004 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3005 (TARGET_RESET_LOCATION_VIEW): New.
3006 * doc/tm.texi: Rebuilt.
3007 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3008 (dwarf2out_default_as_locview_support): New.
3009 (output_asm_line_debug_info): Use option variables.
3010 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3011 (output_loc_list): Likewise.
3012 (add_high_low_attributes): Check option variables.
3013 Don't output entry view attribute in strict mode.
3014 (gen_inlined_subroutine_die): Check option variables.
3015 (dwarf2out_inline_entry): Likewise.
3016 (init_sections_and_labels): Likewise.
3017 (dwarf2out_early_finish): Likewise.
3018 (maybe_reset_location_view): New, from...
3019 (dwarf2out_var_location): ... here. Call it.
3020 * debug.h (dwarf2out_default_as_loc_support): Declare.
3021 (dwarf2out_default_as_locview_support): Declare.
3022 * hooks.c (hook_int_rtx_insn_0): New.
3023 * hooks.h (hook_int_rtx_insn_0): Declare.
3024 * toplev.c (process_options): Take -gas-loc-support and
3025 -gas-locview-support from dwarf2out. Enable
3026 -gvariable-location-views by default only with locview
3027 assembler support. Enable -ginternal-reset-location-views by
3028 default only if the target defines the corresponding hook.
3029 Enable -ginline-points by default if location views are
3030 enabled; force it disabled if statement frontiers are
3032 * tree-inline.c (expand_call_inline): Check option variables.
3033 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3035 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3037 PR tree-optimization/84321
3038 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3039 handling. Also check whether the anti-range contains any values
3040 that satisfy the mask; switch to a VR_RANGE if not.
3042 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3045 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3047 2018-02-13 Martin Jambor <mjambor@suse.cz>
3050 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3051 of call statements, also set location of a load to a temporary.
3053 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3055 * config/rl78/rl78.c (add_vector_labels): New function.
3056 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3057 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3058 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3059 which checks that no arguments are passed.
3060 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3061 * doc/extend.texi: Documentation for the new attribute.
3063 2018-02-13 Andreas Schwab <schwab@suse.de>
3065 * config/riscv/linux.h (CPP_SPEC): Define.
3067 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3070 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3071 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3072 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3073 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3074 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3075 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3076 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3077 temporarily for AES and PCLMUL builtins.
3079 PR tree-optimization/84339
3080 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3081 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3085 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3086 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3087 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3089 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3091 * omp-simd-clone.h: New file.
3092 * omp-simd-clone.c: Include omp-simd-clone.h.
3093 (expand_simd_clones): No longer static.
3094 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3095 cgraph.h and omp-simd-clone.h.
3096 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3097 (vect_recog_widen_shift_pattern): Formatting fix.
3098 (vect_pattern_recog_1): Don't check optab for calls.
3101 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3102 operands[2] into a REG before using gen_lowpart on it.
3104 2018-02-12 Jeff Law <law@redhat.com>
3107 * config/sh/sh.c (find_barrier): Consider a sibling call
3110 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3111 successfully back substituting a reg.
3113 2018-02-12 Richard Biener <rguenther@suse.de>
3115 PR tree-optimization/84037
3116 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3117 parameter, move visited init to caller.
3118 (vect_slp_analyze_operations): Separate cost from validity
3119 check, initialize visited once for all instances.
3120 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3122 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3124 (vect_model_promotion_demotion_cost): Likewise.
3125 (vectorizable_bswap): Guard cost modeling with !slp_node
3126 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3128 (vectorizable_call): Likewise.
3129 (vectorizable_conversion): Likewise.
3130 (vectorizable_assignment): Likewise.
3131 (vectorizable_shift): Likewise.
3132 (vectorizable_operation): Likewise.
3133 (vectorizable_store): Likewise.
3134 (vectorizable_load): Likewise.
3135 (vectorizable_condition): Likewise.
3136 (vectorizable_comparison): Likewise.
3138 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3141 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3142 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3143 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3145 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3148 * params.def (inline-min-speedup): Increase from 8 to 15.
3149 (max-inline-insns-auto): Decrease from 40 to 30.
3150 * ipa-split.c (consider_split): Add some buffer for function to
3151 be considered inlining candidate.
3152 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3155 2018-02-12 Richard Biener <rguenther@suse.de>
3157 PR tree-optimization/84037
3158 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3159 matched stmts if we cannot swap the non-matched ones.
3161 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3163 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3164 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3165 _mm_maskz_scalef_round_ss): New intrinsics.
3166 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3167 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3168 __builtin_ia32_scalefss_round): Remove.
3169 (__builtin_ia32_scalefsd_mask_round,
3170 __builtin_ia32_scalefss_mask_round): New intrinsics.
3171 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3172 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3173 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3174 "<round_constraint>")): Changed to ...
3175 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3176 "<round_scalar_constraint>")): ... this.
3177 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3178 %0, %1, %2<round_op3>}"): Changed to ...
3179 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3180 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3181 %2<round_scalar_mask_op3>}"): ... this.
3182 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3184 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3186 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3187 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3188 (_mm_maskz_sqrt_round_ss): New intrinsics.
3189 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3190 (__builtin_ia32_sqrtsd_mask_round)
3191 (__builtin_ia32_sqrtss_mask_round): New builtins.
3192 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3193 (__builtin_ia32_sqrtss_round): Remove.
3194 (__builtin_ia32_sqrtsd_mask_round)
3195 (__builtin_ia32_sqrtss_mask_round): New builtins.
3196 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3197 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3198 ((match_operand:VF_128 1 "vector_operand"
3199 "xBm,<round_constraint>")): Changed to ...
3200 ((match_operand:VF_128 1 "vector_operand"
3201 "xBm,<round_scalar_constraint>")): ... this.
3202 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3203 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3204 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3205 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3206 %<iptr>1<round_scalar_mask_op3>}): ... this.
3207 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3208 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3210 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3213 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3214 Cast vec_cmpeq result to correct type.
3215 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3216 Cast vec_cmpgt result to correct type.
3218 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3220 * final.c (final_scan_insn_1): Renamed from...
3221 (final_scan_insn): ... this. New wrapper, to recover
3222 seen from the outermost call in recursive ones.
3223 * config/sparc/sparc.c (output_return): Drop seen from call.
3224 (output_sibcall): Likewise.
3225 * config/visium/visium.c (output_branch): Likewise.
3227 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3229 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3232 2018-02-10 Alan Modra <amodra@gmail.com>
3235 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3236 Specify LR as an input.
3238 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3241 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3242 remove_member_access_dummy_vars): New functions.
3243 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3244 lower_omp_1, execute_lower_omp): Use them.
3246 PR rtl-optimization/84308
3247 * shrink-wrap.c (spread_components): Release todo vector.
3249 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3251 PR rtl-optimization/57193
3252 * ira-color.c (struct allocno_color_data): Add member
3253 conflict_allocno_hard_prefs.
3254 (update_conflict_allocno_hard_prefs): New.
3255 (bucket_allocno_compare_func): Add a preference based on
3256 conflict_allocno_hard_prefs.
3257 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3258 (color_allocnos): Remove a dead code. Initiate
3259 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3261 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3264 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3265 constraint from =wa to wa. Avoid a subreg on the output operand,
3266 instead use a pseudo and subreg it in a move.
3267 (p9_xxbrd_<mode>): Changed to ...
3268 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3269 (p9_xxbrd_v2df): New expander.
3270 (p9_xxbrw_<mode>): Changed to ...
3271 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3272 (p9_xxbrw_v4sf): New expander.
3274 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3276 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3278 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3281 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3282 multiply in 32-bit mode.
3283 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3284 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3287 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3289 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3290 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3291 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3292 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3294 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3297 * dwarf2out.c (is_trivial_indirect_ref): New function.
3298 (dwarf2out_late_global_decl): Do not generate a location
3299 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3300 and that are not defined in the current unit.
3302 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3304 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3305 instead of a libcall for UNORDERED.
3307 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3310 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3311 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3313 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3316 * config/s390/s390.c (s390_set_current_function): Invoke
3317 s390_indirect_branch_settings also if fndecl didn't change.
3319 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3321 * config/rs6000/rs6000.md (blockage): Set length to zero.
3323 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3325 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3327 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3330 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3331 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3335 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3336 PARALLEL incoming that failed vt_get_decl_and_offset check.
3339 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3340 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3342 (get_variable_section): For decls in named .bss* sections pass true as
3343 second argument to bss_initializer_p.
3345 2018-02-09 Marek Polacek <polacek@redhat.com>
3346 Jakub Jelinek <jakub@redhat.com>
3349 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3350 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3351 Sync some changes from cxx_fold_indirect_ref.
3353 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3355 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3357 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3358 (BLOCK_INLINE_ENTRY_LABEL): New.
3359 (dwarf2out_var_location): Disregard inline entry markers.
3360 (inline_entry_data): New struct.
3361 (inline_entry_data_hasher): New hashtable type.
3362 (inline_entry_data_hasher::hash): New.
3363 (inline_entry_data_hasher::equal): New.
3364 (inline_entry_data_table): New variable.
3365 (add_high_low_attributes): Add DW_AT_entry_pc and
3366 DW_AT_GNU_entry_view attributes if a pending entry is found
3367 in inline_entry_data_table. Add old entry_pc attribute only
3368 if debug nonbinding markers are disabled.
3369 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3370 markers are enabled.
3371 (block_within_block_p, dwarf2out_inline_entry): New.
3372 (dwarf2out_finish): Check that no entries remained in
3373 inline_entry_data_table.
3374 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3375 (final_scan_insn, notice_source_line): Likewise.
3376 (rest_of_clean_state): Skip inline entry markers.
3377 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3379 * gimple.c (gimple_build_debug_inline_entry): New.
3380 * gimple.h (enum gimple_debug_subcode): Add
3381 GIMPLE_DEBUG_INLINE_ENTRY.
3382 (gimple_build_debug_inline_entry): Declare.
3383 (gimple_debug_inline_entry_p): New.
3384 (gimple_debug_nonbind_marker_p): Adjust.
3385 * insn-notes.def (INLINE_ENTRY): New.
3386 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3387 inline entry marker notes.
3388 (print_insn): Likewise.
3389 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3390 (INSN_DEBUG_MARKER_KIND): Likewise.
3391 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3392 * tree-inline.c (expand_call_inline): Build and insert
3393 debug_inline_entry stmt.
3394 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3395 inline entry blocks early, if nonbind markers are enabled.
3396 (dump_scope_block): Dump fragment info.
3397 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3398 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3399 (gimple_build_debug_inline_entry): New.
3400 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3401 Enable/disable inline entry points too.
3402 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3403 (DEBUG_INSN): Describe inline entry markers.
3405 * common.opt (gvariable-location-views): New.
3406 (gvariable-location-views=incompat5): New.
3407 * config.in: Rebuilt.
3408 * configure: Rebuilt.
3409 * configure.ac: Test assembler for view support.
3410 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3411 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3412 * dwarf2out.c (var_loc_view): New typedef.
3413 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3414 (dwarf2out_locviews_in_attribute): New.
3415 (dwarf2out_locviews_in_loclist): New.
3416 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3417 (enum dw_line_info_opcode): Add LI_adv_address.
3418 (struct dw_line_info_table): Add view.
3419 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3420 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3421 (zero_view_p): New variable.
3422 (ZERO_VIEW_P): New macro.
3423 (output_asm_line_debug_info): New.
3424 (struct var_loc_node): Add view.
3425 (add_AT_view_list, AT_loc_list): New.
3426 (add_var_loc_to_decl): Add view param. Test it against last.
3427 (new_loc_list): Add view params. Record them.
3428 (AT_loc_list_ptr): Handle loc and view lists.
3429 (view_list_to_loc_list_val_node): New.
3430 (print_dw_val): Handle dw_val_class_view_list.
3431 (size_of_die): Likewise.
3432 (value_format): Likewise.
3433 (loc_list_has_views): New.
3434 (gen_llsym): Set vl_symbol too.
3435 (maybe_gen_llsym, skip_loc_list_entry): New.
3436 (dwarf2out_maybe_output_loclist_view_pair): New.
3437 (output_loc_list): Output view list or entries too.
3438 (output_view_list_offset): New.
3439 (output_die): Handle dw_val_class_view_list.
3440 (output_dwarf_version): New.
3441 (output_compilation_unit_header): Use it.
3442 (output_skeleton_debug_sections): Likewise.
3443 (output_rnglists, output_line_info): Likewise.
3444 (output_pubnames, output_aranges): Update version comments.
3445 (output_one_line_info_table): Output view numbers in asm comments.
3446 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3447 Call maybe_gen_llsym.
3448 (loc_list_from_tree_1): Adjust.
3449 (add_AT_location_description): Create view list attribute if
3450 needed, check it's absent otherwise.
3451 (convert_cfa_to_fb_loc_list): Adjust.
3452 (maybe_emit_file): Call output_asm_line_debug_info for test.
3453 (dwarf2out_var_location): Reset views as needed. Precompute
3454 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3455 attribute. Set view.
3456 (new_line_info_table): Reset next view.
3457 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3458 (dwarf2out_source_line): Likewise. Output view resets and labels to
3459 the assembler, or select appropriate line info opcodes.
3460 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3461 (optimize_string_length): Catch it. Adjust.
3462 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3463 dw_val_class_view_list, and remove it if no longer needed.
3464 (hash_loc_list): Hash view numbers.
3465 (loc_list_hasher::equal): Compare them.
3466 (optimize_location_lists): Check whether a view list symbol is
3467 needed, and whether the locview attribute is present, and
3468 whether they match. Remove the locview attribute if no longer
3470 (index_location_lists): Call skip_loc_list_entry for test.
3471 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3472 Use output_dwarf_version.
3473 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3474 (struct dw_val_node): Add val_view_list.
3475 * final.c (SEEN_NEXT_VIEW): New.
3476 (set_next_view_needed): New.
3477 (clear_next_view_needed): New.
3478 (maybe_output_next_view): New.
3479 (final_start_function): Rename to...
3480 (final_start_function_1): ... this. Take pointer to FIRST,
3481 add SEEN parameter. Emit param bindings in the initial view.
3482 (final_start_function): Reintroduce SEEN-less interface.
3483 (final): Rename to...
3484 (final_1): ... this. Take SEEN parameter. Output final pending
3485 next view at the end.
3486 (final): Reintroduce seen-less interface.
3487 (final_scan_insn): Output pending next view before switching
3488 sections or ending a block. Mark the next view as needed when
3489 outputting variable locations. Notify debug backend of section
3490 changes, and of location view changes.
3491 (rest_of_handle_final): Adjust.
3492 * toplev.c (process_options): Autodetect value for debug variable
3493 location views option. Warn on incompat5 without -gdwarf-5.
3494 * doc/invoke.texi (gvariable-location-views): New.
3495 (gvariable-location-views=incompat5): New.
3496 (gno-variable-location-views): New.
3498 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3500 PR tree-optimization/84136
3501 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3502 that the result of find_edge is non-NULL.
3504 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3507 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3508 storing integer register in SImode. Fix cost of 256 and 512
3509 byte aligned SSE register store.
3511 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3513 * config/i386/i386.c (ix86_multiplication_cost): Fix
3514 multiplication cost for TARGET_AVX512DQ.
3516 2018-02-08 Marek Polacek <polacek@redhat.com>
3518 PR tree-optimization/84238
3519 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3522 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3524 PR tree-optimization/84265
3525 * tree-vect-stmts.c (vectorizable_store): Don't treat
3526 VMAT_CONTIGUOUS accesses as grouped.
3527 (vectorizable_load): Likewise.
3529 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3531 PR tree-optimization/81635
3532 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3533 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3534 (test_round_for_mask): New functions.
3535 (wide_int_cc_tests): Call test_round_for_mask.
3536 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3537 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3538 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3539 range returned by get_range_info.
3541 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3544 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3545 * symtab.c: Include builtins.h
3546 (symtab_node::output_to_lto_symbol_table_p): Move here
3547 from lto-streamer-out.c:output_symbol_p.
3548 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3549 (output_symbol_p): Move all logic to symtab.c
3550 (produce_symtab): Update.
3552 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3554 * config/s390/s390-opts.h (enum indirect_branch): Define.
3555 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3556 (s390_indirect_branch_via_thunk)
3557 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3558 (enum s390_indirect_branch_type): Define.
3559 * config/s390/s390.c (struct s390_frame_layout, struct
3560 machine_function): Remove.
3561 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3562 (indirect_branch_table_label_no, indirect_branch_table_name):
3564 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3565 (enum s390_indirect_branch_option): Define.
3566 (s390_return_addr_from_memory): New function.
3567 (s390_handle_string_attribute): New function.
3568 (s390_attribute_table): Add new attribute handler.
3569 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3570 (s390_indirect_branch_via_thunk): New function.
3571 (s390_indirect_branch_via_inline_thunk): New function.
3572 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3573 sibling call optimization for non z10 compiles.
3574 (s390_emit_call): Force indirect branch target to be a single
3575 register. Add r1 clobber for non-z10 compiles.
3576 (s390_emit_epilogue): Emit return jump via return_use expander.
3577 (s390_reorg): Handle JUMP_INSNs as execute targets.
3578 (s390_option_override_internal): Perform validity checks for the
3579 new command line options.
3580 (s390_indirect_branch_attrvalue): New function.
3581 (s390_indirect_branch_settings): New function.
3582 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3583 (s390_output_indirect_thunk_function): New function.
3584 (s390_code_end): Implement target hook.
3585 (s390_case_values_threshold): Implement target hook.
3586 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3588 * config/s390/s390.h (struct s390_frame_layout)
3589 (struct machine_function): Move here from s390.c.
3590 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3591 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3592 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3593 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3594 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3595 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3596 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3597 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3598 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3599 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3600 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3601 (mnemonic attribute): Add values which aren't recognized
3603 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3604 pattern for branch conversion. Fix mnemonic attribute.
3605 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3606 indirect branch via thunk if requested.
3607 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3608 ("*indirect_jump"): Disable for branch conversion using out of
3610 ("indirect_jump_via_thunk<mode>_z10")
3611 ("indirect_jump_via_thunk<mode>")
3612 ("indirect_jump_via_inlinethunk<mode>_z10")
3613 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3614 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3615 ("casesi_jump_via_inlinethunk<mode>_z10")
3616 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3617 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3618 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3619 ("*indirect2_jump"): Disable for branch conversion.
3620 ("casesi_jump"): Turn into expander and expand patterns for branch
3622 ("return_use"): New expander.
3623 ("*return"): Emit return via thunk and rename it to ...
3624 ("*return<mode>"): ... this one.
3625 * config/s390/s390.opt: Add new options and and enum for the
3628 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3630 * lra-constraints.c (match_reload): Unconditionally use
3631 gen_lowpart_SUBREG, rather than selecting between that
3632 and equivalent gen_rtx_SUBREG code.
3634 2018-02-08 Richard Biener <rguenther@suse.de>
3636 PR tree-optimization/84233
3637 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3638 changed flag instead of boguously re-using phi_inserted.
3640 2018-02-08 Martin Jambor <mjambor@suse.cz>
3642 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3643 static local variables.
3645 2018-02-08 Richard Biener <rguenther@suse.de>
3647 PR tree-optimization/84278
3648 * tree-vect-stmts.c (vectorizable_store): When looking for
3649 smaller vector types to perform grouped strided loads/stores
3650 make sure the mode is supported by the target.
3651 (vectorizable_load): Likewise.
3653 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3655 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3656 Increase LDP/STP opportunities by adding adjacent callee-saves.
3658 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3660 PR rtl-optimization/84068
3661 PR rtl-optimization/83459
3662 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3664 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3666 PR tree-optimization/84224
3667 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3668 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3671 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3674 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3675 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3676 count, remove one USE.
3678 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3680 * doc/install.texi (Configuration): Document the
3681 --with-long-double-format={ibm,ieee} PowerPC configuration
3685 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3686 Convert from define_expand to be define_insn_and_split. Rework
3687 float/double/_Float128 conversions to QI/HI/SImode to work with
3688 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3689 conversions to QI/HImode types did a store and then a load to
3690 truncate the value. For conversions to VSX registers, don't split
3691 the insn, instead emit the code directly. Use the code iterator
3692 any_fix to combine signed and unsigned conversions.
3693 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3694 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3695 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3696 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3697 (fix_<mode>di2_hw): Likewise.
3698 (fixuns_<mode>di2_hw): Likewise.
3699 (fix_<mode>si2_hw): Likewise.
3700 (fixuns_<mode>si2_hw): Likewise.
3701 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3702 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3703 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3704 fix<uns>_trunc<SFDF:mode>si2_p8.
3705 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3707 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3708 (fix<uns>_<mode>_mem): Likewise.
3709 (fctiw<u>z_<mode>_mem): Likewise.
3710 (fix<uns>_<mode>_mem): Likewise.
3711 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3712 the register allocator from doing a direct move to the GPRs to do
3713 a store, and instead use the ISA 3.0 store byte/half-word from
3714 vector register instruction. For IEEE 128-bit floating point,
3715 also optimize stores of 32-bit ints.
3716 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3718 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3720 * genextract.c (push_pathstr_operand): New function to support
3722 (walk_rtx): Call push_pathstr_operand.
3723 (print_path): Support [a-zA-Z].
3725 2018-02-07 Richard Biener <rguenther@suse.de>
3727 PR tree-optimization/84037
3728 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3729 (cse_and_gimplify_to_preheader): Declare.
3730 (vect_get_place_in_interleaving_chain): Likewise.
3731 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3733 (_loop_vec_info::~_loop_vec_info): Delete it.
3734 (cse_and_gimplify_to_preheader): New function.
3735 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3736 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3737 (vectorizable_load): Likewise. For grouped stores always base
3738 the IV on the first element.
3739 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3740 condition before gimplifying.
3742 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3744 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3745 *DIV_EXPR and *MOD_EXPR.
3747 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3750 * config/i386/i386.c (ix86_option_override_internal): Mask out
3751 the CF_SET bit when checking -fcf-protection.
3753 2018-02-07 Tom de Vries <tom@codesourcery.com>
3756 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3759 2018-02-07 Richard Biener <rguenther@suse.de>
3761 PR tree-optimization/84204
3762 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3765 PR tree-optimization/84205
3766 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3767 special-case isl_ast_op_zdiv_r.
3769 PR tree-optimization/84223
3770 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3771 Only add conditions from within the region.
3772 (gather_bbs::after_dom_children): Adjust.
3774 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3777 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3778 * config/avr/avr.md: Only post-reload split REG-REG moves if
3779 either register is GENERAL_REG_P.
3781 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3783 PR tree-optimization/84235
3784 * tree-ssa-scopedtables.c
3785 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3786 if the subtraction is performed in floating point type where NaNs are
3787 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3788 build 1. Formatting fix.
3790 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3793 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3794 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3795 and skip it regardless of bb boundaries. Use CALL_P macro,
3796 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3797 unnecessarily, formatting fix.
3799 2018-02-06 Michael Collison <michael.collison@arm.com>
3801 * config/arm/thumb2.md:
3802 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3803 (*thumb_mov_notscc): Ditto.
3805 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3808 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3809 unsigned_fix, not "s".
3811 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3813 * configure.ac (gcc_fn_eh_frame_ro): New function.
3814 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3815 correct .eh_frame permissions.
3816 * configure: Regenerate.
3818 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3820 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3823 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3825 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3826 Display warning message for -mno-speculate-indirect-jumps.
3828 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3830 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3832 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3834 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3836 PR tree-optimization/84225
3837 * tree-eh.c (find_trapping_overflow): Only call
3838 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3840 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3843 * config/i386/i386.c: Reimplement the check of possible options
3844 -mibt/-mshstk conbination. Change error messages.
3845 * doc/invoke.texi: Fix a typo: remove extra '='.
3847 2018-02-06 Marek Polacek <polacek@redhat.com>
3849 PR tree-optimization/84228
3850 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3852 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3855 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3856 emitted arch directives.
3857 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3858 __ARM_FEATURE_COPROC before changing architectures.
3860 2018-02-06 Richard Biener <rguenther@suse.de>
3862 * config/i386/i386.c (print_reg): Fix typo.
3863 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3865 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3867 * configure: Regenerate.
3869 2018-02-05 Martin Sebor <msebor@redhat.com>
3871 PR tree-optimization/83369
3872 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3875 2018-02-05 Martin Liska <mliska@suse.cz>
3877 * doc/invoke.texi: Cherry-pick upstream r323995.
