1 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6 2018-04-05 Tom de Vries <tom@codesourcery.com>
9 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
12 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13 Kito Cheng <kito.cheng@gmail.com>
15 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
17 * config/nds32/nds32.h (nds32_16bit_address_type): Add
18 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20 2018-04-05 Shiva Chen <shiva0217@gmail.com>
21 Kito Cheng <kito.cheng@gmail.com>
23 * config/nds32/constraints.md (Ufe): New memory constraint.
24 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
25 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
26 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
28 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
29 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
31 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
33 * config/nds32/nds32.md: Use optimize_size in the condition for
34 alu-shift instructions.
36 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
38 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
40 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
42 * config/nds32/nds32.md (negsi2): Refine pattern.
44 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
45 Chung-Ju Wu <jasonwucj@gmail.com>
47 * config/nds32/iterators.md (shift_rotate): New code iterator.
48 (shift): New code attribute.
49 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
50 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
51 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
52 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
54 (andsi3, *andsi3): Ditto.
55 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
56 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
57 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
58 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
59 nds32_ior_operand, nds32_xor_operand): New predicates.
61 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
63 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
64 (addsi3, subsi3): ... this.
66 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
68 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
70 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
72 * config/nds32/nds32.md: Adjust indention.
74 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
76 * config/nds32/nds32.md (feature): New attribute.
78 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
80 * config/nds32/nds32.md (subtype): New attribute.
82 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
85 * config/arm/arm-builtins.c (arm_expand_builtin): Change
86 expansion to perform a bitwise AND of the argument followed by a
87 boolean negation of the result.
89 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
91 PR rtl-optimization/84878
92 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
93 the basic block. Assert the use reference is not artificial and that
94 it has an associated insn.
96 2018-04-04 Michael Matz <matz@suse.de>
98 * builtins.c (compute_objsize): Pass correct operand
99 to array_at_struct_end_p.
101 2018-04-04 Richard Biener <rguenther@suse.de>
104 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
105 from contexts for DINFO_LEVEL_TERSE and below.
107 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
109 * config/nds32/nds32-doubleword.md (move_<mode>): Require
110 resiter_operand condition.
111 * config/nds32/nds32.md (*move<mode>): Ditto.
113 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
114 Monk Chiang <sh.chiang04@gmail.com>
116 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
118 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
120 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
122 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
123 Kito Cheng <kito.cheng@gmail.com>
125 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
126 nds32_cond_code_str, output_cond_branch,
127 output_cond_branch_compare_zero, nds32_expand_cbranch,
128 nds32_expand_cstore, nds32_expand_movcc,
129 nds32_output_cbranchsi4_equality_zero,
130 nds32_output_cbranchsi4_equality_reg,
131 nds32_output_cbranchsi4_equality_reg_or_const_int,
132 nds32_output_cbranchsi4_greater_less_zero: New functions.
133 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
134 nds32_expand_cstore, nds32_expand_movcc,
135 nds32_output_cbranchsi4_equality_zero,
136 nds32_output_cbranchsi4_equality_reg,
137 nds32_output_cbranchsi4_equality_reg_or_const_int,
138 nds32_output_cbranchsi4_greater_less_zero): Declare.
139 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
140 nds32_rimm11s_operand): New predicates.
141 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
142 * config/nds32/nds32.md: Rewrite all the branch and conditional move
145 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
147 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
148 * config/nds32/nds32.md: Ditto.
149 * config/nds32/pipelines.md: Ditto.
151 2018-04-04 Richard Biener <rguenther@suse.de>
153 PR tree-optimization/85168
154 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
155 propagating abnormals.
157 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
159 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
161 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
162 Kito Cheng <kito.cheng@gmail.com>
164 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
165 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
166 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
167 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
168 * config/nds32/nds32.md (sibcall_internal): New.
169 (sibcall_register): Remove.
170 (sibcall_immediate): Remove.
171 (sibcall_value_internal): New.
172 (sibcall_value_register): Remove.
173 (sibcall_value_immediate): Remove.
174 * config/nds32/predicates.md (nds32_general_register_operand): New.
175 (nds32_call_address_operand): New.
177 2018-04-03 Jakub Jelinek <jakub@redhat.com>
179 PR rtl-optimization/85167
180 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
181 bb_defs if *split_p, instead preinitialize it to NULL.
183 PR tree-optimization/85156
184 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
185 evaluating the argument multiple times.
187 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
189 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
191 (_mm_cvtpd_ps): Likewise.
192 (_mm_cvttpd_epi32): Likewise.
193 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
194 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
195 vector, pixel, and bool following altivec.h include.
197 2018-04-03 Martin Sebor <msebor@redhat.com>
199 * doc/extend.texi (Common Function Attributes): Clarify.
200 (const attribute): Likewise.
201 (pure attribute): Likewise.
203 2018-04-03 Jakub Jelinek <jakub@redhat.com>
206 * config/i386/i386.c (ix86_expand_vector_set): Use
207 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
209 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
211 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
212 instructions when changing rounding bits to preserve precision bits
213 in the x87 control word.
215 2018-04-03 Martin Liska <mliska@suse.cz>
217 PR tree-optimization/82491
218 * rtl.h (strip_offset_and_add): Replace += suboffset with
219 poly_uint64 () + suboffset.
221 2018-03-29 Martin Liska <mliska@suse.cz>
222 Martin Jambor <mjambor@suse.cz>
225 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
226 param_type is not an integral or pointer type.
228 2018-04-03 Richard Biener <rguenther@suse.de>
230 * sese.h (recompute_all_dominators): Remove.
232 2018-04-02 Martin Sebor <msebor@redhat.com>
234 * doc/invoke.texi (-Wrestrict): Fix typos.
236 2018-04-02 Jim Wilson <jimw@sifive.com>
238 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
239 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
240 (<optab>di3, <optab>si3_extend): Likewise.
241 (<optab>si3_mask, <optab>si3_mask_1): New.
242 (<optab>di3_mask, <optab>di3_mask_1): New.
243 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
244 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
245 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
247 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
249 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
252 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
254 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
255 (nds32_canonicalize_comparison): New function.
257 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
258 Kito Cheng <kito.cheng@gmail.com>
259 Kuan-Lin Chen <kuanlinchentw@gmail.com>
261 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
262 * config/nds32/constants.md (unspec_volatile_element): Add
263 UNSPEC_VOLATILE_RELAX_GROUP.
264 * config/nds32/nds32-relax-opt.c: New file.
265 * config/nds32/nds32-predicates.c
266 (nds32_symbol_load_store_p): New function.
267 * config/nds32/nds32-protos.h
268 (nds32_symbol_load_store_p): Declare function.
269 (make_pass_nds32_relax_opt): Declare new rtl pass function.
270 * config/nds32/nds32.c
271 (nds32_register_pass): New function to register pass.
272 (nds32_register_passes): New function to register passes.
273 * config/nds32/nds32.md (relax_group): New pattern.
274 * config/nds32/nds32.opt (mrelax-hint): New option.
275 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
277 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
279 * config/nds32/t-nds32: Modify files dependency.
281 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
283 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
284 (PROFILE_HOOK): Define its implementation.
286 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
288 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
289 type and 32-bit size.
291 2018-04-01 Jakub Jelinek <jakub@redhat.com>
294 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
295 (V_128_256): New mode iterator.
296 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
297 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
298 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
300 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
301 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
303 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
306 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
307 NaN inputs correctly.
309 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
312 * config/rs6000/vsx.md (??r): New mode attribute.
313 (*vsx_mov<mode>_64bit): Use it.
314 (*vsx_mov<mode>_32bit): Likewise.
316 2018-03-30 Martin Sebor <msebor@redhat.com>
318 PR tree-optimization/84818
319 * builtins.c (check_access): Use warning_n.
321 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
324 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
326 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
329 2018-03-30 Julia Koval <julia.koval@intel.com>
332 * x86-tune.def (movx, partial_reg_dependency): Enable for
335 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
338 * lra-constraints.c (process_alt_operands): Move setting
339 this_alternative_matches below.
341 2018-03-29 Martin Liska <mliska@suse.cz>
344 * doc/invoke.texi: Document how LTO works with debug info.
345 Describe auto-load support of binutils. Mention 'x86-64'
346 as valid option value of -march option.
348 2018-03-29 Jakub Jelinek <jakub@redhat.com>
350 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
353 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
354 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
355 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
358 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
361 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
362 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
363 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
364 for RS6000_BTM_POWERPC64.
365 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
366 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
367 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
372 2018-03-28 Carl Love <cel@us.ibm.com>
375 2017-09-27 Carl Love <cel@us.ibm.com>
377 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
378 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
379 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
382 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
384 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
385 instead of __vector bool.
386 (_mm_max_pu8): Likewise.
387 (_mm_min_pi16): Likewise.
389 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
392 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
396 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
398 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
399 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
400 (div_extend): Likewise.
401 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
403 (__builtin_divweuo): Likewise.
404 (__builtin_divdeo): Likewise.
405 (__builtin_divdeuo): Likewise.
407 2018-03-28 Jakub Jelinek <jakub@redhat.com>
410 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
411 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
413 PR tree-optimization/82004
414 * gimple-match-head.c (optimize_pow_to_exp): New function.
415 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
416 Don't fold to exp if optimize_pow_to_exp is false.
418 2018-03-28 Martin Liska <mliska@suse.cz>
421 * calls.c (initialize_argument_information): Fix trailing space.
422 * common.opt: Fix typo and provide better explanation for
423 -fsanitize-coverage option.
424 * config/i386/i386.opt: Fix typo.
426 2018-03-28 Jakub Jelinek <jakub@redhat.com>
427 Martin Liska <mliska@suse.cz>
430 * gimplify.c (asan_poison_variable): Don't do the check for
431 gimplify_omp_ctxp here.
432 (gimplify_decl_expr): Do it here.
433 (gimplify_target_expr): Likewise.
435 2018-03-28 Martin Liska <mliska@suse.cz>
438 * config/i386/i386.c (ix86_function_arg_advance): Do not call
439 chkp_type_bounds_count if MPX is not enabled.
441 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
443 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
445 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
448 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
449 function to create the function decl for complex long double
450 multiply and divide for -mabi=ieeelongdouble.
451 (init_float128_ieee): Call it.
453 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
456 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
457 -fcf-protection=branch -mibt.
458 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
460 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
463 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
465 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
468 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
469 extern array declarations.
471 2018-03-27 Richard Biener <rguenther@suse.de>
474 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
475 explicit single_use checks.
477 2018-03-27 Richard Biener <rguenther@suse.de>
479 PR tree-optimization/85082
480 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
483 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
485 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
486 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
487 Turn on fasynchronous-unwind-tables and funwind-tables.
489 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
492 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
493 (*bmi_blsr_<mode>_ccz): Ditto.
495 2018-03-26 Tom de Vries <tom@codesourcery.com>
497 PR tree-optimization/85063
498 * omp-general.c (offloading_function_p): New function. Factor out
500 * omp-offload.c (pass_omp_target_link::gate): ... here.
501 * omp-general.h (offloading_function_p): Declare.
502 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
503 with attribute omp declare target for offloading functions.
505 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
507 PR tree-optimization/84005
508 * tree-data-ref.h (get_base_for_alignment): Declare.
509 * tree-data-ref.c (get_base_for_alignment_1): New function.
510 (get_base_for_alignment): Likewise.
511 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
512 get_base_for_alignment to find a suitable base object, instead
513 of always using drb->base_address.
515 2018-03-23 Jakub Jelinek <jakub@redhat.com>
518 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
519 known size by default.
521 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
524 * lra-constraints.c (process_alt_operands): Don't match BLKmode
525 and non BLKmode operands.
527 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
530 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
533 2018-03-23 Richard Biener <rguenther@suse.de>
536 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
537 we are going to emit early debug for LTO.
539 2018-03-23 Jakub Jelinek <jakub@redhat.com>
542 * function.c (match_asm_constraints_1): Don't optimize if input
543 doesn't satisfy general_operand predicate for output's mode.
546 * alias.c (write_dependence_p): Don't require for x_canonicalized
547 non-VOIDmode if x has VOIDmode.
550 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
551 just don't try to optimize it rather than assert it never happens.
553 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
555 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
556 macro expansions for definition of ST_INTERNAL_<mode> and
557 LD_INTERNAL_<mode> builtins.
558 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
560 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
562 (altivec_expand_st_builtin): Likewise.
563 (altivec_expand_builtin): Remove calls to deleted functions.
564 (rs6000_address_for_altivec): Delete this function.
565 * config/rs6000/vector.md: Remove expands for
566 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
568 2018-03-22 Sudakshina Das <sudi.das@arm.com>
571 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
572 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
573 re-computing once computed.
574 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
575 (arm_init_machine_status): Initialize
576 machine->static_chain_stack_bytes.
578 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
581 * doc/extend.texi: Add four new prototypes for vec_ld.
582 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
583 definitions for more logical presentation.
584 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
585 entries for V1TI variants of __builtin_altivec_ld builtin.
586 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
587 handling of V1TI variant of LVX icode pattern.
588 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
589 (rs6000_gimple_fold_builtin): Likewise.
590 (altivec_init_builtins): Add code to define
591 __builtin_altivec_lvx_v1ti function.
593 2018-03-22 Jakub Jelinek <jakub@redhat.com>
596 * function.c (match_asm_constraints_1): Don't do the optimization
597 if input isn't a REG, SUBREG, MEM or constant.
599 2018-03-22 Tom de Vries <tom@codesourcery.com>
601 PR tree-optimization/84956
602 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
603 bb_has_abnormal_pred.
605 2018-03-22 Jakub Jelinek <jakub@redhat.com>
608 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
609 DECL_INITIAL (decl) to decl at the end.
610 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
613 2018-03-21 Joseph Myers <joseph@codesourcery.com>
615 * doc/extend.texi (__builtin_tgmath): Document when complex
616 integer types are treated as _Complex _Float64.
618 2018-03-21 Tom de Vries <tom@codesourcery.com>
620 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
622 2018-03-21 Jakub Jelinek <jakub@redhat.com>
624 PR tree-optimization/84960
625 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
626 if it is ENTRY block, move them into single succ of ENTRY in that case.
628 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
630 PR tree-optimization/84811
631 * poly-int.h (poly_span_traits): Remove the T3 parameter and
632 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
633 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
634 (known_subrange_p): Update accordingly. Cast each value involved
635 in the size comparison, rather than casting the result of the
638 2018-03-21 Jakub Jelinek <jakub@redhat.com>
640 PR tree-optimization/84982
641 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
642 by flipping the least significant bit rather than all bits from
643 bitpos to bitpos + bitsize - 1.
645 2018-03-21 Nathan Sidwell <nathan@acm.org>
647 * doc/extend.texi (Deprecated Features): Remove mention of
648 long-deleted deprecations.
650 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
653 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
654 * configure: Regenerate.
656 2018-03-21 Tom de Vries <tom@codesourcery.com>
658 PR tree-optimization/83126
659 * tree-parloops.c (num_phis): New function.
660 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
662 2018-03-21 Nathan Sidwell <nathan@acm.org>
664 * doc/extend.texi (Deprecated Features): Update deprecated flags,
665 mention anon-struct/union members and trailing attributes.
667 2018-03-21 Bin Cheng <bin.cheng@arm.com>
669 PR tree-optimization/84969
670 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
671 builtin memset partitions if they set different rhs values.
673 2018-03-21 Jakub Jelinek <jakub@redhat.com>
675 PR rtl-optimization/84989
676 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
677 VEC_DUPLICATE with scalar result mode.
679 2018-03-21 Martin Liska <mliska@suse.cz>
682 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
683 not intended return statement.
685 2018-03-21 Martin Liska <mliska@suse.cz>
688 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
689 (chkp_find_bound_slots_1): Limit number of iterations.
691 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
694 * Minor grammar fixes for x86 options.
696 2018-03-20 Jakub Jelinek <jakub@redhat.com>
699 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
700 holding REG_CFA_RESTORE notes, instead turn them into a USE.
702 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
705 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
706 (altivec_lvx_<mode>_1op): Likewise.
707 (altivec_stvx_<mode>_2op): Likewise.
708 (altivec_stvx_<mode>_1op): Likewise.
709 (altivec_lvx_<VM2:mode>): New define_expand.
710 (altivec_stvx_<VM2:mode>): Likewise.
711 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
712 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
713 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
714 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
715 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
716 (rs6000_gen_lvx): Likewise.
717 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
718 (altivec_expand_stv_builtin): Likewise.
719 (altivec_expand_builtin): Likewise.
720 * config/rs6000/vector.md: Likewise.
722 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
725 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
728 2018-03-20 Richard Biener <rguenther@suse.de>
731 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
732 sign-conversions as zero, fall back to standard scalar_stmt
735 2018-03-20 Martin Liska <mliska@suse.cz>
738 * predict.c (rebuild_frequencies): Handle case when we have
739 PROFILE_ABSENT, but flag_guess_branch_prob is false.
741 2018-03-20 Jakub Jelinek <jakub@redhat.com>
744 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
745 flag_section_anchors.
746 * varasm.c (use_blocks_for_decl_p): Remove hack for
750 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
752 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
753 be created, use lowpart_subreg of operands[0] rather than operands[0]
755 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
756 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
757 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
758 and n constraint instead of aarch64_shift_imm_di and Usd.
759 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
760 (*aarch64_<optab>_reg_minus<mode>3): ... this.
762 2018-03-20 Sudakshina Das <sudi.das@arm.com>
765 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
766 to favor GPR over NEON registers.
767 (<shift>di3_neon): Likewise.
769 2018-03-20 Tom de Vries <tom@codesourcery.com>
772 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
773 (nvptx_process_pars): Emit bar.sync asap and alap.
775 2018-03-20 Tom de Vries <tom@codesourcery.com>
778 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
779 seen_label if seen_label is already set.
781 2018-03-20 Jakub Jelinek <jakub@redhat.com>
784 * config/i386/i386.c (fold_builtin_cpu): For features above 31
785 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
786 Use 1U instead of 1. Formatting fixes.
789 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
790 instead of TREE_TYPE (s1) for the return value.
792 2018-03-19 Jakub Jelinek <jakub@redhat.com>
794 PR tree-optimization/84946
795 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
796 bitsize + bitsize in poly_uint64 rather than poly_int64.
799 * dwarf2asm.c: Include fold-const.c.
800 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
801 of decl rather than decl itself.
803 PR rtl-optimization/84643
804 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
806 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
809 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
810 calling assemble_variable.
812 2018-03-19 Sudakshina Das <sudi.das@arm.com>
815 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
816 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
818 2018-03-19 Jim Wilson <jimw@sifive.com>
821 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
822 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
823 (riscv_first_stack_step): Likewise.
824 (riscv_option_override): Use STACK_BOUNDARY instead of
826 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
828 (BIGGEST_ALIGNMENT): Set to 128.
829 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
830 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
833 2018-03-19 Richard Biener <rguenther@suse.de>
835 PR tree-optimization/84933
836 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
837 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
839 2018-03-19 Richard Biener <rguenther@suse.de>
841 PR tree-optimization/84859
842 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
843 (cond_if_else_store_replacement): Perform sinking operation on
844 single-store BBs regardless of MAX_STORES_TO_SINK setting.
845 Generalize what a BB with a single eligible store is.
847 2018-03-19 Richard Biener <rguenther@suse.de>
849 PR tree-optimization/84929
850 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
851 chrec_is_positive against non-chrec arg.
853 2018-03-19 Tamar Christina <tamar.christina@arm.com>
856 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
858 2018-03-18 Martin Liska <mliska@suse.cz>
860 PR rtl-optimization/84635
861 * regrename.c (build_def_use): Use matches_mode only when
864 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
866 PR tree-optimization/84913
867 * tree-vect-loop.c (vectorizable_reduction): Don't try to
868 vectorize chains of COND_EXPRs.
870 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
872 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
874 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
876 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
878 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
880 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
882 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
883 Kito Cheng <kito.cheng@gmail.com>
885 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
886 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
887 (nds32_adjust_reg_alloc_order): New function.
888 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
890 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
892 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
893 nds32_print_operand, nds32_print_operand_address): Use
894 HOST_WIDE_INT_PRINT_DEC instead.
896 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
898 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
900 2018-03-17 Jakub Jelinek <jakub@redhat.com>
903 * config/i386/i386.c (initial_ix86_tune_features,
904 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
906 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
907 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
908 rather than 1u << ix86_tune. Formatting fix.
909 (ix86_option_override_internal): Change ix86_arch_mask from
910 unsigned int to unsigned HOST_WIDE_INT, initialize to
911 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
912 (ix86_function_specific_restore): Likewise.
914 2018-03-16 Jakub Jelinek <jakub@redhat.com>
917 * postreload.c (reload_combine_recognize_pattern): Perform
918 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
919 truncate_int_for_mode the result for the destination's mode.