3879 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3881 * ira.c (ira_init_register_move_cost): Adjust comment.
3883 2018-02-05 Martin Liska <mliska@suse.cz>
3885 PR gcov-profile/84137
3886 * doc/gcov.texi: Fix typo in documentation.
3888 2018-02-05 Martin Liska <mliska@suse.cz>
3890 PR gcov-profile/83879
3891 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3892 using dlopen functionality.
3894 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3896 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3897 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3898 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3899 _mm_maskz_range_round_ss): New intrinsics.
3900 (__builtin_ia32_rangesd128_round)
3901 (__builtin_ia32_rangess128_round): Remove.
3902 (__builtin_ia32_rangesd128_mask_round,
3903 __builtin_ia32_rangess128_mask_round): New builtins.
3904 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3905 __builtin_ia32_rangess128_round): Remove.
3906 (__builtin_ia32_rangesd128_mask_round,
3907 __builtin_ia32_rangess128_mask_round): New builtins.
3908 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3909 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3910 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3911 "<round_saeonly_constraint>")): Changed to ...
3912 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3913 "<round_saeonly_scalar_constraint>")): ... this.
3914 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3915 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3916 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3917 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3918 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3920 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3922 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3924 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3925 Remove all values except native, 8540 and 8548.
3927 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3929 * config/i386/i386.c (ix86_output_function_return): Pass
3930 INVALID_REGNUM, instead of -1, as invalid register number to
3931 indirect_thunk_name and output_indirect_thunk.
3933 2018-02-02 Julia Koval <julia.koval@intel.com>
3935 * config.gcc: Add -march=icelake.
3936 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3937 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3938 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3939 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3940 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3941 (processor_target_table): Add icelake.
3942 (ix86_option_override_internal): Handle new PTAs.
3943 (get_builtin_code_for_version): Handle icelake.
3944 (M_INTEL_COREI7_ICELAKE): New.
3945 (fold_builtin_cpu): Handle icelake.
3946 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3947 * doc/invoke.texi: Add -march=icelake.
3949 2018-02-02 Julia Koval <julia.koval@intel.com>
3951 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3952 to wide_int_bitmask.
3953 * wide-int-bitmask.h: New.
3955 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3958 * config/i386/i386.md: Replace Pmode with word_mode in
3959 builtin_setjmp_setup and builtin_longjmp to support x32.
3961 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3965 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3967 (rs6000_supported_cpu_names): New static variable.
3968 (linux_cpu_translation_table): Likewise.
3969 (elf_platform) <cpu>: Define new static variable and use it.
3970 Translate kernel AT_PLATFORM name to canonical name if needed.
3971 Error if platform name is unknown.
3973 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3976 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3978 2018-02-01 Jeff Law <law@redhat.com>
3981 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3982 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3983 the scratch if RELEASE_VIA_POP is false.
3984 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3985 If we have to save a temporary register, decrement SIZE appropriately.
3986 Pass new arguments to release_scratch_register_on_entry.
3987 (ix86_adjust_stack_and_probe): Likewise.
3988 (ix86_emit_probe_stack_range): Pass new arguments to
3989 release_scratch_register_on_entry.
3991 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3993 PR rtl-optimization/84157
3994 * combine.c (change_zero_ext): Use REG_P predicate in
3995 front of HARD_REGISTER_P predicate.
3997 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3999 * config/avr/avr.c (avr_option_override): Move disabling of
4000 -fdelete-null-pointer-checks to...
4001 * common/config/avr/avr-common.c (avr_option_optimization_table):
4004 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4006 PR tree-optimization/81635
4007 * tree-data-ref.c (split_constant_offset_1): For types that
4008 wrap on overflow, try to use range info to prove that wrapping
4011 2018-02-01 Renlin Li <renlin.li@arm.com>
4014 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4016 (aarch64_register_move_cost): Likewise.
4017 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4019 (REG_CLASS_NAMES): Likewise.
4020 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4021 TAILCALL_ADDR_REGS. Remove IP registers.
4022 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4024 2018-02-01 Richard Biener <rguenther@suse.de>
4026 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4027 for specifying RPO order and allow NULL for that.
4028 * domwalk.c (dom_walker::dom_walker): Likewise.
4029 (dom_walker::walk): Handle NULL RPO order.
4030 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4032 (rewrite_update_dom_walker): Likewise.
4033 (mark_def_dom_walker): Likewise.
4035 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4037 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4038 (aarch64_maybe_expand_sve_subreg_move): Declare.
4039 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4040 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4042 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4043 that are semantically a reverse operation.
4044 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4045 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4046 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4048 (aarch64_can_change_mode_class): For big-endian, forbid changes
4049 between two SVE modes if they have different element sizes.
4051 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4053 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4054 the TImode handling for big-endian targets.
4056 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4058 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4059 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4061 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4062 Remove BSWAP handing for big-endian targets and use the form of
4063 LD1RQ appropariate for the mode.
4065 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4067 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4068 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4071 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4074 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4075 check for operands that need to go through aarch64_sve_reload_be.
4077 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4079 PR tree-optimization/81661
4080 PR tree-optimization/84117
4081 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4082 * tree-eh.c: Include gimplify.h.
4083 (find_trapping_overflow, replace_trapping_overflow,
4084 rewrite_to_non_trapping_overflow): New functions.
4085 * tree-vect-loop.c: Include tree-eh.h.
4086 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4087 * tree-data-ref.c: Include tree-eh.h.
4088 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4090 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4092 PR rtl-optimization/84123
4093 * combine.c (change_zero_ext): Check if hard register satisfies
4094 can_change_dest_mode before calling gen_lowpart_SUBREG.
4096 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4099 * ira.c (ira_init_register_move_cost): Remove assert.
4101 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4103 PR rtl-optimization/84071
4104 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4105 * doc/tm.texi: Regenerate.
4107 2018-01-31 Richard Biener <rguenther@suse.de>
4109 PR tree-optimization/84132
4110 * tree-data-ref.c (analyze_miv_subscript): Properly
4111 check whether evolution_function_is_affine_multivariate_p
4112 before calling gcd_of_steps_may_divide_p.
4114 2018-01-31 Julia Koval <julia.koval@intel.com>
4117 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4118 * config/i386/i386.md (rdpid_rex64) New.
4119 (rdpid): Make 32bit only.
4121 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4124 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4125 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4127 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4130 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4132 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4134 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4136 PR rtl-optimization/84071
4137 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4138 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4140 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4142 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4143 (arc_attribute_table): Add 'aux' attribute.
4144 (arc_in_small_data_p): Consider aux like variables.
4145 (arc_is_aux_reg_p): New function.
4146 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4147 (arc_get_aux_arg): New function.
4148 (prepare_move_operands): Handle aux-register access.
4149 (arc_handle_aux_attribute): New function.
4150 * doc/extend.texi (ARC Variable attributes): Add subsection.
4152 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4154 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4155 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4156 (arc_attribute_table): Add 'uncached' attribute.
4157 (arc_print_operand): Print '.di' flag for uncached memory
4159 (arc_in_small_data_p): Do not consider for small data the uncached
4161 (arc_is_uncached_mem_p): New function.
4162 * config/arc/predicates.md (compact_store_memory_operand): Check
4163 for uncached memory accesses.
4164 (nonvol_nonimm_operand): Likewise.
4165 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4167 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4170 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4171 falign-loops=): Add Optimization flag.
4173 2018-01-30 Jeff Law <law@redhat.com>
4176 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4177 INT_REGISTERS_SAVED. Check it prior to calling
4178 get_scratch_register_on_entry.
4179 (ix86_adjust_stack_and_probe): Similarly.
4180 (ix86_emit_probe_stack_range): Similarly.
4181 (ix86_expand_prologue): Corresponding changes.
4183 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4186 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4187 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4189 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4192 * lra-constraints.c (curr_insn_transform): Process AND in the
4195 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4197 PR rtl-optimization/83986
4198 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4199 dependence against last_pending_memory_flush in addition to
4202 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4204 PR tree-optimization/81611
4205 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4208 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4211 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4214 2018-01-30 Richard Biener <rguenther@suse.de>
4215 Jakub Jelinek <jakub@redhat.com>
4217 PR tree-optimization/84111
4218 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4219 inner loops added during recursion, as they don't have up-to-date
4222 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4225 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4226 (can_inline_edge_by_limits_p): ... here.
4227 (can_early_inline_edge_p, check_callers,
4228 update_caller_keys, update_callee_keys, recursive_inlining,
4229 add_new_edges_to_heap, speculation_useful_p,
4230 inline_small_functions,
4231 inline_small_functions, flatten_function,
4232 inline_to_all_callers_1): Update.
4234 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4236 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4239 2018-01-30 Richard Biener <rguenther@suse.de>
4241 PR tree-optimization/83008
4242 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4243 invariant and constant vector uses in stmts when they need
4246 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4249 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4250 * configure: Regenerate.
4252 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4254 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4256 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4257 Use gen_rtx_REG rather than gen_lowpart.
4259 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4261 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4262 rather than 0 when creating partial subregs.
4264 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4266 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4269 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4272 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4273 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4274 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4275 flags. This restores the settings used before the 2017-07-24.
4276 Turning off pre increment/decrement/modify allows IVOPTS to
4277 optimize DF/SF loops where the index is an int.
4279 2018-01-29 Richard Biener <rguenther@suse.de>
4280 Kelvin Nilsen <kelvin@gcc.gnu.org>
4283 * tree-vect-stmts.c (vectorizable_call): Don't call
4284 targetm.vectorize_builtin_md_vectorized_function if callee is
4287 2018-01-22 Carl Love <cel@us.ibm.com>
4289 * doc/extend.tex: Fix typo in second arg in
4290 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4292 2018-01-29 Richard Biener <rguenther@suse.de>
4294 PR tree-optimization/84086
4295 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4296 (flush_ssaname_freelist): When SSA names were released reset
4297 the SCEV hash table.
4299 2018-01-29 Richard Biener <rguenther@suse.de>
4301 PR tree-optimization/84057
4302 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4303 removed paths when removing edges.
4305 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4307 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4308 -mfunction-return=@var{choice}.
4310 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4313 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4315 (layout::print_source_line): Likewise.
4316 (test_get_line_width_without_trailing_whitespace): Add test cases.
4318 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4321 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4324 2018-01-26 Jim Wilson <jimw@sifive.com>
4326 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4328 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4331 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4333 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4334 and CMP + SUB-immediate -> SUBS.
4336 2018-01-26 Martin Sebor <msebor@redhat.com>
4338 PR tree-optimization/83896
4339 * tree-ssa-strlen.c (get_string_len): Rename...
4340 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4341 Avoid assuming length is constant.
4342 (handle_char_store): Use HOST_WIDE_INT for string length.
4344 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4347 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4348 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4350 2018-01-26 Richard Biener <rguenther@suse.de>
4352 PR rtl-optimization/84003
4353 * dse.c (record_store): Only record redundant stores when
4354 the earlier store aliases at least all accesses the later one does.
4356 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4358 PR rtl-optimization/83985
4359 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4360 REG_CFA_RESTORE insns.
4361 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4362 REG_CFA_RESTORE insns here.
4365 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4366 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4368 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4370 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4371 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4372 (arc_init): Likewise.
4373 (arc_override_options): Likewise.
4374 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4376 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4378 * config/arc/arc.h (TARGET_DBNZ): Define.
4379 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4380 properly set the tune attribute.
4381 (dbnz): Use TARGET_DBNZ guard.
4382 * config/arc/arc.opt (mtune): Add core3 option.
4384 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4386 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4387 recognize new pic like addresses.
4388 (arc_delegitimize_address): Clean up.
4390 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4392 * config/arc/arc-arches.def: Option mrf16 valid for all
4394 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4395 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4396 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4397 * config/arc/arc-tables.opt: Regenerate.
4398 * config/arc/arc.c (arc_conditional_register_usage): Handle
4399 reduced register file case.
4400 (arc_file_start): Set must have build attributes.
4401 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4403 * config/arc/arc.opt (mrf16): Add new option.
4404 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4405 * config/arc/genmultilib.awk: Handle new mrf16 option.
4406 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4407 * config/arc/t-multilib: Regenerate.
4408 * doc/invoke.texi (ARC Options): Document mrf16 option.
4410 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4412 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4413 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4414 (arc_attribute_table): Add 'secure_call' attribute.
4415 (arc_print_operand): Print secure call operand.
4416 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4418 (arc_is_secure_call_p): New function. * config/arc/arc.md
4419 (call_i): Add support for sjli instruction.
4420 (call_value_i): Likewise.
4421 * config/arc/constraints.md (Csc): New constraint.
4423 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4424 John Eric Martin <John.Martin@emmicro-us.com>
4426 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4427 * config/arc/arc.c (_arc_jli_section): New struct.
4428 (arc_jli_section): New type.
4429 (rc_jli_sections): New static variable.
4430 (arc_handle_jli_attribute): New function.
4431 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4432 (arc_file_end): New function.
4433 (TARGET_ASM_FILE_END): Define.
4434 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4435 (arc_add_jli_section): New function.
4436 (jli_call_scan): Likewise.
4437 (arc_reorg): Call jli_call_scan.
4438 (arc_output_addsi): Remove 'S' from printing asm operand.
4439 (arc_is_jli_call_p): New function.
4440 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4442 (movhi_insn): Likewise.
4443 (movsi_insn): Likewise.
4444 (movsi_set_cc_insn): Likewise.
4445 (loadqi_update): Likewise.
4446 (load_zeroextendqisi_update): Likewise.
4447 (load_signextendqisi_update): Likewise.
4448 (loadhi_update): Likewise.
4449 (load_zeroextendhisi_update): Likewise.
4450 (load_signextendhisi_update): Likewise.
4451 (loadsi_update): Likewise.
4452 (loadsf_update): Likewise.
4453 (movsicc_insn): Likewise.
4454 (bset_insn): Likewise.
4455 (bxor_insn): Likewise.
4456 (bclr_insn): Likewise.
4457 (bmsk_insn): Likewise.
4458 (bicsi3_insn): Likewise.
4459 (cmpsi_cc_c_insn): Likewise.
4460 (movsi_ne): Likewise.
4461 (movsi_cond_exec): Likewise.
4462 (clrsbsi2): Likewise.
4473 (call_i): Remove 'S' asm letter, add jli instruction.
4474 (call_value_i): Likewise.
4475 * config/arc/arc.op (mjli-always): New option.
4476 * config/arc/constraints.md (Cji): New constraint.
4477 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4479 (subsf3_fpx): Likewise.
4480 (mulsf3_fpx): Likewise.
4481 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4483 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4485 * doc/invoke.texi (ARC): Document mjli-always option.
4487 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4489 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4490 avoid addition with 0 and use incw and decw where possible.
4492 2018-01-26 Richard Biener <rguenther@suse.de>
4494 PR tree-optimization/81082
4495 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4496 association if it requires casting to unsigned.
4497 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4498 from fold_plusminus_mult_expr to catch important cases late when
4499 range info is available.
4501 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4503 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4504 * configure.ac (hidden_linkonce): New test.
4505 * configure: Regenerate.
4506 * config.in: Regenerate.
4508 2018-01-26 Julia Koval <julia.koval@intel.com>
4510 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4511 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4512 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4513 _mm_mask_bitshuffle_epi64_mask): Fix type.
4514 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4515 USI_FTYPE_V4DI_V4DI_USI): Remove.
4516 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4517 __builtin_ia32_vpshufbitqmb256_mask,
4518 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4519 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4520 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4522 2018-01-26 Alan Modra <amodra@gmail.com>
4525 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4526 UNSPEC_VBPERMQ. Sort other unspecs.
4528 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4530 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4532 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4535 * predict.c (drop_profile): Do not push/pop cfun; update also
4537 (handle_missing_profiles): Fix logic looking for zero profiles.
4539 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4542 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4543 on functions with #pragma omp declare simd or functions with simd
4545 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4546 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4547 Remove trailing \n from warning_at calls.
4549 2018-01-25 Tom de Vries <tom@codesourcery.com>
4552 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4553 for neutered workers.
4555 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4558 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4559 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4561 2018-01-24 Jeff Law <law@redhat.com>
4564 * i386.c (get_probe_interval): Move to earlier point.
4565 (ix86_compute_frame_layout): If -fstack-clash-protection and
4566 the frame is larger than the probe interval, then use pushes
4567 to save registers rather than reg->mem moves.
4568 (ix86_expand_prologue): Remove conditional for int_registers_saved
4571 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4574 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4575 min/max for never referenced object.
4577 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4580 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4582 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4583 attributes from DECL_ATTRIBUTES (decl) without affecting
4584 DECL_ATTRIBUTES (current_function_decl).
4585 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4586 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4588 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4590 PR tree-optimization/83979
4591 * fold-const.c (fold_comparison): Use constant_boolean_node
4592 instead of boolean_{true,false}_node.
4594 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4596 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4599 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4601 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4602 Simplify the clause that sets the length attribute.
4603 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4604 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4605 clause that sets the length attribute.
4606 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4608 2018-01-24 Tom de Vries <tom@codesourcery.com>
4611 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4612 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4613 Add strict parameter.
4614 (prevent_branch_around_nothing): Insert dummy insn between branch to
4615 label and label with no ptx insn inbetween.
4616 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4618 2018-01-24 Tom de Vries <tom@codesourcery.com>
4621 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4622 for neutered threads in warp.
4623 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4625 2018-01-24 Richard Biener <rguenther@suse.de>
4627 PR tree-optimization/83176
4628 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4631 2018-01-24 Richard Biener <rguenther@suse.de>
4633 PR tree-optimization/82819
4634 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4635 code generating pluses that are no-ops in the target precision.
4637 2018-01-24 Richard Biener <rguenther@suse.de>
4640 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4642 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4644 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4645 to merge probabilities.
4646 * predict.c (probably_never_executed): Also mark as cold functions
4647 with global 0 profile and guessed local profile.
4648 * profile-count.c (profile_probability::combine_with_count): New
4650 * profile-count.h (profile_probability::operator*,
4651 profile_probability::operator*=, profile_probability::operator/,
4652 profile_probability::operator/=): Reduce precision to adjusted
4653 and set value to guessed on contradictory divisions.
4654 (profile_probability::combine_with_freq): Remove.
4655 (profile_probability::combine_wiht_count): Declare.
4656 (profile_count::force_nonzero):: Set to adjusted.
4657 (profile_count::probability_in):: Set quality to adjusted.
4658 * tree-ssa-tail-merge.c (replace_block_by): Use
4661 2018-01-23 Andrew Waterman <andrew@sifive.com>
4662 Jim Wilson <jimw@sifive.com>
4664 * config/riscv/riscv.c (riscv_stack_boundary): New.
4665 (riscv_option_override): Set riscv_stack_boundary. Handle
4666 riscv_preferred_stack_boundary_arg.
4667 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4668 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4669 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4670 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4671 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4672 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4674 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4677 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4678 of struct ix86_frame.
4679 (ix86_expand_epilogue): Likewise. Add a local variable for
4680 the reg_save_offset field in struct ix86_frame.
4682 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4684 PR tree-optimization/82604
4685 * tree-loop-distribution.c (enum partition_kind): New enum item
4686 PKIND_PARTIAL_MEMSET.
4687 (partition_builtin_p): Support above new enum item.
4688 (generate_code_for_partition): Ditto.
4689 (compute_access_range): Differentiate cases that equality can be
4690 proven at all loops, the innermost loops or no loops.
4691 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4692 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4693 (finalize_partitions, distribute_loop): Don't fuse partition of
4694 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4695 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4698 2018-01-23 Martin Liska <mliska@suse.cz>
4700 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4701 order to ignore the predictor.
4702 (PRED_POLYMORPHIC_CALL): Likewise.
4703 (PRED_RECURSIVE_CALL): Likewise.
4705 2018-01-23 Martin Liska <mliska@suse.cz>
4707 * tree-profile.c (tree_profiling): Print function header to
4708 aware reader which function we are working on.
4709 * value-prof.c (gimple_find_values_to_profile): Do not print
4710 not interesting value histograms.
4712 2018-01-23 Martin Liska <mliska@suse.cz>
4714 * profile-count.h (enum profile_quality): Add
4715 profile_uninitialized as the first value. Do not number values
4716 as they are zero based.
4717 (profile_count::verify): Update sanity check.
4718 (profile_probability::verify): Likewise.
4720 2018-01-23 Nathan Sidwell <nathan@acm.org>
4722 * doc/invoke.texi (ffor-scope): Deprecate.
4724 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4726 PR tree-optimization/83510
4727 * domwalk.c (set_all_edges_as_executable): New function.
4728 (dom_walker::dom_walker): Convert bool param
4729 "skip_unreachable_blocks" to enum reachability. Move setup of
4730 edge flags to set_all_edges_as_executable and only do it when
4731 reachability is REACHABLE_BLOCKS.
4732 * domwalk.h (enum dom_walker::reachability): New enum.
4733 (dom_walker::dom_walker): Convert bool param
4734 "skip_unreachable_blocks" to enum reachability.
4735 (set_all_edges_as_executable): New decl.
4736 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4737 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4739 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4740 but converting true to REACHABLE_BLOCKS.
4741 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4743 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4744 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4745 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4747 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4748 if check_all_array_refs will be called.
4750 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4752 * tree.c (selftest::test_location_wrappers): Add more test
4755 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4757 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4758 (selftest::test_bit_in_range): Likewise.
4760 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4763 * doc/sourcebuild.texi (vect_float): Say that the selector
4764 only describes the situation when -funsafe-math-optimizations is on.
4765 (vect_float_strict): Document.
4767 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4769 PR tree-optimization/83965
4770 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4771 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4772 instead of checking only for a reduction.
4773 (vect_recog_widen_sum_pattern): Likewise.
4775 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4777 * predict.c (probably_never_executed): Only use precise profile info.
4778 (compute_function_frequency): Skip after inlining hack since we now
4779 have quality checking.
4781 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4783 * profile-count.h (profile_probability::very_unlikely,
4784 profile_probability::unlikely, profile_probability::even): Set
4785 precision to guessed.
4787 2018-01-23 Richard Biener <rguenther@suse.de>
4789 PR tree-optimization/83963
4790 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4791 Properly terminate dominator walk when crossing the exit edge not
4792 when visiting its source block.
4794 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4797 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4798 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4800 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4802 PR tree-optimization/83957
4803 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4804 semicolon after for body surrounded by braces.
4806 PR tree-optimization/83081
4807 * profile-count.h (profile_probability::split): New method.
4808 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4809 Use profile_probability::split.
4810 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4811 when splitting a single conditional jump into 2.
4813 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4815 PR tree-optimization/69452
4816 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4819 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4821 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4822 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4823 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4825 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4827 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4828 * config/rl78/rl78.md (movdi): New define_expand.
4829 * config/rl78/rl78.c (rl78_split_movdi): New function.
4831 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4834 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4836 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4837 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4838 128-bit to produce an UNSPEC move to get the double word with the
4839 signbit and then a shift directly to do signbit.
4840 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4841 implementation with a new version that just does either a direct
4842 move or a regular move. Move memory interface to separate insns.
4843 Move insns so they are next to the expander.
4844 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4845 with signbit move. Split big and little endian case.
4846 (signbit<mode>2_dm_mem_le): Likewise.
4847 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4848 (signbit<mode>2_dm2): Likewise.
4850 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4852 * config/rl78/rl78.md (anddi3): New define_expand.
4854 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4856 * config/rl78/rl78.md (umindi3): New define_expand.
4858 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4860 * config/rl78/rl78.md (smindi3): New define_expand.
4862 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4864 * config/rl78/rl78.md (smaxdi3): New define_expand.
4866 2018-01-22 Carl Love <cel@us.ibm.com>
4868 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4869 LVX_V1TI): Add macro expansion.
4870 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4871 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4872 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4873 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4874 Change check to determine if the instruction is a byte reversing
4875 entry. Fix typo in comment.
4876 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4877 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4878 Add def_builtin calls for new builtins.
4879 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4880 Add define_insn expansion.
4882 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4884 * config/rl78/rl78.md (umaxdi3): New define_expand.
4886 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4888 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4889 for non-QImode registers.
4891 2018-01-22 Richard Biener <rguenther@suse.de>
4893 PR tree-optimization/83963
4894 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4895 including the loop exit block.
4896 (scop_detection::merge_sese): Likewise.
4897 (scop_detection::add_scop): Do it here instead.
4899 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4901 * doc/sourcebuild.texi (arm_softfloat): Document.
4903 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4906 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4907 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4908 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4910 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4911 David Edelsohn <dje.gcc@gmail.com>
4914 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4915 Change "crset eq" to "crset 2".
4916 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4917 (*call_indirect_aix<mode>_nospec): Likewise.
4918 (*call_value_indirect_aix<mode>_nospec): Likewise.
4919 (*call_indirect_elfv2<mode>_nospec): Likewise.
4920 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4921 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4922 change assembly output from . to $.
4923 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4924 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4925 (*tablejump<mode>_internal1_nospec): Likewise.
4927 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4930 * config/sh/sh_optimize_sett_clrt.cc:
4931 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4933 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4935 PR tree-optimization/83940
4936 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4937 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4938 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4939 use it to pass back the definition type.
4940 (vect_check_store_rhs): Likewise rhs_dt_out.
4941 (vect_build_gather_load_calls): Add a mask_dt argument and use
4942 it instead of a call to vect_is_simple_use.
4943 (vectorizable_store): Update calls to vect_check_load_store_mask
4944 and vect_check_store_rhs. Use the dt returned by the latter instead
4945 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4946 instead of calls to vect_is_simple_use. Pass the scalar rather
4947 than the vector operand to vect_is_simple_use when handling
4948 second and subsequent copies of an rhs value.
4949 (vectorizable_load): Update calls to vect_check_load_store_mask
4950 and vect_build_gather_load_calls. Use the cached mask_dt and
4951 gs_info.offset_dt instead of calls to vect_is_simple_use.
4953 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4956 * tree-emutls.c: Include gimplify.h.
4957 (lower_emutls_2): New function.
4958 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4959 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4960 it before further processing.
4963 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4964 UINTVAL (trueop1) instead of INTVAL (op1).
4966 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4970 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4971 INCOMING_FRAME_SP_OFFSET if not defined.
4972 (scan_trace): Add ENTRY argument. If true and
4973 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4974 emit a note to adjust the CFA offset.
4975 (create_cfi_notes): Adjust scan_trace callers.
4976 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4977 INCOMING_FRAME_SP_OFFSET in the CIE.
4978 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4979 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4981 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4982 * doc/tm.texi: Regenerated.
4984 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4986 PR rtl-optimization/83147
4987 * lra-constraints.c (remove_inheritance_pseudos): Use
4988 lra_substitute_pseudo_within_insn.
4990 2018-01-19 Tom de Vries <tom@codesourcery.com>
4991 Cesar Philippidis <cesar@codesourcery.com>
4994 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4996 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4999 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5000 spaces for function labels.
5002 2018-01-19 Martin Liska <mliska@suse.cz>
5004 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5005 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5006 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5007 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5008 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5009 (PRED_CONST_RETURN): Change from 69 to 65.
5010 (PRED_NULL_RETURN): Change from 91 to 71.
5011 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5012 (PRED_LOOP_GUARD): Change from 66 to 73.
5014 2018-01-19 Martin Liska <mliska@suse.cz>
5016 * predict.c (predict_insn_def): Add new assert.
5017 (struct branch_predictor): Change type to signed integer.
5018 (test_prediction_value_range): Amend test to cover
5020 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5021 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5022 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5023 (PRED_LOOP_IV_COMPARE): Likewise.
5024 * predict.h (PROB_UNINITIALIZED): Define new constant.
5026 2018-01-19 Martin Liska <mliska@suse.cz>
5028 * predict.c (dump_prediction): Add new format for
5029 analyze_brprob.py script which is enabled with -details
5031 * profile-count.h (precise_p): New function.
5033 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5035 PR tree-optimization/83922
5036 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5037 there are no statements that need masking.
5038 (vect_active_double_reduction_p): New function.
5039 (vect_analyze_loop_operations): Use it when handling phis that
5040 are not in the loop header.
5042 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5044 PR tree-optimization/83914
5045 * tree-vect-loop.c (vectorizable_induction): Don't convert
5046 init_expr or apply the peeling adjustment for inductions
5047 that are nested within the vectorized loop.
5049 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5051 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5054 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5058 * function.h (gimplify_parameters): Add gimple_seq * argument.
5059 * function.c: Include gimple.h and options.h.
5060 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5061 for the added local temporaries if needed.
5062 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5063 if there are any parameter cleanups, wrap whole body into a
5064 try/finally with the cleanups.
5066 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5069 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5070 Use GET_MODE_CLASS for scalar floating point.
5072 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5076 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5077 Fix call of call_cgraph_insertion_hooks.
5079 2018-01-18 Martin Sebor <msebor@redhat.com>
5081 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5083 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5086 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5089 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5092 * common.opt: (-ffile-prefix-map): New option.
5093 * opts.c (common_handle_option): Defer it.
5094 * opts-global.c (handle_common_deferred_options): Handle it.
5095 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5096 * file-prefix-map.h: New file.
5097 (remap_debug_filename, add_debug_prefix_map): ...here.
5098 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5099 * final.c (debug_prefix_map, add_debug_prefix_map
5100 remap_debug_filename): Move to...
5101 * file-prefix-map.c: New file.
5102 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5103 generalize, get rid of alloca(), use strrchr() instead of strchr().
5104 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5105 Implement in terms of add_prefix_map().
5106 (remap_macro_filename, remap_debug_filename): Implement in term of
5108 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5109 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5110 * dbxout.c: Include file-prefix-map.h.
5111 * varasm.c: Likewise.
5112 * vmsdbgout.c: Likewise.
5113 * xcoffout.c: Likewise.
5114 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5115 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5116 * doc/invoke.texi (-ffile-prefix-map): Document.
5117 (-fdebug-prefix-map): Update description.
5119 2018-01-18 Martin Liska <mliska@suse.cz>
5121 * config/i386/i386.c (indirect_thunk_name): Document that also
5123 (output_indirect_thunk): Document why both instructions
5124 (pause and lfence) are generated.
5126 2018-01-18 Richard Biener <rguenther@suse.de>
5128 PR tree-optimization/83887
5129 * graphite-scop-detection.c
5130 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5131 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5132 (scop_detection::merge_sese): Re-implement with a flood-fill
5133 algorithm that properly finds a SESE region if it exists.
5135 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5138 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5139 pointer_diff optimizations use view_convert instead of convert.
5141 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5143 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5144 Generate different code for -mno-speculate-indirect-jumps.
5145 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5146 (*call_indirect_aix<mode>): Disable for
5147 -mno-speculate-indirect-jumps.
5148 (*call_indirect_aix<mode>_nospec): New define_insn.
5149 (*call_value_indirect_aix<mode>): Disable for
5150 -mno-speculate-indirect-jumps.
5151 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5152 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5153 -mno-speculate-indirect-jumps.
5154 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5156 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5158 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5159 long double type, set the flags for noting the default long double
5160 type, even if we don't pass or return a long double type.
5162 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5165 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5168 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5170 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5171 support for merge[hl].
5172 (fold_mergehl_helper): New helper function.
5173 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5174 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5175 (altivec_vmrglw_direct): Add xxmrglw insn.
5177 2018-01-17 Andrew Waterman <andrew@sifive.com>
5179 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5180 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5182 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5185 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5186 call the lto_location_cache before reading the
5187 DECL_SOURCE_LOCATION of the types.
5189 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5190 Richard Sandiford <richard.sandiford@linaro.org>
5192 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5193 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5194 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5195 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5196 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5198 * config/aarch64/constraints.md (aarch64_movti_operand):
5200 * config/aarch64/predicates.md (Uti): Add new constraint.
5202 2018-01-17 Carl Love <cel@us.ibm.com>
5204 * config/rs6000/vsx.md (define_expand xl_len_r,
5205 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5206 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5208 (define_expand, define_insn): Move the shift left from the
5209 define_insn to the define_expand for lxvl and stxvl instructions.
5210 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5211 and XL_LEN_R definitions to PURE.
5213 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5215 * config/i386/i386.c (indirect_thunk_name): Declare regno
5216 as unsigned int. Compare regno with INVALID_REGNUM.
5217 (output_indirect_thunk): Ditto.
5218 (output_indirect_thunk_function): Ditto.
5219 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5220 in the call to output_indirect_thunk_function.
5222 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5225 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5226 rather than the size of inner_type to determine the stack slot size
5227 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5229 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5232 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5235 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5237 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5238 endian Linux systems to optionally enable multilibs for selecting
5239 the long double type if the user configured an explicit type.
5240 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5241 have no long double multilibs if not defined.
5242 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5243 warn if the user used -mabi={ieee,ibm}longdouble and we built
5244 multilibs for long double.
5245 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5246 appropriate multilib option.
5247 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5249 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5250 for building long double multilibs.
5251 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5253 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5255 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5258 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5260 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5263 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5266 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5269 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5271 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5272 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5274 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5276 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5277 different rtl trees depending on TARGET_64BIT.
5278 (rs6000_gen_lvx): Likewise.
5280 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5282 * config/visium/visium.md (nop): Tweak comment.
5283 (hazard_nop): Likewise.
5285 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5287 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5288 -mspeculate-indirect-jumps.
5289 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5290 for -mno-speculate-indirect-jumps.
5291 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5292 (*call_value_indirect_elfv2<mode>): Disable for
5293 -mno-speculate-indirect-jumps.
5294 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5295 (indirect_jump): Emit different RTL for
5296 -mno-speculate-indirect-jumps.
5297 (*indirect_jump<mode>): Disable for
5298 -mno-speculate-indirect-jumps.
5299 (*indirect_jump<mode>_nospec): New define_insn.
5300 (tablejump): Emit different RTL for
5301 -mno-speculate-indirect-jumps.
5302 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5303 (tablejumpsi_nospec): New define_expand.
5304 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5305 (tablejumpdi_nospec): New define_expand.
5306 (*tablejump<mode>_internal1): Disable for
5307 -mno-speculate-indirect-jumps.
5308 (*tablejump<mode>_internal1_nospec): New define_insn.
5309 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5312 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5314 * caller-save.c (insert_save): Drop unnecessary parameter. All
5317 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5318 Richard Biener <rguenth@suse.de>
5321 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5322 return early, inline manually is_gimple_sizepos. Make sure if we
5323 call gimplify_expr we don't end up with a gimple constant.
5324 * tree.c (variably_modified_type_p): Don't return true for
5325 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5326 * gimplify.h (is_gimple_sizepos): Remove.
5328 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5330 PR tree-optimization/83857
5331 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5332 vectorizable_live_operation for pure SLP statements.
5333 (vectorizable_live_operation): Handle PHIs.
5335 2018-01-16 Richard Biener <rguenther@suse.de>
5337 PR tree-optimization/83867
5338 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5339 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5341 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5344 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5345 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5346 If off is not INTEGER_CST, issue a may not be aligned warning
5347 rather than isn't aligned. Use isn%'t rather than isn't.
5348 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5350 <case MULT_EXPR>: Improve the case when bottom and one of the
5351 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5352 operand, in that case check if the other operand is multiple of
5353 bottom divided by the INTEGER_CST operand.
5355 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5358 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5359 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5360 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5361 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5362 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5363 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5364 (pa_function_arg_size): New function.
5366 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5368 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5369 in a separate statement.
5371 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5373 PR tree-optimization/83847
5374 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5375 group gathers and scatters.
5377 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5379 PR rtl-optimization/86620
5380 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5382 PR rtl-optimization/83213
5383 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5384 to last if both are JUMP_INSNs.
5386 PR tree-optimization/83843
5387 * gimple-ssa-store-merging.c
5388 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5389 store_immediate_info for bswap/nop orig_stores.
5391 2018-01-15 Andrew Waterman <andrew@sifive.com>
5393 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5395 <UDIV>: Increase cost if !TARGET_DIV.
5397 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5399 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5400 (define_attr "cr_logical_3op"): New.
5401 (cceq_ior_compare): Adjust.
5402 (cceq_ior_compare_complement): Adjust.
5403 (*cceq_rev_compare): Adjust.
5404 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5405 (is_cracked_insn): Adjust.
5406 (insn_must_be_first_in_group): Adjust.
5407 * config/rs6000/40x.md: Adjust.
5408 * config/rs6000/440.md: Adjust.
5409 * config/rs6000/476.md: Adjust.
5410 * config/rs6000/601.md: Adjust.
5411 * config/rs6000/603.md: Adjust.
5412 * config/rs6000/6xx.md: Adjust.
5413 * config/rs6000/7450.md: Adjust.
5414 * config/rs6000/7xx.md: Adjust.
5415 * config/rs6000/8540.md: Adjust.
5416 * config/rs6000/cell.md: Adjust.
5417 * config/rs6000/e300c2c3.md: Adjust.
5418 * config/rs6000/e500mc.md: Adjust.
5419 * config/rs6000/e500mc64.md: Adjust.
5420 * config/rs6000/e5500.md: Adjust.
5421 * config/rs6000/e6500.md: Adjust.
5422 * config/rs6000/mpc.md: Adjust.
5423 * config/rs6000/power4.md: Adjust.
5424 * config/rs6000/power5.md: Adjust.
5425 * config/rs6000/power6.md: Adjust.
5426 * config/rs6000/power7.md: Adjust.
5427 * config/rs6000/power8.md: Adjust.
5428 * config/rs6000/power9.md: Adjust.
5429 * config/rs6000/rs64.md: Adjust.
5430 * config/rs6000/titan.md: Adjust.
5432 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5434 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5435 ix86_indirect_branch_register logic.
5437 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5439 * config/i386/constraints.md (Bs): Update
5440 ix86_indirect_branch_register check. Don't check
5441 ix86_indirect_branch_register with GOT_memory_operand.
5443 * config/i386/predicates.md (GOT_memory_operand): Don't check
5444 ix86_indirect_branch_register here.
5445 (GOT32_symbol_operand): Likewise.
5447 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5449 * config/i386/predicates.md (constant_call_address_operand):
5450 Rewrite ix86_indirect_branch_register logic.
5451 (sibcall_insn_operand): Likewise.
5453 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5455 * config/i386/constraints.md (Bs): Replace
5456 ix86_indirect_branch_thunk_register with
5457 ix86_indirect_branch_register.
5459 * config/i386/i386.md (indirect_jump): Likewise.
5460 (tablejump): Likewise.
5461 (*sibcall_memory): Likewise.
5462 (*sibcall_value_memory): Likewise.
5463 Peepholes of indirect call and jump via memory: Likewise.
5464 * config/i386/i386.opt: Likewise.
5465 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5466 (GOT_memory_operand): Likewise.
5467 (call_insn_operand): Likewise.
5468 (sibcall_insn_operand): Likewise.
5469 (GOT32_symbol_operand): Likewise.
5471 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5474 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5475 type rather than type addr's type points to.
5476 (expand_omp_atomic_mutex): Likewise.
5477 (expand_omp_atomic): Likewise.
5479 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5482 * config/i386/i386.c (output_indirect_thunk_function): Use
5483 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5484 for __x86_return_thunk.
5486 2018-01-15 Richard Biener <rguenther@suse.de>
5489 * expmed.c (extract_bit_field_1): Fix typo.
5491 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5494 * config/arm/iterators.md (VF): New mode iterator.
5495 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5496 Remove integer-related logic from pattern.
5497 (neon_vabd<mode>_3): Likewise.
5499 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5502 * common.opt (fstrict-overflow): No longer an alias.
5503 (fwrapv-pointer): New option.
5504 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5505 also for pointer types based on flag_wrapv_pointer.
5506 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5507 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5508 opts->x_flag_wrapv got set.
5509 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5510 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5511 POINTER_TYPE_OVERFLOW_UNDEFINED.
5512 * match.pd: Likewise in address comparison pattern.
5513 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5515 2018-01-15 Richard Biener <rguenther@suse.de>
5518 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5519 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5520 Reset type names to their identifier if their TYPE_DECL doesn't
5521 have linkage (and thus is used for ODR and devirt).
5522 (save_debug_info_for_decl): Remove.
5523 (save_debug_info_for_type): Likewise.
5524 (add_tree_to_fld_list): Adjust.
5525 * tree-pretty-print.c (dump_generic_node): Make dumping of
5526 type names more robust.
5528 2018-01-15 Richard Biener <rguenther@suse.de>
5530 * BASE-VER: Bump to 8.0.1.
5532 2018-01-14 Martin Sebor <msebor@redhat.com>
5535 * builtins.c (check_access): Avoid warning when the no-warning bit
5538 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5540 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5541 * ira-color (allocno_hard_regs_compare): Likewise.
5543 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5546 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5547 Use .pushsection/.popsection.
5549 2018-01-14 Martin Sebor <msebor@redhat.com>
5552 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5554 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5556 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5557 entry from extra_headers.
5558 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5559 extra_headers, make the list bitwise identical to the i?86-*-* one.
5561 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5563 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5564 -mcmodel=large with -mindirect-branch=thunk,
5565 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5566 -mfunction-return=thunk-extern.
5567 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5568 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5569 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5573 * config/i386/i386.c (print_reg): Print the name of the full
5574 integer register without '%'.
5575 (ix86_print_operand): Handle 'V'.
5576 * doc/extend.texi: Document 'V' modifier.
5578 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5580 * config/i386/constraints.md (Bs): Disallow memory operand for
5581 -mindirect-branch-register.
5583 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5584 (GOT_memory_operand): Likewise.
5585 (call_insn_operand): Likewise.
5586 (sibcall_insn_operand): Likewise.
5587 (GOT32_symbol_operand): Likewise.
5588 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5589 for -mindirect-branch-register.
5590 (tablejump): Likewise.
5591 (*sibcall_memory): Likewise.
5592 (*sibcall_value_memory): Likewise.
5593 Disallow peepholes of indirect call and jump via memory for
5594 -mindirect-branch-register.
5595 (*call_pop): Replace m with Bw.
5596 (*call_value_pop): Likewise.
5597 (*sibcall_pop_memory): Replace m with Bs.
5598 * config/i386/i386.opt (mindirect-branch-register): New option.
5599 * doc/invoke.texi: Document -mindirect-branch-register option.
5601 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5603 * config/i386/i386-protos.h (ix86_output_function_return): New.
5604 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5605 set function_return_type.
5606 (indirect_thunk_name): Add ret_p to indicate thunk for function
5608 (output_indirect_thunk_function): Pass false to
5609 indirect_thunk_name.
5610 (ix86_output_indirect_branch_via_reg): Likewise.
5611 (ix86_output_indirect_branch_via_push): Likewise.
5612 (output_indirect_thunk_function): Create alias for function
5613 return thunk if regno < 0.
5614 (ix86_output_function_return): New function.
5615 (ix86_handle_fndecl_attribute): Handle function_return.
5616 (ix86_attribute_table): Add function_return.
5617 * config/i386/i386.h (machine_function): Add
5618 function_return_type.
5619 * config/i386/i386.md (simple_return_internal): Use
5620 ix86_output_function_return.
5621 (simple_return_internal_long): Likewise.
5622 * config/i386/i386.opt (mfunction-return=): New option.
5623 (indirect_branch): Mention -mfunction-return=.
5624 * doc/extend.texi: Document function_return function attribute.
5625 * doc/invoke.texi: Document -mfunction-return= option.
5627 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5629 * config/i386/i386-opts.h (indirect_branch): New.
5630 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5631 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5632 with local indirect jump when converting indirect call and jump.
5633 (ix86_set_indirect_branch_type): New.
5634 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5635 (indirectlabelno): New.
5636 (indirect_thunk_needed): Likewise.
5637 (indirect_thunk_bnd_needed): Likewise.
5638 (indirect_thunks_used): Likewise.
5639 (indirect_thunks_bnd_used): Likewise.
5640 (INDIRECT_LABEL): Likewise.
5641 (indirect_thunk_name): Likewise.
5642 (output_indirect_thunk): Likewise.
5643 (output_indirect_thunk_function): Likewise.
5644 (ix86_output_indirect_branch_via_reg): Likewise.
5645 (ix86_output_indirect_branch_via_push): Likewise.
5646 (ix86_output_indirect_branch): Likewise.
5647 (ix86_output_indirect_jmp): Likewise.
5648 (ix86_code_end): Call output_indirect_thunk_function if needed.
5649 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5651 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5652 (ix86_attribute_table): Add indirect_branch.
5653 * config/i386/i386.h (machine_function): Add indirect_branch_type
5654 and has_local_indirect_jump.
5655 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5657 (tablejump): Likewise.
5658 (*indirect_jump): Use ix86_output_indirect_jmp.
5659 (*tablejump_1): Likewise.
5660 (simple_return_indirect_internal): Likewise.
5661 * config/i386/i386.opt (mindirect-branch=): New option.
5662 (indirect_branch): New.
5665 (thunk-inline): Likewise.
5666 (thunk-extern): Likewise.
5667 * doc/extend.texi: Document indirect_branch function attribute.
5668 * doc/invoke.texi: Document -mindirect-branch= option.
5670 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5673 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5675 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5677 * ipa-inline.c (want_inline_small_function_p): Return false if
5678 inlining has already failed with CIF_FINAL_ERROR.
5679 (update_caller_keys): Call want_inline_small_function_p before
5681 (update_callee_keys): Likewise.
5683 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5685 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5687 (rs6000_quadword_masked_address_p): Likewise.
5688 (quad_aligned_load_p): Likewise.
5689 (quad_aligned_store_p): Likewise.
5690 (const_load_sequence_p): Add comment to describe the outer-most loop.
5691 (mimic_memory_attributes_and_flags): New function.
5692 (rs6000_gen_stvx): Likewise.
5693 (replace_swapped_aligned_store): Likewise.
5694 (rs6000_gen_lvx): Likewise.
5695 (replace_swapped_aligned_load): Likewise.
5696 (replace_swapped_load_constant): Capitalize argument name in
5697 comment describing this function.
5698 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5699 and stores that access quad-word aligned addresses and replace
5700 with stvx or lvx instructions when appropriate.
5701 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5702 New function prototype.
5703 (rs6000_quadword_masked_address_p): Likewise.
5704 (rs6000_gen_lvx): Likewise.
5705 (rs6000_gen_stvx): Likewise.
5706 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5707 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5708 when memory address is aligned.
5709 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5710 this split to select lvx instruction when memory address is aligned.
5711 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5712 instruction when memory address is aligned.
5713 (*vsx_le_perm_load_v16qi): Likewise.
5714 (four unnamed splitters): Modify to select the stvx instruction
5715 when memory is aligned.
5717 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5719 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5720 which appears in the queue multiple times.
5722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5723 Alan Hayward <alan.hayward@arm.com>
5724 David Sherwood <david.sherwood@arm.com>
5726 * tree-vectorizer.h (vec_lower_bound): New structure.
5727 (_loop_vec_info): Add check_nonzero and lower_bounds.
5728 (LOOP_VINFO_CHECK_NONZERO): New macro.
5729 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5730 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5731 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5732 fields. Make seg_len the distance travelled, not including the
5734 (dr_direction_indicator): Declare.
5735 (dr_zero_step_indicator): Likewise.
5736 (dr_known_forward_stride_p): Likewise.
5737 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5739 (runtime_alias_check_p): Allow runtime alias checks with
5741 (operator ==): Compare access_size and align.
5742 (prune_runtime_alias_test_list): Rework for new distinction between
5743 the access_size and seg_len.
5744 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5746 (get_segment_min_max): New function.
5747 (create_intersect_range_checks): Use it.
5748 (dr_step_indicator): New function.
5749 (dr_direction_indicator): Likewise.
5750 (dr_zero_step_indicator): Likewise.
5751 (dr_known_forward_stride_p): Likewise.
5752 * tree-loop-distribution.c (data_ref_segment_size): Return
5753 DR_STEP * (niters - 1).
5754 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5756 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5757 (vect_preserves_scalar_order_p): New function, split out from...
5758 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5759 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5760 (vect_vfa_access_size): New function.
5761 (vect_vfa_align): Likewise.
5762 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5763 (dump_lower_bound): New function.
5764 (vect_check_lower_bound): Likewise.
5765 (vect_small_gap_p): Likewise.
5766 (vectorizable_with_step_bound_p): Likewise.
5767 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5768 depencies if the vectorization factor is 1. Convert the checks
5769 for nonzero steps into checks on the bounds of DR_STEP. Try using
5770 a bunds check for variable steps if the minimum required step is
5771 relatively small. Update calls to the dr_with_seg_len
5772 constructor and to vect_compile_time_alias.
5773 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5775 (vect_loop_versioning): Call it.