922 * hsa-gen.c (mem_type_for_type): Fix comment typo.
923 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
925 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
928 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
931 * lra-assigns.c (lra_split_hard_reg_for): Don't use
932 regno_allocno_class_array and sorted_pseudos.
933 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
934 insns where regno is used.
936 2018-03-16 Martin Liska <mliska@suse.cz>
939 * multiple_target.c (create_dispatcher_calls): Redirect
940 reference in the symbol table.
942 2018-03-16 Martin Liska <mliska@suse.cz>
945 * multiple_target.c (create_dispatcher_calls): Redirect also
948 2018-03-16 Jakub Jelinek <jakub@redhat.com>
952 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
953 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
954 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
956 2018-03-16 Julia Koval <julia.koval@intel.com>
958 * doc/invoke.texi (Skylake Server): Add CLWB.
959 Cannonlake): Remove CLWB.
961 2018-03-16 Jakub Jelinek <jakub@redhat.com>
963 PR tree-optimization/84841
964 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
966 (FLOAT_ONE_CONST_TYPE): Define.
967 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
968 (sort_by_operand_rank): Put entries with higher constant_type last
969 rather than first to match comments.
971 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
973 * config/nios2/nios2.md (movsi_internal): Fix thinko in
976 2018-03-15 Jakub Jelinek <jakub@redhat.com>
979 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
980 check and use address of target always.
982 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
985 * config/i386/i386.c (indirect_thunk_needed): Update comments.
986 (indirect_thunk_bnd_needed): Likewise.
987 (indirect_thunks_used): Likewise.
988 (indirect_thunks_bnd_used): Likewise.
989 (indirect_return_needed): New.
990 (indirect_return_bnd_needed): Likewise.
991 (output_indirect_thunk_function): Add a bool argument for
993 (output_indirect_thunk_function): Don't generate alias for
994 function return thunk.
995 (ix86_code_end): Call output_indirect_thunk_function to generate
996 function return thunks.
997 (ix86_output_function_return): Set indirect_return_bnd_needed
998 and indirect_return_needed instead of indirect_thunk_bnd_needed
999 and indirect_thunk_needed.
1001 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1003 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1004 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1005 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1007 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1008 Paul Hua <paul.hua.gm@gmail.com>
1011 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1013 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1015 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1016 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1019 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1022 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1023 instead of GET_MODE_SIZE when comparing Units.
1025 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1028 * varasm.c (hash_section): Return an unchangeble hash value
1029 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1030 Return !aarch64_can_use_per_function_literal_pools_p ().
1032 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1035 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1036 rather than address of cmode as last argument to prepare_cmp_insn.
1038 2018-03-15 Julia Koval <julia.koval@intel.com>
1040 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1041 F_AVX512VNNI, F_AVX512BITALG): New.
1043 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1046 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1047 insn for floating-point loads and stores.
1049 2018-03-14 Carl Love <cel@us.ibm.com>
1051 * config/rs6000/rs6000-c.c: Add macro definitions for
1052 ALTIVEC_BUILTIN_VEC_PERMXOR.
1053 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1054 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1055 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1056 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1058 * config/doc/extend.texi: Add prototypes for vec_permxor.
1060 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1063 * diagnostic-show-locus.c (class layout_point): Convert m_line
1064 from int to linenum_type.
1065 (line_span::comparator): Use linenum "compare" function when
1066 comparing line numbers.
1067 (test_line_span): New function.
1068 (layout_range::contains_point): Convert param "row" from int to
1070 (layout_range::intersects_line_p): Likewise.
1071 (layout::will_show_line_p): Likewise.
1072 (layout::print_source_line): Likewise.
1073 (layout::should_print_annotation_line_p): Likewise.
1074 (layout::print_annotation_line): Likewise.
1075 (layout::print_leading_fixits): Likewise.
1076 (layout::annotation_line_showed_range_p): Likewise.
1077 (struct line_corrections): Likewise for field m_row.
1078 (line_corrections::line_corrections): Likewise for param "row".
1079 (layout::print_trailing_fixits): Likewise.
1080 (layout::get_state_at_point): Likewise.
1081 (layout::get_x_bound_for_row): Likewise.
1082 (layout::print_line): Likewise.
1083 (diagnostic_show_locus): Likewise for locals "last_line" and
1085 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1086 * input.c (selftest::test_linenum_comparisons): New function.
1087 (selftest::input_c_tests): Call it.
1088 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1089 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1090 * selftest.h (ASSERT_GT): New macro.
1091 (ASSERT_GT_AT): New macro.
1092 (ASSERT_LT): New macro.
1093 (ASSERT_LT_AT): New macro.
1095 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1097 PR rtl-optimization/84780
1098 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1100 2018-03-14 Martin Liska <mliska@suse.cz>
1102 * tree.c (record_node_allocation_statistics): Use
1103 get_stats_node_kind.
1104 (get_stats_node_kind): New function extracted from
1105 record_node_allocation_statistics.
1106 (free_node): Use get_stats_node_kind.
1108 2018-03-14 Richard Biener <rguenther@suse.de>
1110 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1111 that the value-set of ANTIC_IN doesn't grow.
1114 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1116 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1117 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1119 2018-03-14 Julia Koval <julia.koval@intel.com>
1121 * config.gcc (icelake-client, icelake-server): New.
1123 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1124 (initial_ix86_arch_features): Ditto.
1125 (PTA_SKYLAKE): Add SGX.
1126 (PTA_ICELAKE): Remove.
1127 (PTA_ICELAKE_CLIENT): New.
1128 (PTA_ICELAKE_SERVER): New.
1129 (ix86_option_override_internal): Split up icelake on icelake client and
1131 (get_builtin_code_for_version): Ditto.
1132 (fold_builtin_cpu): Ditto.
1133 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1134 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1135 * config/i386/i386.h (processor_type): Ditto.
1136 * doc/invoke.texi: Ditto.
1138 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1141 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1142 INTEGER_CST offset, add it together with bitpos / 8 and
1143 sign extend based on POINTER_SIZE.
1147 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1150 * config/i386/constraints.md (Yc): New register constraint.
1151 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1152 Use Yc constraint for alternative 2 of operand 0. Remove
1153 preferred_for_speed attribute.
1155 2018-03-14 Richard Biener <rguenther@suse.de>
1157 PR tree-optimization/84830
1158 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1159 with the old one to avoid oscillations.
1161 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1164 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1166 (assign_by_spills): Return a flag of reload assignment failure.
1167 Do not process the reload assignment failures. Do not spill other
1168 reload pseudos if they has the same reg class. Update n if
1170 (lra_assign): Add a return arg. Set up from the result of
1171 assign_by_spills call.
1172 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1173 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1174 usage_insns if it is not NULL.
1175 (spill_hard_reg_in_range): New function.
1176 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1177 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1178 function prototypes.
1179 (lra_assign): Change prototype.
1180 * lra.c (lra): Add code to deal with fails by splitting hard reg
1183 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1185 * config/riscv/riscv.opt (mrelax): New option.
1186 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1187 "norelax" when riscv_mrelax is disabled.
1188 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1190 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1193 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1194 reassociation for int modes.
1196 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1198 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1199 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1201 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1202 * config/aarch64/aarch64-sve.md
1203 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1204 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1205 (*extend<mode><Vwide>2): Rename to...
1206 (aarch64_sve_extend<mode><Vwide>2): ...this.
1207 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1208 renaming the old pattern to...
1209 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1211 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1212 define_expand, renaming the old pattern to...
1213 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1214 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1215 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1216 account when deciding which SVE instruction the optab should use.
1217 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1219 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1221 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1222 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1223 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1224 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1225 (tlsdesc_small_<mode>): Turn a define_expand and use
1226 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1227 (tlsdesc_small_advsimd_<mode>): ...this.
1228 (tlsdesc_small_sve_<mode>): New pattern.
1230 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1232 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1233 (UNSPEC_UMUL_HIGHPART): New constants.
1234 (MUL_HIGHPART): New int iteraor.
1235 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1236 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1238 (*<su>mul<mode>3_highpart): New define_insn.
1240 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1243 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1246 2018-03-13 Martin Liska <mliska@suse.cz>
1249 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1251 (sem_item_optimizer::~sem_item_optimizer): Release it.
1252 (sem_item_optimizer::merge_classes): Register variable aliases.
1253 (sem_item_optimizer::fixup_pt_set): New function.
1254 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1255 * ipa-icf.h: Declare new variables and functions.
1257 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1260 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1261 integer_pow2p@2 and test integer_pow2p in condition.
1262 (A < 0 ? C : 0): Similarly for @1.
1265 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1266 characters starting at p contain '\0' character, don't look beyond
1270 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1271 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1274 * reg-stack.c (change_stack): Change update_end var from int to
1275 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1276 also call set_block_for_insn on the newly added insns and rescan.
1279 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1280 on the last operand.
1283 * tree.c (stabilize_reference_1): Return save_expr (e) for
1284 STATEMENT_LIST even if it doesn't have side-effects.
1286 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1288 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1290 2018-03-12 Renlin Li <renlin.li@arm.com>
1292 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1293 aarch64_output_scalar_simd_mov_immediate.
1295 2018-03-12 Martin Sebor <msebor@redhat.com>
1297 PR tree-optimization/83456
1298 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1299 for perfectly overlapping calls to memcpy.
1300 (gimple_fold_builtin_memory_chk): Same.
1301 (gimple_fold_builtin_strcpy): Handle no-warning.
1302 (gimple_fold_builtin_stxcpy_chk): Same.
1303 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1305 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1307 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1308 parameter. Use it for SFmode.
1309 (rs6000_function_arg_advance_1): Adjust.
1310 (rs6000_function_arg): Adjust.
1311 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1313 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1315 PR rtl-optimization/84169
1316 PR rtl-optimization/84780
1317 * combine.c (can_combine_p): Check for a 2-insn combination whether
1318 the destination register is used between the two insns, too.
1320 2018-03-12 Richard Biener <rguenther@suse.de>
1322 PR tree-optimization/84803
1323 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1324 for refs DR analysis didn't process.
1326 2018-03-12 Richard Biener <rguenther@suse.de>
1328 PR tree-optimization/84777
1329 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1330 force-vectorize loops ignore whether we are optimizing for size.
1332 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1334 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1335 (TARGET_MD_ASM_ADJUST): Define.
1337 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1338 Kito Cheng <kito.cheng@gmail.com>
1339 Chung-Ju Wu <jasonwucj@gmail.com>
1341 * config/nds32/nds32.c (nds32_compute_stack_frame,
1342 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1343 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1344 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1345 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1346 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1347 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1348 * config/nds32/nds32.md (prologue, epilogue): Use macro
1349 NDS32_V3PUSH_AVAILABLE_P to do checking.
1351 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1354 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1355 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1356 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1357 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1358 addition of most attributes on !orig_type_die or the attribute not
1359 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1361 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1362 Chung-Ju Wu <jasonwucj@gmail.com>
1364 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1366 * config/nds32/nds32.opt (mvh): New option.
1368 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1369 Chung-Ju Wu <jasonwucj@gmail.com>
1371 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1373 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1374 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1377 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1378 Chung-Ju Wu <jasonwucj@gmail.com>
1380 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1382 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1383 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1385 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1386 Kito Cheng <kito.cheng@gmail.com>
1387 Chung-Ju Wu <jasonwucj@gmail.com>
1389 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1390 UNSPEC_FFMISM and UNSPEC_FLMISM.
1391 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1392 for ffb, ffmism and flmism.
1393 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1394 (unspec_ffmism): Ditto.
1395 (unspec_flmism): Ditto.
1396 (nds32_expand_builtin_impl): Check if string extension is available.
1397 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1398 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1400 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1403 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1406 * lra-assigns.c (assign_by_spills): Return a flag of reload
1407 assignment failure. Do not process the reload assignment
1408 failures. Do not spill other reload pseudos if they has the same
1410 (lra_assign): Add a return arg. Set up from the result of
1411 assign_by_spills call.
1412 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1413 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1414 usage_insns if it is not NULL.
1415 (spill_hard_reg_in_range): New function.
1416 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1417 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1418 function prototypes.
1419 (lra_assign): Change prototype.
1420 * lra.c (lra): Add code to deal with fails by splitting hard reg
1423 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1426 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1428 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1431 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1432 (dw_val_node): Add val_symbolic_view.
1433 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1434 (symview_upper_bound): New.
1435 (new_line_info_table): Initialize symviews_since_reset.
1436 (dwarf2out_source_line): Count symviews_since_reset and set
1437 symview_upper_bound.
1438 (dw_val_equal_p): Handle symview.
1439 (add_AT_symview): New.
1440 (print_dw_val): Handle symview.
1441 (attr_checksum, attr_checksum_ordered): Likewise.
1442 (same_dw_val_p, size_of_die): Likewise.
1443 (value_format, output_die): Likewise.
1444 (add_high_low_attributes): Use add_AT_symview for entry_view.
1445 (dwarf2out_finish): Reset symview_upper_bound, clear
1448 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1451 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1452 Add strict argument and use it.
1453 (rs6000_split_multireg_move): Update for new strict argument.
1454 (mem_operand_gpr): Disallow all non-offsettable addresses.
1455 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1457 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1460 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1461 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1462 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1465 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1466 decl, use remap_type if we want to use the type.
1468 2018-03-09 Martin Sebor <msebor@redhat.com>
1470 PR tree-optimization/84526
1471 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1473 (builtin_access::generic_overlap): Be prepared to handle non-array
1476 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1478 PR rtl-optimization/84682
1479 * lra-constraints.c (process_address_1): Check is_address flag
1480 for address constraints.
1481 (process_alt_operands): Likewise.
1482 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1483 preprocess_constraints.
1484 * recog.h (preprocess_constraints): Add oploc parameter.
1486 * recog.c (preprocess_constraints): Test address_operand for
1487 CT_ADDRESS constraints.
1489 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1492 * lra-assigns.c (assign_by_spills): Return a flag of reload
1493 assignment failure. Do not process the reload assignment
1494 failures. Do not spill other reload pseudos if they has the same
1496 (lra_assign): Add a return arg. Set up from the result of
1497 assign_by_spills call.
1498 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1499 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1500 usage_insns if it is not NULL.
1501 (spill_hard_reg_in_range): New function.
1502 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1503 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1504 function prototypes.
1505 (lra_assign): Change prototype.
1506 * lra.c (lra): Add code to deal with fails by splitting hard reg
1509 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1512 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1513 Accept complain bool parameter. Only emit errors if it is true.
1514 (arm_parse_cpu_option_name): Likewise.
1515 (arm_target_thumb_only): Adjust callers of the above.
1516 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1517 prototype to take a default true bool parameter.
1518 (arm_parse_arch_option_name): Likewise.
1520 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1521 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1525 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1526 * configure: Regenerate.
1527 * configure.ac ("linker --version-script option"): New.
1528 ("linker soname option"): New.
1530 2018-03-09 Richard Biener <rguenther@suse.de>
1532 PR tree-optimization/84775
1533 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1534 immediate uses of predicate stmts and mark them modified.
1537 PR tree-optimization/84178
1538 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1540 (version_loop_for_if_conversion): Delay update_ssa call.
1541 (tree_if_conversion): Delay update_ssa until after predicate
1544 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1547 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1548 when the function accesses prior frames.
1550 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1553 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1554 gen_llsym, otherwise call maybe_gen_llsym.
1557 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1558 has ',' character inside of it.
1560 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1563 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1564 as clobbering CC_REGNUM.
1566 2018-03-08 Richard Biener <rguenther@suse.de>
1569 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1570 (follow_copies_to_constant): Do not follow SSA names registered
1573 2018-03-08 Richard Biener <rguenther@suse.de>
1575 PR tree-optimization/84178
1576 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1578 (version_loop_for_if_conversion): Delay update_ssa call.
1579 (tree_if_conversion): Delay update_ssa until after predicate
1582 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1584 PR tree-optimization/84178
1585 * tree-if-conv.c (release_bb_predicate): Remove the
1586 the assertion that the stmts have NULL use_ops.
1587 Discard the statements, asserting that they haven't
1588 yet been added to a BB.
1590 2018-03-08 Richard Biener <rguenther@suse.de>
1592 PR tree-optimization/84746
1593 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1594 (phi_translate): Pass in destination ANTIC_OUT set.
1595 (phi_translate_1): Likewise. For a simplified result lookup
1596 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1597 (phi_translate_set): Adjust.
1598 (do_pre_regular_insertion): Likewise.
1599 (do_pre_partial_partial_insertion): Likewise.
1601 2018-03-08 Martin Liska <mliska@suse.cz>
1603 PR gcov-profile/84735
1604 * doc/gcov.texi: Document usage of profile files.
1605 * gcov-io.h: Document changes in the format.
1607 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1611 * dwarf2out.c (struct dw_line_info_table): Update comments for
1613 (FORCE_RESET_NEXT_VIEW): New.
1614 (FORCE_RESETTING_VIEW_P): New.
1615 (RESETTING_VIEW_P): Check for -1 too.
1616 (ZERO_VIEW_P): Likewise.
1617 (new_line_info_table): Force-reset next view.
1618 (dwarf2out_begin_function): Likewise.
1619 (dwarf2out_source_line): Simplify zero_view_p initialization.
1620 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1621 view directly. Omit view when omitting .loc at line 0.
1623 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1625 PR tree-optimization/84740
1626 * tree-switch-conversion.c (process_switch): Call build_constructors
1627 only if info.phi_count is non-zero.
1629 PR tree-optimization/84739
1630 * tree-tailcall.c (find_tail_calls): Check call arguments against
1631 DECL_ARGUMENTS (current_function_decl) rather than
1632 DECL_ARGUMENTS (func) when checking for tail recursion.
1634 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1636 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1637 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1638 Volker Reichelt's entry and add entries for people that perform
1639 GCC fuzzy testing and report numerous bugs.
1641 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1644 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1645 readonly data in sdata, if that is disabled.
1646 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1647 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1648 -mreadonly-in-sdata option.
1650 2018-03-07 Martin Sebor <msebor@redhat.com>
1652 PR tree-optimization/84468
1653 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1654 basic block when looking for nul assignment.
1656 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1659 * except.h (output_function_exception_table): Adjust prototype.
1660 * except.c (output_function_exception_table): Remove FNNAME parameter
1661 and add SECTION parameter. Ouput one part of the table at a time.
1662 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1663 the first part of the exception table and emit unwind directives.
1664 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1665 (i386_pe_seh_cold_init): Likewise.
1666 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1667 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1668 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1669 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1670 * config/i386/winnt.c: Include except.h.
1671 (struct seh_frame_state): Add reg_offset, after_prologue and
1672 in_cold_section fields.
1673 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1674 (i386_pe_seh_cold_init): New function.
1675 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1676 to seh->in_cold_section.
1677 (seh_emit_push): Record the offset of the push.
1678 (seh_emit_save): Record the offet of the save.
1679 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1680 Test seh->after_prologue to disregard the epilogue.
1681 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1682 (i386_pe_end_cold_function): New function.
1684 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1687 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1688 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1691 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1693 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1694 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1697 * multiple_target.c: Include tree-inline.h and intl.h.
1698 (expand_target_clones): Diagnose and fail if node->definition and
1699 !tree_versionable_function_p (node->decl).
1701 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1703 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1705 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1706 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1707 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1709 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1712 * combine.c (try_combine): Use reg_or_subregno instead of handling
1713 just paradoxical SUBREGs and REGs.
1715 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1717 * config/arc/arc.c (arc_finalize_pic): Remove function.
1718 (arc_must_save_register): We use single base PIC register, remove
1719 checks to save/restore the PIC register.
1720 (arc_expand_prologue): Likewise.
1721 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1723 (arc_verify_short): Likewise.
1724 (arc_attr_type): Likewise.
1725 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1726 (walk_stores): Likewise.
1727 (arc_address_cost): Make it static.
1728 (arc_verify_short): Likewise.
1729 (branch_dest): Likewise.
1730 (arc_attr_type): Likewise.
1731 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1732 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1733 (arc_final_prescan_insn): Remove inserting the nops due to
1734 hardware hazards. It is done in reorg step.
1735 (insn_length_variant_t): Remove.
1736 (insn_length_parameters_t): Likewise.
1737 (arc_insn_length_parameters): Likewise.
1738 (arc_get_insn_variants): Likewise.
1739 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1741 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1744 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1747 PR tree-optimization/84687
1748 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1750 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1752 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1754 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1755 Rename to ppc_speculation_barrier.
1756 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1757 __builtin_ppc_speculation_barrier.
1759 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1762 * combine.c (combine_simplify_rtx): Don't try to simplify if
1763 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1766 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1768 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1769 to 32 bytes when compiling for POWER9.