5776 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5778 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5781 Alan Hayward <alan.hayward@arm.com>
5782 David Sherwood <david.sherwood@arm.com>
5784 * doc/sourcebuild.texi (vect_scatter_store): Document.
5785 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5787 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5789 * genopinit.c (main): Add supports_vec_scatter_store and
5790 supports_vec_scatter_store_cached to target_optabs.
5791 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5792 IFN_MASK_SCATTER_STORE.
5793 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5795 * internal-fn.h (internal_store_fn_p): Declare.
5796 (internal_fn_stored_value_index): Likewise.
5797 * internal-fn.c (scatter_store_direct): New macro.
5798 (expand_scatter_store_optab_fn): New function.
5799 (direct_scatter_store_optab_supported_p): New macro.
5800 (internal_store_fn_p): New function.
5801 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5802 IFN_MASK_SCATTER_STORE.
5803 (internal_fn_mask_index): Likewise.
5804 (internal_fn_stored_value_index): New function.
5805 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5807 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5808 * optabs-query.c (supports_vec_scatter_store_p): New function.
5809 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5810 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5811 true for scatter stores.
5812 (vect_gather_scatter_fn_p): Handle scatter stores too.
5813 (vect_check_gather_scatter): Consider using scatter stores if
5814 supports_vec_scatter_store_p.
5815 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5817 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5818 internal_fn_stored_value_index.
5819 (check_load_store_masking): Handle scatter stores too.
5820 (vect_get_store_rhs): Make public.
5821 (vectorizable_call): Use internal_store_fn_p.
5822 (vectorizable_store): Handle scatter store internal functions.
5823 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5824 when deciding whether the end of the group has been reached.
5825 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5826 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5827 (mask_scatter_store<mode>): New insns.
5829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5830 Alan Hayward <alan.hayward@arm.com>
5831 David Sherwood <david.sherwood@arm.com>
5833 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5834 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5835 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5837 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5838 Use vect_truncate_gather_scatter_offset if we can't treat the
5839 operation as a normal gather load or scatter store.
5840 (get_group_load_store_type): Take the gather_scatter_info
5841 as argument. Try using a gather load or scatter store for
5842 single-element groups.
5843 (get_load_store_type): Update calls to get_group_load_store_type
5844 and vect_use_strided_gather_scatters_p.
5846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5847 Alan Hayward <alan.hayward@arm.com>
5848 David Sherwood <david.sherwood@arm.com>
5850 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5851 optional tree argument.
5852 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5854 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5855 but continue to use the current value as a fallback.
5856 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5857 to compare the updates.
5858 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5859 (get_load_store_type): Use it when handling a strided access.
5860 (vect_get_strided_load_store_ops): New function.
5861 (vect_get_data_ptr_increment): Likewise.
5862 (vectorizable_load): Handle strided gather loads. Always pass
5863 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5866 Alan Hayward <alan.hayward@arm.com>
5867 David Sherwood <david.sherwood@arm.com>
5869 * doc/md.texi (gather_load@var{m}): Document.
5870 (mask_gather_load@var{m}): Likewise.
5871 * genopinit.c (main): Add supports_vec_gather_load and
5872 supports_vec_gather_load_cached to target_optabs.
5873 * optabs-tree.c (init_tree_optimization_optabs): Use
5874 ggc_cleared_alloc to allocate target_optabs.
5875 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5876 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5878 * internal-fn.h (internal_load_fn_p): Declare.
5879 (internal_gather_scatter_fn_p): Likewise.
5880 (internal_fn_mask_index): Likewise.
5881 (internal_gather_scatter_fn_supported_p): Likewise.
5882 * internal-fn.c (gather_load_direct): New macro.
5883 (expand_gather_load_optab_fn): New function.
5884 (direct_gather_load_optab_supported_p): New macro.
5885 (direct_internal_fn_optab): New function.
5886 (internal_load_fn_p): Likewise.
5887 (internal_gather_scatter_fn_p): Likewise.
5888 (internal_fn_mask_index): Likewise.
5889 (internal_gather_scatter_fn_supported_p): Likewise.
5890 * optabs-query.c (supports_at_least_one_mode_p): New function.
5891 (supports_vec_gather_load_p): Likewise.
5892 * optabs-query.h (supports_vec_gather_load_p): Declare.
5893 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5894 and memory_type field.
5895 (NUM_PATTERNS): Bump to 15.
5896 * tree-vect-data-refs.c: Include internal-fn.h.
5897 (vect_gather_scatter_fn_p): New function.
5898 (vect_describe_gather_scatter_call): Likewise.
5899 (vect_check_gather_scatter): Try using internal functions for
5900 gather loads. Recognize existing calls to a gather load function.
5901 (vect_analyze_data_refs): Consider using gather loads if
5902 supports_vec_gather_load_p.
5903 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5904 (vect_get_gather_scatter_offset_type): Likewise.
5905 (vect_convert_mask_for_vectype): Likewise.
5906 (vect_add_conversion_to_patterm): Likewise.
5907 (vect_try_gather_scatter_pattern): Likewise.
5908 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5909 (vect_vect_recog_func_ptrs): Add it.
5910 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5911 internal_fn_mask_index and internal_gather_scatter_fn_p.
5912 (check_load_store_masking): Take the gather_scatter_info as an
5913 argument and handle gather loads.
5914 (vect_get_gather_scatter_ops): New function.
5915 (vectorizable_call): Check internal_load_fn_p.
5916 (vectorizable_load): Likewise. Handle gather load internal
5918 (vectorizable_store): Update call to check_load_store_masking.
5919 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5920 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5921 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5922 (aarch64_gather_scale_operand_d): New predicates.
5923 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5924 (mask_gather_load<mode>): New insns.
5926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5927 Alan Hayward <alan.hayward@arm.com>
5928 David Sherwood <david.sherwood@arm.com>
5930 * optabs.def (fold_left_plus_optab): New optab.
5931 * doc/md.texi (fold_left_plus_@var{m}): Document.
5932 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5933 * internal-fn.c (fold_left_direct): Define.
5934 (expand_fold_left_optab_fn): Likewise.
5935 (direct_fold_left_optab_supported_p): Likewise.
5936 * fold-const-call.c (fold_const_fold_left): New function.
5937 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5938 * tree-parloops.c (valid_reduction_p): New function.
5939 (gather_scalar_reductions): Use it.
5940 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5941 (vect_finish_replace_stmt): Declare.
5942 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5943 (needs_fold_left_reduction_p): New function, split out from...
5944 (vect_is_simple_reduction): ...here. Accept reductions that
5945 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5946 (vect_force_simple_reduction): Also store the reduction type in
5947 the assignment's STMT_VINFO_REDUC_TYPE.
5948 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5949 (merge_with_identity): New function.
5950 (vect_expand_fold_left): Likewise.
5951 (vectorize_fold_left_reduction): Likewise.
5952 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5953 scalar phi in place for it. Check for target support and reject
5954 cases that would reassociate the operation. Defer the transform
5955 phase to vectorize_fold_left_reduction.
5956 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5957 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5958 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5962 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5963 call to ifc_temp_var.
5965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5966 Alan Hayward <alan.hayward@arm.com>
5967 David Sherwood <david.sherwood@arm.com>
5969 * target.def (legitimize_address_displacement): Take the original
5970 offset as a poly_int.
5971 * targhooks.h (default_legitimize_address_displacement): Update
5973 * targhooks.c (default_legitimize_address_displacement): Likewise.
5974 * doc/tm.texi: Regenerate.
5975 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5976 as an argument, moving assert of ad->disp == ad->disp_term to...
5977 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5978 Try calling targetm.legitimize_address_displacement before expanding
5979 the address rather than afterwards, and adjust for the new interface.
5980 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5981 Match the new hook interface. Handle SVE addresses.
5982 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5987 * Makefile.in (OBJS): Add early-remat.o.
5988 * target.def (select_early_remat_modes): New hook.
5989 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5990 * doc/tm.texi: Regenerate.
5991 * targhooks.h (default_select_early_remat_modes): Declare.
5992 * targhooks.c (default_select_early_remat_modes): New function.
5993 * timevar.def (TV_EARLY_REMAT): New timevar.
5994 * passes.def (pass_early_remat): New pass.
5995 * tree-pass.h (make_pass_early_remat): Declare.
5996 * early-remat.c: New file.
5997 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5999 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6002 Alan Hayward <alan.hayward@arm.com>
6003 David Sherwood <david.sherwood@arm.com>
6005 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6006 vfm1 with a bound_epilog parameter.
6007 (vect_do_peeling): Update calls accordingly, and move the prologue
6008 call earlier in the function. Treat the base bound_epilog as 0 for
6009 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6010 this base when peeling for gaps.
6011 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6012 with fully-masked loops.
6013 (vect_estimate_min_profitable_iters): Handle the single peeled
6014 iteration in that case.
6016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6017 Alan Hayward <alan.hayward@arm.com>
6018 David Sherwood <david.sherwood@arm.com>
6020 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6021 single-element interleaving even if the size is not a power of 2.
6022 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6023 accesses for single-element interleaving if the group size is
6026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6027 Alan Hayward <alan.hayward@arm.com>
6028 David Sherwood <david.sherwood@arm.com>
6030 * doc/md.texi (fold_extract_last_@var{m}): Document.
6031 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6032 * optabs.def (fold_extract_last_optab): New optab.
6033 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6034 * internal-fn.c (fold_extract_direct): New macro.
6035 (expand_fold_extract_optab_fn): Likewise.
6036 (direct_fold_extract_optab_supported_p): Likewise.
6037 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6038 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6039 EXTRACT_LAST_REDUCTION.
6040 (get_initial_def_for_reduction): Do not create an initial vector
6041 for EXTRACT_LAST_REDUCTION reductions.
6042 (vectorizable_reduction): Leave the scalar phi in place for
6043 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6044 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6045 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6046 transform phase to vectorizable_condition.
6047 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6049 (vect_finish_stmt_generation): ...here.
6050 (vect_finish_replace_stmt): New function.
6051 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6052 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6054 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6057 Alan Hayward <alan.hayward@arm.com>
6058 David Sherwood <david.sherwood@arm.com>
6060 * doc/md.texi (extract_last_@var{m}): Document.
6061 * optabs.def (extract_last_optab): New optab.
6062 * internal-fn.def (EXTRACT_LAST): New internal function.
6063 * internal-fn.c (cond_unary_direct): New macro.
6064 (expand_cond_unary_optab_fn): Likewise.
6065 (direct_cond_unary_optab_supported_p): Likewise.
6066 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6067 loops using EXTRACT_LAST.
6068 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6069 (extract_last_<mode>): ...this optab.
6070 (vec_extract<mode><Vel>): Update accordingly.
6072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6073 Alan Hayward <alan.hayward@arm.com>
6074 David Sherwood <david.sherwood@arm.com>
6076 * target.def (empty_mask_is_expensive): New hook.
6077 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6078 * doc/tm.texi: Regenerate.
6079 * targhooks.h (default_empty_mask_is_expensive): Declare.
6080 * targhooks.c (default_empty_mask_is_expensive): New function.
6081 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6082 if the target says that empty masks are expensive.
6083 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6085 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6088 Alan Hayward <alan.hayward@arm.com>
6089 David Sherwood <david.sherwood@arm.com>
6091 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6092 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6093 (vect_use_loop_mask_for_alignment_p): New function.
6094 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6095 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6096 niters_skip argument. Make sure that the first niters_skip elements
6097 of the first iteration are inactive.
6098 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6099 Update call to vect_set_loop_masks_directly.
6100 (get_misalign_in_elems): New function, split out from...
6101 (vect_gen_prolog_loop_niters): ...here.
6102 (vect_update_init_of_dr): Take a code argument that specifies whether
6103 the adjustment should be added or subtracted.
6104 (vect_update_init_of_drs): Likewise.
6105 (vect_prepare_for_masked_peels): New function.
6106 (vect_do_peeling): Skip prologue peeling if we're using a mask
6107 instead. Update call to vect_update_inits_of_drs.
6108 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6110 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6111 alignment. Do not include the number of peeled iterations in
6112 the minimum threshold in that case.
6113 (vectorizable_induction): Adjust the start value down by
6114 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6115 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6116 Take the number of skipped iterations into account when calculating
6118 * tree-vect-stmts.c (vect_gen_while_not): New function.
6120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6121 Alan Hayward <alan.hayward@arm.com>
6122 David Sherwood <david.sherwood@arm.com>
6124 * doc/sourcebuild.texi (vect_fully_masked): Document.
6125 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6127 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6129 (vect_analyze_loop_2): ...here. Don't check the vectorization
6130 factor against the number of loop iterations if the loop is
6133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6134 Alan Hayward <alan.hayward@arm.com>
6135 David Sherwood <david.sherwood@arm.com>
6137 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6138 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6139 (dump_groups): Update accordingly.
6140 (iv_use::mem_type): New member variable.
6141 (address_p): New function.
6142 (record_use): Add a mem_type argument and initialize the new
6144 (record_group_use): Add a mem_type argument. Use address_p.
6145 Remove obsolete null checks of base_object. Update call to record_use.
6146 (find_interesting_uses_op): Update call to record_group_use.
6147 (find_interesting_uses_cond): Likewise.
6148 (find_interesting_uses_address): Likewise.
6149 (get_mem_type_for_internal_fn): New function.
6150 (find_address_like_use): Likewise.
6151 (find_interesting_uses_stmt): Try find_address_like_use before
6152 calling find_interesting_uses_op.
6153 (addr_offset_valid_p): Use the iv mem_type field as the type
6154 of the addressed memory.
6155 (add_autoinc_candidates): Likewise.
6156 (get_address_cost): Likewise.
6157 (split_small_address_groups_p): Use address_p.
6158 (split_address_groups): Likewise.
6159 (add_iv_candidate_for_use): Likewise.
6160 (autoinc_possible_for_pair): Likewise.
6161 (rewrite_groups): Likewise.
6162 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6163 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6164 (get_alias_ptr_type_for_ptr_address): New function.
6165 (rewrite_use_address): Rewrite address uses in calls that were
6166 identified by find_address_like_use.
6168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6169 Alan Hayward <alan.hayward@arm.com>
6170 David Sherwood <david.sherwood@arm.com>
6172 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6174 * gimple-expr.h (is_gimple_addressable: Likewise.
6175 * gimple-expr.c (is_gimple_address): Likewise.
6176 * internal-fn.c (expand_call_mem_ref): New function.
6177 (expand_mask_load_optab_fn): Use it.
6178 (expand_mask_store_optab_fn): Likewise.
6180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6181 Alan Hayward <alan.hayward@arm.com>
6182 David Sherwood <david.sherwood@arm.com>
6184 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6185 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6186 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6187 (cond_umax@var{mode}): Document.
6188 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6189 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6190 (cond_umin_optab, cond_umax_optab): New optabs.
6191 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6192 (COND_IOR, COND_XOR): New internal functions.
6193 * internal-fn.h (get_conditional_internal_fn): Declare.
6194 * internal-fn.c (cond_binary_direct): New macro.
6195 (expand_cond_binary_optab_fn): Likewise.
6196 (direct_cond_binary_optab_supported_p): Likewise.
6197 (get_conditional_internal_fn): New function.
6198 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6199 Cope with reduction statements that are vectorized as calls rather
6201 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6202 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6203 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6204 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6205 (UNSPEC_COND_EOR): New unspecs.
6206 (optab): Add mappings for them.
6207 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6208 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
6215 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6216 * internal-fn.def (WHILE_ULT): New internal function.
6217 * internal-fn.h (direct_internal_fn_supported_p): New override
6218 that takes two types as argument.
6219 * internal-fn.c (while_direct): New macro.
6220 (expand_while_optab_fn): New function.
6221 (convert_optab_supported_p): Likewise.
6222 (direct_while_optab_supported_p): New macro.
6223 * wide-int.h (wi::udiv_ceil): New function.
6224 * tree-vectorizer.h (rgroup_masks): New structure.
6225 (vec_loop_masks): New typedef.
6226 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6228 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6229 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6230 (vect_max_vf): New function.
6231 (slpeel_make_loop_iterate_ntimes): Delete.
6232 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6233 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6234 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6235 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6236 internal-fn.h, stor-layout.h and optabs-query.h.
6237 (vect_set_loop_mask): New function.
6238 (add_preheader_seq): Likewise.
6239 (add_header_seq): Likewise.
6240 (interleave_supported_p): Likewise.
6241 (vect_maybe_permute_loop_masks): Likewise.
6242 (vect_set_loop_masks_directly): Likewise.
6243 (vect_set_loop_condition_masked): Likewise.
6244 (vect_set_loop_condition_unmasked): New function, split out from
6245 slpeel_make_loop_iterate_ntimes.
6246 (slpeel_make_loop_iterate_ntimes): Rename to..
6247 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6248 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6249 (vect_do_peeling): Update call accordingly.
6250 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6252 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6253 mask_compare_type, can_fully_mask_p and fully_masked_p.
6254 (release_vec_loop_masks): New function.
6255 (_loop_vec_info): Use it to free the loop masks.
6256 (can_produce_all_loop_masks_p): New function.
6257 (vect_get_max_nscalars_per_iter): Likewise.
6258 (vect_verify_full_masking): Likewise.
6259 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6260 retries, and free the mask rgroups before retrying. Check loop-wide
6261 reasons for disallowing fully-masked loops. Make the final decision
6262 about whether use a fully-masked loop or not.
6263 (vect_estimate_min_profitable_iters): Do not assume that peeling
6264 for the number of iterations will be needed for fully-masked loops.
6265 (vectorizable_reduction): Disable fully-masked loops.
6266 (vectorizable_live_operation): Likewise.
6267 (vect_halve_mask_nunits): New function.
6268 (vect_double_mask_nunits): Likewise.
6269 (vect_record_loop_mask): Likewise.
6270 (vect_get_loop_mask): Likewise.
6271 (vect_transform_loop): Handle the case in which the final loop
6272 iteration might handle a partial vector. Call vect_set_loop_condition
6273 instead of slpeel_make_loop_iterate_ntimes.
6274 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6275 (check_load_store_masking): New function.
6276 (prepare_load_store_mask): Likewise.
6277 (vectorizable_store): Handle fully-masked loops.
6278 (vectorizable_load): Likewise.
6279 (supportable_widening_operation): Use vect_halve_mask_nunits for
6281 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6282 (vect_gen_while): New function.
6283 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6284 (aarch64_uqdec<mode>): New insn.
6286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6287 Alan Hayward <alan.hayward@arm.com>
6288 David Sherwood <david.sherwood@arm.com>
6290 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6291 (reduc_xor_scal_optab): New optabs.
6292 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6293 (reduc_xor_scal_@var{m}): Document.
6294 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6295 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6297 * fold-const-call.c (fold_const_call): Handle them.
6298 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6299 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6300 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6301 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6302 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6303 (UNSPEC_XORV): New unspecs.
6304 (optab): Add entries for them.
6305 (BITWISEV): New int iterator.
6306 (bit_reduc_op): New int attributes.
6308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6309 Alan Hayward <alan.hayward@arm.com>
6310 David Sherwood <david.sherwood@arm.com>
6312 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6313 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6314 * optabs.def (vec_shl_insert_optab): New optab.
6315 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6316 (duplicate_and_interleave): Likewise.
6317 * tree-vect-loop.c: Include internal-fn.h.
6318 (neutral_op_for_slp_reduction): New function, split out from
6319 get_initial_defs_for_reduction.
6320 (get_initial_def_for_reduction): Handle option 2 for variable-length
6321 vectors by loading the neutral value into a vector and then shifting
6322 the initial value into element 0.
6323 (get_initial_defs_for_reduction): Replace the code argument with
6324 the neutral value calculated by neutral_op_for_slp_reduction.
6325 Use gimple_build_vector for constant-length vectors.
6326 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6327 but the first group_size elements have a neutral value.
6328 Use duplicate_and_interleave otherwise.
6329 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6330 Update call to get_initial_defs_for_reduction. Handle SLP
6331 reductions for variable-length vectors by creating one vector
6332 result for each scalar result, with the elements associated
6333 with other scalar results stubbed out with the neutral value.
6334 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6335 Require IFN_VEC_SHL_INSERT for double reductions on
6336 variable-length vectors, or SLP reductions that have
6337 a neutral value. Require can_duplicate_and_interleave_p
6338 support for variable-length unchained SLP reductions if there
6339 is no neutral value, such as for MIN/MAX reductions. Also require
6340 the number of vector elements to be a multiple of the number of
6341 SLP statements when doing variable-length unchained SLP reductions.
6342 Update call to vect_create_epilog_for_reduction.
6343 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6344 and remove initial values.
6345 (duplicate_and_interleave): Make public.
6346 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6347 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6350 Alan Hayward <alan.hayward@arm.com>
6351 David Sherwood <david.sherwood@arm.com>
6353 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6354 (can_duplicate_and_interleave_p): New function.
6355 (vect_get_and_check_slp_defs): Take the vector of statements
6356 rather than just the current one. Remove excess parentheses.
6357 Restriction rejectinon of vect_constant_def and vect_external_def
6358 for variable-length vectors to boolean types, or types for which
6359 can_duplicate_and_interleave_p is false.
6360 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6361 (duplicate_and_interleave): New function.
6362 (vect_get_constant_vectors): Use gimple_build_vector for
6363 constant-length vectors and suitable variable-length constant
6364 vectors. Use duplicate_and_interleave for other variable-length
6365 vectors. Don't defer the update when inserting new statements.
6367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6368 Alan Hayward <alan.hayward@arm.com>
6369 David Sherwood <david.sherwood@arm.com>
6371 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6372 min_profitable_iters doesn't go negative.
6374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6375 Alan Hayward <alan.hayward@arm.com>
6376 David Sherwood <david.sherwood@arm.com>
6378 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6379 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6380 * optabs.def (vec_mask_load_lanes_optab): New optab.
6381 (vec_mask_store_lanes_optab): Likewise.
6382 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6383 (MASK_STORE_LANES): Likewise.
6384 * internal-fn.c (mask_load_lanes_direct): New macro.
6385 (mask_store_lanes_direct): Likewise.
6386 (expand_mask_load_optab_fn): Handle masked operations.
6387 (expand_mask_load_lanes_optab_fn): New macro.
6388 (expand_mask_store_optab_fn): Handle masked operations.
6389 (expand_mask_store_lanes_optab_fn): New macro.
6390 (direct_mask_load_lanes_optab_supported_p): Likewise.
6391 (direct_mask_store_lanes_optab_supported_p): Likewise.
6392 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6394 (vect_load_lanes_supported): Likewise.
6395 * tree-vect-data-refs.c (strip_conversion): New function.
6396 (can_group_stmts_p): Likewise.
6397 (vect_analyze_data_ref_accesses): Use it instead of checking
6398 for a pair of assignments.
6399 (vect_store_lanes_supported): Take a masked_p parameter.
6400 (vect_load_lanes_supported): Likewise.
6401 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6402 vect_store_lanes_supported and vect_load_lanes_supported.
6403 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6404 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6405 parameter. Don't allow gaps for masked accesses.
6406 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6407 and vect_load_lanes_supported.
6408 (get_load_store_type): Take a masked_p parameter and update
6409 call to get_group_load_store_type.
6410 (vectorizable_store): Update call to get_load_store_type.
6411 Handle IFN_MASK_STORE_LANES.
6412 (vectorizable_load): Update call to get_load_store_type.
6413 Handle IFN_MASK_LOAD_LANES.
6415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6416 Alan Hayward <alan.hayward@arm.com>
6417 David Sherwood <david.sherwood@arm.com>
6419 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6421 * config/aarch64/aarch64-protos.h
6422 (aarch64_sve_struct_memory_operand_p): Declare.
6423 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6424 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6425 (VPRED, vpred): Handle SVE structure modes.
6426 * config/aarch64/constraints.md (Utx): New constraint.
6427 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6428 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6429 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6430 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6431 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6432 structure modes. Split into pieces after RA.
6433 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6434 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6436 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6437 SVE structure modes.
6438 (aarch64_classify_address): Likewise.
6439 (sizetochar): Move earlier in file.
6440 (aarch64_print_operand): Handle SVE register lists.
6441 (aarch64_array_mode): New function.
6442 (aarch64_sve_struct_memory_operand_p): Likewise.
6443 (TARGET_ARRAY_MODE): Redefine.
6445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6446 Alan Hayward <alan.hayward@arm.com>
6447 David Sherwood <david.sherwood@arm.com>
6449 * target.def (array_mode): New target hook.
6450 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6451 * doc/tm.texi: Regenerate.
6452 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6453 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6454 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6456 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6460 Alan Hayward <alan.hayward@arm.com>
6461 David Sherwood <david.sherwood@arm.com>
6463 * fold-const.c (fold_binary_loc): Check the argument types
6464 rather than the result type when testing for a vector operation.
6466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6468 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6469 * doc/tm.texi: Regenerate.
6471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6472 Alan Hayward <alan.hayward@arm.com>
6473 David Sherwood <david.sherwood@arm.com>
6475 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6476 (sve): Document new AArch64 extension.
6477 * doc/md.texi (w): Extend the description of the AArch64
6478 constraint to include SVE vectors.
6479 (Upl, Upa): Document new AArch64 predicate constraints.
6480 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6482 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6483 (msve-vector-bits=): New option.
6484 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6485 SVE when these are disabled.
6486 (sve): New extension.
6487 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6488 modes. Adjust their number of units based on aarch64_sve_vg.
6489 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6490 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6491 aarch64_addr_query_type.
6492 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6493 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6494 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6495 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6496 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6497 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6498 (aarch64_simd_imm_zero_p): Delete.
6499 (aarch64_check_zero_based_sve_index_immediate): Declare.
6500 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6501 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6502 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6503 (aarch64_sve_float_mul_immediate_p): Likewise.
6504 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6506 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6507 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6508 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6509 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6510 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6511 (aarch64_regmode_natural_size): Likewise.
6512 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6513 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6515 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6516 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6517 for VG and the SVE predicate registers.
6518 (V_ALIASES): Add a "z"-prefixed alias.
6519 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6520 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6521 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6522 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6523 (REG_CLASS_NAMES): Add entries for them.
6524 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6525 and the predicate registers.
6526 (aarch64_sve_vg): Declare.
6527 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6528 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6529 (REGMODE_NATURAL_SIZE): Define.
6530 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6532 * config/aarch64/aarch64.c: Include cfgrtl.h.
6533 (simd_immediate_info): Add a constructor for series vectors,
6534 and an associated step field.
6535 (aarch64_sve_vg): New variable.
6536 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6537 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6538 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6539 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6540 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6541 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6542 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6543 (aarch64_get_mask_mode): New functions.
6544 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6545 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6546 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6547 predicate modes and predicate registers. Explicitly restrict
6548 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6549 to store a vector mode if it is recognized by
6550 aarch64_classify_vector_mode.
6551 (aarch64_regmode_natural_size): New function.
6552 (aarch64_hard_regno_caller_save_mode): Return the original mode
6554 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6555 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6556 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6557 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6559 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6560 does not overlap dest if the function is frame-related. Handle
6562 (aarch64_split_add_offset): New function.
6563 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6564 them aarch64_add_offset.
6565 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6566 and update call to aarch64_sub_sp.
6567 (aarch64_add_cfa_expression): New function.
6568 (aarch64_expand_prologue): Pass extra temporary registers to the
6569 functions above. Handle the case in which we need to emit new
6570 DW_CFA_expressions for registers that were originally saved
6571 relative to the stack pointer, but now have to be expressed
6572 relative to the frame pointer.
6573 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6575 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6576 IP0 and IP1 values for SVE frames.
6577 (aarch64_expand_vec_series): New function.
6578 (aarch64_expand_sve_widened_duplicate): Likewise.
6579 (aarch64_expand_sve_const_vector): Likewise.
6580 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6581 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6582 into the register, rather than emitting a SET directly.
6583 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6584 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6585 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6586 (offset_9bit_signed_scaled_p): New functions.
6587 (aarch64_replicate_bitmask_imm): New function.
6588 (aarch64_bitmask_imm): Use it.
6589 (aarch64_cannot_force_const_mem): Reject expressions involving
6590 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6591 (aarch64_classify_index): Handle SVE indices, by requiring
6592 a plain register index with a scale that matches the element size.
6593 (aarch64_classify_address): Handle SVE addresses. Assert that
6594 the mode of the address is VOIDmode or an integer mode.
6595 Update call to aarch64_classify_symbol.
6596 (aarch64_classify_symbolic_expression): Update call to
6597 aarch64_classify_symbol.
6598 (aarch64_const_vec_all_in_range_p): New function.
6599 (aarch64_print_vector_float_operand): Likewise.
6600 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6601 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6602 and the FP immediates 1.0 and 0.5.
6603 (aarch64_print_address_internal): Handle SVE addresses.
6604 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6605 (aarch64_regno_regclass): Handle predicate registers.
6606 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6608 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6609 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6610 (aarch64_convert_sve_vector_bits): New function.
6611 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6612 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6614 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6615 Handle SVE vector and predicate modes. Accept VL-based constants
6616 that need only one temporary register, and VL offsets that require
6617 no temporary registers.
6618 (aarch64_conditional_register_usage): Mark the predicate registers
6619 as fixed if SVE isn't available.
6620 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6621 Return true for SVE vector and predicate modes.
6622 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6623 rather than an unsigned int. Handle SVE modes.
6624 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6626 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6628 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6629 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6630 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6631 (aarch64_sve_float_mul_immediate_p): New functions.
6632 (aarch64_sve_valid_immediate): New function.
6633 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6634 Explicitly reject structure modes. Check for INDEX constants.
6635 Handle PTRUE and PFALSE constants.
6636 (aarch64_check_zero_based_sve_index_immediate): New function.
6637 (aarch64_simd_imm_zero_p): Delete.
6638 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6639 vector modes. Accept constants in the range of CNT[BHWD].
6640 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6641 ask for an Advanced SIMD mode.
6642 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6643 (aarch64_simd_vector_alignment): Handle SVE predicates.
6644 (aarch64_vectorize_preferred_vector_alignment): New function.
6645 (aarch64_simd_vector_alignment_reachable): Use it instead of
6647 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6648 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6650 (MAX_VECT_LEN): Delete.
6651 (expand_vec_perm_d): Add a vec_flags field.
6652 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6653 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6654 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6656 (aarch64_evpc_rev): Rename to...
6657 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6658 (aarch64_evpc_rev_global): New function.
6659 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6660 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6662 (aarch64_evpc_sve_tbl): New function.
6663 (aarch64_expand_vec_perm_const_1): Update after rename of
6664 aarch64_evpc_rev. Handle SVE permutes too, trying
6665 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6666 than aarch64_evpc_tbl.
6667 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6668 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6669 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6670 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6671 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6672 (aarch64_expand_sve_vcond): New functions.
6673 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6674 of aarch64_vector_mode_p.
6675 (aarch64_dwarf_poly_indeterminate_value): New function.
6676 (aarch64_compute_pressure_classes): Likewise.
6677 (aarch64_can_change_mode_class): Likewise.
6678 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6679 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6680 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6681 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6682 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6683 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6684 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6685 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6687 (Dn, Dl, Dr): Accept const as well as const_vector.
6688 (Dz): Likewise. Compare against CONST0_RTX.
6689 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6690 of "vector" where appropriate.
6691 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6692 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6693 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6694 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6695 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6696 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6697 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6698 (v_int_equiv): Extend to SVE modes.
6699 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6701 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6702 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6703 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6704 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6705 (SVE_COND_FP_CMP): New int iterators.
6706 (perm_hilo): Handle the new unpack unspecs.
6707 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6709 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6710 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6711 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6712 (aarch64_equality_operator, aarch64_constant_vector_operand)
6713 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6714 (aarch64_sve_nonimmediate_operand): Likewise.
6715 (aarch64_sve_general_operand): Likewise.
6716 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6717 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6718 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6719 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6720 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6721 (aarch64_sve_float_arith_immediate): Likewise.
6722 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6723 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6724 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6725 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6726 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6727 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6728 (aarch64_sve_float_arith_operand): Likewise.
6729 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6730 (aarch64_sve_float_mul_operand): Likewise.
6731 (aarch64_sve_vec_perm_operand): Likewise.
6732 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6733 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6734 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6735 as well as const_vector.
6736 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6737 in file. Use CONST0_RTX and CONSTM1_RTX.
6738 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6739 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6740 Use aarch64_simd_imm_zero.
6741 * config/aarch64/aarch64-sve.md: New file.
6742 * config/aarch64/aarch64.md: Include it.
6743 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6744 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6745 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6746 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6747 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6748 (sve): New attribute.
6749 (enabled): Disable instructions with the sve attribute unless
6751 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6752 aarch64_expand_mov_immediate.
6753 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6754 CNT[BHSD] immediates.
6755 (movti): Split CONST_POLY_INT moves into two halves.
6756 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6757 Split additions that need a temporary here if the destination
6758 is the stack pointer.
6759 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6760 (*add<mode>3_poly_1): New instruction.
6761 (set_clobber_cc): New expander.
6763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6765 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6766 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6767 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6768 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6769 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6770 Change innermode from fixed_mode_size to machine_mode.
6771 (simplify_subreg): Update call accordingly. Handle a constant-sized
6772 subreg of a variable-length CONST_VECTOR.
6774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6775 Alan Hayward <alan.hayward@arm.com>
6776 David Sherwood <david.sherwood@arm.com>
6778 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6779 (add_offset_to_base): New function, split out from...
6780 (create_mem_ref): ...here. When handling a scale other than 1,
6781 check first whether the address is valid without the offset.
6782 Add it into the base if so, leaving the index and scale as-is.
6784 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6787 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6788 fold_for_warn before checking if arg2 is INTEGER_CST.
6790 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6792 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6793 (store_multiple_operation): Delete.
6794 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6795 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6796 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6797 guarded by TARGET_STRING.
6798 (rs6000_output_load_multiple): Delete.
6799 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6800 OPTION_MASK_STRING / TARGET_STRING handling.
6801 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6802 (const rs6000_opt_masks) <"string">: Change mask to 0.
6803 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6804 (MASK_STRING): Delete.
6805 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6807 (load_multiple): Delete.
6814 (store_multiple): Delete.
6821 (movmemsi_8reg): Delete.
6822 (corresponding unnamed define_insn): Delete.
6823 (movmemsi_6reg): Delete.
6824 (corresponding unnamed define_insn): Delete.
6825 (movmemsi_4reg): Delete.
6826 (corresponding unnamed define_insn): Delete.
6827 (movmemsi_2reg): Delete.
6828 (corresponding unnamed define_insn): Delete.
6829 (movmemsi_1reg): Delete.
6830 (corresponding unnamed define_insn): Delete.
6831 * config/rs6000/rs6000.opt (mno-string): New.
6832 (mstring): Replace by deprecation warning stub.
6833 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6835 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6837 * regrename.c (regrename_do_replace): If replacing the same
6838 reg multiple times, try to reuse last created gen_raw_REG.
6841 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6842 main to workaround a bug in GDB.
6844 2018-01-12 Tom de Vries <tom@codesourcery.com>
6847 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6849 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6851 PR rtl-optimization/80481
6852 * ira-color.c (get_cap_member): New function.
6853 (allocnos_conflict_by_live_ranges_p): Use it.
6854 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6855 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6857 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6860 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6861 (*saddl_se_1): Ditto.
6863 (*ssubl_se_1): Ditto.
6865 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6867 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6868 rather than wi::to_widest for DR_INITs.
6869 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6870 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6871 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6873 (vect_analyze_group_access_1): Note that here.
6875 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6877 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6878 polynomial type sizes.
6880 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6882 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6883 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6884 (gimple_add_tmp_var): Likewise.
6886 2018-01-12 Martin Liska <mliska@suse.cz>
6888 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6889 (gimple_alloc_sizes): Likewise.
6890 (dump_gimple_statistics): Use PRIu64 in printf format.
6891 * gimple.h: Change uint64_t to int.
6893 2018-01-12 Martin Liska <mliska@suse.cz>
6895 * tree-core.h: Use uint64_t instead of int.
6896 * tree.c (tree_node_counts): Likewise.
6897 (tree_node_sizes): Likewise.
6898 (dump_tree_statistics): Use PRIu64 in printf format.
6900 2018-01-12 Martin Liska <mliska@suse.cz>
6902 * Makefile.in: As qsort_chk is implemented in vec.c, add
6903 vec.o to linkage of gencfn-macros.
6904 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6905 passing the info to record_node_allocation_statistics.
6906 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6908 * ggc-common.c (struct ggc_usage): Add operator== and use
6909 it in operator< and compare function.
6910 * mem-stats.h (struct mem_usage): Likewise.
6911 * vec.c (struct vec_usage): Remove operator< and compare
6912 function. Can be simply inherited.
6914 2018-01-12 Martin Jambor <mjambor@suse.cz>
6917 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6918 * tree-ssa-math-opts.c: Include domwalk.h.
6919 (convert_mult_to_fma_1): New function.
6920 (fma_transformation_info): New type.
6921 (fma_deferring_state): Likewise.
6922 (cancel_fma_deferring): New function.
6923 (result_of_phi): Likewise.
6924 (last_fma_candidate_feeds_initial_phi): Likewise.
6925 (convert_mult_to_fma): Added deferring logic, split actual
6926 transformation to convert_mult_to_fma_1.
6927 (math_opts_dom_walker): New type.
6928 (math_opts_dom_walker::after_dom_children): New method, body moved
6929 here from pass_optimize_widening_mul::execute, added deferring logic
6931 (pass_optimize_widening_mul::execute): Moved most of code to
6932 math_opts_dom_walker::after_dom_children.
6933 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6934 * config/i386/i386.c (ix86_option_override_internal): Added
6935 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6937 2018-01-12 Richard Biener <rguenther@suse.de>
6940 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6941 inline instance vars.
6943 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6946 * config/rx/rx.c (rx_is_restricted_memory_address):
6949 2018-01-12 Richard Biener <rguenther@suse.de>
6951 PR tree-optimization/80846
6952 * target.def (split_reduction): New target hook.
6953 * targhooks.c (default_split_reduction): New function.
6954 * targhooks.h (default_split_reduction): Declare.
6955 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6956 target requests first reduce vectors by combining low and high
6958 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6959 (get_vectype_for_scalar_type_and_size): Export.
6960 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6961 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6962 * doc/tm.texi: Regenerate.
6963 * config/i386/i386.c (ix86_split_reduction): Implement
6964 TARGET_VECTORIZE_SPLIT_REDUCTION.
6966 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6969 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6970 in PIC mode except for TARGET_VXWORKS_RTP.
6971 * config/sparc/sparc.c: Include cfgrtl.h.
6972 (TARGET_INIT_PIC_REG): Define.
6973 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6974 (sparc_pic_register_p): New predicate.
6975 (sparc_legitimate_address_p): Use it.
6976 (sparc_legitimize_pic_address): Likewise.
6977 (sparc_delegitimize_address): Likewise.
6978 (sparc_mode_dependent_address_p): Likewise.
6979 (gen_load_pcrel_sym): Remove 4th parameter.
6980 (load_got_register): Adjust call to above. Remove obsolete stuff.
6981 (sparc_expand_prologue): Do not call load_got_register here.
6982 (sparc_flat_expand_prologue): Likewise.
6983 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6984 (sparc_use_pseudo_pic_reg): New function.
6985 (sparc_init_pic_reg): Likewise.
6986 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6987 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6989 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6991 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6992 Add item for branch_cost.
6994 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6996 PR rtl-optimization/83565
6997 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6998 not extend the result to a larger mode for rotate operations.
6999 (num_sign_bit_copies1): Likewise.
7001 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7004 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7006 Use values-Xc.o for -pedantic.
7007 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7009 2018-01-12 Martin Liska <mliska@suse.cz>
7012 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7014 (possible_polymorphic_call_targets): Use it.
7015 (ipa_devirt): Likewise.
7017 2018-01-12 Martin Liska <mliska@suse.cz>
7019 * profile-count.h (enum profile_quality): Use 0 as invalid
7020 enum value of profile_quality.
7022 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7024 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7025 -mext-string options.
7027 2018-01-12 Richard Biener <rguenther@suse.de>
7029 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7030 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7031 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7033 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7035 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7037 * configure.ac (--with-long-double-format): Add support for the
7038 configuration option to change the default long double format on
7040 * config.gcc (powerpc*-linux*-*): Likewise.
7041 * configure: Regenerate.
7042 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7043 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7044 used without modification.
7046 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7048 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7049 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7050 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7051 MISC_BUILTIN_SPEC_BARRIER.
7052 (rs6000_init_builtins): Likewise.
7053 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7055 (speculation_barrier): New define_insn.
7056 * doc/extend.texi: Document __builtin_speculation_barrier.
7058 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7061 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7062 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7063 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7065 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7066 integral modes instead of "ss" and "sd".
7067 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7068 vectors with 32-bit and 64-bit elements.
7069 (vecdupssescalarmodesuffix): New mode attribute.
7070 (vec_dup<mode>): Use it.
7072 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7075 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7076 frame if argument is passed on stack.
7078 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7081 * ree.c (combine_reaching_defs): Optimize also
7082 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7083 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7085 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7088 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7090 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7093 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7094 after they are computed.
7096 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7098 PR tree-optimization/83695
7099 * gimple-loop-linterchange.cc
7100 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7101 reset cached scev information after interchange.
7102 (pass_linterchange::execute): Remove call to scev_reset_htab.
7104 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7106 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7107 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7108 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7109 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7110 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7111 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7112 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7113 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7114 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7115 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7116 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7117 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7118 (V_lane_reg): Likewise.
7119 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7121 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7122 (vfmal_lane_low<mode>_intrinsic,
7123 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7124 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7125 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7126 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7127 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7128 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7130 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7132 * config/arm/arm-cpus.in (fp16fml): New feature.
7133 (ALL_SIMD): Add fp16fml.
7134 (armv8.2-a): Add fp16fml as an option.
7135 (armv8.3-a): Likewise.
7136 (armv8.4-a): Add fp16fml as part of fp16.
7137 * config/arm/arm.h (TARGET_FP16FML): Define.
7138 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7140 * config/arm/arm-modes.def (V2HF): Define.
7141 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7142 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7143 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7144 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7145 vfmsl_low, vfmsl_high): New set of builtins.
7146 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7147 (vfml_op): New code attribute.
7148 (VFMLHALVES): New int iterator.
7149 (VFML, VFMLSEL): New mode attributes.
7150 (V_reg): Define mapping for V2HF.
7151 (V_hi, V_lo): New mode attributes.
7152 (VF_constraint): Likewise.
7153 (vfml_half, vfml_half_selector): New int attributes.
7154 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7156 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7157 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7159 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7160 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7161 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7162 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7164 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7165 Document new effective target and option set.
7167 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7169 * config/arm/arm-cpus.in (armv8_4): New feature.
7170 (ARMv8_4a): New fgroup.
7171 (armv8.4-a): New arch.
7172 * config/arm/arm-tables.opt: Regenerate.
7173 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7174 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7175 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7176 Add matching rules for -march=armv8.4-a and extensions.
7177 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7179 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7182 * config/rx/rx.md (BW): New mode attribute.
7183 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7185 2018-01-11 Richard Biener <rguenther@suse.de>
7187 PR tree-optimization/83435
7188 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7189 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7190 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7192 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7193 Alan Hayward <alan.hayward@arm.com>
7194 David Sherwood <david.sherwood@arm.com>
7196 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7198 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7199 (aarch64_print_address_internal): Use it to check for a zero offset.
7201 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7202 Alan Hayward <alan.hayward@arm.com>
7203 David Sherwood <david.sherwood@arm.com>
7205 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7206 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7207 Return a poly_int64 rather than a HOST_WIDE_INT.
7208 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7209 rather than a HOST_WIDE_INT.
7210 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7211 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7212 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7213 final_offset from HOST_WIDE_INT to poly_int64.
7214 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7215 to_constant when getting the number of units in an Advanced SIMD
7217 (aarch64_builtin_vectorized_function): Check for a constant number
7219 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7221 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7222 attribute instead of GET_MODE_NUNITS.
7223 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7224 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7225 GET_MODE_SIZE for fixed-size registers.
7226 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7227 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7228 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7229 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7230 (aarch64_print_operand, aarch64_print_address_internal)
7231 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7232 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7233 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7234 Handle polynomial GET_MODE_SIZE.
7235 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7236 wider than SImode without modification.
7237 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7238 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7239 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7240 passing and returning SVE modes.
7241 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7242 rather than GEN_INT.
7243 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7244 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7245 (aarch64_allocate_and_probe_stack_space): Likewise.
7246 (aarch64_layout_frame): Cope with polynomial offsets.
7247 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7248 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7250 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7251 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7252 poly_int64 rather than a HOST_WIDE_INT.
7253 (aarch64_get_separate_components, aarch64_process_components)
7254 (aarch64_expand_prologue, aarch64_expand_epilogue)
7255 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7256 (aarch64_anchor_offset): New function, split out from...
7257 (aarch64_legitimize_address): ...here.
7258 (aarch64_builtin_vectorization_cost): Handle polynomial
7259 TYPE_VECTOR_SUBPARTS.
7260 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7262 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7263 number of elements from the PARALLEL rather than the mode.
7264 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7265 rather than GET_MODE_BITSIZE.
7266 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7267 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7268 (aarch64_expand_vec_perm_const_1): Handle polynomial
7269 d->perm.length () and d->perm elements.
7270 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7271 Apply to_constant to d->perm elements.
7272 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7273 polynomial CONST_VECTOR_NUNITS.
7274 (aarch64_move_pointer): Take amount as a poly_int64 rather
7276 (aarch64_progress_pointer): Avoid temporary variable.
7277 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7278 the mode attribute instead of GET_MODE.
7280 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7281 Alan Hayward <alan.hayward@arm.com>
7282 David Sherwood <david.sherwood@arm.com>
7284 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7285 x exists before using it.
7286 (aarch64_add_constant_internal): Rename to...
7287 (aarch64_add_offset_1): ...this. Replace regnum with separate
7288 src and dest rtxes. Handle the case in which they're different,
7289 including when the offset is zero. Replace scratchreg with an rtx.
7290 Use 2 additions if there is no spare register into which we can
7291 move a 16-bit constant.
7292 (aarch64_add_constant): Delete.
7293 (aarch64_add_offset): Replace reg with separate src and dest
7294 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7295 Use aarch64_add_offset_1.
7296 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7297 an rtx rather than an int. Take the delta as a poly_int64
7298 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7299 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7300 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7301 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7302 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7304 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7305 aarch64_add_constant.
7307 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7309 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7310 Use scalar_float_mode.
7312 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7314 * config/aarch64/aarch64-simd.md
7315 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7316 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7317 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7318 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7319 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7320 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7321 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7322 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7323 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7324 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7326 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7329 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7330 targ_options->x_arm_arch_string is non NULL.
7332 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7334 * config/aarch64/aarch64.h
7335 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7337 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7340 * expmed.c (emit_store_flag_force): Swap if const op0
7341 and change VOIDmode to mode of op0.
7343 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7345 PR rtl-optimization/83761
7346 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7347 than bytes to mode_for_size.
7349 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7352 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7353 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7356 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7359 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7360 when in layout mode.
7361 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7362 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7365 2018-01-10 Michael Collison <michael.collison@arm.com>
7367 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7368 * config/aarch64/aarch64-option-extension.def: Add
7369 AARCH64_OPT_EXTENSION of 'fp16fml'.
7370 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7371 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7372 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7373 * config/aarch64/constraints.md (Ui7): New constraint.
7374 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7375 (VFMLA_SEL_W): Ditto.
7378 (VFMLA16_LOW): New int iterator.
7379 (VFMLA16_HIGH): Ditto.
7380 (UNSPEC_FMLAL): New unspec.
7381 (UNSPEC_FMLSL): Ditto.
7382 (UNSPEC_FMLAL2): Ditto.
7383 (UNSPEC_FMLSL2): Ditto.
7384 (f16mac): New code attribute.
7385 * config/aarch64/aarch64-simd-builtins.def
7386 (aarch64_fmlal_lowv2sf): Ditto.
7387 (aarch64_fmlsl_lowv2sf): Ditto.
7388 (aarch64_fmlalq_lowv4sf): Ditto.
7389 (aarch64_fmlslq_lowv4sf): Ditto.
7390 (aarch64_fmlal_highv2sf): Ditto.
7391 (aarch64_fmlsl_highv2sf): Ditto.
7392 (aarch64_fmlalq_highv4sf): Ditto.
7393 (aarch64_fmlslq_highv4sf): Ditto.
7394 (aarch64_fmlal_lane_lowv2sf): Ditto.