1771 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1774 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1775 regparm >= 3 with no arg reg available also for calls with
1776 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1779 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1781 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1783 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1786 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1788 2018-03-05 Richard Biener <rguenther@suse.de>
1790 PR tree-optimization/84486
1791 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1792 When inserting a __builtin_assume_aligned call set the LHS
1793 SSA name alignment info accordingly.
1795 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1797 PR tree-optimization/84114
1798 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1799 Avoid reassociation of FLOAT_MODE addition.
1801 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1803 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1804 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1805 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1806 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1807 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1808 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1809 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1811 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1812 __builtin_ia32_wbinvd): New builtins.
1813 (SPECIAL_ARGS2): New.
1814 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1815 (SPECIAL_ARGS2): New.
1816 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1817 (ix86_valid_target_attribute_inner_p): Ditto.
1818 (ix86_init_mmx_sse_builtins): Add special_args2.
1819 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1820 TARGET_WBNOINVD_P): New.
1821 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1822 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1823 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1824 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1825 * config/i386/pconfigintrin.h: New file.
1826 * config/i386/wbnoinvdintrin.h: Ditto.
1827 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1828 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1830 2018-03-05 Richard Biener <rguenther@suse.de>
1832 PR tree-optimization/84670
1833 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1835 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1836 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1837 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1838 doesn't grow if all successors (recursively) were visited at least
1841 2018-03-05 Richard Biener <rguenther@suse.de>
1843 PR tree-optimization/84650
1844 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1845 if executed in the loop pipeline.
1847 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1849 * doc/configfiles.texi (Configuration Files): Move info about
1850 conditionalizing $target-protos.h to...
1851 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1852 differs from $target-protos.h.
1854 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1855 Chung-Ju Wu <jasonwucj@gmail.com>
1857 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1858 * config/nds32/nds32-multiple.md (setmemsi): Define.
1859 * config/nds32/nds32-memory-manipulation.c
1860 (nds32_gen_dup_4_byte_to_word_value): New.
1861 (emit_setmem_word_loop): New.
1862 (emit_setmem_byte_loop): New.
1863 (nds32_expand_setmem_loop): New.
1864 (nds32_expand_setmem_loop_v3m): New.
1865 (nds32_expand_setmem_unroll): New.
1866 (nds32_expand_setmem): New.
1868 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1869 Chung-Ju Wu <jasonwucj@gmail.com>
1871 * config/nds32/nds32-memory-manipulation.c
1872 (nds32_emit_load_store): New.
1873 (nds32_emit_post_inc_load_store): New.
1874 (nds32_emit_mem_move): New.
1875 (nds32_emit_mem_move_block): New.
1876 (nds32_expand_movmemsi_loop_unknown_size): New.
1877 (nds32_expand_movmemsi_loop_known_size): New.
1878 (nds32_expand_movmemsi_loop): New.
1879 (nds32_expand_movmemsi_unroll): New.
1880 (nds32_expand_movmemqi): Rename ...
1881 (nds32_expand_movmemsi): ... to this.
1882 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1883 (movmemsi): ... to this.
1884 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1885 (nds32_expand_movmemsi): ... to this.
1887 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1888 Monk Chiang <sh.chiang04@gmail.com>
1889 Chung-Ju Wu <jasonwucj@gmail.com>
1891 * config/nds32/nds32-protos.h
1892 (nds32_expand_load_multiple): New arguments.
1893 (nds32_expand_store_multiple): Ditto.
1894 (nds32_valid_multiple_load_store): Rename ...
1895 (nds32_valid_multiple_load_store_p): ... to this.
1896 * config/nds32/nds32-memory-manipulation.c
1897 (nds32_expand_load_multiple): Refine implementation.
1898 (nds32_expand_store_multiple): Ditto.
1899 * config/nds32/nds32-multiple.md
1900 (load_multiple): Update nds32_expand_load_multiple interface.
1901 (store_multiple): Update nds32_expand_store_multiple interface.
1902 * config/nds32/nds32-predicates.c
1903 (nds32_valid_multiple_load_store): Rename ...
1904 (nds32_valid_multiple_load_store_p): ... to this and refine
1906 * config/nds32/predicates.md
1907 (nds32_load_multiple_and_update_address_operation): New predicate.
1908 (nds32_store_multiple_and_update_address_operation): New predicate.
1910 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1911 Chung-Ju Wu <jasonwucj@gmail.com>
1913 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1914 (combo): New attribute.
1915 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1917 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1919 * config/nds32/nds32.opt: Change -mcmodel= default value.
1921 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1922 Monk Chiang <sh.chiang04@gmail.com>
1923 Chung-Ju Wu <jasonwucj@gmail.com>
1925 * config/nds32/constants.md (unspec_element): New enum.
1926 * config/nds32/constraints.md (Umw): New constraint.
1927 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1928 * config/nds32/nds32-intrinsic.md: Likewise.
1929 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1930 (nds32_valid_smw_lwm_base_p): New.
1931 (nds32_output_smw_single_word): New.
1932 (nds32_output_lmw_single_word): New.
1933 (nds32_expand_unaligned_load): New.
1934 (nds32_expand_unaligned_store): New.
1935 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1936 (nds32_output_smw_single_word): Declare.
1937 (nds32_output_lmw_single_word): Declare.
1938 (nds32_expand_unaligned_load): Declare.
1939 (nds32_expand_unaligned_store): Declare.
1940 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1941 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1942 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1943 NDS32_BUILTIN_UASTORE_DW.
1944 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1947 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1948 Kito Cheng <kito.cheng@gmail.com>
1949 Chung-Ju Wu <jasonwucj@gmail.com>
1951 * config/nds32/nds32-intrinsic.c
1952 (nds32_expand_builtin_null_ftype_reg): Delete.
1953 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1954 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1955 (nds32_read_argument): New.
1956 (nds32_legitimize_target): Ditto.
1957 (nds32_legitimize_argument): Ditto.
1958 (nds32_check_constant_argument): Ditto.
1959 (nds32_expand_unop_builtin): Ditto.
1960 (nds32_expand_unopimm_builtin): Ditto.
1961 (nds32_expand_binop_builtin): Ditto.
1962 (nds32_builtin_decl_impl): Ditto.
1963 (builtin_description): Ditto.
1964 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1965 (nds32_init_builtins_impl): Ditto.
1966 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1967 (nds32_builtin_decl): New.
1968 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1969 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1971 2018-03-02 Jeff Law <law@redhat.com>
1973 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1974 (redundant_insn, fill_simple_delay_slots): Likewise.
1975 (fill_slots_from_thread): Likewise.
1976 * resource.c (mark_referenced_resources): Likewise.
1977 (mark_set_resources, find_dead_or_set_registers): Likewise.
1979 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1981 * substring-locations.h (format_warning_va): Formatting fix for
1983 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1985 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1986 * substring-locations.c: Include intl.h.
1987 (format_warning_va): Turned into small wrapper around
1988 format_warning_n_va, renamed to ...
1989 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1990 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1992 (format_warning_at_substring_n): New function.
1993 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1994 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1995 format_warning_at_substring with just a shorter name instead of
1996 const function pointer.
1997 (fmtwarn_n): New function.
1998 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1999 appropriate, get rid of all the fmtstr temporaries, move conditionals
2000 with G_() wrapped string literals directly into fmtwarn arguments,
2001 cast dir.len to (int), formatting fixes.
2003 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2005 * doc/invoke.texi: Remove "Cilk Plus" references.
2007 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2008 Richard Biener <rguenther@suse.de>
2011 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2012 for error or warning attributes if CALL_FROM_THUNK_P is set.
2015 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2018 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2019 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2022 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2023 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2025 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2026 instead of -1U in last predictors element's probability member.
2028 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2031 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2032 arguments if they are comparable.
2034 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2036 PR tree-optimization/84634
2037 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2038 masks and masked_loop_p with a single loop_masks, making sure it's
2039 null for bb vectorization.
2041 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2043 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2044 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2045 loop->force_vectorize to check whether there is no alias.
2047 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2050 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2052 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2054 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2055 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2056 instead of a loop around prev_real_insn.
2057 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2061 * config/i386/i386.c (ix86_print_operand): Use conditional
2062 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2065 2018-03-02 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/84427
2068 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2069 (bitmap_set_subtract_values): Rewrite to handle multiple
2072 (prune_clobbered_mems): Likewise.
2073 (phi_translate): Take edge instead of pred/phiblock.
2074 (phi_translate_1): Likewise.
2075 (phi_translate_set): Likewise. Insert all translated
2076 exprs for a value into the set, keeping possibly multiple
2077 expressions per value.
2078 (compute_antic_aux): Adjust for phi_translate changes.
2079 When intersecting union the expressions and prune those
2080 not in the final value set, keeping possibly multiple
2081 expressions per value. Do not use value-insertion
2082 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2083 all expressions. Add verification that the value-sets
2084 only shrink during iteration.
2085 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2086 (do_pre_regular_insertion): Likewise.
2087 (do_pre_partial_partial_insertion): Likewise.
2089 2018-03-02 Richard Biener <rguenther@suse.de>
2092 * config/darwin.c (saved_debug_info_level): New static global.
2093 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2094 (darwin_asm_lto_end): Restore debug info generation settings.
2096 2018-03-01 Martin Liska <mliska@suse.cz>
2099 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2102 2018-03-01 Richard Biener <rguenther@suse.de>
2105 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2106 type annotation with LTO when debug was disabled at compile-time.
2108 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2110 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2112 (mips_final_postscan_insn): Likewise.
2114 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2116 PR rtl-optimization/84528
2117 * alias.c (init_alias_target): Add commentary.
2118 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2119 a unique base value if the frame pointer is not eliminated
2120 to the stack pointer.
2122 2018-03-01 Tom de Vries <tom@codesourcery.com>
2124 PR rtl-optimization/83327
2125 * lra-int.h (hard_regs_spilled_into): Declare.
2126 * lra.c (hard_regs_spilled_into): Define.
2127 (init_reg_info): Init hard_regs_spilled_into.
2128 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2129 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2130 (process_bb_lives): Handle hard_regs_spilled_into.
2131 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2132 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2134 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2136 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2137 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2138 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2139 * config/rs6000/aix72.h: New file.
2141 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2143 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2144 instead of warning_at with conditional singular and plural messages
2148 * stor-layout.c (update_alignment_for_field): For
2149 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2150 && !DECL_PACKED (field), do the alignment update, just use
2151 only desired_align instead of MAX (type_align, desired_align)
2153 (place_field): Don't do known_align < desired_align handling
2154 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2155 is non-NULL, instead do it after rli->prev_field handling and
2156 only if not within a bitfield word. For DECL_PACKED (field)
2157 use type_align of BITS_PER_UNIT.
2159 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2161 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2162 superfluous parentheses and trailing spaces.
2164 2018-02-28 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/84584
2167 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2168 SCoPs with fake exit edge.
2170 2018-02-28 Martin Liska <mliska@suse.cz>
2173 * timevar.c (timer::print): Fix format to properly print 100%
2176 2018-02-28 Richard Biener <rguenther@suse.de>
2179 * genmatch.c (capture_info::walk_match): Do not mark
2180 captured expressions without operands as expr_p given
2181 they act more like predicates and should be subject to
2182 "lost tail" side-effect preserving.
2184 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2186 PR rtl-optimization/81611
2187 * auto-inc-dec.c (attempt_change): Move dead note from
2188 mem_insn if it's the next use of regno
2189 (find_address): Take address use of reg holding
2190 non-incremented value. Add parm to limit search to the named
2192 (merge_in_block): Attempt to use a mem insn that is the next
2193 use of the original regno.
2195 2018-02-27 Martin Sebor <msebor@redhat.com>
2198 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2199 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2201 2018-02-27 Martin Sebor <msebor@redhat.com>
2203 PR translation/84207
2204 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2205 n argument to unsigned HOST_WIDE_INT.
2206 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2207 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2208 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2209 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2211 2018-02-27 Richard Biener <rguenther@suse.de>
2213 PR tree-optimization/84512
2214 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2215 Do not use the estimate returned from record_stmt_cost for
2216 the scalar iteration cost but sum properly using add_stmt_cost.
2218 2018-02-27 Richard Biener <rguenther@suse.de>
2220 PR tree-optimization/84466
2221 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2222 Adjust last change to less strictly validate use operands.
2224 2018-02-27 Martin Liska <mliska@suse.cz>
2226 PR gcov-profile/84548
2227 * gcov.c (process_file): Allow partial overlap and consider it
2228 also as group functions.
2229 (output_lines): Properly calculate range of lines for a group.
2231 2018-02-27 Martin Liska <mliska@suse.cz>
2233 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2234 'ggc' suffixes. Change first column width.
2235 (timer::print): Fix formatting of the column.
2237 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2239 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2240 preserve inline entry blocks for the sake of debug inline
2241 entry point markers alone.
2242 (remove_unused_locals): Suggest in comments a better place to
2243 force the preservation of inline entry blocks that are
2244 otherwise unused, but do not preserve them.
2246 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2248 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2250 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2253 * config/i386/constraints.md (Bs): Replace
2254 ix86_indirect_branch_register with
2255 TARGET_INDIRECT_BRANCH_REGISTER.
2257 * config/i386/i386.md (indirect_jump): Likewise.
2258 (tablejump): Likewise.
2259 (*sibcall_memory): Likewise.
2260 (*sibcall_value_memory): Likewise.
2261 Peepholes of indirect call and jump via memory: Likewise.
2262 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2263 (*sibcall_value_GOT_32): Likewise.
2264 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2265 (GOT_memory_operand): Likewise.
2266 (call_insn_operand): Likewise.
2267 (sibcall_insn_operand): Likewise.
2268 (GOT32_symbol_operand): Likewise.
2269 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2271 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2273 PR rtl-optimization/83496
2274 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2275 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2276 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2277 redundant insn, if any.
2278 (relax_delay_slots): Likewise.
2279 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2281 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2283 PR tree-optimization/83965
2284 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2285 that grouped statements are part of a reduction chain. Return
2286 true if the statement is not marked as a reduction itself but
2288 (vect_recog_dot_prod_pattern): Don't check whether the statement
2289 is part of a group here.
2290 (vect_recog_sad_pattern): Likewise.
2291 (vect_recog_widen_sum_pattern): Likewise.
2293 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2296 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2298 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2301 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2303 (ix86_output_indirect_function_return): New prototype.
2304 (ix86_split_simple_return_pop_internal): Likewise.
2305 * config/i386/i386.c (indirect_return_via_cx): New.
2306 (indirect_return_via_cx_bnd): Likewise.
2307 (indirect_thunk_name): Handle return va CX_REG.
2308 (output_indirect_thunk_function): Create alias for
2309 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2310 (ix86_output_indirect_jmp): Remove the bool argument.
2311 (ix86_output_indirect_function_return): New function.
2312 (ix86_split_simple_return_pop_internal): Likewise.
2313 * config/i386/i386.md (*indirect_jump): Don't pass false
2314 to ix86_output_indirect_jmp.
2315 (*tablejump_1): Likewise.
2316 (simple_return_pop_internal): Change it to define_insn_and_split.
2317 Call ix86_split_simple_return_pop_internal to split it for
2319 (simple_return_indirect_internal): Call
2320 ix86_output_indirect_function_return instead of
2321 ix86_output_indirect_jmp.
2323 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2326 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2327 memset and value initialization afterwards.
2329 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2331 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2333 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2336 * common/config/aarch64/aarch64-common.c
2337 (aarch_option_optimization_table[]): Switch
2338 off fomit-frame-pointer
2340 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2341 Chung-Ju Wu <jasonwucj@gmail.com>
2343 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2345 (store_multiple): Ditto.
2347 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2349 * config.gcc: Add --with-cpu support for nds32 target.
2350 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2351 * config/nds32/nds32.opt: Add -mcpu= option.
2353 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2355 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2356 isel=yes): Warn for these deprecated options.
2358 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2360 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2361 ISA_2_5_MASKS_EMBEDDED.
2363 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2365 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2366 p->max as pointers rather than using iterative_hash_expr.
2368 2018-02-23 Carl Love <cel@us.ibm.com>
2370 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2371 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2373 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2374 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2375 P8V_BUILTIN_VEC_VUNSIGNED2.
2377 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2380 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2381 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2383 * lra-constraints.c (curr_insn_transform): Set up
2384 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2386 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2387 * lra-eliminations.c (spill_pseudos): Ditto.
2388 (process_insn_for_elimination): Ditto.
2389 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2390 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2391 LRA_NON_CLOBBERED_ALT.
2393 2018-02-22 Martin Sebor <msebor@redhat.com>
2395 PR tree-optimization/84480
2396 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2397 to maybe_diag_stxncpy_trunc. Call it.
2398 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2399 from gimple_fold_builtin_strcpy. Print inlining stack.
2400 (handle_builtin_stxncpy): Print inlining stack.
2401 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2403 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2406 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2407 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2408 and -fcheck-pointer-bounds are used together.
2409 (indirect_thunk_prefix): New enum.
2410 (indirect_thunk_need_prefix): New function.
2411 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2412 "_nt" instead of "_bnd" for NOTRACK prefix.
2413 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2414 (output_indirect_thunk_function): Likewise.
2416 (ix86_code_end): Update output_indirect_thunk_function calls.
2417 (ix86_output_indirect_branch_via_reg): Replace
2418 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2419 (ix86_output_indirect_branch_via_push): Likewise.
2420 (ix86_output_function_return): Likewise.
2421 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2422 incompatible with -fcf-protection=branch and
2423 -fcheck-pointer-bounds.
2425 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2428 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2429 Change gcc_assert call to output_operand_lossage.
2431 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2433 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2435 2018-02-22 DJ Delorie <dj@redhat.com>
2436 Sebastian Perta <sebastian.perta@renesas.com>
2437 Oleg Endo <olegendo@gcc.gnu.org>
2439 * config/rx/rx.c (rx_rtx_costs): New function.
2440 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2442 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2444 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2446 2018-02-22 Martin Liska <mliska@suse.cz>
2449 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2450 Add "native" as a possible value.
2452 2018-02-22 Martin Liska <mliska@suse.cz>
2455 * config/i386/i386.c (ix86_option_override_internal):
2456 Add "native" as a possible value for -march and -mtune.
2458 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2461 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2462 to all type variants.
2464 PR tree-optimization/84503
2465 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2466 width as info->bitpos + info->bitsize - start.
2467 (merged_store_group::merge_overlapping): Simplify width computation.
2468 (check_no_overlap): New function.
2469 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2470 start + width and last_order of the group, fail if check_no_overlap
2472 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2473 to group if check_no_overlap fails.
2475 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2477 * config/rs6000/altivec.md: Delete contraint arguments to
2478 define_expand, define_split, and define_peephole2, and in
2479 define_insn_and_split if always unused.
2480 * config/rs6000/darwin.md: Ditto.
2481 * config/rs6000/dfp.md: Ditto.
2482 * config/rs6000/rs6000.md: Ditto.
2483 * config/rs6000/sync.md: Ditto.
2484 * config/rs6000/vector.md: Ditto.
2485 * config/rs6000/vsx.md: Ditto.
2487 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2489 * config/rs6000/altivec.md: Write output control strings as braced
2490 blocks instead of double-quoted strings.
2491 * config/rs6000/darwin.md: Ditto.
2492 * config/rs6000/rs6000.md: Ditto.
2493 * config/rs6000/vector.md: Ditto.
2494 * config/rs6000/vsx.md: Ditto.
2496 2018-02-21 Jason Merrill <jason@redhat.com>
2498 PR c++/84314 - ICE with templates and fastcall attribute.
2499 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2501 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2503 * ipa-cp.c (determine_versionability): Fix comment typos.
2505 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2508 * ipa-cp.c (determine_versionability): Do not version functions caling
2511 2018-02-21 Martin Liska <mliska@suse.cz>
2514 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2515 Add "native" as a possible value.
2516 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2517 the macro when native cpu detection is available.
2519 2018-02-21 Martin Liska <mliska@suse.cz>
2522 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2523 Add "native" as a possible value.
2524 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2525 when native cpu detection is available.
2527 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2528 Martin Sebor <msebor@redhat.com>
2530 PR tree-optimization/84478
2531 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2533 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2534 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2535 support which is conservatively correct, for 2 only stay conservative
2536 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2537 argument to the 2 argument get_range_strlen, adjust 6 arg
2538 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2540 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2541 (gimple_fold_builtin_strlen): Pass true as last argument to
2544 2018-02-20 Martin Sebor <msebor@redhat.com>
2547 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2548 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2549 (builtin_memref::builtin_memref): Factor out parts into
2550 set_base_and_offset and call it.
2552 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2555 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2556 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2557 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2558 search at the associated MODE_INT.
2560 2018-02-20 Jeff Law <law@redhat.com>
2563 PR tree-optimization/81592
2565 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2566 for range data rather than using global data.
2567 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2568 range data rather than using global data.
2569 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2570 pass it to children as needed.
2571 (struct directive::fmtresult): Similarly.
2572 (struct directive::set_width): Similarly.
2573 (struct directive::set_precision): Similarly.
2574 (format_integer, format_directive, parse_directive): Similarly.
2575 (format_none): Accept unnamed vr_values parameter.
2576 (format_percent, format_floating, format_character): Similarly.
2577 (format_string, format_plain): Similarly.
2578 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2579 the EVRP range analyzer for range data rather than using global data.
2580 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2581 gimple-ssa-evrp-analyze.h
2582 (class sprintf_dom_walker): Add after_dom_children member function.
2583 Add evrp_range_analyzer member.
2584 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2585 range analyzer as needed.
2586 (sprintf_dom_walker::after_dom_children): New member function.
2587 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2589 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2590 (evrp_range_analyzer::pop_to_marker): Likewise.
2592 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2594 PR tree-optimization/84419
2595 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2596 with the required type if its current type is compatible but
2599 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2602 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2603 after vectorization.
2605 2018-02-20 Martin Liska <mliska@suse.cz>
2608 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2609 possible values if we don't have a hint.
2611 2018-02-20 Martin Liska <mliska@suse.cz>
2615 * final.c (shorten_branches): Build align_tab array with one
2617 * opts.c (finish_options): Add alignment option limit check.
2618 (MAX_CODE_ALIGN): Likewise.
2619 (MAX_CODE_ALIGN_VALUE): Likewise.
2620 * doc/invoke.texi: Document maximum allowed option value for
2621 all -falign-* options.
2623 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2626 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2627 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2628 * var-tracking.c (emit_note_insn_var_location): Remove all references
2629 to NOTE_INSN_CALL_ARG_LOCATION.
2630 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2631 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2632 Use copy_rtx_if_shared.
2633 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2634 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2635 (dwarf2out_var_location): Remove handling of
2636 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2638 * final.c (final_scan_insn): Remove all references to
2639 NOTE_INSN_CALL_ARG_LOCATION.
2640 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2641 before dumping final insns.
2642 * except.c (emit_note_eh_region_end): Remove all references to
2643 NOTE_INSN_CALL_ARG_LOCATION.
2644 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2645 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2646 * config/arc/arc.c (hwloop_optimize): Likewise.
2647 * config/arm/arm.c (create_fix_barrier): Likewise.
2648 * config/s390/s390.c (s390_chunkify_start): Likewise.
2649 * config/sh/sh.c (find_barrier): Likewise.
2650 * config/i386/i386.c (rest_of_insert_endbranch,
2651 ix86_seh_fixup_eh_fallthru): Likewise.
2652 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2653 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2654 * config/frv/frv.c (frv_function_prologue): Likewise.
2655 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2657 (note_outside_basic_block_p): Remove all references to
2658 NOTE_INSN_CALL_ARG_LOCATION.
2659 * gengtype.c (adjust_field_rtx_def): Likewise.
2660 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2662 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2663 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2666 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2669 PR tree-optimization/84452
2670 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2671 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2674 2018-02-19 Martin Liska <mliska@suse.cz>
2677 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2679 2018-02-19 Martin Liska <mliska@suse.cz>
2680 Richard Sandiford <richard.sandiford@linaro.org>
2682 PR tree-optimization/82491
2683 * gimple-fold.c (get_base_constructor): Make earlier bail out
2686 2018-02-19 Carl Love <cel@us.ibm.com>
2688 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2689 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2691 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2692 P8V_BUILTIN_VEC_NEG.
2694 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2696 * config/rl78/rl78.md (movdf): New define expand.
2698 2018-02-19 Martin Liska <mliska@suse.cz>
2701 * doc/invoke.texi: Fix typo.
2702 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2704 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2706 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2707 handle rs6000_single_float and rs6000_double_float specially for
2710 2018-02-16 Jeff Law <law@redhat.com>
2712 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2715 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2718 * ipa-inline.c (inline_small_functions): Fix a typo.
2720 2018-02-16 Nathan Sidwell <nathan@acm.org>
2722 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2724 2018-02-16 Carl Love <cel@us.ibm.com>
2726 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2727 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2728 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2729 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2730 expansion to P8V_BUILTIN_VEC_FLOAT2.
2732 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2734 PR rtl-optimization/70023
2735 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2736 src_regno into account.
2738 2018-02-16 Carl Love <cel@us.ibm.com>
2740 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2741 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2742 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2743 * config/rs6000/rs6000.c: Remove case statements for
2744 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2745 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2746 and P9V_BUILTIN_VEC_VINSERT4B.
2747 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2748 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2749 * config/rs6000/vsx.md:
2750 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2753 2018-02-16 Carl Love <cel@us.ibm.com>
2755 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2757 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2759 * config/rs6000/rs6000-c.c: Add the definitions for
2760 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2761 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2762 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2763 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2764 definition for insert4b and define insn *insert3b_internal.
2765 * doc/extend.texi: Add documentation for vec_extract4b.
2767 2018-02-16 Nathan Sidwell <nathan@acm.org>
2769 * doc/extend.texi (Backwards Compatibility): Mention friend
2770 injection. Note for-scope is deprecated.
2771 * doc/invoke.texi (-ffriend-injection): Deprecate.
2773 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2775 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2776 that moved to I2, also allow destinations that are a paradoxical
2777 subreg (instead of a normal reg).
2779 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2782 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2785 2018-02-16 Richard Biener <rguenther@suse.de>
2787 PR tree-optimization/84037
2788 PR tree-optimization/84016
2790 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2791 Adjust vec_construct for the fact we need additional higher latency
2792 128bit inserts for AVX256 and AVX512 vector builds.
2793 (ix86_add_stmt_cost): Scale vector construction cost for
2796 2018-02-16 Richard Biener <rguenther@suse.de>
2798 PR tree-optimization/84417
2799 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2800 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2801 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2803 2018-02-16 Martin Liska <mliska@suse.cz>
2806 * internal-fn.def (ASAN_CHECK): Set proper flags.
2807 (ASAN_MARK): Likewise.
2809 2018-02-16 Julia Koval <julia.koval@intel.com>
2811 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2812 from PTA_CANNONLAKE.
2814 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2817 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2818 Use ++iter rather than iter++ for std::list iterators.
2819 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2820 defer deleting them until all nodes in the forest are processed. Do
2821 free even leaf nodes. Change to_process into auto_vec.
2824 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2825 * vec.h (vec_default_construct): Use memset instead of placement new
2826 if BROKEN_VALUE_INITIALIZATION is defined.
2827 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2828 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2831 PR rtl-optimization/83723
2832 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2833 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2834 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2836 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2838 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2840 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2842 PR rtl-optimization/81443
2843 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2844 from inner REGs to paradoxical SUBREGs.
2846 2018-02-16 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/84399
2849 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2850 For operands we can analyze at their definition make sure we can
2851 analyze them at each use as well.
2853 2018-02-16 Richard Biener <rguenther@suse.de>
2855 PR tree-optimization/84190
2856 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2857 volatile accesses if the decl isn't volatile.
2859 2018-02-15 Jason Merrill <jason@redhat.com>
2861 PR c++/84314 - ICE with templates and fastcall attribute.
2862 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2863 TYPE_CANONICAL on an existing type.
2865 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2867 PR tree-optimization/84383
2868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2869 dstoff nor call operand_equal_p if dstbase is NULL.
2871 PR tree-optimization/84334
2872 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2873 also a CONSTANT_CLASS_P, punt.
2875 2018-02-14 Jim Wilson <jimw@sifive.com>
2877 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2878 first SMALL_OPERAND check. New local min_second_step. Move assert
2879 to where locals are set. Add TARGET_RVC support.
2880 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2882 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2884 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2886 2018-02-14 Martin Sebor <msebor@redhat.com>
2888 PR tree-optimization/83698
2889 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2890 arrays constrain the offset range to their bounds.
2891 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2892 (builtin_access::overlap): Avoid setting the size of overlap if it's
2894 (maybe_diag_overlap): Also consider arrays when deciding what values
2895 of offsets to include in diagnostics.
2897 2018-02-14 Martin Sebor <msebor@redhat.com>
2900 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2901 that correspond to the kind of a declaration.
2903 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2906 * config/pa/pa.md: Load address of PIC label using the linkage table
2907 if the label is nonlocal.
2909 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2911 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2912 warning message if user requests -maltivec=be.
2913 * doc/invoke.texi: Document deprecation of -maltivec=be.
2915 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2918 * config/rs6000/rs6000-c.c: Update definitions for
2919 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2920 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2922 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2925 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2926 add _get_ssp intrinsics. Remove argument from
2927 __builtin_ia32_rdssp[d|q].
2928 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2929 * config/i386/i386-builtin.def: Remove argument from
2930 __builtin_ia32_rdssp[d|q].
2931 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2932 ix86_expand_special_args_builtin for _rdssp[d|q].
2933 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2934 Clear register before usage.
2935 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2936 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2938 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2940 PR tree-optimization/84357
2941 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2942 operand 1 of an ARRAY_REF too.
2944 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2947 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2948 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2950 (set_of_reg): New struct.
2951 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2952 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2953 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2955 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2956 Split into bitclr, bitset, bitinvert patterns if appropriate.
2957 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2958 use rx_fuse_in_memory_bitop.
2959 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2960 to named insn, correct maximum insn length.
2962 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2965 * machmode.def: Define a complex mode for PARTIAL_INT.
2966 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2968 * doc/rtl.texi: Document CSPImode.
2969 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2971 (msp430_hard_regno_nregs_with_padding): Likewise.
2973 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2976 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2978 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2980 PR rtl-optimization/84169
2981 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2982 we generated a parallel as new i3 and we split that to new i2 and i3
2983 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2984 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2985 those to i2, not i1. Partially rewrite this scan code.
2987 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2990 * stor-layout.c (place_field): For variable length fields, adjust
2991 offset_align afterwards not just based on the field's alignment,
2992 but also on the size.
2995 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2996 of exps and logs in the use_exp2 case.
2998 2018-02-13 Jeff Law <law@redhat.com>
3000 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3003 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3006 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3010 * common.opt (gas-loc-support, gas-locview-support): New.
3011 (ginline-points, ginternal-reset-location-views): New.
3012 * doc/invoke.texi: Document them. Use @itemx where intended.
3013 (gvariable-location-views): Adjust.
3014 * target.def (reset_location_view): New.
3015 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3016 (TARGET_RESET_LOCATION_VIEW): New.
3017 * doc/tm.texi: Rebuilt.
3018 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3019 (dwarf2out_default_as_locview_support): New.
3020 (output_asm_line_debug_info): Use option variables.
3021 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3022 (output_loc_list): Likewise.
3023 (add_high_low_attributes): Check option variables.
3024 Don't output entry view attribute in strict mode.
3025 (gen_inlined_subroutine_die): Check option variables.
3026 (dwarf2out_inline_entry): Likewise.
3027 (init_sections_and_labels): Likewise.
3028 (dwarf2out_early_finish): Likewise.
3029 (maybe_reset_location_view): New, from...
3030 (dwarf2out_var_location): ... here. Call it.
3031 * debug.h (dwarf2out_default_as_loc_support): Declare.
3032 (dwarf2out_default_as_locview_support): Declare.
3033 * hooks.c (hook_int_rtx_insn_0): New.
3034 * hooks.h (hook_int_rtx_insn_0): Declare.
3035 * toplev.c (process_options): Take -gas-loc-support and
3036 -gas-locview-support from dwarf2out. Enable
3037 -gvariable-location-views by default only with locview
3038 assembler support. Enable -ginternal-reset-location-views by
3039 default only if the target defines the corresponding hook.
3040 Enable -ginline-points by default if location views are
3041 enabled; force it disabled if statement frontiers are
3043 * tree-inline.c (expand_call_inline): Check option variables.
3044 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3046 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3048 PR tree-optimization/84321
3049 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3050 handling. Also check whether the anti-range contains any values
3051 that satisfy the mask; switch to a VR_RANGE if not.
3053 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3056 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3058 2018-02-13 Martin Jambor <mjambor@suse.cz>
3061 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3062 of call statements, also set location of a load to a temporary.
3064 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3066 * config/rl78/rl78.c (add_vector_labels): New function.
3067 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3068 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3069 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3070 which checks that no arguments are passed.
3071 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3072 * doc/extend.texi: Documentation for the new attribute.
3074 2018-02-13 Andreas Schwab <schwab@suse.de>
3076 * config/riscv/linux.h (CPP_SPEC): Define.
3078 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3081 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3082 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3083 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3084 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3085 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3086 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3087 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3088 temporarily for AES and PCLMUL builtins.
3090 PR tree-optimization/84339
3091 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3092 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3096 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3097 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3098 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3100 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3102 * omp-simd-clone.h: New file.
3103 * omp-simd-clone.c: Include omp-simd-clone.h.
3104 (expand_simd_clones): No longer static.
3105 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3106 cgraph.h and omp-simd-clone.h.
3107 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3108 (vect_recog_widen_shift_pattern): Formatting fix.
3109 (vect_pattern_recog_1): Don't check optab for calls.
3112 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3113 operands[2] into a REG before using gen_lowpart on it.
3115 2018-02-12 Jeff Law <law@redhat.com>
3118 * config/sh/sh.c (find_barrier): Consider a sibling call
3121 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3122 successfully back substituting a reg.
3124 2018-02-12 Richard Biener <rguenther@suse.de>
3126 PR tree-optimization/84037
3127 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3128 parameter, move visited init to caller.
3129 (vect_slp_analyze_operations): Separate cost from validity
3130 check, initialize visited once for all instances.
3131 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3133 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3135 (vect_model_promotion_demotion_cost): Likewise.
3136 (vectorizable_bswap): Guard cost modeling with !slp_node
3137 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3139 (vectorizable_call): Likewise.
3140 (vectorizable_conversion): Likewise.
3141 (vectorizable_assignment): Likewise.
3142 (vectorizable_shift): Likewise.
3143 (vectorizable_operation): Likewise.
3144 (vectorizable_store): Likewise.
3145 (vectorizable_load): Likewise.
3146 (vectorizable_condition): Likewise.
3147 (vectorizable_comparison): Likewise.
3149 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3152 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3153 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3154 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3156 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3159 * params.def (inline-min-speedup): Increase from 8 to 15.
3160 (max-inline-insns-auto): Decrease from 40 to 30.
3161 * ipa-split.c (consider_split): Add some buffer for function to
3162 be considered inlining candidate.
3163 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3166 2018-02-12 Richard Biener <rguenther@suse.de>
3168 PR tree-optimization/84037
3169 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3170 matched stmts if we cannot swap the non-matched ones.
3172 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3174 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3175 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3176 _mm_maskz_scalef_round_ss): New intrinsics.
3177 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3178 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3179 __builtin_ia32_scalefss_round): Remove.
3180 (__builtin_ia32_scalefsd_mask_round,
3181 __builtin_ia32_scalefss_mask_round): New intrinsics.
3182 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3183 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3184 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3185 "<round_constraint>")): Changed to ...
3186 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3187 "<round_scalar_constraint>")): ... this.
3188 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3189 %0, %1, %2<round_op3>}"): Changed to ...
3190 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3191 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3192 %2<round_scalar_mask_op3>}"): ... this.
3193 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3195 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3197 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3198 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3199 (_mm_maskz_sqrt_round_ss): New intrinsics.
3200 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3201 (__builtin_ia32_sqrtsd_mask_round)
3202 (__builtin_ia32_sqrtss_mask_round): New builtins.
3203 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3204 (__builtin_ia32_sqrtss_round): Remove.
3205 (__builtin_ia32_sqrtsd_mask_round)
3206 (__builtin_ia32_sqrtss_mask_round): New builtins.
3207 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3208 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3209 ((match_operand:VF_128 1 "vector_operand"
3210 "xBm,<round_constraint>")): Changed to ...
3211 ((match_operand:VF_128 1 "vector_operand"
3212 "xBm,<round_scalar_constraint>")): ... this.
3213 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3214 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3215 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3216 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3217 %<iptr>1<round_scalar_mask_op3>}): ... this.
3218 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3219 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3221 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3224 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3225 Cast vec_cmpeq result to correct type.
3226 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3227 Cast vec_cmpgt result to correct type.
3229 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3231 * final.c (final_scan_insn_1): Renamed from...
3232 (final_scan_insn): ... this. New wrapper, to recover
3233 seen from the outermost call in recursive ones.
3234 * config/sparc/sparc.c (output_return): Drop seen from call.
3235 (output_sibcall): Likewise.
3236 * config/visium/visium.c (output_branch): Likewise.
3238 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3240 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3243 2018-02-10 Alan Modra <amodra@gmail.com>
3246 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3247 Specify LR as an input.
3249 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3252 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3253 remove_member_access_dummy_vars): New functions.
3254 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3255 lower_omp_1, execute_lower_omp): Use them.
3257 PR rtl-optimization/84308
3258 * shrink-wrap.c (spread_components): Release todo vector.
3260 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3262 PR rtl-optimization/57193
3263 * ira-color.c (struct allocno_color_data): Add member
3264 conflict_allocno_hard_prefs.
3265 (update_conflict_allocno_hard_prefs): New.
3266 (bucket_allocno_compare_func): Add a preference based on
3267 conflict_allocno_hard_prefs.
3268 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3269 (color_allocnos): Remove a dead code. Initiate
3270 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3272 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3275 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3276 constraint from =wa to wa. Avoid a subreg on the output operand,
3277 instead use a pseudo and subreg it in a move.
3278 (p9_xxbrd_<mode>): Changed to ...
3279 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3280 (p9_xxbrd_v2df): New expander.
3281 (p9_xxbrw_<mode>): Changed to ...
3282 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3283 (p9_xxbrw_v4sf): New expander.
3285 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3287 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3289 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3292 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3293 multiply in 32-bit mode.
3294 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3295 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3298 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3300 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3301 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3302 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3303 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3305 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3308 * dwarf2out.c (is_trivial_indirect_ref): New function.
3309 (dwarf2out_late_global_decl): Do not generate a location
3310 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3311 and that are not defined in the current unit.
3313 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3315 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3316 instead of a libcall for UNORDERED.
3318 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3321 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3322 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3324 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3327 * config/s390/s390.c (s390_set_current_function): Invoke
3328 s390_indirect_branch_settings also if fndecl didn't change.
3330 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3332 * config/rs6000/rs6000.md (blockage): Set length to zero.
3334 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3336 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3338 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3341 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3342 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3346 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3347 PARALLEL incoming that failed vt_get_decl_and_offset check.
3350 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3351 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3353 (get_variable_section): For decls in named .bss* sections pass true as
3354 second argument to bss_initializer_p.
3356 2018-02-09 Marek Polacek <polacek@redhat.com>
3357 Jakub Jelinek <jakub@redhat.com>
3360 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3361 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3362 Sync some changes from cxx_fold_indirect_ref.
3364 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3366 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3368 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3369 (BLOCK_INLINE_ENTRY_LABEL): New.
3370 (dwarf2out_var_location): Disregard inline entry markers.
3371 (inline_entry_data): New struct.
3372 (inline_entry_data_hasher): New hashtable type.
3373 (inline_entry_data_hasher::hash): New.
3374 (inline_entry_data_hasher::equal): New.
3375 (inline_entry_data_table): New variable.
3376 (add_high_low_attributes): Add DW_AT_entry_pc and
3377 DW_AT_GNU_entry_view attributes if a pending entry is found
3378 in inline_entry_data_table. Add old entry_pc attribute only
3379 if debug nonbinding markers are disabled.
3380 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3381 markers are enabled.
3382 (block_within_block_p, dwarf2out_inline_entry): New.
3383 (dwarf2out_finish): Check that no entries remained in
3384 inline_entry_data_table.
3385 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3386 (final_scan_insn, notice_source_line): Likewise.
3387 (rest_of_clean_state): Skip inline entry markers.
3388 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3390 * gimple.c (gimple_build_debug_inline_entry): New.
3391 * gimple.h (enum gimple_debug_subcode): Add
3392 GIMPLE_DEBUG_INLINE_ENTRY.
3393 (gimple_build_debug_inline_entry): Declare.
3394 (gimple_debug_inline_entry_p): New.
3395 (gimple_debug_nonbind_marker_p): Adjust.
3396 * insn-notes.def (INLINE_ENTRY): New.
3397 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3398 inline entry marker notes.
3399 (print_insn): Likewise.
3400 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3401 (INSN_DEBUG_MARKER_KIND): Likewise.
3402 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3403 * tree-inline.c (expand_call_inline): Build and insert
3404 debug_inline_entry stmt.
3405 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3406 inline entry blocks early, if nonbind markers are enabled.
3407 (dump_scope_block): Dump fragment info.
3408 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3409 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3410 (gimple_build_debug_inline_entry): New.