7395 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7396 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7397 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7398 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7399 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7400 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7401 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7402 (aarch64_fmlal_lane_highv2sf): Ditto.
7403 (aarch64_fmlsl_lane_highv2sf): Ditto.
7404 (aarch64_fmlal_laneq_highv2sf): Ditto.
7405 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7406 (aarch64_fmlalq_lane_highv4sf): Ditto.
7407 (aarch64_fmlsl_lane_highv4sf): Ditto.
7408 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7409 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7410 * config/aarch64/aarch64-simd.md:
7411 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7412 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7413 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7414 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7415 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7416 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7417 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7418 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7419 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7420 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7421 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7422 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7423 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7424 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7425 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7426 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7427 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7428 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7429 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7430 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7431 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7432 (vfmlsl_low_u32): Ditto.
7433 (vfmlalq_low_u32): Ditto.
7434 (vfmlslq_low_u32): Ditto.
7435 (vfmlal_high_u32): Ditto.
7436 (vfmlsl_high_u32): Ditto.
7437 (vfmlalq_high_u32): Ditto.
7438 (vfmlslq_high_u32): Ditto.
7439 (vfmlal_lane_low_u32): Ditto.
7440 (vfmlsl_lane_low_u32): Ditto.
7441 (vfmlal_laneq_low_u32): Ditto.
7442 (vfmlsl_laneq_low_u32): Ditto.
7443 (vfmlalq_lane_low_u32): Ditto.
7444 (vfmlslq_lane_low_u32): Ditto.
7445 (vfmlalq_laneq_low_u32): Ditto.
7446 (vfmlslq_laneq_low_u32): Ditto.
7447 (vfmlal_lane_high_u32): Ditto.
7448 (vfmlsl_lane_high_u32): Ditto.
7449 (vfmlal_laneq_high_u32): Ditto.
7450 (vfmlsl_laneq_high_u32): Ditto.
7451 (vfmlalq_lane_high_u32): Ditto.
7452 (vfmlslq_lane_high_u32): Ditto.
7453 (vfmlalq_laneq_high_u32): Ditto.
7454 (vfmlslq_laneq_high_u32): Ditto.
7455 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7456 (AARCH64_FL_FOR_ARCH8_4): New.
7457 (AARCH64_ISA_F16FML): New ISA flag.
7458 (TARGET_F16FML): New feature flag for fp16fml.
7459 (doc/invoke.texi): Document new fp16fml option.
7461 2018-01-10 Michael Collison <michael.collison@arm.com>
7463 * config/aarch64/aarch64-builtins.c:
7464 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7465 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7466 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7467 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7468 (AARCH64_ISA_SHA3): New ISA flag.
7469 (TARGET_SHA3): New feature flag for sha3.
7470 * config/aarch64/iterators.md (sha512_op): New int attribute.
7471 (CRYPTO_SHA512): New int iterator.
7472 (UNSPEC_SHA512H): New unspec.
7473 (UNSPEC_SHA512H2): Ditto.
7474 (UNSPEC_SHA512SU0): Ditto.
7475 (UNSPEC_SHA512SU1): Ditto.
7476 * config/aarch64/aarch64-simd-builtins.def
7477 (aarch64_crypto_sha512hqv2di): New builtin.
7478 (aarch64_crypto_sha512h2qv2di): Ditto.
7479 (aarch64_crypto_sha512su0qv2di): Ditto.
7480 (aarch64_crypto_sha512su1qv2di): Ditto.
7481 (aarch64_eor3qv8hi): Ditto.
7482 (aarch64_rax1qv2di): Ditto.
7483 (aarch64_xarqv2di): Ditto.
7484 (aarch64_bcaxqv8hi): Ditto.
7485 * config/aarch64/aarch64-simd.md:
7486 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7487 (aarch64_crypto_sha512su0qv2di): Ditto.
7488 (aarch64_crypto_sha512su1qv2di): Ditto.
7489 (aarch64_eor3qv8hi): Ditto.
7490 (aarch64_rax1qv2di): Ditto.
7491 (aarch64_xarqv2di): Ditto.
7492 (aarch64_bcaxqv8hi): Ditto.
7493 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7494 (vsha512h2q_u64): Ditto.
7495 (vsha512su0q_u64): Ditto.
7496 (vsha512su1q_u64): Ditto.
7497 (veor3q_u16): Ditto.
7498 (vrax1q_u64): Ditto.
7500 (vbcaxq_u16): Ditto.
7501 * config/arm/types.md (crypto_sha512): New type attribute.
7502 (crypto_sha3): Ditto.
7503 (doc/invoke.texi): Document new sha3 option.
7505 2018-01-10 Michael Collison <michael.collison@arm.com>
7507 * config/aarch64/aarch64-builtins.c:
7508 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7509 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7510 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7511 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7512 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7513 (AARCH64_ISA_SM4): New ISA flag.
7514 (TARGET_SM4): New feature flag for sm4.
7515 * config/aarch64/aarch64-simd-builtins.def
7516 (aarch64_sm3ss1qv4si): Ditto.
7517 (aarch64_sm3tt1aq4si): Ditto.
7518 (aarch64_sm3tt1bq4si): Ditto.
7519 (aarch64_sm3tt2aq4si): Ditto.
7520 (aarch64_sm3tt2bq4si): Ditto.
7521 (aarch64_sm3partw1qv4si): Ditto.
7522 (aarch64_sm3partw2qv4si): Ditto.
7523 (aarch64_sm4eqv4si): Ditto.
7524 (aarch64_sm4ekeyqv4si): Ditto.
7525 * config/aarch64/aarch64-simd.md:
7526 (aarch64_sm3ss1qv4si): Ditto.
7527 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7528 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7529 (aarch64_sm4eqv4si): Ditto.
7530 (aarch64_sm4ekeyqv4si): Ditto.
7531 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7532 (sm3part_op): Ditto.
7533 (CRYPTO_SM3TT): Ditto.
7534 (CRYPTO_SM3PART): Ditto.
7535 (UNSPEC_SM3SS1): New unspec.
7536 (UNSPEC_SM3TT1A): Ditto.
7537 (UNSPEC_SM3TT1B): Ditto.
7538 (UNSPEC_SM3TT2A): Ditto.
7539 (UNSPEC_SM3TT2B): Ditto.
7540 (UNSPEC_SM3PARTW1): Ditto.
7541 (UNSPEC_SM3PARTW2): Ditto.
7542 (UNSPEC_SM4E): Ditto.
7543 (UNSPEC_SM4EKEY): Ditto.
7544 * config/aarch64/constraints.md (Ui2): New constraint.
7545 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7546 * config/arm/types.md (crypto_sm3): New type attribute.
7547 (crypto_sm4): Ditto.
7548 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7549 (vsm3tt1aq_u32): Ditto.
7550 (vsm3tt1bq_u32): Ditto.
7551 (vsm3tt2aq_u32): Ditto.
7552 (vsm3tt2bq_u32): Ditto.
7553 (vsm3partw1q_u32): Ditto.
7554 (vsm3partw2q_u32): Ditto.
7555 (vsm4eq_u32): Ditto.
7556 (vsm4ekeyq_u32): Ditto.
7557 (doc/invoke.texi): Document new sm4 option.
7559 2018-01-10 Michael Collison <michael.collison@arm.com>
7561 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7562 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7563 (AARCH64_FL_FOR_ARCH8_4): New.
7564 (AARCH64_FL_V8_4): New flag.
7565 (doc/invoke.texi): Document new armv8.4-a option.
7567 2018-01-10 Michael Collison <michael.collison@arm.com>
7569 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7570 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7571 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7572 * config/aarch64/aarch64-option-extension.def: Add
7573 AARCH64_OPT_EXTENSION of 'sha2'.
7574 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7575 (crypto): Disable sha2 and aes if crypto disabled.
7576 (crypto): Enable aes and sha2 if enabled.
7577 (simd): Disable sha2 and aes if simd disabled.
7578 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7580 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7581 (TARGET_SHA2): New feature flag for sha2.
7582 (TARGET_AES): New feature flag for aes.
7583 * config/aarch64/aarch64-simd.md:
7584 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7585 conditional on TARGET_AES.
7586 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7587 (aarch64_crypto_sha1hsi): Make pattern conditional
7589 (aarch64_crypto_sha1hv4si): Ditto.
7590 (aarch64_be_crypto_sha1hv4si): Ditto.
7591 (aarch64_crypto_sha1su1v4si): Ditto.
7592 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7593 (aarch64_crypto_sha1su0v4si): Ditto.
7594 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7595 (aarch64_crypto_sha256su0v4si): Ditto.
7596 (aarch64_crypto_sha256su1v4si): Ditto.
7597 (doc/invoke.texi): Document new aes and sha2 options.
7599 2018-01-10 Martin Sebor <msebor@redhat.com>
7601 PR tree-optimization/83781
7602 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7605 2018-01-11 Martin Sebor <msebor@gmail.com>
7606 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7608 PR tree-optimization/83501
7609 PR tree-optimization/81703
7611 * tree-ssa-strlen.c (get_string_cst): Rename...
7612 (get_string_len): ...to this. Handle global constants.
7613 (handle_char_store): Adjust.
7615 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7616 Jim Wilson <jimw@sifive.com>
7618 * config/riscv/riscv-protos.h (riscv_output_return): New.
7619 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7620 (riscv_attribute_table, riscv_output_return),
7621 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7622 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7623 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7625 (riscv_expand_prologue): Add early return for naked function.
7626 (riscv_expand_epilogue): Likewise.
7627 (riscv_function_ok_for_sibcall): Return false for naked function.
7628 (riscv_set_current_function): New.
7629 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7630 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7631 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7632 * doc/extend.texi (RISC-V Function Attributes): New.
7634 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7636 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7637 check for 128-bit long double before checking TCmode.
7638 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7639 128-bit long doubles before checking TFmode or TCmode.
7640 (FLOAT128_IBM_P): Likewise.
7642 2018-01-10 Martin Sebor <msebor@redhat.com>
7644 PR tree-optimization/83671
7645 * builtins.c (c_strlen): Unconditionally return zero for the empty
7647 Use -Warray-bounds for warnings.
7648 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7649 for non-constant array indices with COMPONENT_REF, arrays of
7650 arrays, and pointers to arrays.
7651 (gimple_fold_builtin_strlen): Determine and set length range for
7652 non-constant character arrays.
7654 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7657 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7660 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7662 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7664 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7667 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7668 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7669 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7670 indexed_or_indirect_operand predicate.
7671 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7672 (*vsx_le_perm_load_v8hi): Likewise.
7673 (*vsx_le_perm_load_v16qi): Likewise.
7674 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7675 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7676 (*vsx_le_perm_store_v8hi): Likewise.
7677 (*vsx_le_perm_store_v16qi): Likewise.
7678 (eight unnamed splitters): Likewise.
7680 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7682 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7683 * config/rs6000/emmintrin.h: Likewise.
7684 * config/rs6000/mmintrin.h: Likewise.
7685 * config/rs6000/xmmintrin.h: Likewise.
7687 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7690 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7692 * tree.c (tree_nop_conversion): Return true for location wrapper
7694 (maybe_wrap_with_location): New function.
7695 (selftest::check_strip_nops): New function.
7696 (selftest::test_location_wrappers): New function.
7697 (selftest::tree_c_tests): Call it.
7698 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7699 (maybe_wrap_with_location): New decl.
7700 (EXPR_LOCATION_WRAPPER_P): New macro.
7701 (location_wrapper_p): New inline function.
7702 (tree_strip_any_location_wrapper): New inline function.
7704 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7707 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7708 stack_realign_offset for the largest alignment of stack slot
7710 (ix86_find_max_used_stack_alignment): New function.
7711 (ix86_finalize_stack_frame_flags): Use it. Set
7712 max_used_stack_alignment if we don't realign stack.
7713 * config/i386/i386.h (machine_function): Add
7714 max_used_stack_alignment.
7716 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7718 * config/arm/arm.opt (-mbranch-cost): New option.
7719 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7722 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7725 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7726 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7728 2018-01-10 Richard Biener <rguenther@suse.de>
7731 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7732 early out so it also covers the case where we have a non-NULL
7735 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7737 PR tree-optimization/83753
7738 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7739 for non-strided grouped accesses if the number of elements is 1.
7741 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7744 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7745 * i386.h (TARGET_USE_GATHER): Define.
7746 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7748 2018-01-10 Martin Liska <mliska@suse.cz>
7751 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7752 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7754 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7755 CLEANUP_NO_PARTITIONING is not set.
7757 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7759 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7760 for vectors, as a partial revert of r254296.
7761 * rtl.h (const_vec_p): Delete.
7762 (const_vec_duplicate_p): Don't test for vector CONSTs.
7763 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7764 * expmed.c (make_tree): Likewise.
7767 * common.md (E, F): Use CONSTANT_P instead of checking for
7769 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7770 checking for CONST_VECTOR.
7772 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7775 * predict.c (force_edge_cold): Handle in more sane way edges
7778 2018-01-09 Carl Love <cel@us.ibm.com>
7780 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7782 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7783 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7784 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7785 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7786 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7787 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7788 * config/rs6000/rs6000-protos.h: Add extern defition for
7789 rs6000_generate_float2_double_code.
7790 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7792 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7793 (float2_v2df): Add define_expand.
7795 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7798 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7799 op_mode in the force_to_mode call.
7801 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7803 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7804 instead of checking each element individually.
7805 (aarch64_evpc_uzp): Likewise.
7806 (aarch64_evpc_zip): Likewise.
7807 (aarch64_evpc_ext): Likewise.
7808 (aarch64_evpc_rev): Likewise.
7809 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7810 instead of checking each element individually. Return true without
7812 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7813 whether all selected elements come from the same input, instead of
7814 checking each element individually. Remove calls to gen_rtx_REG,
7815 start_sequence and end_sequence and instead assert that no rtl is
7818 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7820 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7821 order of HIGH and CONST checks.
7823 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7825 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7826 if the destination isn't an SSA_NAME.
7828 2018-01-09 Richard Biener <rguenther@suse.de>
7830 PR tree-optimization/83668
7831 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7833 (canonicalize_loop_form): ... here, renamed from ...
7834 (canonicalize_loop_closed_ssa_form): ... this and amended to
7835 swap successor edges for loop exit blocks to make us use
7836 the RPO order we need for initial schedule generation.
7838 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7840 PR tree-optimization/64811
7841 * match.pd: When optimizing comparisons with Inf, avoid
7842 introducing or losing exceptions from comparisons with NaN.
7844 2018-01-09 Martin Liska <mliska@suse.cz>
7847 * asan.c (shadow_mem_size): Add gcc_assert.
7849 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7851 Don't save registers in main().
7854 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7855 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7856 * config/avr/avr.c (avr_set_current_function): Don't error if
7857 naked, OS_task or OS_main are specified at the same time.
7858 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7860 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7862 * common/config/avr/avr-common.c (avr_option_optimization_table):
7863 Switch on -mmain-is-OS_task for optimizing compilations.
7865 2018-01-09 Richard Biener <rguenther@suse.de>
7867 PR tree-optimization/83572
7868 * graphite.c: Include cfganal.h.
7869 (graphite_transform_loops): Connect infinite loops to exit
7870 and remove fake edges at the end.
7872 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7874 * ipa-inline.c (edge_badness): Revert accidental checkin.
7876 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7879 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7880 symbols; not inline clones.
7882 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7885 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7886 hard registers. Formatting fixes.
7888 PR preprocessor/83722
7889 * gcc.c (try_generate_repro): Pass
7890 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7891 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7894 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7895 Kito Cheng <kito.cheng@gmail.com>
7897 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7898 (riscv_leaf_function_p): Delete.
7899 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7901 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7903 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7905 (do_ifelse): New function.
7906 (do_isel): New function.
7907 (do_sub3): New function.
7908 (do_add3): New function.
7909 (do_load_mask_compare): New function.
7910 (do_overlap_load_compare): New function.
7911 (expand_compare_loop): New function.
7912 (expand_block_compare): Call expand_compare_loop() when appropriate.
7913 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7915 (-mblock-compare-inline-loop-limit): New option.
7917 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7920 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7921 Reverse order of second and third operands in first alternative.
7922 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7923 of first and second elements in UNSPEC_VPERMR vector.
7924 (altivec_expand_vec_perm_le): Likewise.
7926 2018-01-08 Jeff Law <law@redhat.com>
7928 PR rtl-optimizatin/81308
7929 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7930 (process_switch): If group_case_labels makes a change, then set
7932 (pass_convert_switch::execute): If a switch is converted, then
7933 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7935 PR rtl-optimization/81308
7936 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7939 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7941 PR target/83663 - Revert r255946
7942 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7943 generation for cases where splatting a value is not useful.
7944 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7945 across a vec_duplicate and a paradoxical subreg forming a vector
7946 mode to a vec_concat.
7948 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7950 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7951 -march=armv8.3-a variants.
7952 * config/arm/t-multilib: Likewise.
7953 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7955 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7957 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7959 (cceq_ior_compare_complement): Give it a name so I can use it, and
7960 change boolean_or_operator predicate to boolean_operator so it can
7961 be used to generate a crand.
7962 (eqne): New code iterator.
7963 (bd/bd_neg): New code_attrs.
7964 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7965 a single define_insn.
7966 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7967 decrement (bdnzt/bdnzf/bdzt/bdzf).
7968 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7969 with the new names of the branch decrement patterns, and added the
7970 names of the branch decrement conditional patterns.
7972 2018-01-08 Richard Biener <rguenther@suse.de>
7974 PR tree-optimization/83563
7975 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7978 2018-01-08 Richard Biener <rguenther@suse.de>
7981 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7983 2018-01-08 Richard Biener <rguenther@suse.de>
7985 PR tree-optimization/83685
7986 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7987 references to abnormals.
7989 2018-01-08 Richard Biener <rguenther@suse.de>
7992 * dwarf2out.c (output_indirect_strings): Handle empty
7993 skeleton_debug_str_hash.
7994 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7996 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7998 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7999 (emit_store_direct): Likewise.
8000 (arc_trampoline_adjust_address): Likewise.
8001 (arc_asm_trampoline_template): New function.
8002 (arc_initialize_trampoline): Use asm_trampoline_template.
8003 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8004 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8005 * config/arc/arc.md (flush_icache): Delete pattern.
8007 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8009 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8010 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8013 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8016 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8017 by not USED_FOR_TARGET.
8018 (make_pass_resolve_sw_modes): Likewise.
8020 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8022 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8025 2018-01-08 Richard Biener <rguenther@suse.de>
8028 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8030 2018-01-08 Richard Biener <rguenther@suse.de>
8033 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8035 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8038 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8039 basic blocks with a small number of successors.
8040 (convert_control_dep_chain_into_preds): Improve handling of
8042 (dump_predicates): Split apart into...
8043 (dump_pred_chain): ...here...
8044 (dump_pred_info): ...and here.
8045 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8046 (can_chain_union_be_invalidated_p): Improve check for invalidation
8048 (uninit_uses_cannot_happen): Avoid unnecessary if
8049 convert_control_dep_chain_into_preds yielded nothing.
8051 2018-01-06 Martin Sebor <msebor@redhat.com>
8053 PR tree-optimization/83640
8054 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8055 subtracting negative offset from size.
8056 (builtin_access::overlap): Adjust offset bounds of the access to fall
8057 within the size of the object if possible.
8059 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8061 PR rtl-optimization/83699
8062 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8063 extract_bit_field_as_subreg to cases in which the extracted
8064 value is also a vector.
8066 * lra-constraints.c (process_alt_operands): Test for the equivalence
8067 substitutions when detecting a possible reload cycle.
8069 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8072 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8073 by default if flag_selective_schedling{,2}. Formatting fixes.
8075 PR rtl-optimization/83682
8076 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8077 if it has non-VECTOR_MODE element mode.
8078 (vec_duplicate_p): Likewise.
8081 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8082 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8084 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8087 * config/i386/i386-builtin.def
8088 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8089 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8090 Require also OPTION_MASK_ISA_AVX512F in addition to
8091 OPTION_MASK_ISA_GFNI.
8092 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8093 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8094 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8095 to OPTION_MASK_ISA_GFNI.
8096 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8097 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8098 OPTION_MASK_ISA_AVX512BW.
8099 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8100 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8101 addition to OPTION_MASK_ISA_GFNI.
8102 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8103 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8104 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8105 to OPTION_MASK_ISA_GFNI.
8106 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8107 a requirement for all ISAs rather than any of them with a few
8109 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8111 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8112 bitmasks to be enabled with 3 exceptions, instead of requiring any
8113 enabled ISA with lots of exceptions.
8114 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8115 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8116 Change avx512bw in isa attribute to avx512f.
8117 * config/i386/sgxintrin.h: Add license boilerplate.
8118 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8119 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8120 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8121 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8123 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8124 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8125 temporarily sse2 rather than sse if not enabled already.
8128 * config/i386/sse.md (VI248_VLBW): Rename to ...
8129 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8130 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8131 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8132 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8133 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8134 mode iterator instead of VI248_VLBW.
8136 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8138 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8139 (record_modified): Skip clobbers; add debug output.
8140 (param_change_prob): Use sreal frequencies.
8142 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8144 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8145 punt for user-aligned variables.
8147 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8149 * tree-chrec.c (chrec_contains_symbols): Return true for
8152 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8155 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8156 of (x|y) == x for BICS pattern.
8158 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8160 PR tree-optimization/83605
8161 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8162 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8165 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8167 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8168 * config/epiphany/rtems.h: New file.
8170 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8171 Uros Bizjak <ubizjak@gmail.com>
8174 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8175 QIreg_operand instead of register_operand predicate.
8176 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8177 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8178 comments instead of -fmitigate[-_]rop.
8180 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8183 * cgraphunit.c (symbol_table::compile): Switch to text_section
8184 before calling assembly_start debug hook.
8185 * run-rtl-passes.c (run_rtl_passes): Likewise.
8188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8190 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8191 range_int_cst_p rather than !symbolic_range_p before calling
8192 extract_range_from_multiplicative_op_1.
8194 2018-01-04 Jeff Law <law@redhat.com>
8196 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8197 redundant test in assertion.
8199 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8201 * doc/rtl.texi: Document machine_mode wrapper classes.
8203 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8205 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8208 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8210 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8211 the VEC_PERM_EXPR fold to fail.
8213 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8216 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8217 to switched_sections.
8219 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8222 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8225 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8228 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8229 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8231 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8234 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8235 is BLKmode and bitpos not zero or mode change is needed.
8237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8240 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8243 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8246 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8247 instead of MULT rtx. Update all corresponding splitters.
8249 (*ssub<modesuffix>): Ditto.
8251 (*cmp_sadd_di): Update split patterns.
8252 (*cmp_sadd_si): Ditto.
8253 (*cmp_sadd_sidi): Ditto.
8254 (*cmp_ssub_di): Ditto.
8255 (*cmp_ssub_si): Ditto.
8256 (*cmp_ssub_sidi): Ditto.
8257 * config/alpha/predicates.md (const23_operand): New predicate.
8258 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8259 Look for ASHIFT, not MULT inner operand.
8260 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8262 2018-01-04 Martin Liska <mliska@suse.cz>
8264 PR gcov-profile/83669
8265 * gcov.c (output_intermediate_file): Add version to intermediate
8267 * doc/gcov.texi: Document new field 'version' in intermediate
8268 file format. Fix location of '-k' option of gcov command.
8270 2018-01-04 Martin Liska <mliska@suse.cz>
8273 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8275 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8277 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8279 2018-01-03 Martin Sebor <msebor@redhat.com>
8281 PR tree-optimization/83655
8282 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8283 checking calls with invalid arguments.
8285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8287 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8288 (vectorizable_mask_load_store): Delete.
8289 (vectorizable_call): Return false for masked loads and stores.
8290 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8291 instead of gimple_assign_rhs1.
8292 (vectorizable_load): Handle IFN_MASK_LOAD.
8293 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8297 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8299 (vectorizable_mask_load_store): ...here.
8300 (vectorizable_load): ...and here.
8302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8304 * tree-vect-stmts.c (vect_build_all_ones_mask)
8305 (vect_build_zero_merge_argument): New functions, split out from...
8306 (vectorizable_load): ...here.
8308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8310 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8312 (vectorizable_mask_load_store): ...here.
8313 (vectorizable_store): ...and here.
8315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8317 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8319 (vectorizable_mask_load_store): ...here.