3411 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3412 Enable/disable inline entry points too.
3413 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3414 (DEBUG_INSN): Describe inline entry markers.
3416 * common.opt (gvariable-location-views): New.
3417 (gvariable-location-views=incompat5): New.
3418 * config.in: Rebuilt.
3419 * configure: Rebuilt.
3420 * configure.ac: Test assembler for view support.
3421 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3422 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3423 * dwarf2out.c (var_loc_view): New typedef.
3424 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3425 (dwarf2out_locviews_in_attribute): New.
3426 (dwarf2out_locviews_in_loclist): New.
3427 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3428 (enum dw_line_info_opcode): Add LI_adv_address.
3429 (struct dw_line_info_table): Add view.
3430 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3431 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3432 (zero_view_p): New variable.
3433 (ZERO_VIEW_P): New macro.
3434 (output_asm_line_debug_info): New.
3435 (struct var_loc_node): Add view.
3436 (add_AT_view_list, AT_loc_list): New.
3437 (add_var_loc_to_decl): Add view param. Test it against last.
3438 (new_loc_list): Add view params. Record them.
3439 (AT_loc_list_ptr): Handle loc and view lists.
3440 (view_list_to_loc_list_val_node): New.
3441 (print_dw_val): Handle dw_val_class_view_list.
3442 (size_of_die): Likewise.
3443 (value_format): Likewise.
3444 (loc_list_has_views): New.
3445 (gen_llsym): Set vl_symbol too.
3446 (maybe_gen_llsym, skip_loc_list_entry): New.
3447 (dwarf2out_maybe_output_loclist_view_pair): New.
3448 (output_loc_list): Output view list or entries too.
3449 (output_view_list_offset): New.
3450 (output_die): Handle dw_val_class_view_list.
3451 (output_dwarf_version): New.
3452 (output_compilation_unit_header): Use it.
3453 (output_skeleton_debug_sections): Likewise.
3454 (output_rnglists, output_line_info): Likewise.
3455 (output_pubnames, output_aranges): Update version comments.
3456 (output_one_line_info_table): Output view numbers in asm comments.
3457 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3458 Call maybe_gen_llsym.
3459 (loc_list_from_tree_1): Adjust.
3460 (add_AT_location_description): Create view list attribute if
3461 needed, check it's absent otherwise.
3462 (convert_cfa_to_fb_loc_list): Adjust.
3463 (maybe_emit_file): Call output_asm_line_debug_info for test.
3464 (dwarf2out_var_location): Reset views as needed. Precompute
3465 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3466 attribute. Set view.
3467 (new_line_info_table): Reset next view.
3468 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3469 (dwarf2out_source_line): Likewise. Output view resets and labels to
3470 the assembler, or select appropriate line info opcodes.
3471 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3472 (optimize_string_length): Catch it. Adjust.
3473 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3474 dw_val_class_view_list, and remove it if no longer needed.
3475 (hash_loc_list): Hash view numbers.
3476 (loc_list_hasher::equal): Compare them.
3477 (optimize_location_lists): Check whether a view list symbol is
3478 needed, and whether the locview attribute is present, and
3479 whether they match. Remove the locview attribute if no longer
3481 (index_location_lists): Call skip_loc_list_entry for test.
3482 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3483 Use output_dwarf_version.
3484 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3485 (struct dw_val_node): Add val_view_list.
3486 * final.c (SEEN_NEXT_VIEW): New.
3487 (set_next_view_needed): New.
3488 (clear_next_view_needed): New.
3489 (maybe_output_next_view): New.
3490 (final_start_function): Rename to...
3491 (final_start_function_1): ... this. Take pointer to FIRST,
3492 add SEEN parameter. Emit param bindings in the initial view.
3493 (final_start_function): Reintroduce SEEN-less interface.
3494 (final): Rename to...
3495 (final_1): ... this. Take SEEN parameter. Output final pending
3496 next view at the end.
3497 (final): Reintroduce seen-less interface.
3498 (final_scan_insn): Output pending next view before switching
3499 sections or ending a block. Mark the next view as needed when
3500 outputting variable locations. Notify debug backend of section
3501 changes, and of location view changes.
3502 (rest_of_handle_final): Adjust.
3503 * toplev.c (process_options): Autodetect value for debug variable
3504 location views option. Warn on incompat5 without -gdwarf-5.
3505 * doc/invoke.texi (gvariable-location-views): New.
3506 (gvariable-location-views=incompat5): New.
3507 (gno-variable-location-views): New.
3509 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3511 PR tree-optimization/84136
3512 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3513 that the result of find_edge is non-NULL.
3515 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3518 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3519 storing integer register in SImode. Fix cost of 256 and 512
3520 byte aligned SSE register store.
3522 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3524 * config/i386/i386.c (ix86_multiplication_cost): Fix
3525 multiplication cost for TARGET_AVX512DQ.
3527 2018-02-08 Marek Polacek <polacek@redhat.com>
3529 PR tree-optimization/84238
3530 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3533 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3535 PR tree-optimization/84265
3536 * tree-vect-stmts.c (vectorizable_store): Don't treat
3537 VMAT_CONTIGUOUS accesses as grouped.
3538 (vectorizable_load): Likewise.
3540 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3542 PR tree-optimization/81635
3543 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3544 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3545 (test_round_for_mask): New functions.
3546 (wide_int_cc_tests): Call test_round_for_mask.
3547 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3548 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3549 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3550 range returned by get_range_info.
3552 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3555 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3556 * symtab.c: Include builtins.h
3557 (symtab_node::output_to_lto_symbol_table_p): Move here
3558 from lto-streamer-out.c:output_symbol_p.
3559 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3560 (output_symbol_p): Move all logic to symtab.c
3561 (produce_symtab): Update.
3563 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3565 * config/s390/s390-opts.h (enum indirect_branch): Define.
3566 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3567 (s390_indirect_branch_via_thunk)
3568 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3569 (enum s390_indirect_branch_type): Define.
3570 * config/s390/s390.c (struct s390_frame_layout, struct
3571 machine_function): Remove.
3572 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3573 (indirect_branch_table_label_no, indirect_branch_table_name):
3575 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3576 (enum s390_indirect_branch_option): Define.
3577 (s390_return_addr_from_memory): New function.
3578 (s390_handle_string_attribute): New function.
3579 (s390_attribute_table): Add new attribute handler.
3580 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3581 (s390_indirect_branch_via_thunk): New function.
3582 (s390_indirect_branch_via_inline_thunk): New function.
3583 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3584 sibling call optimization for non z10 compiles.
3585 (s390_emit_call): Force indirect branch target to be a single
3586 register. Add r1 clobber for non-z10 compiles.
3587 (s390_emit_epilogue): Emit return jump via return_use expander.
3588 (s390_reorg): Handle JUMP_INSNs as execute targets.
3589 (s390_option_override_internal): Perform validity checks for the
3590 new command line options.
3591 (s390_indirect_branch_attrvalue): New function.
3592 (s390_indirect_branch_settings): New function.
3593 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3594 (s390_output_indirect_thunk_function): New function.
3595 (s390_code_end): Implement target hook.
3596 (s390_case_values_threshold): Implement target hook.
3597 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3599 * config/s390/s390.h (struct s390_frame_layout)
3600 (struct machine_function): Move here from s390.c.
3601 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3602 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3603 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3604 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3605 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3606 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3607 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3608 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3609 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3610 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3611 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3612 (mnemonic attribute): Add values which aren't recognized
3614 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3615 pattern for branch conversion. Fix mnemonic attribute.
3616 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3617 indirect branch via thunk if requested.
3618 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3619 ("*indirect_jump"): Disable for branch conversion using out of
3621 ("indirect_jump_via_thunk<mode>_z10")
3622 ("indirect_jump_via_thunk<mode>")
3623 ("indirect_jump_via_inlinethunk<mode>_z10")
3624 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3625 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3626 ("casesi_jump_via_inlinethunk<mode>_z10")
3627 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3628 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3629 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3630 ("*indirect2_jump"): Disable for branch conversion.
3631 ("casesi_jump"): Turn into expander and expand patterns for branch
3633 ("return_use"): New expander.
3634 ("*return"): Emit return via thunk and rename it to ...
3635 ("*return<mode>"): ... this one.
3636 * config/s390/s390.opt: Add new options and and enum for the
3639 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3641 * lra-constraints.c (match_reload): Unconditionally use
3642 gen_lowpart_SUBREG, rather than selecting between that
3643 and equivalent gen_rtx_SUBREG code.
3645 2018-02-08 Richard Biener <rguenther@suse.de>
3647 PR tree-optimization/84233
3648 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3649 changed flag instead of boguously re-using phi_inserted.
3651 2018-02-08 Martin Jambor <mjambor@suse.cz>
3653 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3654 static local variables.
3656 2018-02-08 Richard Biener <rguenther@suse.de>
3658 PR tree-optimization/84278
3659 * tree-vect-stmts.c (vectorizable_store): When looking for
3660 smaller vector types to perform grouped strided loads/stores
3661 make sure the mode is supported by the target.
3662 (vectorizable_load): Likewise.
3664 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3666 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3667 Increase LDP/STP opportunities by adding adjacent callee-saves.
3669 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3671 PR rtl-optimization/84068
3672 PR rtl-optimization/83459
3673 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3675 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3677 PR tree-optimization/84224
3678 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3679 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3682 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3685 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3686 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3687 count, remove one USE.
3689 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3691 * doc/install.texi (Configuration): Document the
3692 --with-long-double-format={ibm,ieee} PowerPC configuration
3696 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3697 Convert from define_expand to be define_insn_and_split. Rework
3698 float/double/_Float128 conversions to QI/HI/SImode to work with
3699 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3700 conversions to QI/HImode types did a store and then a load to
3701 truncate the value. For conversions to VSX registers, don't split
3702 the insn, instead emit the code directly. Use the code iterator
3703 any_fix to combine signed and unsigned conversions.
3704 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3705 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3706 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3707 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3708 (fix_<mode>di2_hw): Likewise.
3709 (fixuns_<mode>di2_hw): Likewise.
3710 (fix_<mode>si2_hw): Likewise.
3711 (fixuns_<mode>si2_hw): Likewise.
3712 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3713 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3714 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3715 fix<uns>_trunc<SFDF:mode>si2_p8.
3716 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3718 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3719 (fix<uns>_<mode>_mem): Likewise.
3720 (fctiw<u>z_<mode>_mem): Likewise.
3721 (fix<uns>_<mode>_mem): Likewise.
3722 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3723 the register allocator from doing a direct move to the GPRs to do
3724 a store, and instead use the ISA 3.0 store byte/half-word from
3725 vector register instruction. For IEEE 128-bit floating point,
3726 also optimize stores of 32-bit ints.
3727 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3729 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3731 * genextract.c (push_pathstr_operand): New function to support
3733 (walk_rtx): Call push_pathstr_operand.
3734 (print_path): Support [a-zA-Z].
3736 2018-02-07 Richard Biener <rguenther@suse.de>
3738 PR tree-optimization/84037
3739 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3740 (cse_and_gimplify_to_preheader): Declare.
3741 (vect_get_place_in_interleaving_chain): Likewise.
3742 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3744 (_loop_vec_info::~_loop_vec_info): Delete it.
3745 (cse_and_gimplify_to_preheader): New function.
3746 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3747 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3748 (vectorizable_load): Likewise. For grouped stores always base
3749 the IV on the first element.
3750 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3751 condition before gimplifying.
3753 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3755 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3756 *DIV_EXPR and *MOD_EXPR.
3758 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3761 * config/i386/i386.c (ix86_option_override_internal): Mask out
3762 the CF_SET bit when checking -fcf-protection.
3764 2018-02-07 Tom de Vries <tom@codesourcery.com>
3767 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3770 2018-02-07 Richard Biener <rguenther@suse.de>
3772 PR tree-optimization/84204
3773 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3776 PR tree-optimization/84205
3777 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3778 special-case isl_ast_op_zdiv_r.
3780 PR tree-optimization/84223
3781 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3782 Only add conditions from within the region.
3783 (gather_bbs::after_dom_children): Adjust.
3785 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3788 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3789 * config/avr/avr.md: Only post-reload split REG-REG moves if
3790 either register is GENERAL_REG_P.
3792 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3794 PR tree-optimization/84235
3795 * tree-ssa-scopedtables.c
3796 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3797 if the subtraction is performed in floating point type where NaNs are
3798 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3799 build 1. Formatting fix.
3801 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3804 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3805 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3806 and skip it regardless of bb boundaries. Use CALL_P macro,
3807 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3808 unnecessarily, formatting fix.
3810 2018-02-06 Michael Collison <michael.collison@arm.com>
3812 * config/arm/thumb2.md:
3813 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3814 (*thumb_mov_notscc): Ditto.
3816 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3819 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3820 unsigned_fix, not "s".
3822 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3824 * configure.ac (gcc_fn_eh_frame_ro): New function.
3825 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3826 correct .eh_frame permissions.
3827 * configure: Regenerate.
3829 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3831 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3834 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3836 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3837 Display warning message for -mno-speculate-indirect-jumps.
3839 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3841 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3843 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3845 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3847 PR tree-optimization/84225
3848 * tree-eh.c (find_trapping_overflow): Only call
3849 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3851 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3854 * config/i386/i386.c: Reimplement the check of possible options
3855 -mibt/-mshstk conbination. Change error messages.
3856 * doc/invoke.texi: Fix a typo: remove extra '='.
3858 2018-02-06 Marek Polacek <polacek@redhat.com>
3860 PR tree-optimization/84228
3861 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3863 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3866 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3867 emitted arch directives.
3868 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3869 __ARM_FEATURE_COPROC before changing architectures.
3871 2018-02-06 Richard Biener <rguenther@suse.de>
3873 * config/i386/i386.c (print_reg): Fix typo.
3874 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3876 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3878 * configure: Regenerate.
3880 2018-02-05 Martin Sebor <msebor@redhat.com>
3882 PR tree-optimization/83369
3883 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3886 2018-02-05 Martin Liska <mliska@suse.cz>
3888 * doc/invoke.texi: Cherry-pick upstream r323995.
3890 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3892 * ira.c (ira_init_register_move_cost): Adjust comment.
3894 2018-02-05 Martin Liska <mliska@suse.cz>
3896 PR gcov-profile/84137
3897 * doc/gcov.texi: Fix typo in documentation.
3899 2018-02-05 Martin Liska <mliska@suse.cz>
3901 PR gcov-profile/83879
3902 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3903 using dlopen functionality.
3905 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3907 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3908 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3909 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3910 _mm_maskz_range_round_ss): New intrinsics.
3911 (__builtin_ia32_rangesd128_round)
3912 (__builtin_ia32_rangess128_round): Remove.
3913 (__builtin_ia32_rangesd128_mask_round,
3914 __builtin_ia32_rangess128_mask_round): New builtins.
3915 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3916 __builtin_ia32_rangess128_round): Remove.
3917 (__builtin_ia32_rangesd128_mask_round,
3918 __builtin_ia32_rangess128_mask_round): New builtins.
3919 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3920 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3921 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3922 "<round_saeonly_constraint>")): Changed to ...
3923 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3924 "<round_saeonly_scalar_constraint>")): ... this.
3925 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3926 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3927 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3928 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3929 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3931 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3933 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3935 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3936 Remove all values except native, 8540 and 8548.
3938 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3940 * config/i386/i386.c (ix86_output_function_return): Pass
3941 INVALID_REGNUM, instead of -1, as invalid register number to
3942 indirect_thunk_name and output_indirect_thunk.
3944 2018-02-02 Julia Koval <julia.koval@intel.com>
3946 * config.gcc: Add -march=icelake.
3947 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3948 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3949 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3950 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3951 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3952 (processor_target_table): Add icelake.
3953 (ix86_option_override_internal): Handle new PTAs.
3954 (get_builtin_code_for_version): Handle icelake.
3955 (M_INTEL_COREI7_ICELAKE): New.
3956 (fold_builtin_cpu): Handle icelake.
3957 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3958 * doc/invoke.texi: Add -march=icelake.
3960 2018-02-02 Julia Koval <julia.koval@intel.com>
3962 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3963 to wide_int_bitmask.
3964 * wide-int-bitmask.h: New.
3966 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3969 * config/i386/i386.md: Replace Pmode with word_mode in
3970 builtin_setjmp_setup and builtin_longjmp to support x32.
3972 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3976 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3978 (rs6000_supported_cpu_names): New static variable.
3979 (linux_cpu_translation_table): Likewise.
3980 (elf_platform) <cpu>: Define new static variable and use it.
3981 Translate kernel AT_PLATFORM name to canonical name if needed.
3982 Error if platform name is unknown.
3984 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3987 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3989 2018-02-01 Jeff Law <law@redhat.com>
3992 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3993 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3994 the scratch if RELEASE_VIA_POP is false.
3995 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3996 If we have to save a temporary register, decrement SIZE appropriately.
3997 Pass new arguments to release_scratch_register_on_entry.
3998 (ix86_adjust_stack_and_probe): Likewise.
3999 (ix86_emit_probe_stack_range): Pass new arguments to
4000 release_scratch_register_on_entry.
4002 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4004 PR rtl-optimization/84157
4005 * combine.c (change_zero_ext): Use REG_P predicate in
4006 front of HARD_REGISTER_P predicate.
4008 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4010 * config/avr/avr.c (avr_option_override): Move disabling of
4011 -fdelete-null-pointer-checks to...
4012 * common/config/avr/avr-common.c (avr_option_optimization_table):
4015 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4017 PR tree-optimization/81635
4018 * tree-data-ref.c (split_constant_offset_1): For types that
4019 wrap on overflow, try to use range info to prove that wrapping
4022 2018-02-01 Renlin Li <renlin.li@arm.com>
4025 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4027 (aarch64_register_move_cost): Likewise.
4028 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4030 (REG_CLASS_NAMES): Likewise.
4031 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4032 TAILCALL_ADDR_REGS. Remove IP registers.
4033 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4035 2018-02-01 Richard Biener <rguenther@suse.de>
4037 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4038 for specifying RPO order and allow NULL for that.
4039 * domwalk.c (dom_walker::dom_walker): Likewise.
4040 (dom_walker::walk): Handle NULL RPO order.
4041 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4043 (rewrite_update_dom_walker): Likewise.
4044 (mark_def_dom_walker): Likewise.
4046 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4048 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4049 (aarch64_maybe_expand_sve_subreg_move): Declare.
4050 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4051 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4053 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4054 that are semantically a reverse operation.
4055 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4056 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4057 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4059 (aarch64_can_change_mode_class): For big-endian, forbid changes
4060 between two SVE modes if they have different element sizes.
4062 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4064 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4065 the TImode handling for big-endian targets.
4067 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4069 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4070 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4072 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4073 Remove BSWAP handing for big-endian targets and use the form of
4074 LD1RQ appropariate for the mode.
4076 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4078 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4079 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4082 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4085 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4086 check for operands that need to go through aarch64_sve_reload_be.
4088 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4090 PR tree-optimization/81661
4091 PR tree-optimization/84117
4092 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4093 * tree-eh.c: Include gimplify.h.
4094 (find_trapping_overflow, replace_trapping_overflow,
4095 rewrite_to_non_trapping_overflow): New functions.
4096 * tree-vect-loop.c: Include tree-eh.h.
4097 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4098 * tree-data-ref.c: Include tree-eh.h.
4099 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4101 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4103 PR rtl-optimization/84123
4104 * combine.c (change_zero_ext): Check if hard register satisfies
4105 can_change_dest_mode before calling gen_lowpart_SUBREG.
4107 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4110 * ira.c (ira_init_register_move_cost): Remove assert.
4112 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4114 PR rtl-optimization/84071
4115 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4116 * doc/tm.texi: Regenerate.
4118 2018-01-31 Richard Biener <rguenther@suse.de>
4120 PR tree-optimization/84132
4121 * tree-data-ref.c (analyze_miv_subscript): Properly
4122 check whether evolution_function_is_affine_multivariate_p
4123 before calling gcd_of_steps_may_divide_p.
4125 2018-01-31 Julia Koval <julia.koval@intel.com>
4128 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4129 * config/i386/i386.md (rdpid_rex64) New.
4130 (rdpid): Make 32bit only.
4132 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4135 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4136 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4138 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4141 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4143 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4145 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4147 PR rtl-optimization/84071
4148 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4149 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4151 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4153 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4154 (arc_attribute_table): Add 'aux' attribute.
4155 (arc_in_small_data_p): Consider aux like variables.
4156 (arc_is_aux_reg_p): New function.
4157 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4158 (arc_get_aux_arg): New function.
4159 (prepare_move_operands): Handle aux-register access.
4160 (arc_handle_aux_attribute): New function.
4161 * doc/extend.texi (ARC Variable attributes): Add subsection.
4163 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4165 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4166 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4167 (arc_attribute_table): Add 'uncached' attribute.
4168 (arc_print_operand): Print '.di' flag for uncached memory
4170 (arc_in_small_data_p): Do not consider for small data the uncached
4172 (arc_is_uncached_mem_p): New function.
4173 * config/arc/predicates.md (compact_store_memory_operand): Check
4174 for uncached memory accesses.
4175 (nonvol_nonimm_operand): Likewise.
4176 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4178 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4181 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4182 falign-loops=): Add Optimization flag.
4184 2018-01-30 Jeff Law <law@redhat.com>
4187 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4188 INT_REGISTERS_SAVED. Check it prior to calling
4189 get_scratch_register_on_entry.
4190 (ix86_adjust_stack_and_probe): Similarly.
4191 (ix86_emit_probe_stack_range): Similarly.
4192 (ix86_expand_prologue): Corresponding changes.
4194 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4197 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4198 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4200 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4203 * lra-constraints.c (curr_insn_transform): Process AND in the
4206 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4208 PR rtl-optimization/83986
4209 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4210 dependence against last_pending_memory_flush in addition to
4213 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4215 PR tree-optimization/81611
4216 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4219 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4222 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4225 2018-01-30 Richard Biener <rguenther@suse.de>
4226 Jakub Jelinek <jakub@redhat.com>
4228 PR tree-optimization/84111
4229 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4230 inner loops added during recursion, as they don't have up-to-date
4233 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4236 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4237 (can_inline_edge_by_limits_p): ... here.
4238 (can_early_inline_edge_p, check_callers,
4239 update_caller_keys, update_callee_keys, recursive_inlining,
4240 add_new_edges_to_heap, speculation_useful_p,
4241 inline_small_functions,
4242 inline_small_functions, flatten_function,
4243 inline_to_all_callers_1): Update.
4245 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4247 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4250 2018-01-30 Richard Biener <rguenther@suse.de>
4252 PR tree-optimization/83008
4253 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4254 invariant and constant vector uses in stmts when they need
4257 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4260 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4261 * configure: Regenerate.
4263 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4265 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4267 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4268 Use gen_rtx_REG rather than gen_lowpart.
4270 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4272 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4273 rather than 0 when creating partial subregs.
4275 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4277 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4280 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4283 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4284 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4285 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4286 flags. This restores the settings used before the 2017-07-24.
4287 Turning off pre increment/decrement/modify allows IVOPTS to
4288 optimize DF/SF loops where the index is an int.
4290 2018-01-29 Richard Biener <rguenther@suse.de>
4291 Kelvin Nilsen <kelvin@gcc.gnu.org>
4294 * tree-vect-stmts.c (vectorizable_call): Don't call
4295 targetm.vectorize_builtin_md_vectorized_function if callee is
4298 2018-01-22 Carl Love <cel@us.ibm.com>
4300 * doc/extend.tex: Fix typo in second arg in
4301 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4303 2018-01-29 Richard Biener <rguenther@suse.de>
4305 PR tree-optimization/84086
4306 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4307 (flush_ssaname_freelist): When SSA names were released reset
4308 the SCEV hash table.
4310 2018-01-29 Richard Biener <rguenther@suse.de>
4312 PR tree-optimization/84057
4313 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4314 removed paths when removing edges.
4316 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4318 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4319 -mfunction-return=@var{choice}.
4321 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4324 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4326 (layout::print_source_line): Likewise.
4327 (test_get_line_width_without_trailing_whitespace): Add test cases.
4329 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4332 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4335 2018-01-26 Jim Wilson <jimw@sifive.com>
4337 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4339 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4342 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4344 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4345 and CMP + SUB-immediate -> SUBS.
4347 2018-01-26 Martin Sebor <msebor@redhat.com>
4349 PR tree-optimization/83896
4350 * tree-ssa-strlen.c (get_string_len): Rename...
4351 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4352 Avoid assuming length is constant.
4353 (handle_char_store): Use HOST_WIDE_INT for string length.
4355 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4358 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4359 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4361 2018-01-26 Richard Biener <rguenther@suse.de>
4363 PR rtl-optimization/84003
4364 * dse.c (record_store): Only record redundant stores when
4365 the earlier store aliases at least all accesses the later one does.
4367 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4369 PR rtl-optimization/83985
4370 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4371 REG_CFA_RESTORE insns.
4372 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4373 REG_CFA_RESTORE insns here.
4376 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4377 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4379 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4381 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4382 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4383 (arc_init): Likewise.
4384 (arc_override_options): Likewise.
4385 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4387 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4389 * config/arc/arc.h (TARGET_DBNZ): Define.
4390 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4391 properly set the tune attribute.
4392 (dbnz): Use TARGET_DBNZ guard.
4393 * config/arc/arc.opt (mtune): Add core3 option.
4395 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4397 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4398 recognize new pic like addresses.
4399 (arc_delegitimize_address): Clean up.
4401 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4403 * config/arc/arc-arches.def: Option mrf16 valid for all
4405 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4406 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4407 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4408 * config/arc/arc-tables.opt: Regenerate.
4409 * config/arc/arc.c (arc_conditional_register_usage): Handle
4410 reduced register file case.
4411 (arc_file_start): Set must have build attributes.
4412 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4414 * config/arc/arc.opt (mrf16): Add new option.
4415 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4416 * config/arc/genmultilib.awk: Handle new mrf16 option.
4417 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4418 * config/arc/t-multilib: Regenerate.
4419 * doc/invoke.texi (ARC Options): Document mrf16 option.
4421 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4423 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4424 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4425 (arc_attribute_table): Add 'secure_call' attribute.
4426 (arc_print_operand): Print secure call operand.
4427 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4429 (arc_is_secure_call_p): New function. * config/arc/arc.md
4430 (call_i): Add support for sjli instruction.
4431 (call_value_i): Likewise.
4432 * config/arc/constraints.md (Csc): New constraint.
4434 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4435 John Eric Martin <John.Martin@emmicro-us.com>
4437 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4438 * config/arc/arc.c (_arc_jli_section): New struct.
4439 (arc_jli_section): New type.
4440 (rc_jli_sections): New static variable.
4441 (arc_handle_jli_attribute): New function.
4442 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4443 (arc_file_end): New function.
4444 (TARGET_ASM_FILE_END): Define.
4445 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4446 (arc_add_jli_section): New function.
4447 (jli_call_scan): Likewise.
4448 (arc_reorg): Call jli_call_scan.
4449 (arc_output_addsi): Remove 'S' from printing asm operand.
4450 (arc_is_jli_call_p): New function.
4451 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4453 (movhi_insn): Likewise.
4454 (movsi_insn): Likewise.
4455 (movsi_set_cc_insn): Likewise.
4456 (loadqi_update): Likewise.
4457 (load_zeroextendqisi_update): Likewise.
4458 (load_signextendqisi_update): Likewise.
4459 (loadhi_update): Likewise.
4460 (load_zeroextendhisi_update): Likewise.
4461 (load_signextendhisi_update): Likewise.
4462 (loadsi_update): Likewise.
4463 (loadsf_update): Likewise.
4464 (movsicc_insn): Likewise.
4465 (bset_insn): Likewise.
4466 (bxor_insn): Likewise.
4467 (bclr_insn): Likewise.
4468 (bmsk_insn): Likewise.
4469 (bicsi3_insn): Likewise.
4470 (cmpsi_cc_c_insn): Likewise.
4471 (movsi_ne): Likewise.
4472 (movsi_cond_exec): Likewise.
4473 (clrsbsi2): Likewise.
4484 (call_i): Remove 'S' asm letter, add jli instruction.
4485 (call_value_i): Likewise.
4486 * config/arc/arc.op (mjli-always): New option.
4487 * config/arc/constraints.md (Cji): New constraint.
4488 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4490 (subsf3_fpx): Likewise.
4491 (mulsf3_fpx): Likewise.
4492 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4494 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4496 * doc/invoke.texi (ARC): Document mjli-always option.
4498 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4500 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4501 avoid addition with 0 and use incw and decw where possible.
4503 2018-01-26 Richard Biener <rguenther@suse.de>
4505 PR tree-optimization/81082
4506 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4507 association if it requires casting to unsigned.
4508 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4509 from fold_plusminus_mult_expr to catch important cases late when
4510 range info is available.
4512 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4514 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4515 * configure.ac (hidden_linkonce): New test.
4516 * configure: Regenerate.
4517 * config.in: Regenerate.
4519 2018-01-26 Julia Koval <julia.koval@intel.com>
4521 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4522 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4523 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4524 _mm_mask_bitshuffle_epi64_mask): Fix type.
4525 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4526 USI_FTYPE_V4DI_V4DI_USI): Remove.
4527 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4528 __builtin_ia32_vpshufbitqmb256_mask,
4529 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4530 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4531 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4533 2018-01-26 Alan Modra <amodra@gmail.com>
4536 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4537 UNSPEC_VBPERMQ. Sort other unspecs.
4539 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4541 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4543 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4546 * predict.c (drop_profile): Do not push/pop cfun; update also
4548 (handle_missing_profiles): Fix logic looking for zero profiles.
4550 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4553 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4554 on functions with #pragma omp declare simd or functions with simd
4556 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4557 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4558 Remove trailing \n from warning_at calls.
4560 2018-01-25 Tom de Vries <tom@codesourcery.com>
4563 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4564 for neutered workers.
4566 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4569 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4570 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4572 2018-01-24 Jeff Law <law@redhat.com>
4575 * i386.c (get_probe_interval): Move to earlier point.
4576 (ix86_compute_frame_layout): If -fstack-clash-protection and
4577 the frame is larger than the probe interval, then use pushes
4578 to save registers rather than reg->mem moves.
4579 (ix86_expand_prologue): Remove conditional for int_registers_saved
4582 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4585 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4586 min/max for never referenced object.
4588 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4591 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4593 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4594 attributes from DECL_ATTRIBUTES (decl) without affecting
4595 DECL_ATTRIBUTES (current_function_decl).
4596 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4597 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4599 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4601 PR tree-optimization/83979
4602 * fold-const.c (fold_comparison): Use constant_boolean_node
4603 instead of boolean_{true,false}_node.
4605 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4607 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4610 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4612 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4613 Simplify the clause that sets the length attribute.
4614 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4615 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4616 clause that sets the length attribute.
4617 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4619 2018-01-24 Tom de Vries <tom@codesourcery.com>
4622 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4623 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4624 Add strict parameter.
4625 (prevent_branch_around_nothing): Insert dummy insn between branch to
4626 label and label with no ptx insn inbetween.
4627 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4629 2018-01-24 Tom de Vries <tom@codesourcery.com>
4632 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4633 for neutered threads in warp.
4634 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4636 2018-01-24 Richard Biener <rguenther@suse.de>
4638 PR tree-optimization/83176
4639 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4642 2018-01-24 Richard Biener <rguenther@suse.de>
4644 PR tree-optimization/82819
4645 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4646 code generating pluses that are no-ops in the target precision.
4648 2018-01-24 Richard Biener <rguenther@suse.de>
4651 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4653 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4655 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4656 to merge probabilities.
4657 * predict.c (probably_never_executed): Also mark as cold functions
4658 with global 0 profile and guessed local profile.
4659 * profile-count.c (profile_probability::combine_with_count): New
4661 * profile-count.h (profile_probability::operator*,
4662 profile_probability::operator*=, profile_probability::operator/,
4663 profile_probability::operator/=): Reduce precision to adjusted
4664 and set value to guessed on contradictory divisions.
4665 (profile_probability::combine_with_freq): Remove.
4666 (profile_probability::combine_wiht_count): Declare.
4667 (profile_count::force_nonzero):: Set to adjusted.
4668 (profile_count::probability_in):: Set quality to adjusted.
4669 * tree-ssa-tail-merge.c (replace_block_by): Use
4672 2018-01-23 Andrew Waterman <andrew@sifive.com>
4673 Jim Wilson <jimw@sifive.com>
4675 * config/riscv/riscv.c (riscv_stack_boundary): New.
4676 (riscv_option_override): Set riscv_stack_boundary. Handle
4677 riscv_preferred_stack_boundary_arg.
4678 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4679 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4680 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4681 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4682 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4683 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4685 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4688 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4689 of struct ix86_frame.
4690 (ix86_expand_epilogue): Likewise. Add a local variable for
4691 the reg_save_offset field in struct ix86_frame.
4693 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4695 PR tree-optimization/82604
4696 * tree-loop-distribution.c (enum partition_kind): New enum item
4697 PKIND_PARTIAL_MEMSET.
4698 (partition_builtin_p): Support above new enum item.
4699 (generate_code_for_partition): Ditto.
4700 (compute_access_range): Differentiate cases that equality can be
4701 proven at all loops, the innermost loops or no loops.
4702 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4703 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4704 (finalize_partitions, distribute_loop): Don't fuse partition of
4705 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4706 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4709 2018-01-23 Martin Liska <mliska@suse.cz>
4711 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4712 order to ignore the predictor.
4713 (PRED_POLYMORPHIC_CALL): Likewise.
4714 (PRED_RECURSIVE_CALL): Likewise.
4716 2018-01-23 Martin Liska <mliska@suse.cz>
4718 * tree-profile.c (tree_profiling): Print function header to
4719 aware reader which function we are working on.
4720 * value-prof.c (gimple_find_values_to_profile): Do not print
4721 not interesting value histograms.
4723 2018-01-23 Martin Liska <mliska@suse.cz>
4725 * profile-count.h (enum profile_quality): Add
4726 profile_uninitialized as the first value. Do not number values
4727 as they are zero based.
4728 (profile_count::verify): Update sanity check.
4729 (profile_probability::verify): Likewise.
4731 2018-01-23 Nathan Sidwell <nathan@acm.org>
4733 * doc/invoke.texi (ffor-scope): Deprecate.
4735 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4737 PR tree-optimization/83510
4738 * domwalk.c (set_all_edges_as_executable): New function.
4739 (dom_walker::dom_walker): Convert bool param
4740 "skip_unreachable_blocks" to enum reachability. Move setup of
4741 edge flags to set_all_edges_as_executable and only do it when
4742 reachability is REACHABLE_BLOCKS.
4743 * domwalk.h (enum dom_walker::reachability): New enum.
4744 (dom_walker::dom_walker): Convert bool param
4745 "skip_unreachable_blocks" to enum reachability.
4746 (set_all_edges_as_executable): New decl.
4747 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4748 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4750 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4751 but converting true to REACHABLE_BLOCKS.
4752 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4754 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4755 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4756 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4758 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4759 if check_all_array_refs will be called.
4761 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4763 * tree.c (selftest::test_location_wrappers): Add more test
4766 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4768 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4769 (selftest::test_bit_in_range): Likewise.
4771 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4774 * doc/sourcebuild.texi (vect_float): Say that the selector
4775 only describes the situation when -funsafe-math-optimizations is on.
4776 (vect_float_strict): Document.
4778 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4780 PR tree-optimization/83965
4781 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4782 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4783 instead of checking only for a reduction.
4784 (vect_recog_widen_sum_pattern): Likewise.
4786 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4788 * predict.c (probably_never_executed): Only use precise profile info.
4789 (compute_function_frequency): Skip after inlining hack since we now
4790 have quality checking.
4792 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4794 * profile-count.h (profile_probability::very_unlikely,
4795 profile_probability::unlikely, profile_probability::even): Set
4796 precision to guessed.
4798 2018-01-23 Richard Biener <rguenther@suse.de>
4800 PR tree-optimization/83963
4801 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4802 Properly terminate dominator walk when crossing the exit edge not
4803 when visiting its source block.
4805 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4808 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4809 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4811 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4813 PR tree-optimization/83957
4814 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4815 semicolon after for body surrounded by braces.
4817 PR tree-optimization/83081
4818 * profile-count.h (profile_probability::split): New method.
4819 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4820 Use profile_probability::split.
4821 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4822 when splitting a single conditional jump into 2.
4824 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4826 PR tree-optimization/69452
4827 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4830 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4832 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4833 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4834 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4836 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4838 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4839 * config/rl78/rl78.md (movdi): New define_expand.
4840 * config/rl78/rl78.c (rl78_split_movdi): New function.
4842 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4845 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4847 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4848 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4849 128-bit to produce an UNSPEC move to get the double word with the
4850 signbit and then a shift directly to do signbit.
4851 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4852 implementation with a new version that just does either a direct
4853 move or a regular move. Move memory interface to separate insns.
4854 Move insns so they are next to the expander.
4855 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4856 with signbit move. Split big and little endian case.
4857 (signbit<mode>2_dm_mem_le): Likewise.
4858 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4859 (signbit<mode>2_dm2): Likewise.
4861 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4863 * config/rl78/rl78.md (anddi3): New define_expand.
4865 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4867 * config/rl78/rl78.md (umindi3): New define_expand.
4869 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4871 * config/rl78/rl78.md (smindi3): New define_expand.
4873 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4875 * config/rl78/rl78.md (smaxdi3): New define_expand.
4877 2018-01-22 Carl Love <cel@us.ibm.com>
4879 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4880 LVX_V1TI): Add macro expansion.
4881 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4882 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4883 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4884 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4885 Change check to determine if the instruction is a byte reversing
4886 entry. Fix typo in comment.
4887 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4888 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4889 Add def_builtin calls for new builtins.
4890 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4891 Add define_insn expansion.
4893 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4895 * config/rl78/rl78.md (umaxdi3): New define_expand.
4897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4899 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4900 for non-QImode registers.
4902 2018-01-22 Richard Biener <rguenther@suse.de>
4904 PR tree-optimization/83963
4905 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4906 including the loop exit block.
4907 (scop_detection::merge_sese): Likewise.
4908 (scop_detection::add_scop): Do it here instead.
4910 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4912 * doc/sourcebuild.texi (arm_softfloat): Document.
4914 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4917 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4918 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4919 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4921 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4922 David Edelsohn <dje.gcc@gmail.com>
4925 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4926 Change "crset eq" to "crset 2".
4927 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4928 (*call_indirect_aix<mode>_nospec): Likewise.
4929 (*call_value_indirect_aix<mode>_nospec): Likewise.
4930 (*call_indirect_elfv2<mode>_nospec): Likewise.
4931 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4932 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4933 change assembly output from . to $.
4934 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4935 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4936 (*tablejump<mode>_internal1_nospec): Likewise.
4938 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4941 * config/sh/sh_optimize_sett_clrt.cc:
4942 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4944 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4946 PR tree-optimization/83940
4947 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4948 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4949 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4950 use it to pass back the definition type.
4951 (vect_check_store_rhs): Likewise rhs_dt_out.
4952 (vect_build_gather_load_calls): Add a mask_dt argument and use
4953 it instead of a call to vect_is_simple_use.
4954 (vectorizable_store): Update calls to vect_check_load_store_mask
4955 and vect_check_store_rhs. Use the dt returned by the latter instead
4956 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4957 instead of calls to vect_is_simple_use. Pass the scalar rather
4958 than the vector operand to vect_is_simple_use when handling
4959 second and subsequent copies of an rhs value.
4960 (vectorizable_load): Update calls to vect_check_load_store_mask
4961 and vect_build_gather_load_calls. Use the cached mask_dt and
4962 gs_info.offset_dt instead of calls to vect_is_simple_use.
4964 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4967 * tree-emutls.c: Include gimplify.h.
4968 (lower_emutls_2): New function.
4969 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4970 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4971 it before further processing.
4974 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4975 UINTVAL (trueop1) instead of INTVAL (op1).
4977 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4981 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4982 INCOMING_FRAME_SP_OFFSET if not defined.
4983 (scan_trace): Add ENTRY argument. If true and
4984 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4985 emit a note to adjust the CFA offset.
4986 (create_cfi_notes): Adjust scan_trace callers.
4987 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4988 INCOMING_FRAME_SP_OFFSET in the CIE.
4989 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4990 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4992 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4993 * doc/tm.texi: Regenerated.
4995 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4997 PR rtl-optimization/83147
4998 * lra-constraints.c (remove_inheritance_pseudos): Use
4999 lra_substitute_pseudo_within_insn.
5001 2018-01-19 Tom de Vries <tom@codesourcery.com>
5002 Cesar Philippidis <cesar@codesourcery.com>
5005 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5007 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5010 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5011 spaces for function labels.
5013 2018-01-19 Martin Liska <mliska@suse.cz>
5015 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5016 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5017 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5018 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5019 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5020 (PRED_CONST_RETURN): Change from 69 to 65.
5021 (PRED_NULL_RETURN): Change from 91 to 71.
5022 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5023 (PRED_LOOP_GUARD): Change from 66 to 73.
5025 2018-01-19 Martin Liska <mliska@suse.cz>
5027 * predict.c (predict_insn_def): Add new assert.
5028 (struct branch_predictor): Change type to signed integer.