8321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8323 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8324 (vect_model_store_cost): Take a vec_load_store_type instead of a
8326 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8327 (vect_model_store_cost): Take a vec_load_store_type instead of a
8329 (vectorizable_mask_load_store): Update accordingly.
8330 (vectorizable_store): Likewise.
8331 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8335 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8336 IFN_MASK_LOAD calls here rather than...
8337 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8340 Alan Hayward <alan.hayward@arm.com>
8341 David Sherwood <david.sherwood@arm.com>
8343 * expmed.c (extract_bit_field_1): For vector extracts,
8344 fall back to extract_bit_field_as_subreg if vec_extract
8347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8348 Alan Hayward <alan.hayward@arm.com>
8349 David Sherwood <david.sherwood@arm.com>
8351 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8352 they are variable or constant sized.
8353 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8354 slots for constant-sized data.
8356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8357 Alan Hayward <alan.hayward@arm.com>
8358 David Sherwood <david.sherwood@arm.com>
8360 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8361 handling COND_EXPRs with boolean comparisons, try to find a better
8362 basis for the mask type than the boolean itself.
8364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8366 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8367 is calculated and how it can be overridden.
8368 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8369 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8371 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8375 Alan Hayward <alan.hayward@arm.com>
8376 David Sherwood <david.sherwood@arm.com>
8378 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8379 Remove the mode argument.
8380 (aarch64_simd_valid_immediate): Remove the mode and inverse
8382 * config/aarch64/iterators.md (bitsize): New iterator.
8383 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8384 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8385 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8386 aarch64_simd_valid_immediate.
8387 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8388 (aarch64_reg_or_bic_imm): Likewise.
8389 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8390 with an insn_type enum and msl with a modifier_type enum.
8391 Replace element_width with a scalar_mode. Change the shift
8392 to unsigned int. Add constructors for scalar_float_mode and
8393 scalar_int_mode elements.
8394 (aarch64_vect_float_const_representable_p): Delete.
8395 (aarch64_can_const_movi_rtx_p)
8396 (aarch64_simd_scalar_immediate_valid_for_move)
8397 (aarch64_simd_make_constant): Update call to
8398 aarch64_simd_valid_immediate.
8399 (aarch64_advsimd_valid_immediate_hs): New function.
8400 (aarch64_advsimd_valid_immediate): Likewise.
8401 (aarch64_simd_valid_immediate): Remove mode and inverse
8402 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8403 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8404 and aarch64_float_const_representable_p on the result.
8405 (aarch64_output_simd_mov_immediate): Remove mode argument.
8406 Update call to aarch64_simd_valid_immediate and use of
8407 simd_immediate_info.
8408 (aarch64_output_scalar_simd_mov_immediate): Update call
8411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8412 Alan Hayward <alan.hayward@arm.com>
8413 David Sherwood <david.sherwood@arm.com>
8415 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8416 (mode_nunits): Likewise CONST_MODE_NUNITS.
8417 * machmode.def (ADJUST_NUNITS): Document.
8418 * genmodes.c (mode_data::need_nunits_adj): New field.
8419 (blank_mode): Update accordingly.
8420 (adj_nunits): New variable.
8421 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8423 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8424 listed in adj_nunits.
8425 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8426 listed in adj_nunits. Don't emit case statements for such modes.
8427 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8428 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8429 nothing if adj_nunits is nonnull.
8430 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8431 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8432 (emit_mode_fbit): Update use of print_maybe_const_decl.
8433 (emit_move_size): Likewise. Treat the array as non-const
8435 (emit_mode_adjustments): Handle adj_nunits.
8437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8439 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8440 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8441 (VECTOR_MODES): Use it.
8442 (make_vector_modes): Take the prefix as an argument.
8444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8445 Alan Hayward <alan.hayward@arm.com>
8446 David Sherwood <david.sherwood@arm.com>
8448 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8449 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8450 for MODE_VECTOR_BOOL.
8451 * machmode.def (VECTOR_BOOL_MODE): Document.
8452 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8453 (make_vector_bool_mode): New function.
8454 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8456 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8457 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8459 * stor-layout.c (int_mode_for_mode): Likewise.
8460 * tree.c (build_vector_type_for_mode): Likewise.
8461 * varasm.c (output_constant_pool_2): Likewise.
8462 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8463 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8464 for MODE_VECTOR_BOOL.
8465 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8466 of mode class checks.
8467 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8468 instead of a list of mode class checks.
8469 (expand_vector_scalar_condition): Likewise.
8470 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8473 Alan Hayward <alan.hayward@arm.com>
8474 David Sherwood <david.sherwood@arm.com>
8476 * machmode.h (mode_size): Change from unsigned short to
8478 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8479 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8480 or if measurement_type is not polynomial.
8481 (fixed_size_mode::includes_p): Check for constant-sized modes.
8482 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8483 return a poly_uint16 rather than an unsigned short.
8484 (emit_mode_size): Change the type of mode_size from unsigned short
8485 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8486 (emit_mode_adjustments): Cope with polynomial vector sizes.
8487 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8489 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8491 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8492 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8493 * caller-save.c (setup_save_areas): Likewise.
8494 (replace_reg_with_saved_mem): Likewise.
8495 * calls.c (emit_library_call_value_1): Likewise.
8496 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8497 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8498 (gen_lowpart_for_combine): Likewise.
8499 * convert.c (convert_to_integer_1): Likewise.
8500 * cse.c (equiv_constant, cse_insn): Likewise.
8501 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8502 (cselib_subst_to_values): Likewise.
8503 * dce.c (word_dce_process_block): Likewise.
8504 * df-problems.c (df_word_lr_mark_ref): Likewise.
8505 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8506 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8507 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8508 (rtl_for_decl_location): Likewise.
8509 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8510 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8511 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8512 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8513 (expand_expr_real_1): Likewise.
8514 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8515 (pad_below): Likewise.
8516 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8517 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8518 * ira.c (get_subreg_tracking_sizes): Likewise.
8519 * ira-build.c (ira_create_allocno_objects): Likewise.
8520 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8521 (ira_sort_regnos_for_alter_reg): Likewise.
8522 * ira-costs.c (record_operand_costs): Likewise.
8523 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8524 (resolve_simple_move): Likewise.
8525 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8526 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8527 (lra_constraints): Likewise.
8528 (CONST_POOL_OK_P): Reject variable-sized modes.
8529 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8530 (add_pseudo_to_slot, lra_spill): Likewise.
8531 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8532 * optabs-query.c (get_best_extraction_insn): Likewise.
8533 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8534 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8535 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8536 * recog.c (offsettable_address_addr_space_p): Likewise.
8537 * regcprop.c (maybe_mode_change): Likewise.
8538 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8539 * regrename.c (build_def_use): Likewise.
8540 * regstat.c (dump_reg_info): Likewise.
8541 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8542 (find_reloads, find_reloads_subreg_address): Likewise.
8543 * reload1.c (eliminate_regs_1): Likewise.
8544 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8545 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8546 (simplify_binary_operation_1, simplify_subreg): Likewise.
8547 * targhooks.c (default_function_arg_padding): Likewise.
8548 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8549 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8550 (verify_gimple_assign_ternary): Likewise.
8551 * tree-inline.c (estimate_move_cost): Likewise.
8552 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8553 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8554 (get_address_cost_ainc): Likewise.
8555 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8556 (vect_supportable_dr_alignment): Likewise.
8557 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8558 (vectorizable_reduction): Likewise.
8559 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8560 (vectorizable_operation, vectorizable_load): Likewise.
8561 * tree.c (build_same_sized_truth_vector_type): Likewise.
8562 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8563 * var-tracking.c (emit_note_insn_var_location): Likewise.
8564 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8565 (ADDR_VEC_ALIGN): Likewise.
8567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8568 Alan Hayward <alan.hayward@arm.com>
8569 David Sherwood <david.sherwood@arm.com>
8571 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8573 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8574 or if measurement_type is polynomial.
8575 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8576 * combine.c (make_extraction): Likewise.
8577 * dse.c (find_shift_sequence): Likewise.
8578 * dwarf2out.c (mem_loc_descriptor): Likewise.
8579 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8580 (extract_bit_field, extract_low_bits): Likewise.
8581 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8582 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8583 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8584 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8585 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8586 * reload.c (find_reloads): Likewise.
8587 * reload1.c (alter_reg): Likewise.
8588 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8589 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8590 * tree-if-conv.c (predicate_mem_writes): Likewise.
8591 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8592 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8593 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8594 * valtrack.c (dead_debug_insert_temp): Likewise.
8595 * varasm.c (mergeable_constant_section): Likewise.
8596 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8599 Alan Hayward <alan.hayward@arm.com>
8600 David Sherwood <david.sherwood@arm.com>
8602 * expr.c (expand_assignment): Cope with polynomial mode sizes
8603 when assigning to a CONCAT.
8605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8606 Alan Hayward <alan.hayward@arm.com>
8607 David Sherwood <david.sherwood@arm.com>
8609 * machmode.h (mode_precision): Change from unsigned short to
8611 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8613 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8614 or if measurement_type is not polynomial.
8615 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8616 in which the mode is already known to be a scalar_int_mode.
8617 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8618 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8620 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8621 for GET_MODE_PRECISION.
8622 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8623 for GET_MODE_PRECISION.
8624 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8626 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8627 (expand_field_assignment, make_extraction): Likewise.
8628 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8629 (get_last_value): Likewise.
8630 * convert.c (convert_to_integer_1): Likewise.
8631 * cse.c (cse_insn): Likewise.
8632 * expr.c (expand_expr_real_1): Likewise.
8633 * lra-constraints.c (simplify_operand_subreg): Likewise.
8634 * optabs-query.c (can_atomic_load_p): Likewise.
8635 * optabs.c (expand_atomic_load): Likewise.
8636 (expand_atomic_store): Likewise.
8637 * ree.c (combine_reaching_defs): Likewise.
8638 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8639 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8640 * tree.h (type_has_mode_precision_p): Likewise.
8641 * ubsan.c (instrument_si_overflow): Likewise.
8643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8644 Alan Hayward <alan.hayward@arm.com>
8645 David Sherwood <david.sherwood@arm.com>
8647 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8648 polynomial numbers of units.
8649 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8650 (valid_vector_subparts_p): New function.
8651 (build_vector_type): Remove temporary shim and take the number
8652 of units as a poly_uint64 rather than an int.
8653 (build_opaque_vector_type): Take the number of units as a
8654 poly_uint64 rather than an int.
8655 * tree.c (build_vector_from_ctor): Handle polynomial
8656 TYPE_VECTOR_SUBPARTS.
8657 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8658 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8659 (build_vector_from_val): If the number of units is variable,
8660 use build_vec_duplicate_cst for constant operands and
8661 VEC_DUPLICATE_EXPR otherwise.
8662 (make_vector_type): Remove temporary is_constant ().
8663 (build_vector_type, build_opaque_vector_type): Take the number of
8664 units as a poly_uint64 rather than an int.
8665 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8667 * cfgexpand.c (expand_debug_expr): Likewise.
8668 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8669 (store_constructor, expand_expr_real_1): Likewise.
8670 (const_scalar_mask_from_tree): Likewise.
8671 * fold-const-call.c (fold_const_reduction): Likewise.
8672 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8673 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8674 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8675 (fold_relational_const): Likewise.
8676 (native_interpret_vector): Likewise. Change the size from an
8677 int to an unsigned int.
8678 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8679 TYPE_VECTOR_SUBPARTS.
8680 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8681 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8682 duplicating a non-constant operand into a variable-length vector.
8683 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8684 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8685 * ipa-icf.c (sem_variable::equals): Likewise.
8686 * match.pd: Likewise.
8687 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8688 * print-tree.c (print_node): Likewise.
8689 * stor-layout.c (layout_type): Likewise.
8690 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8691 * tree-cfg.c (verify_gimple_comparison): Likewise.
8692 (verify_gimple_assign_binary): Likewise.
8693 (verify_gimple_assign_ternary): Likewise.
8694 (verify_gimple_assign_single): Likewise.
8695 * tree-pretty-print.c (dump_generic_node): Likewise.
8696 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8697 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8698 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8699 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8700 (vect_shift_permute_load_chain): Likewise.
8701 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8702 (expand_vector_condition, optimize_vector_constructor): Likewise.
8703 (lower_vec_perm, get_compute_type): Likewise.
8704 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8705 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8706 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8707 (vect_recog_mask_conversion_pattern): Likewise.
8708 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8709 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8710 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8711 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8712 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8713 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8714 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8715 (supportable_widening_operation): Likewise.
8716 (supportable_narrowing_operation): Likewise.
8717 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8719 * varasm.c (output_constant): Likewise.
8721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8722 Alan Hayward <alan.hayward@arm.com>
8723 David Sherwood <david.sherwood@arm.com>
8725 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8726 so that both the length == 3 and length != 3 cases set up their
8727 own permute vectors. Add comments explaining why we know the
8728 number of elements is constant.
8729 (vect_permute_load_chain): Likewise.
8731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8732 Alan Hayward <alan.hayward@arm.com>
8733 David Sherwood <david.sherwood@arm.com>
8735 * machmode.h (mode_nunits): Change from unsigned char to
8737 (ONLY_FIXED_SIZE_MODES): New macro.
8738 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8739 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8740 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8742 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8743 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8744 or if measurement_type is not polynomial.
8745 * genmodes.c (ZERO_COEFFS): New macro.
8746 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8748 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8749 Use ZERO_COEFFS when emitting initializers.
8750 * data-streamer.h (bp_pack_poly_value): New function.
8751 (bp_unpack_poly_value): Likewise.
8752 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8753 for GET_MODE_NUNITS.
8754 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8755 for GET_MODE_NUNITS.
8756 * tree.c (make_vector_type): Remove temporary shim and make
8757 the real function take the number of units as a poly_uint64
8759 (build_vector_type_for_mode): Handle polynomial nunits.
8760 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8761 * emit-rtl.c (const_vec_series_p_1): Likewise.
8762 (gen_rtx_CONST_VECTOR): Likewise.
8763 * fold-const.c (test_vec_duplicate_folding): Likewise.
8764 * genrecog.c (validate_pattern): Likewise.
8765 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8766 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8767 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8768 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8769 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8770 * rtlanal.c (subreg_get_info): Likewise.
8771 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8772 (vect_grouped_load_supported): Likewise.
8773 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8774 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8775 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8776 (simplify_const_unary_operation, simplify_binary_operation_1)
8777 (simplify_const_binary_operation, simplify_ternary_operation)
8778 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8779 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8780 instead of CONST_VECTOR_NUNITS.
8781 * varasm.c (output_constant_pool_2): Likewise.
8782 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8783 explicit-encoded elements in the XVEC for variable-length vectors.
8785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8787 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8790 Alan Hayward <alan.hayward@arm.com>
8791 David Sherwood <david.sherwood@arm.com>
8793 * coretypes.h (fixed_size_mode): Declare.
8794 (fixed_size_mode_pod): New typedef.
8795 * builtins.h (target_builtins::x_apply_args_mode)
8796 (target_builtins::x_apply_result_mode): Change type to
8797 fixed_size_mode_pod.
8798 * builtins.c (apply_args_size, apply_result_size, result_vector)
8799 (expand_builtin_apply_args_1, expand_builtin_apply)
8800 (expand_builtin_return): Update accordingly.
8802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8804 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8805 * cselib.c (cselib_hash_rtx): Likewise.
8806 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8807 CONST_VECTOR encoding.
8809 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8810 Jeff Law <law@redhat.com>
8813 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8814 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8815 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8816 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8819 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8820 explicitly probe *sp in a noreturn function if there were any callee
8821 register saves or frame pointer is needed.
8823 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8826 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8827 BLKmode for ternary, binary or unary expressions.
8830 * var-tracking.c (delete_vta_debug_insn): New inline function.
8831 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8832 insns from get_insns () to NULL instead of each bb separately.
8833 Use delete_vta_debug_insn. No longer static.
8834 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8835 delete_vta_debug_insns callers.
8836 * rtl.h (delete_vta_debug_insns): Declare.
8837 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8838 instead of variable_tracking_main.
8840 2018-01-03 Martin Sebor <msebor@redhat.com>
8842 PR tree-optimization/83603
8843 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8844 arguments past the endof the argument list in functions declared
8845 without a prototype.
8846 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8847 Avoid checking when arguments are null.
8849 2018-01-03 Martin Sebor <msebor@redhat.com>
8852 * doc/extend.texi (attribute const): Fix a typo.
8853 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8854 issuing -Wsuggest-attribute for void functions.
8856 2018-01-03 Martin Sebor <msebor@redhat.com>
8858 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8859 offset_int::from instead of wide_int::to_shwi.
8860 (maybe_diag_overlap): Remove assertion.
8861 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8862 * gimple-ssa-sprintf.c (format_directive): Same.
8863 (parse_directive): Same.
8864 (sprintf_dom_walker::compute_format_length): Same.
8865 (try_substitute_return_value): Same.
8867 2018-01-03 Jeff Law <law@redhat.com>
8870 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8871 non-constant residual for zero at runtime and avoid probing in
8872 that case. Reorganize code for trailing problem to mirror handling
8875 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8877 PR tree-optimization/83501
8878 * tree-ssa-strlen.c (get_string_cst): New.
8879 (handle_char_store): Call get_string_cst.
8881 2018-01-03 Martin Liska <mliska@suse.cz>
8883 PR tree-optimization/83593
8884 * tree-ssa-strlen.c: Include tree-cfg.h.
8885 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8886 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8887 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8889 (strlen_dom_walker::before_dom_children): Call
8890 gimple_purge_dead_eh_edges. Dump tranformation with details
8892 (strlen_dom_walker::before_dom_children): Update call by adding
8893 new argument cleanup_eh.
8894 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8896 2018-01-03 Martin Liska <mliska@suse.cz>
8899 * cif-code.def (VARIADIC_THUNK): New enum value.
8900 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8903 2018-01-03 Jan Beulich <jbeulich@suse.com>
8905 * sse.md (mov<mode>_internal): Tighten condition for when to use
8906 vmovdqu<ssescalarsize> for TI and OI modes.
8908 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8910 Update copyright years.
8912 2018-01-03 Martin Liska <mliska@suse.cz>
8915 * ipa-visibility.c (function_and_variable_visibility): Skip
8916 functions with noipa attribure.
8918 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8920 * gcc.c (process_command): Update copyright notice dates.
8921 * gcov-dump.c (print_version): Ditto.
8922 * gcov.c (print_version): Ditto.
8923 * gcov-tool.c (print_version): Ditto.
8924 * gengtype.c (create_file): Ditto.
8925 * doc/cpp.texi: Bump @copying's copyright year.
8926 * doc/cppinternals.texi: Ditto.
8927 * doc/gcc.texi: Ditto.
8928 * doc/gccint.texi: Ditto.
8929 * doc/gcov.texi: Ditto.
8930 * doc/install.texi: Ditto.
8931 * doc/invoke.texi: Ditto.
8933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8935 * vector-builder.h (vector_builder::m_full_nelts): Change from
8936 unsigned int to poly_uint64.
8937 (vector_builder::full_nelts): Update prototype accordingly.
8938 (vector_builder::new_vector): Likewise.
8939 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8940 (vector_builder::operator ==): Likewise.
8941 (vector_builder::finalize): Likewise.
8942 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8943 Take the number of elements as a poly_uint64 rather than an
8945 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8946 from unsigned int to poly_uint64.
8947 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8948 (vec_perm_indices::new_vector): Likewise.
8949 (vec_perm_indices::length): Likewise.
8950 (vec_perm_indices::nelts_per_input): Likewise.
8951 (vec_perm_indices::input_nelts): Likewise.
8952 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8953 number of elements per input as a poly_uint64 rather than an
8954 unsigned int. Use the original encoding for variable-length
8955 vectors, rather than clamping each individual element.
8956 For the second and subsequent elements in each pattern,
8957 clamp the step and base before clamping their sum.
8958 (vec_perm_indices::series_p): Handle polynomial element counts.
8959 (vec_perm_indices::all_in_range_p): Likewise.
8960 (vec_perm_indices_to_tree): Likewise.
8961 (vec_perm_indices_to_rtx): Likewise.
8962 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8963 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8964 (tree_vector_builder::new_binary_operation): Handle polynomial
8965 element counts. Return false if we need to know the number
8966 of elements at compile time.
8967 * fold-const.c (fold_vec_perm): Punt if the number of elements
8968 isn't known at compile time.
8970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8972 * vec-perm-indices.h (vec_perm_builder): Change element type
8973 from HOST_WIDE_INT to poly_int64.
8974 (vec_perm_indices::element_type): Update accordingly.
8975 (vec_perm_indices::clamp): Handle polynomial element_types.
8976 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8977 (vec_perm_indices::all_in_range_p): Likewise.
8978 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8980 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8981 polynomial vec_perm_indices element types.
8982 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8983 * fold-const.c (fold_vec_perm): Likewise.
8984 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8985 * tree-vect-generic.c (lower_vec_perm): Likewise.
8986 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8987 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8988 element type to HOST_WIDE_INT.
8990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8991 Alan Hayward <alan.hayward@arm.com>
8992 David Sherwood <david.sherwood@arm.com>
8994 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8995 rather than an int. Use plus_constant.
8996 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8997 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9000 Alan Hayward <alan.hayward@arm.com>
9001 David Sherwood <david.sherwood@arm.com>
9003 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9004 a HOST_WIDE_INT to a poly_int64.
9006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9007 Alan Hayward <alan.hayward@arm.com>
9008 David Sherwood <david.sherwood@arm.com>
9010 * calls.c (load_register_parameters): Cope with polynomial
9011 mode sizes. Require a constant size for BLKmode parameters
9012 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9013 forces a parameter to be padded at the lsb end in order to
9014 fill a complete number of words, require the parameter size
9015 to be ordered wrt UNITS_PER_WORD.
9017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9018 Alan Hayward <alan.hayward@arm.com>
9019 David Sherwood <david.sherwood@arm.com>
9021 * reload1.c (spill_stack_slot_width): Change element type
9022 from unsigned int to poly_uint64_pod.
9023 (alter_reg): Treat mode sizes as polynomial.
9025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9026 Alan Hayward <alan.hayward@arm.com>
9027 David Sherwood <david.sherwood@arm.com>
9029 * reload.c (complex_word_subreg_p): New function.
9030 (reload_inner_reg_of_subreg, push_reload): Use it.
9032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9033 Alan Hayward <alan.hayward@arm.com>
9034 David Sherwood <david.sherwood@arm.com>
9036 * lra-constraints.c (process_alt_operands): Reject matched
9037 operands whose sizes aren't ordered.
9038 (match_reload): Refer to this check here.
9040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9041 Alan Hayward <alan.hayward@arm.com>
9042 David Sherwood <david.sherwood@arm.com>
9044 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9045 that the mode size is in the set {1, 2, 4, 8, 16}.
9047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9048 Alan Hayward <alan.hayward@arm.com>
9049 David Sherwood <david.sherwood@arm.com>
9051 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9052 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9059 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9060 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9061 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9062 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9063 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9064 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9065 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9066 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9067 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9069 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9070 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9071 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9072 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9073 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9074 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9075 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9076 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9077 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9078 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9080 * expr.c (emit_move_resolve_push): Treat the input and result
9081 of PUSH_ROUNDING as a poly_int64.
9082 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9083 (emit_push_insn): Likewise.
9084 * lra-eliminations.c (mark_not_eliminable): Likewise.
9085 * recog.c (push_operand): Likewise.
9086 * reload1.c (elimination_effects): Likewise.
9087 * rtlanal.c (nonzero_bits1): Likewise.
9088 * calls.c (store_one_arg): Likewise. Require the padding to be
9089 known at compile time.
9091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9092 Alan Hayward <alan.hayward@arm.com>
9093 David Sherwood <david.sherwood@arm.com>
9095 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9096 Use plus_constant instead of gen_rtx_PLUS.
9098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9099 Alan Hayward <alan.hayward@arm.com>
9100 David Sherwood <david.sherwood@arm.com>
9102 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9106 Alan Hayward <alan.hayward@arm.com>
9107 David Sherwood <david.sherwood@arm.com>
9109 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9110 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9111 via stack temporaries. Treat the mode size as polynomial too.
9113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9114 Alan Hayward <alan.hayward@arm.com>
9115 David Sherwood <david.sherwood@arm.com>
9117 * expr.c (expand_expr_real_2): When handling conversions involving
9118 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9119 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9120 as a poly_uint64 too.