5029 (test_prediction_value_range): Amend test to cover
5031 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5032 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5033 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5034 (PRED_LOOP_IV_COMPARE): Likewise.
5035 * predict.h (PROB_UNINITIALIZED): Define new constant.
5037 2018-01-19 Martin Liska <mliska@suse.cz>
5039 * predict.c (dump_prediction): Add new format for
5040 analyze_brprob.py script which is enabled with -details
5042 * profile-count.h (precise_p): New function.
5044 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5046 PR tree-optimization/83922
5047 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5048 there are no statements that need masking.
5049 (vect_active_double_reduction_p): New function.
5050 (vect_analyze_loop_operations): Use it when handling phis that
5051 are not in the loop header.
5053 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5055 PR tree-optimization/83914
5056 * tree-vect-loop.c (vectorizable_induction): Don't convert
5057 init_expr or apply the peeling adjustment for inductions
5058 that are nested within the vectorized loop.
5060 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5062 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5065 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5069 * function.h (gimplify_parameters): Add gimple_seq * argument.
5070 * function.c: Include gimple.h and options.h.
5071 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5072 for the added local temporaries if needed.
5073 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5074 if there are any parameter cleanups, wrap whole body into a
5075 try/finally with the cleanups.
5077 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5080 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5081 Use GET_MODE_CLASS for scalar floating point.
5083 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5087 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5088 Fix call of call_cgraph_insertion_hooks.
5090 2018-01-18 Martin Sebor <msebor@redhat.com>
5092 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5094 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5097 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5100 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5103 * common.opt: (-ffile-prefix-map): New option.
5104 * opts.c (common_handle_option): Defer it.
5105 * opts-global.c (handle_common_deferred_options): Handle it.
5106 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5107 * file-prefix-map.h: New file.
5108 (remap_debug_filename, add_debug_prefix_map): ...here.
5109 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5110 * final.c (debug_prefix_map, add_debug_prefix_map
5111 remap_debug_filename): Move to...
5112 * file-prefix-map.c: New file.
5113 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5114 generalize, get rid of alloca(), use strrchr() instead of strchr().
5115 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5116 Implement in terms of add_prefix_map().
5117 (remap_macro_filename, remap_debug_filename): Implement in term of
5119 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5120 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5121 * dbxout.c: Include file-prefix-map.h.
5122 * varasm.c: Likewise.
5123 * vmsdbgout.c: Likewise.
5124 * xcoffout.c: Likewise.
5125 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5126 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5127 * doc/invoke.texi (-ffile-prefix-map): Document.
5128 (-fdebug-prefix-map): Update description.
5130 2018-01-18 Martin Liska <mliska@suse.cz>
5132 * config/i386/i386.c (indirect_thunk_name): Document that also
5134 (output_indirect_thunk): Document why both instructions
5135 (pause and lfence) are generated.
5137 2018-01-18 Richard Biener <rguenther@suse.de>
5139 PR tree-optimization/83887
5140 * graphite-scop-detection.c
5141 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5142 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5143 (scop_detection::merge_sese): Re-implement with a flood-fill
5144 algorithm that properly finds a SESE region if it exists.
5146 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5149 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5150 pointer_diff optimizations use view_convert instead of convert.
5152 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5154 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5155 Generate different code for -mno-speculate-indirect-jumps.
5156 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5157 (*call_indirect_aix<mode>): Disable for
5158 -mno-speculate-indirect-jumps.
5159 (*call_indirect_aix<mode>_nospec): New define_insn.
5160 (*call_value_indirect_aix<mode>): Disable for
5161 -mno-speculate-indirect-jumps.
5162 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5163 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5164 -mno-speculate-indirect-jumps.
5165 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5167 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5169 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5170 long double type, set the flags for noting the default long double
5171 type, even if we don't pass or return a long double type.
5173 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5176 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5179 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5181 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5182 support for merge[hl].
5183 (fold_mergehl_helper): New helper function.
5184 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5185 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5186 (altivec_vmrglw_direct): Add xxmrglw insn.
5188 2018-01-17 Andrew Waterman <andrew@sifive.com>
5190 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5191 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5193 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5196 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5197 call the lto_location_cache before reading the
5198 DECL_SOURCE_LOCATION of the types.
5200 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5201 Richard Sandiford <richard.sandiford@linaro.org>
5203 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5204 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5205 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5206 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5207 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5209 * config/aarch64/constraints.md (aarch64_movti_operand):
5211 * config/aarch64/predicates.md (Uti): Add new constraint.
5213 2018-01-17 Carl Love <cel@us.ibm.com>
5215 * config/rs6000/vsx.md (define_expand xl_len_r,
5216 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5217 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5219 (define_expand, define_insn): Move the shift left from the
5220 define_insn to the define_expand for lxvl and stxvl instructions.
5221 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5222 and XL_LEN_R definitions to PURE.
5224 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5226 * config/i386/i386.c (indirect_thunk_name): Declare regno
5227 as unsigned int. Compare regno with INVALID_REGNUM.
5228 (output_indirect_thunk): Ditto.
5229 (output_indirect_thunk_function): Ditto.
5230 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5231 in the call to output_indirect_thunk_function.
5233 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5236 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5237 rather than the size of inner_type to determine the stack slot size
5238 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5240 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5243 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5246 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5248 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5249 endian Linux systems to optionally enable multilibs for selecting
5250 the long double type if the user configured an explicit type.
5251 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5252 have no long double multilibs if not defined.
5253 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5254 warn if the user used -mabi={ieee,ibm}longdouble and we built
5255 multilibs for long double.
5256 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5257 appropriate multilib option.
5258 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5260 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5261 for building long double multilibs.
5262 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5264 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5266 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5269 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5271 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5274 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5277 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5280 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5282 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5283 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5285 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5287 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5288 different rtl trees depending on TARGET_64BIT.
5289 (rs6000_gen_lvx): Likewise.
5291 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5293 * config/visium/visium.md (nop): Tweak comment.
5294 (hazard_nop): Likewise.
5296 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5298 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5299 -mspeculate-indirect-jumps.
5300 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5301 for -mno-speculate-indirect-jumps.
5302 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5303 (*call_value_indirect_elfv2<mode>): Disable for
5304 -mno-speculate-indirect-jumps.
5305 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5306 (indirect_jump): Emit different RTL for
5307 -mno-speculate-indirect-jumps.
5308 (*indirect_jump<mode>): Disable for
5309 -mno-speculate-indirect-jumps.
5310 (*indirect_jump<mode>_nospec): New define_insn.
5311 (tablejump): Emit different RTL for
5312 -mno-speculate-indirect-jumps.
5313 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5314 (tablejumpsi_nospec): New define_expand.
5315 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5316 (tablejumpdi_nospec): New define_expand.
5317 (*tablejump<mode>_internal1): Disable for
5318 -mno-speculate-indirect-jumps.
5319 (*tablejump<mode>_internal1_nospec): New define_insn.
5320 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5323 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5325 * caller-save.c (insert_save): Drop unnecessary parameter. All
5328 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5329 Richard Biener <rguenth@suse.de>
5332 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5333 return early, inline manually is_gimple_sizepos. Make sure if we
5334 call gimplify_expr we don't end up with a gimple constant.
5335 * tree.c (variably_modified_type_p): Don't return true for
5336 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5337 * gimplify.h (is_gimple_sizepos): Remove.
5339 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5341 PR tree-optimization/83857
5342 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5343 vectorizable_live_operation for pure SLP statements.
5344 (vectorizable_live_operation): Handle PHIs.
5346 2018-01-16 Richard Biener <rguenther@suse.de>
5348 PR tree-optimization/83867
5349 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5350 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5352 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5355 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5356 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5357 If off is not INTEGER_CST, issue a may not be aligned warning
5358 rather than isn't aligned. Use isn%'t rather than isn't.
5359 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5361 <case MULT_EXPR>: Improve the case when bottom and one of the
5362 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5363 operand, in that case check if the other operand is multiple of
5364 bottom divided by the INTEGER_CST operand.
5366 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5369 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5370 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5371 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5372 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5373 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5374 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5375 (pa_function_arg_size): New function.
5377 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5379 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5380 in a separate statement.
5382 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5384 PR tree-optimization/83847
5385 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5386 group gathers and scatters.
5388 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5390 PR rtl-optimization/86620
5391 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5393 PR rtl-optimization/83213
5394 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5395 to last if both are JUMP_INSNs.
5397 PR tree-optimization/83843
5398 * gimple-ssa-store-merging.c
5399 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5400 store_immediate_info for bswap/nop orig_stores.
5402 2018-01-15 Andrew Waterman <andrew@sifive.com>
5404 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5406 <UDIV>: Increase cost if !TARGET_DIV.
5408 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5410 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5411 (define_attr "cr_logical_3op"): New.
5412 (cceq_ior_compare): Adjust.
5413 (cceq_ior_compare_complement): Adjust.
5414 (*cceq_rev_compare): Adjust.
5415 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5416 (is_cracked_insn): Adjust.
5417 (insn_must_be_first_in_group): Adjust.
5418 * config/rs6000/40x.md: Adjust.
5419 * config/rs6000/440.md: Adjust.
5420 * config/rs6000/476.md: Adjust.
5421 * config/rs6000/601.md: Adjust.
5422 * config/rs6000/603.md: Adjust.
5423 * config/rs6000/6xx.md: Adjust.
5424 * config/rs6000/7450.md: Adjust.
5425 * config/rs6000/7xx.md: Adjust.
5426 * config/rs6000/8540.md: Adjust.
5427 * config/rs6000/cell.md: Adjust.
5428 * config/rs6000/e300c2c3.md: Adjust.
5429 * config/rs6000/e500mc.md: Adjust.
5430 * config/rs6000/e500mc64.md: Adjust.
5431 * config/rs6000/e5500.md: Adjust.
5432 * config/rs6000/e6500.md: Adjust.
5433 * config/rs6000/mpc.md: Adjust.
5434 * config/rs6000/power4.md: Adjust.
5435 * config/rs6000/power5.md: Adjust.
5436 * config/rs6000/power6.md: Adjust.
5437 * config/rs6000/power7.md: Adjust.
5438 * config/rs6000/power8.md: Adjust.
5439 * config/rs6000/power9.md: Adjust.
5440 * config/rs6000/rs64.md: Adjust.
5441 * config/rs6000/titan.md: Adjust.
5443 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5445 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5446 ix86_indirect_branch_register logic.
5448 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5450 * config/i386/constraints.md (Bs): Update
5451 ix86_indirect_branch_register check. Don't check
5452 ix86_indirect_branch_register with GOT_memory_operand.
5454 * config/i386/predicates.md (GOT_memory_operand): Don't check
5455 ix86_indirect_branch_register here.
5456 (GOT32_symbol_operand): Likewise.
5458 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5460 * config/i386/predicates.md (constant_call_address_operand):
5461 Rewrite ix86_indirect_branch_register logic.
5462 (sibcall_insn_operand): Likewise.
5464 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5466 * config/i386/constraints.md (Bs): Replace
5467 ix86_indirect_branch_thunk_register with
5468 ix86_indirect_branch_register.
5470 * config/i386/i386.md (indirect_jump): Likewise.
5471 (tablejump): Likewise.
5472 (*sibcall_memory): Likewise.
5473 (*sibcall_value_memory): Likewise.
5474 Peepholes of indirect call and jump via memory: Likewise.
5475 * config/i386/i386.opt: Likewise.
5476 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5477 (GOT_memory_operand): Likewise.
5478 (call_insn_operand): Likewise.
5479 (sibcall_insn_operand): Likewise.
5480 (GOT32_symbol_operand): Likewise.
5482 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5485 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5486 type rather than type addr's type points to.
5487 (expand_omp_atomic_mutex): Likewise.
5488 (expand_omp_atomic): Likewise.
5490 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5493 * config/i386/i386.c (output_indirect_thunk_function): Use
5494 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5495 for __x86_return_thunk.
5497 2018-01-15 Richard Biener <rguenther@suse.de>
5500 * expmed.c (extract_bit_field_1): Fix typo.
5502 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5505 * config/arm/iterators.md (VF): New mode iterator.
5506 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5507 Remove integer-related logic from pattern.
5508 (neon_vabd<mode>_3): Likewise.
5510 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5513 * common.opt (fstrict-overflow): No longer an alias.
5514 (fwrapv-pointer): New option.
5515 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5516 also for pointer types based on flag_wrapv_pointer.
5517 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5518 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5519 opts->x_flag_wrapv got set.
5520 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5521 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5522 POINTER_TYPE_OVERFLOW_UNDEFINED.
5523 * match.pd: Likewise in address comparison pattern.
5524 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5526 2018-01-15 Richard Biener <rguenther@suse.de>
5529 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5530 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5531 Reset type names to their identifier if their TYPE_DECL doesn't
5532 have linkage (and thus is used for ODR and devirt).
5533 (save_debug_info_for_decl): Remove.
5534 (save_debug_info_for_type): Likewise.
5535 (add_tree_to_fld_list): Adjust.
5536 * tree-pretty-print.c (dump_generic_node): Make dumping of
5537 type names more robust.
5539 2018-01-15 Richard Biener <rguenther@suse.de>
5541 * BASE-VER: Bump to 8.0.1.
5543 2018-01-14 Martin Sebor <msebor@redhat.com>
5546 * builtins.c (check_access): Avoid warning when the no-warning bit
5549 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5551 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5552 * ira-color (allocno_hard_regs_compare): Likewise.
5554 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5557 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5558 Use .pushsection/.popsection.
5560 2018-01-14 Martin Sebor <msebor@redhat.com>
5563 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5565 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5567 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5568 entry from extra_headers.
5569 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5570 extra_headers, make the list bitwise identical to the i?86-*-* one.
5572 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5574 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5575 -mcmodel=large with -mindirect-branch=thunk,
5576 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5577 -mfunction-return=thunk-extern.
5578 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5579 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5580 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5582 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5584 * config/i386/i386.c (print_reg): Print the name of the full
5585 integer register without '%'.
5586 (ix86_print_operand): Handle 'V'.
5587 * doc/extend.texi: Document 'V' modifier.
5589 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5591 * config/i386/constraints.md (Bs): Disallow memory operand for
5592 -mindirect-branch-register.
5594 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5595 (GOT_memory_operand): Likewise.
5596 (call_insn_operand): Likewise.
5597 (sibcall_insn_operand): Likewise.
5598 (GOT32_symbol_operand): Likewise.
5599 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5600 for -mindirect-branch-register.
5601 (tablejump): Likewise.
5602 (*sibcall_memory): Likewise.
5603 (*sibcall_value_memory): Likewise.
5604 Disallow peepholes of indirect call and jump via memory for
5605 -mindirect-branch-register.
5606 (*call_pop): Replace m with Bw.
5607 (*call_value_pop): Likewise.
5608 (*sibcall_pop_memory): Replace m with Bs.
5609 * config/i386/i386.opt (mindirect-branch-register): New option.
5610 * doc/invoke.texi: Document -mindirect-branch-register option.
5612 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5614 * config/i386/i386-protos.h (ix86_output_function_return): New.
5615 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5616 set function_return_type.
5617 (indirect_thunk_name): Add ret_p to indicate thunk for function
5619 (output_indirect_thunk_function): Pass false to
5620 indirect_thunk_name.
5621 (ix86_output_indirect_branch_via_reg): Likewise.
5622 (ix86_output_indirect_branch_via_push): Likewise.
5623 (output_indirect_thunk_function): Create alias for function
5624 return thunk if regno < 0.
5625 (ix86_output_function_return): New function.
5626 (ix86_handle_fndecl_attribute): Handle function_return.
5627 (ix86_attribute_table): Add function_return.
5628 * config/i386/i386.h (machine_function): Add
5629 function_return_type.
5630 * config/i386/i386.md (simple_return_internal): Use
5631 ix86_output_function_return.
5632 (simple_return_internal_long): Likewise.
5633 * config/i386/i386.opt (mfunction-return=): New option.
5634 (indirect_branch): Mention -mfunction-return=.
5635 * doc/extend.texi: Document function_return function attribute.
5636 * doc/invoke.texi: Document -mfunction-return= option.
5638 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5640 * config/i386/i386-opts.h (indirect_branch): New.
5641 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5642 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5643 with local indirect jump when converting indirect call and jump.
5644 (ix86_set_indirect_branch_type): New.
5645 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5646 (indirectlabelno): New.
5647 (indirect_thunk_needed): Likewise.
5648 (indirect_thunk_bnd_needed): Likewise.
5649 (indirect_thunks_used): Likewise.
5650 (indirect_thunks_bnd_used): Likewise.
5651 (INDIRECT_LABEL): Likewise.
5652 (indirect_thunk_name): Likewise.
5653 (output_indirect_thunk): Likewise.
5654 (output_indirect_thunk_function): Likewise.
5655 (ix86_output_indirect_branch_via_reg): Likewise.
5656 (ix86_output_indirect_branch_via_push): Likewise.
5657 (ix86_output_indirect_branch): Likewise.
5658 (ix86_output_indirect_jmp): Likewise.
5659 (ix86_code_end): Call output_indirect_thunk_function if needed.
5660 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5662 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5663 (ix86_attribute_table): Add indirect_branch.
5664 * config/i386/i386.h (machine_function): Add indirect_branch_type
5665 and has_local_indirect_jump.
5666 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5668 (tablejump): Likewise.
5669 (*indirect_jump): Use ix86_output_indirect_jmp.
5670 (*tablejump_1): Likewise.
5671 (simple_return_indirect_internal): Likewise.
5672 * config/i386/i386.opt (mindirect-branch=): New option.
5673 (indirect_branch): New.
5676 (thunk-inline): Likewise.
5677 (thunk-extern): Likewise.
5678 * doc/extend.texi: Document indirect_branch function attribute.
5679 * doc/invoke.texi: Document -mindirect-branch= option.
5681 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5684 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5686 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5688 * ipa-inline.c (want_inline_small_function_p): Return false if
5689 inlining has already failed with CIF_FINAL_ERROR.
5690 (update_caller_keys): Call want_inline_small_function_p before
5692 (update_callee_keys): Likewise.
5694 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5696 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5698 (rs6000_quadword_masked_address_p): Likewise.
5699 (quad_aligned_load_p): Likewise.
5700 (quad_aligned_store_p): Likewise.
5701 (const_load_sequence_p): Add comment to describe the outer-most loop.
5702 (mimic_memory_attributes_and_flags): New function.
5703 (rs6000_gen_stvx): Likewise.
5704 (replace_swapped_aligned_store): Likewise.
5705 (rs6000_gen_lvx): Likewise.
5706 (replace_swapped_aligned_load): Likewise.
5707 (replace_swapped_load_constant): Capitalize argument name in
5708 comment describing this function.
5709 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5710 and stores that access quad-word aligned addresses and replace
5711 with stvx or lvx instructions when appropriate.
5712 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5713 New function prototype.
5714 (rs6000_quadword_masked_address_p): Likewise.
5715 (rs6000_gen_lvx): Likewise.
5716 (rs6000_gen_stvx): Likewise.
5717 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5718 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5719 when memory address is aligned.
5720 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5721 this split to select lvx instruction when memory address is aligned.
5722 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5723 instruction when memory address is aligned.
5724 (*vsx_le_perm_load_v16qi): Likewise.
5725 (four unnamed splitters): Modify to select the stvx instruction
5726 when memory is aligned.
5728 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5730 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5731 which appears in the queue multiple times.
5733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5734 Alan Hayward <alan.hayward@arm.com>
5735 David Sherwood <david.sherwood@arm.com>
5737 * tree-vectorizer.h (vec_lower_bound): New structure.
5738 (_loop_vec_info): Add check_nonzero and lower_bounds.
5739 (LOOP_VINFO_CHECK_NONZERO): New macro.
5740 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5741 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5742 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5743 fields. Make seg_len the distance travelled, not including the
5745 (dr_direction_indicator): Declare.
5746 (dr_zero_step_indicator): Likewise.
5747 (dr_known_forward_stride_p): Likewise.
5748 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5750 (runtime_alias_check_p): Allow runtime alias checks with
5752 (operator ==): Compare access_size and align.
5753 (prune_runtime_alias_test_list): Rework for new distinction between
5754 the access_size and seg_len.
5755 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5757 (get_segment_min_max): New function.
5758 (create_intersect_range_checks): Use it.
5759 (dr_step_indicator): New function.
5760 (dr_direction_indicator): Likewise.
5761 (dr_zero_step_indicator): Likewise.
5762 (dr_known_forward_stride_p): Likewise.
5763 * tree-loop-distribution.c (data_ref_segment_size): Return
5764 DR_STEP * (niters - 1).
5765 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5767 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5768 (vect_preserves_scalar_order_p): New function, split out from...
5769 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5770 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5771 (vect_vfa_access_size): New function.
5772 (vect_vfa_align): Likewise.