9122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9123 Alan Hayward <alan.hayward@arm.com>
9124 David Sherwood <david.sherwood@arm.com>
9126 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9129 Alan Hayward <alan.hayward@arm.com>
9130 David Sherwood <david.sherwood@arm.com>
9132 * combine.c (can_change_dest_mode): Handle polynomial
9133 REGMODE_NATURAL_SIZE.
9134 * expmed.c (store_bit_field_1): Likewise.
9135 * expr.c (store_constructor): Likewise.
9136 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9137 and polynomial REGMODE_NATURAL_SIZE.
9138 (gen_lowpart_common): Likewise.
9139 * reginfo.c (record_subregs_of_mode): Likewise.
9140 * rtlanal.c (read_modify_subreg_p): Likewise.
9142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9143 Alan Hayward <alan.hayward@arm.com>
9144 David Sherwood <david.sherwood@arm.com>
9146 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9147 numbers of elements.
9149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9150 Alan Hayward <alan.hayward@arm.com>
9151 David Sherwood <david.sherwood@arm.com>
9153 * match.pd: Cope with polynomial numbers of vector elements.
9155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9156 Alan Hayward <alan.hayward@arm.com>
9157 David Sherwood <david.sherwood@arm.com>
9159 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9160 in a POINTER_PLUS_EXPR.
9162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9163 Alan Hayward <alan.hayward@arm.com>
9164 David Sherwood <david.sherwood@arm.com>
9166 * omp-simd-clone.c (simd_clone_subparts): New function.
9167 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9168 (ipa_simd_modify_function_body): Likewise.
9170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9171 Alan Hayward <alan.hayward@arm.com>
9172 David Sherwood <david.sherwood@arm.com>
9174 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9175 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9176 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9177 (expand_vector_condition, vector_element): Likewise.
9178 (subparts_gt): New function.
9179 (get_compute_type): Use subparts_gt.
9180 (count_type_subparts): Delete.
9181 (expand_vector_operations_1): Use subparts_gt instead of
9182 count_type_subparts.
9184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9185 Alan Hayward <alan.hayward@arm.com>
9186 David Sherwood <david.sherwood@arm.com>
9188 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9189 (vect_compile_time_alias): ...this new function. Do the calculation
9190 on poly_ints rather than trees.
9191 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9199 (vect_schedule_slp_instance): Likewise.
9201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9202 Alan Hayward <alan.hayward@arm.com>
9203 David Sherwood <david.sherwood@arm.com>
9205 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9206 constant and extern definitions for variable-length vectors.
9207 (vect_get_constant_vectors): Note that the number of units
9208 is known to be constant.
9210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9211 Alan Hayward <alan.hayward@arm.com>
9212 David Sherwood <david.sherwood@arm.com>
9214 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9215 of units as polynomial. Choose between WIDE and NARROW based
9218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9219 Alan Hayward <alan.hayward@arm.com>
9220 David Sherwood <david.sherwood@arm.com>
9222 * tree-vect-stmts.c (simd_clone_subparts): New function.
9223 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9226 Alan Hayward <alan.hayward@arm.com>
9227 David Sherwood <david.sherwood@arm.com>
9229 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9230 vectors as polynomial. Use build_index_vector for
9233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9234 Alan Hayward <alan.hayward@arm.com>
9235 David Sherwood <david.sherwood@arm.com>
9237 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9238 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9239 for variable-length vectors.
9240 (vectorizable_mask_load_store): Treat the number of units as
9241 polynomial, asserting that it is constant if the condition has
9242 already been enforced.
9243 (vectorizable_store, vectorizable_load): Likewise.
9245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9246 Alan Hayward <alan.hayward@arm.com>
9247 David Sherwood <david.sherwood@arm.com>
9249 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9250 of units as polynomial. Punt if we can't tell at compile time
9251 which vector contains the final result.
9253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9254 Alan Hayward <alan.hayward@arm.com>
9255 David Sherwood <david.sherwood@arm.com>
9257 * tree-vect-loop.c (vectorizable_induction): Treat the number
9258 of units as polynomial. Punt on SLP inductions. Use an integer
9259 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9260 cast of such a series for variable-length floating-point
9263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9264 Alan Hayward <alan.hayward@arm.com>
9265 David Sherwood <david.sherwood@arm.com>
9267 * tree.h (build_index_vector): Declare.
9268 * tree.c (build_index_vector): New function.
9269 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9270 of units as polynomial, forcibly converting it to a constant if
9271 vectorizable_reduction has already enforced the condition.
9272 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9273 to create a {1,2,3,...} vector.
9274 (vectorizable_reduction): Treat the number of units as polynomial.
9275 Choose vectype_in based on the largest scalar element size rather
9276 than the smallest number of units. Enforce the restrictions
9279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9280 Alan Hayward <alan.hayward@arm.com>
9281 David Sherwood <david.sherwood@arm.com>
9283 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9284 number of units as polynomial.
9286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9287 Alan Hayward <alan.hayward@arm.com>
9288 David Sherwood <david.sherwood@arm.com>
9290 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9291 * target.def (autovectorize_vector_sizes): Return the vector sizes
9292 by pointer, using vector_sizes rather than a bitmask.
9293 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9294 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9295 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9297 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9298 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9299 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9300 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9301 * omp-general.c (omp_max_vf): Likewise.
9302 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9303 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9304 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9305 * tree-vect-slp.c (vect_slp_bb): Likewise.
9306 * doc/tm.texi: Regenerate.
9307 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9309 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9310 the vector size as a poly_uint64 rather than an unsigned int.
9311 (current_vector_size): Change from an unsigned int to a poly_uint64.
9312 (get_vectype_for_scalar_type): Update accordingly.
9313 * tree.h (build_truth_vector_type): Take the size and number of
9314 units as a poly_uint64 rather than an unsigned int.
9315 (build_vector_type): Add a temporary overload that takes
9316 the number of units as a poly_uint64 rather than an unsigned int.
9317 * tree.c (make_vector_type): Likewise.
9318 (build_truth_vector_type): Take the number of units as a poly_uint64
9319 rather than an unsigned int.
9321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9322 Alan Hayward <alan.hayward@arm.com>
9323 David Sherwood <david.sherwood@arm.com>
9325 * target.def (get_mask_mode): Take the number of units and length
9326 as poly_uint64s rather than unsigned ints.
9327 * targhooks.h (default_get_mask_mode): Update accordingly.
9328 * targhooks.c (default_get_mask_mode): Likewise.
9329 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9330 * doc/tm.texi: Regenerate.
9332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9333 Alan Hayward <alan.hayward@arm.com>
9334 David Sherwood <david.sherwood@arm.com>
9336 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9337 * omp-general.c (omp_max_vf): Likewise.
9338 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9339 (expand_omp_simd): Handle polynomial safelen.
9340 * omp-low.c (omplow_simd_context): Add a default constructor.
9341 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9342 (lower_rec_simd_input_clauses): Update accordingly.
9343 (lower_rec_input_clauses): Likewise.
9345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9346 Alan Hayward <alan.hayward@arm.com>
9347 David Sherwood <david.sherwood@arm.com>
9349 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9350 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9351 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9352 (vect_analyze_slp_cost): Likewise.
9353 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9354 (vect_model_load_cost): Likewise.
9356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9357 Alan Hayward <alan.hayward@arm.com>
9358 David Sherwood <david.sherwood@arm.com>
9360 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9361 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9362 from an unsigned int * to a poly_uint64_pod *.
9363 (calculate_unrolling_factor): New function.
9364 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9367 Alan Hayward <alan.hayward@arm.com>
9368 David Sherwood <david.sherwood@arm.com>
9370 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9371 from an unsigned int to a poly_uint64.
9372 (_loop_vec_info::slp_unrolling_factor): Likewise.
9373 (_loop_vec_info::vectorization_factor): Change from an int
9375 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9376 (vect_get_num_vectors): New function.
9377 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9378 (vect_get_num_copies): Use vect_get_num_vectors.
9379 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9380 to an unsigned int *.
9381 (vect_analyze_data_refs): Change min_vf from an int * to a
9383 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9384 than an unsigned HOST_WIDE_INT.
9385 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9386 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9387 to an unsigned int *.
9388 (vect_analyze_data_ref_dependences): Likewise.
9389 (vect_compute_data_ref_alignment): Handle polynomial vf.
9390 (vect_enhance_data_refs_alignment): Likewise.
9391 (vect_prune_runtime_alias_test_list): Likewise.
9392 (vect_shift_permute_load_chain): Likewise.
9393 (vect_supportable_dr_alignment): Likewise.
9394 (dependence_distance_ge_vf): Take the vectorization factor as a
9395 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9396 (vect_analyze_data_refs): Change min_vf from an int * to a
9398 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9399 vfm1 as a poly_uint64 rather than an int. Make the same change
9400 for the returned bound_scalar.
9401 (vect_gen_vector_loop_niters): Handle polynomial vf.
9402 (vect_do_peeling): Likewise. Update call to
9403 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9404 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9406 * tree-vect-loop.c (vect_determine_vectorization_factor)
9407 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9408 (vect_get_known_peeling_cost): Likewise.
9409 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9410 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9411 (vect_transform_loop): Likewise. Use the lowest possible VF when
9412 updating the upper bounds of the loop.
9413 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9415 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9416 polynomial unroll factors.
9417 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9418 (vect_make_slp_decision): Likewise.
9419 (vect_supported_load_permutation_p): Likewise, and polynomial
9421 (vect_analyze_slp_cost): Handle polynomial vf.
9422 (vect_slp_analyze_node_operations): Likewise.
9423 (vect_slp_analyze_bb_1): Likewise.
9424 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9425 than an unsigned HOST_WIDE_INT.
9426 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9427 (vectorizable_load): Handle polynomial vf.
9428 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9430 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9433 Alan Hayward <alan.hayward@arm.com>
9434 David Sherwood <david.sherwood@arm.com>
9436 * match.pd: Handle bit operations involving three constants
9437 and try to fold one pair.
9439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9441 * tree-vect-loop-manip.c: Include gimple-fold.h.
9442 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9443 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9444 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9445 Add a path that uses a step of VF instead of 1, but disable it
9447 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9448 and niters_no_overflow parameters. Update calls to
9449 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9450 Create a new SSA name if the latter choses to use a ste other
9451 than zero, and return it via niters_vector_mult_vf_var.
9452 * tree-vect-loop.c (vect_transform_loop): Update calls to
9453 vect_do_peeling, vect_gen_vector_loop_niters and
9454 slpeel_make_loop_iterate_ntimes.
9455 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9456 (vect_gen_vector_loop_niters): Update declarations after above changes.
9458 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9460 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9461 128-bit round to integer instructions.
9462 (ceil<mode>2): Likewise.
9463 (btrunc<mode>2): Likewise.
9464 (round<mode>2): Likewise.
9466 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9468 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9469 unaligned VSX load/store on P8/P9.
9470 (expand_block_clear): Allow the use of unaligned VSX
9471 load/store on P8/P9.
9473 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9475 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9477 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9478 swap associated with both a load and a store.
9480 2018-01-02 Andrew Waterman <andrew@sifive.com>
9482 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9483 * config/riscv/riscv.md (clear_cache): Use it.
9485 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9487 * web.c: Remove out-of-date comment.
9489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9491 * expr.c (fixup_args_size_notes): Check that any existing
9492 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9493 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9494 (emit_single_push_insn): ...here.
9496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9498 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9499 (const_vector_encoded_nelts): New function.
9500 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9501 (const_vector_int_elt, const_vector_elt): Declare.
9502 * emit-rtl.c (const_vector_int_elt_1): New function.
9503 (const_vector_elt): Likewise.
9504 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9505 of CONST_VECTOR_ELT.
9507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9509 * expr.c: Include rtx-vector-builder.h.
9510 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9511 directly on the tree encoding.
9512 (const_vector_from_tree): Likewise.
9513 * optabs.c: Include rtx-vector-builder.h.
9514 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9515 sequence of "u" values.
9516 * vec-perm-indices.c: Include rtx-vector-builder.h.
9517 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9518 directly on the vec_perm_indices encoding.
9520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9522 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9523 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9524 * rtx-vector-builder.h: New file.
9525 * rtx-vector-builder.c: Likewise.
9526 * rtl.h (rtx_def::u2): Add a const_vector field.
9527 (CONST_VECTOR_NPATTERNS): New macro.
9528 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9529 (CONST_VECTOR_DUPLICATE_P): Likewise.
9530 (CONST_VECTOR_STEPPED_P): Likewise.
9531 (CONST_VECTOR_ENCODED_ELT): Likewise.
9532 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9533 (unwrap_const_vec_duplicate): Likewise.
9534 (const_vec_series_p): Check for a non-duplicated vector encoding.
9535 Say that the function only returns true for integer vectors.
9536 * emit-rtl.c: Include rtx-vector-builder.h.
9537 (gen_const_vec_duplicate_1): Delete.
9538 (gen_const_vector): Call gen_const_vec_duplicate instead of
9539 gen_const_vec_duplicate_1.
9540 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9541 (gen_const_vec_duplicate): Use rtx_vector_builder.
9542 (gen_const_vec_series): Likewise.
9543 (gen_rtx_CONST_VECTOR): Likewise.
9544 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9545 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9546 Build a new vector rather than modifying a CONST_VECTOR in-place.
9547 (handle_special_swappables): Update call accordingly.
9548 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9549 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9550 Build a new vector rather than modifying a CONST_VECTOR in-place.
9551 (handle_special_swappables): Update call accordingly.
9553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9555 * simplify-rtx.c (simplify_const_binary_operation): Use
9556 CONST_VECTOR_ELT instead of XVECEXP.
9558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9560 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9561 the selector elements to be different from the data elements
9562 if the selector is a VECTOR_CST.
9563 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9564 ssizetype for the selector.
9566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9568 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9569 before testing each element individually.
9570 * tree-vect-generic.c (lower_vec_perm): Likewise.
9572 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9574 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9575 * selftest-run-tests.c (selftest::run_tests): Call it.
9576 * vector-builder.h (vector_builder::operator ==): New function.
9577 (vector_builder::operator !=): Likewise.
9578 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9579 (vec_perm_indices::all_from_input_p): New function.
9580 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9581 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9582 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9583 instead of reading the VECTOR_CST directly. Detect whether both
9584 vector inputs are the same before constructing the vec_perm_indices,
9585 and update the number of inputs argument accordingly. Use the
9586 utility functions added above. Only construct sel2 if we need to.
9588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9590 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9591 the broadcast of the low byte.
9592 (expand_mult_highpart): Use an explicit encoding for the permutes.
9593 * optabs-query.c (can_mult_highpart_p): Likewise.
9594 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9595 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9596 (vectorizable_bswap): Likewise.
9597 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9598 explicit encoding for the power-of-2 permutes.
9599 (vect_permute_store_chain): Likewise.
9600 (vect_grouped_load_supported): Likewise.
9601 (vect_permute_load_chain): Likewise.
9603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9605 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9606 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9607 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9608 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9609 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9610 (vect_gen_perm_mask_any): Likewise.
9612 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9614 * int-vector-builder.h: New file.
9615 * vec-perm-indices.h: Include int-vector-builder.h.
9616 (vec_perm_indices): Redefine as an int_vector_builder.
9617 (auto_vec_perm_indices): Delete.
9618 (vec_perm_builder): Redefine as a stand-alone class.
9619 (vec_perm_indices::vec_perm_indices): New function.
9620 (vec_perm_indices::clamp): Likewise.
9621 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9622 (vec_perm_indices::new_vector): New function.
9623 (vec_perm_indices::new_expanded_vector): Update for new
9624 vec_perm_indices class.
9625 (vec_perm_indices::rotate_inputs): New function.
9626 (vec_perm_indices::all_in_range_p): Operate directly on the
9627 encoded form, without computing elided elements.
9628 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9629 encoding. Update for new vec_perm_indices class.
9630 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9631 the given vec_perm_builder.
9632 (expand_vec_perm_var): Update vec_perm_builder constructor.
9633 (expand_mult_highpart): Use vec_perm_builder instead of
9634 auto_vec_perm_indices.
9635 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9636 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9637 or double series encoding as appropriate.
9638 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9639 vec_perm_indices instead of auto_vec_perm_indices.
9640 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9641 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9642 (vect_permute_store_chain): Likewise.
9643 (vect_grouped_load_supported): Likewise.
9644 (vect_permute_load_chain): Likewise.
9645 (vect_shift_permute_load_chain): Likewise.
9646 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9647 (vect_transform_slp_perm_load): Likewise.
9648 (vect_schedule_slp_instance): Likewise.
9649 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9650 (vectorizable_mask_load_store): Likewise.
9651 (vectorizable_bswap): Likewise.
9652 (vectorizable_store): Likewise.
9653 (vectorizable_load): Likewise.
9654 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9655 vec_perm_indices instead of auto_vec_perm_indices. Use
9656 tree_to_vec_perm_builder to read the vector from a tree.
9657 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9658 vec_perm_builder instead of a vec_perm_indices.
9659 (have_whole_vector_shift): Use vec_perm_builder and
9660 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9661 truncation to calc_vec_perm_mask_for_shift.
9662 (vect_create_epilog_for_reduction): Likewise.
9663 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9664 from auto_vec_perm_indices to vec_perm_indices.
9665 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9666 instead of changing individual elements.
9667 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9668 the vector in d.perm.
9669 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9670 from auto_vec_perm_indices to vec_perm_indices.
9671 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9672 instead of changing individual elements.
9673 (arm_vectorize_vec_perm_const): Use new_vector to install
9674 the vector in d.perm.
9675 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9676 Update vec_perm_builder constructor.
9677 (rs6000_expand_interleave): Likewise.
9678 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9679 (rs6000_expand_interleave): Likewise.
9681 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9683 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9684 to qimode could truncate the indices.
9685 * optabs.c (expand_vec_perm_var): Likewise.
9687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9689 * Makefile.in (OBJS): Add vec-perm-indices.o.
9690 * vec-perm-indices.h: New file.
9691 * vec-perm-indices.c: Likewise.
9692 * target.h (vec_perm_indices): Replace with a forward class
9694 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9695 * optabs.h: Include vec-perm-indices.h.
9696 (expand_vec_perm): Delete.
9697 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9698 (expand_vec_perm_const): Declare.
9699 * target.def (vec_perm_const_ok): Replace with...
9700 (vec_perm_const): ...this new hook.
9701 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9702 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9703 * doc/tm.texi: Regenerate.
9704 * optabs.def (vec_perm_const): Delete.
9705 * doc/md.texi (vec_perm_const): Likewise.
9706 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9707 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9708 expand_vec_perm for constant permutation vectors. Assert that
9709 the mode of variable permutation vectors is the integer equivalent
9710 of the mode that is being permuted.
9711 * optabs-query.h (selector_fits_mode_p): Declare.
9712 * optabs-query.c: Include vec-perm-indices.h.
9713 (selector_fits_mode_p): New function.
9714 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9715 is defined, instead of checking whether the vec_perm_const_optab
9716 exists. Use targetm.vectorize.vec_perm_const instead of
9717 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9718 fit in the vector mode before using a variable permute.
9719 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9720 vec_perm_indices instead of an rtx.
9721 (expand_vec_perm): Replace with...
9722 (expand_vec_perm_const): ...this new function. Take the selector
9723 as a vec_perm_indices rather than an rtx. Also take the mode of
9724 the selector. Update call to shift_amt_for_vec_perm_mask.
9725 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9726 Use vec_perm_indices::new_expanded_vector to expand the original
9727 selector into bytes. Check whether the indices fit in the vector
9728 mode before using a variable permute.
9729 (expand_vec_perm_var): Make global.
9730 (expand_mult_highpart): Use expand_vec_perm_const.
9731 * fold-const.c: Includes vec-perm-indices.h.
9732 * tree-ssa-forwprop.c: Likewise.
9733 * tree-vect-data-refs.c: Likewise.
9734 * tree-vect-generic.c: Likewise.
9735 * tree-vect-loop.c: Likewise.
9736 * tree-vect-slp.c: Likewise.
9737 * tree-vect-stmts.c: Likewise.
9738 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9740 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9741 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9742 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9743 (aarch64_vectorize_vec_perm_const): ...this new function.
9744 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9745 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9746 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9747 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9748 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9749 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9750 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9752 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9753 check for NEON modes.
9754 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9755 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9756 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9757 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9759 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9760 the old VEC_PERM_CONST conditions.
9761 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9762 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9763 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9764 (ia64_vectorize_vec_perm_const_ok): Merge into...
9765 (ia64_vectorize_vec_perm_const): ...this new function.
9766 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9767 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9768 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9769 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9770 * config/mips/mips.c (mips_expand_vec_perm_const)
9771 (mips_vectorize_vec_perm_const_ok): Merge into...
9772 (mips_vectorize_vec_perm_const): ...this new function.
9773 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9774 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9775 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9776 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9777 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9778 (rs6000_expand_vec_perm_const): Delete.
9779 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9781 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9782 (altivec_expand_vec_perm_const_le): Take each operand individually.
9783 Operate on constant selectors rather than rtxes.
9784 (altivec_expand_vec_perm_const): Likewise. Update call to
9785 altivec_expand_vec_perm_const_le.
9786 (rs6000_expand_vec_perm_const): Delete.
9787 (rs6000_vectorize_vec_perm_const_ok): Delete.
9788 (rs6000_vectorize_vec_perm_const): New function.
9789 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9790 an element count and rtx array.
9791 (rs6000_expand_extract_even): Update call accordingly.
9792 (rs6000_expand_interleave): Likewise.
9793 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9794 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9795 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9796 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9797 (rs6000_expand_vec_perm_const): Delete.
9798 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9799 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9800 (altivec_expand_vec_perm_const_le): Take each operand individually.
9801 Operate on constant selectors rather than rtxes.
9802 (altivec_expand_vec_perm_const): Likewise. Update call to
9803 altivec_expand_vec_perm_const_le.
9804 (rs6000_expand_vec_perm_const): Delete.
9805 (rs6000_vectorize_vec_perm_const_ok): Delete.
9806 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9807 reference to the SPE evmerge intructions.
9808 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9809 an element count and rtx array.
9810 (rs6000_expand_extract_even): Update call accordingly.
9811 (rs6000_expand_interleave): Likewise.
9812 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9813 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9815 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9819 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9820 vector mode and that that mode matches the mode of the data
9822 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9823 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9824 directly using expand_vec_perm_1 when forcing selectors into
9826 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9830 * optabs-query.h (can_vec_perm_p): Delete.
9831 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9832 * optabs-query.c (can_vec_perm_p): Split into...
9833 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9834 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9835 particular selector is valid.
9836 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9837 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9838 (vect_grouped_load_supported): Likewise.
9839 (vect_shift_permute_load_chain): Likewise.
9840 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9841 (vect_transform_slp_perm_load): Likewise.
9842 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9843 (vectorizable_bswap): Likewise.
9844 (vect_gen_perm_mask_checked): Likewise.
9845 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9846 implementations of variable permutation vectors into account
9847 when deciding which selector to use.
9848 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9849 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9850 with a false third argument.
9851 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9852 to test whether the constant selector is valid and can_vec_perm_var_p
9853 to test whether a variable selector is valid.
9855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9857 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9858 * optabs-query.c (can_vec_perm_p): Likewise.
9859 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9860 instead of vec_perm_indices.
9861 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9862 (vect_gen_perm_mask_checked): Likewise,
9863 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9864 (vect_gen_perm_mask_checked): Likewise,
9866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9868 * optabs-query.h (qimode_for_vec_perm): Declare.
9869 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9870 (qimode_for_vec_perm): ...this new function.
9871 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9873 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9875 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9876 does not have a conditional at the top.
9878 2018-01-02 Richard Biener <rguenther@suse.de>
9880 * ipa-inline.c (big_speedup_p): Fix expression.
9882 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9885 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9888 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9892 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9893 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9894 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9895 cond_taken_branch_cost 3->4.
9897 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9899 PR tree-optimization/83581
9900 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9901 TODO_cleanup_cfg if any changes have been made.
9904 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9905 convert_modes if target mode has the right side, but different mode
9909 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9910 last argument when extracting from CONCAT. If either from_real or
9911 from_imag is NULL, use expansion through memory. If result is not
9912 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9913 the parts directly to inner mode, if even that fails, use expansion
9917 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9918 check for bswap in mode rather than HImode and use that in expand_unop
9921 Copyright (C) 2018 Free Software Foundation, Inc.
9923 Copying and distribu