5773 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5774 (dump_lower_bound): New function.
5775 (vect_check_lower_bound): Likewise.
5776 (vect_small_gap_p): Likewise.
5777 (vectorizable_with_step_bound_p): Likewise.
5778 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5779 depencies if the vectorization factor is 1. Convert the checks
5780 for nonzero steps into checks on the bounds of DR_STEP. Try using
5781 a bunds check for variable steps if the minimum required step is
5782 relatively small. Update calls to the dr_with_seg_len
5783 constructor and to vect_compile_time_alias.
5784 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5786 (vect_loop_versioning): Call it.
5787 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5789 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5792 Alan Hayward <alan.hayward@arm.com>
5793 David Sherwood <david.sherwood@arm.com>
5795 * doc/sourcebuild.texi (vect_scatter_store): Document.
5796 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5798 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5800 * genopinit.c (main): Add supports_vec_scatter_store and
5801 supports_vec_scatter_store_cached to target_optabs.
5802 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5803 IFN_MASK_SCATTER_STORE.
5804 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5806 * internal-fn.h (internal_store_fn_p): Declare.
5807 (internal_fn_stored_value_index): Likewise.
5808 * internal-fn.c (scatter_store_direct): New macro.
5809 (expand_scatter_store_optab_fn): New function.
5810 (direct_scatter_store_optab_supported_p): New macro.
5811 (internal_store_fn_p): New function.
5812 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5813 IFN_MASK_SCATTER_STORE.
5814 (internal_fn_mask_index): Likewise.
5815 (internal_fn_stored_value_index): New function.
5816 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5818 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5819 * optabs-query.c (supports_vec_scatter_store_p): New function.
5820 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5821 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5822 true for scatter stores.
5823 (vect_gather_scatter_fn_p): Handle scatter stores too.
5824 (vect_check_gather_scatter): Consider using scatter stores if
5825 supports_vec_scatter_store_p.
5826 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5828 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5829 internal_fn_stored_value_index.
5830 (check_load_store_masking): Handle scatter stores too.
5831 (vect_get_store_rhs): Make public.
5832 (vectorizable_call): Use internal_store_fn_p.
5833 (vectorizable_store): Handle scatter store internal functions.
5834 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5835 when deciding whether the end of the group has been reached.
5836 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5837 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5838 (mask_scatter_store<mode>): New insns.
5840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5841 Alan Hayward <alan.hayward@arm.com>
5842 David Sherwood <david.sherwood@arm.com>
5844 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5845 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5846 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5848 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5849 Use vect_truncate_gather_scatter_offset if we can't treat the
5850 operation as a normal gather load or scatter store.
5851 (get_group_load_store_type): Take the gather_scatter_info
5852 as argument. Try using a gather load or scatter store for
5853 single-element groups.
5854 (get_load_store_type): Update calls to get_group_load_store_type
5855 and vect_use_strided_gather_scatters_p.
5857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5858 Alan Hayward <alan.hayward@arm.com>
5859 David Sherwood <david.sherwood@arm.com>
5861 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5862 optional tree argument.
5863 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5865 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5866 but continue to use the current value as a fallback.
5867 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5868 to compare the updates.
5869 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5870 (get_load_store_type): Use it when handling a strided access.
5871 (vect_get_strided_load_store_ops): New function.
5872 (vect_get_data_ptr_increment): Likewise.
5873 (vectorizable_load): Handle strided gather loads. Always pass
5874 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5877 Alan Hayward <alan.hayward@arm.com>
5878 David Sherwood <david.sherwood@arm.com>
5880 * doc/md.texi (gather_load@var{m}): Document.
5881 (mask_gather_load@var{m}): Likewise.
5882 * genopinit.c (main): Add supports_vec_gather_load and
5883 supports_vec_gather_load_cached to target_optabs.
5884 * optabs-tree.c (init_tree_optimization_optabs): Use
5885 ggc_cleared_alloc to allocate target_optabs.
5886 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5887 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5889 * internal-fn.h (internal_load_fn_p): Declare.
5890 (internal_gather_scatter_fn_p): Likewise.
5891 (internal_fn_mask_index): Likewise.
5892 (internal_gather_scatter_fn_supported_p): Likewise.
5893 * internal-fn.c (gather_load_direct): New macro.
5894 (expand_gather_load_optab_fn): New function.
5895 (direct_gather_load_optab_supported_p): New macro.
5896 (direct_internal_fn_optab): New function.
5897 (internal_load_fn_p): Likewise.
5898 (internal_gather_scatter_fn_p): Likewise.
5899 (internal_fn_mask_index): Likewise.
5900 (internal_gather_scatter_fn_supported_p): Likewise.
5901 * optabs-query.c (supports_at_least_one_mode_p): New function.
5902 (supports_vec_gather_load_p): Likewise.
5903 * optabs-query.h (supports_vec_gather_load_p): Declare.
5904 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5905 and memory_type field.
5906 (NUM_PATTERNS): Bump to 15.
5907 * tree-vect-data-refs.c: Include internal-fn.h.
5908 (vect_gather_scatter_fn_p): New function.
5909 (vect_describe_gather_scatter_call): Likewise.
5910 (vect_check_gather_scatter): Try using internal functions for
5911 gather loads. Recognize existing calls to a gather load function.
5912 (vect_analyze_data_refs): Consider using gather loads if
5913 supports_vec_gather_load_p.
5914 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5915 (vect_get_gather_scatter_offset_type): Likewise.
5916 (vect_convert_mask_for_vectype): Likewise.
5917 (vect_add_conversion_to_patterm): Likewise.
5918 (vect_try_gather_scatter_pattern): Likewise.
5919 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5920 (vect_vect_recog_func_ptrs): Add it.
5921 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5922 internal_fn_mask_index and internal_gather_scatter_fn_p.
5923 (check_load_store_masking): Take the gather_scatter_info as an
5924 argument and handle gather loads.
5925 (vect_get_gather_scatter_ops): New function.
5926 (vectorizable_call): Check internal_load_fn_p.
5927 (vectorizable_load): Likewise. Handle gather load internal
5929 (vectorizable_store): Update call to check_load_store_masking.
5930 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5931 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5932 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5933 (aarch64_gather_scale_operand_d): New predicates.
5934 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5935 (mask_gather_load<mode>): New insns.
5937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5938 Alan Hayward <alan.hayward@arm.com>
5939 David Sherwood <david.sherwood@arm.com>
5941 * optabs.def (fold_left_plus_optab): New optab.
5942 * doc/md.texi (fold_left_plus_@var{m}): Document.
5943 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5944 * internal-fn.c (fold_left_direct): Define.
5945 (expand_fold_left_optab_fn): Likewise.
5946 (direct_fold_left_optab_supported_p): Likewise.
5947 * fold-const-call.c (fold_const_fold_left): New function.
5948 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5949 * tree-parloops.c (valid_reduction_p): New function.
5950 (gather_scalar_reductions): Use it.
5951 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5952 (vect_finish_replace_stmt): Declare.
5953 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5954 (needs_fold_left_reduction_p): New function, split out from...
5955 (vect_is_simple_reduction): ...here. Accept reductions that
5956 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5957 (vect_force_simple_reduction): Also store the reduction type in
5958 the assignment's STMT_VINFO_REDUC_TYPE.
5959 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5960 (merge_with_identity): New function.
5961 (vect_expand_fold_left): Likewise.
5962 (vectorize_fold_left_reduction): Likewise.
5963 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5964 scalar phi in place for it. Check for target support and reject
5965 cases that would reassociate the operation. Defer the transform
5966 phase to vectorize_fold_left_reduction.
5967 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5968 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5969 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5973 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5974 call to ifc_temp_var.
5976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5977 Alan Hayward <alan.hayward@arm.com>
5978 David Sherwood <david.sherwood@arm.com>
5980 * target.def (legitimize_address_displacement): Take the original
5981 offset as a poly_int.
5982 * targhooks.h (default_legitimize_address_displacement): Update
5984 * targhooks.c (default_legitimize_address_displacement): Likewise.
5985 * doc/tm.texi: Regenerate.
5986 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5987 as an argument, moving assert of ad->disp == ad->disp_term to...
5988 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5989 Try calling targetm.legitimize_address_displacement before expanding
5990 the address rather than afterwards, and adjust for the new interface.
5991 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5992 Match the new hook interface. Handle SVE addresses.
5993 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5998 * Makefile.in (OBJS): Add early-remat.o.
5999 * target.def (select_early_remat_modes): New hook.
6000 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6001 * doc/tm.texi: Regenerate.
6002 * targhooks.h (default_select_early_remat_modes): Declare.
6003 * targhooks.c (default_select_early_remat_modes): New function.
6004 * timevar.def (TV_EARLY_REMAT): New timevar.
6005 * passes.def (pass_early_remat): New pass.
6006 * tree-pass.h (make_pass_early_remat): Declare.
6007 * early-remat.c: New file.
6008 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6010 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6013 Alan Hayward <alan.hayward@arm.com>
6014 David Sherwood <david.sherwood@arm.com>
6016 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6017 vfm1 with a bound_epilog parameter.
6018 (vect_do_peeling): Update calls accordingly, and move the prologue
6019 call earlier in the function. Treat the base bound_epilog as 0 for
6020 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6021 this base when peeling for gaps.
6022 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6023 with fully-masked loops.
6024 (vect_estimate_min_profitable_iters): Handle the single peeled
6025 iteration in that case.
6027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6028 Alan Hayward <alan.hayward@arm.com>
6029 David Sherwood <david.sherwood@arm.com>
6031 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6032 single-element interleaving even if the size is not a power of 2.
6033 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6034 accesses for single-element interleaving if the group size is
6037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6038 Alan Hayward <alan.hayward@arm.com>
6039 David Sherwood <david.sherwood@arm.com>
6041 * doc/md.texi (fold_extract_last_@var{m}): Document.
6042 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6043 * optabs.def (fold_extract_last_optab): New optab.
6044 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6045 * internal-fn.c (fold_extract_direct): New macro.
6046 (expand_fold_extract_optab_fn): Likewise.
6047 (direct_fold_extract_optab_supported_p): Likewise.
6048 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6049 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6050 EXTRACT_LAST_REDUCTION.
6051 (get_initial_def_for_reduction): Do not create an initial vector
6052 for EXTRACT_LAST_REDUCTION reductions.
6053 (vectorizable_reduction): Leave the scalar phi in place for
6054 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6055 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6056 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6057 transform phase to vectorizable_condition.
6058 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6060 (vect_finish_stmt_generation): ...here.
6061 (vect_finish_replace_stmt): New function.
6062 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6063 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6065 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6068 Alan Hayward <alan.hayward@arm.com>
6069 David Sherwood <david.sherwood@arm.com>
6071 * doc/md.texi (extract_last_@var{m}): Document.
6072 * optabs.def (extract_last_optab): New optab.
6073 * internal-fn.def (EXTRACT_LAST): New internal function.
6074 * internal-fn.c (cond_unary_direct): New macro.
6075 (expand_cond_unary_optab_fn): Likewise.
6076 (direct_cond_unary_optab_supported_p): Likewise.
6077 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6078 loops using EXTRACT_LAST.
6079 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6080 (extract_last_<mode>): ...this optab.
6081 (vec_extract<mode><Vel>): Update accordingly.
6083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6084 Alan Hayward <alan.hayward@arm.com>
6085 David Sherwood <david.sherwood@arm.com>
6087 * target.def (empty_mask_is_expensive): New hook.
6088 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6089 * doc/tm.texi: Regenerate.
6090 * targhooks.h (default_empty_mask_is_expensive): Declare.
6091 * targhooks.c (default_empty_mask_is_expensive): New function.
6092 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6093 if the target says that empty masks are expensive.
6094 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6096 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6099 Alan Hayward <alan.hayward@arm.com>
6100 David Sherwood <david.sherwood@arm.com>
6102 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6103 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6104 (vect_use_loop_mask_for_alignment_p): New function.
6105 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6106 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6107 niters_skip argument. Make sure that the first niters_skip elements
6108 of the first iteration are inactive.
6109 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6110 Update call to vect_set_loop_masks_directly.
6111 (get_misalign_in_elems): New function, split out from...
6112 (vect_gen_prolog_loop_niters): ...here.
6113 (vect_update_init_of_dr): Take a code argument that specifies whether
6114 the adjustment should be added or subtracted.
6115 (vect_update_init_of_drs): Likewise.
6116 (vect_prepare_for_masked_peels): New function.
6117 (vect_do_peeling): Skip prologue peeling if we're using a mask
6118 instead. Update call to vect_update_inits_of_drs.
6119 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6121 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6122 alignment. Do not include the number of peeled iterations in
6123 the minimum threshold in that case.
6124 (vectorizable_induction): Adjust the start value down by
6125 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6126 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6127 Take the number of skipped iterations into account when calculating
6129 * tree-vect-stmts.c (vect_gen_while_not): New function.
6131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6132 Alan Hayward <alan.hayward@arm.com>
6133 David Sherwood <david.sherwood@arm.com>
6135 * doc/sourcebuild.texi (vect_fully_masked): Document.
6136 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6138 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6140 (vect_analyze_loop_2): ...here. Don't check the vectorization
6141 factor against the number of loop iterations if the loop is
6144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6145 Alan Hayward <alan.hayward@arm.com>
6146 David Sherwood <david.sherwood@arm.com>
6148 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6149 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6150 (dump_groups): Update accordingly.
6151 (iv_use::mem_type): New member variable.
6152 (address_p): New function.
6153 (record_use): Add a mem_type argument and initialize the new
6155 (record_group_use): Add a mem_type argument. Use address_p.
6156 Remove obsolete null checks of base_object. Update call to record_use.
6157 (find_interesting_uses_op): Update call to record_group_use.
6158 (find_interesting_uses_cond): Likewise.
6159 (find_interesting_uses_address): Likewise.
6160 (get_mem_type_for_internal_fn): New function.
6161 (find_address_like_use): Likewise.
6162 (find_interesting_uses_stmt): Try find_address_like_use before
6163 calling find_interesting_uses_op.
6164 (addr_offset_valid_p): Use the iv mem_type field as the type
6165 of the addressed memory.
6166 (add_autoinc_candidates): Likewise.
6167 (get_address_cost): Likewise.
6168 (split_small_address_groups_p): Use address_p.
6169 (split_address_groups): Likewise.
6170 (add_iv_candidate_for_use): Likewise.
6171 (autoinc_possible_for_pair): Likewise.
6172 (rewrite_groups): Likewise.
6173 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6174 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6175 (get_alias_ptr_type_for_ptr_address): New function.
6176 (rewrite_use_address): Rewrite address uses in calls that were
6177 identified by find_address_like_use.
6179 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6180 Alan Hayward <alan.hayward@arm.com>
6181 David Sherwood <david.sherwood@arm.com>
6183 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6185 * gimple-expr.h (is_gimple_addressable: Likewise.
6186 * gimple-expr.c (is_gimple_address): Likewise.
6187 * internal-fn.c (expand_call_mem_ref): New function.
6188 (expand_mask_load_optab_fn): Use it.
6189 (expand_mask_store_optab_fn): Likewise.
6191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6192 Alan Hayward <alan.hayward@arm.com>
6193 David Sherwood <david.sherwood@arm.com>
6195 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6196 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6197 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6198 (cond_umax@var{mode}): Document.
6199 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6200 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6201 (cond_umin_optab, cond_umax_optab): New optabs.
6202 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6203 (COND_IOR, COND_XOR): New internal functions.
6204 * internal-fn.h (get_conditional_internal_fn): Declare.
6205 * internal-fn.c (cond_binary_direct): New macro.
6206 (expand_cond_binary_optab_fn): Likewise.
6207 (direct_cond_binary_optab_supported_p): Likewise.
6208 (get_conditional_internal_fn): New function.
6209 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6210 Cope with reduction statements that are vectorized as calls rather
6212 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6213 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6214 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6215 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6216 (UNSPEC_COND_EOR): New unspecs.
6217 (optab): Add mappings for them.
6218 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6219 (sve_int_op, sve_fp_op): New int attributes.
6221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6222 Alan Hayward <alan.hayward@arm.com>
6223 David Sherwood <david.sherwood@arm.com>
6225 * optabs.def (while_ult_optab): New optab.
6226 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6227 * internal-fn.def (WHILE_ULT): New internal function.
6228 * internal-fn.h (direct_internal_fn_supported_p): New override
6229 that takes two types as argument.
6230 * internal-fn.c (while_direct): New macro.
6231 (expand_while_optab_fn): New function.
6232 (convert_optab_supported_p): Likewise.
6233 (direct_while_optab_supported_p): New macro.
6234 * wide-int.h (wi::udiv_ceil): New function.
6235 * tree-vectorizer.h (rgroup_masks): New structure.
6236 (vec_loop_masks): New typedef.
6237 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6239 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6240 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6241 (vect_max_vf): New function.
6242 (slpeel_make_loop_iterate_ntimes): Delete.
6243 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6244 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6245 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6246 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6247 internal-fn.h, stor-layout.h and optabs-query.h.
6248 (vect_set_loop_mask): New function.
6249 (add_preheader_seq): Likewise.
6250 (add_header_seq): Likewise.
6251 (interleave_supported_p): Likewise.
6252 (vect_maybe_permute_loop_masks): Likewise.
6253 (vect_set_loop_masks_directly): Likewise.
6254 (vect_set_loop_condition_masked): Likewise.
6255 (vect_set_loop_condition_unmasked): New function, split out from
6256 slpeel_make_loop_iterate_ntimes.
6257 (slpeel_make_loop_iterate_ntimes): Rename to..
6258 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6259 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6260 (vect_do_peeling): Update call accordingly.
6261 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6263 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6264 mask_compare_type, can_fully_mask_p and fully_masked_p.
6265 (release_vec_loop_masks): New function.
6266 (_loop_vec_info): Use it to free the loop masks.
6267 (can_produce_all_loop_masks_p): New function.
6268 (vect_get_max_nscalars_per_iter): Likewise.
6269 (vect_verify_full_masking): Likewise.
6270 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6271 retries, and free the mask rgroups before retrying. Check loop-wide
6272 reasons for disallowing fully-masked loops. Make the final decision
6273 about whether use a fully-masked loop or not.
6274 (vect_estimate_min_profitable_iters): Do not assume that peeling
6275 for the number of iterations will be needed for fully-masked loops.
6276 (vectorizable_reduction): Disable fully-masked loops.
6277 (vectorizable_live_operation): Likewise.
6278 (vect_halve_mask_nunits): New function.
6279 (vect_double_mask_nunits): Likewise.
6280 (vect_record_loop_mask): Likewise.
6281 (vect_get_loop_mask): Likewise.
6282 (vect_transform_loop): Handle the case in which the final loop
6283 iteration might handle a partial vector. Call vect_set_loop_condition
6284 instead of slpeel_make_loop_iterate_ntimes.
6285 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6286 (check_load_store_masking): New function.
6287 (prepare_load_store_mask): Likewise.
6288 (vectorizable_store): Handle fully-masked loops.
6289 (vectorizable_load): Likewise.
6290 (supportable_widening_operation): Use vect_halve_mask_nunits for
6292 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6293 (vect_gen_while): New function.
6294 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6295 (aarch64_uqdec<mode>): New insn.
6297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6298 Alan Hayward <alan.hayward@arm.com>
6299 David Sherwood <david.sherwood@arm.com>
6301 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6302 (reduc_xor_scal_optab): New optabs.
6303 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6304 (reduc_xor_scal_@var{m}): Document.
6305 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6306 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6308 * fold-const-call.c (fold_const_call): Handle them.
6309 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6310 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6311 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6312 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6313 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6314 (UNSPEC_XORV): New unspecs.
6315 (optab): Add entries for them.
6316 (BITWISEV): New int iterator.
6317 (bit_reduc_op): New int attributes.
6319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6320 Alan Hayward <alan.hayward@arm.com>
6321 David Sherwood <david.sherwood@arm.com>
6323 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6324 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6325 * optabs.def (vec_shl_insert_optab): New optab.
6326 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6327 (duplicate_and_interleave): Likewise.
6328 * tree-vect-loop.c: Include internal-fn.h.
6329 (neutral_op_for_slp_reduction): New function, split out from
6330 get_initial_defs_for_reduction.
6331 (get_initial_def_for_reduction): Handle option 2 for variable-length
6332 vectors by loading the neutral value into a vector and then shifting
6333 the initial value into element 0.
6334 (get_initial_defs_for_reduction): Replace the code argument with
6335 the neutral value calculated by neutral_op_for_slp_reduction.
6336 Use gimple_build_vector for constant-length vectors.
6337 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6338 but the first group_size elements have a neutral value.
6339 Use duplicate_and_interleave otherwise.
6340 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6341 Update call to get_initial_defs_for_reduction. Handle SLP
6342 reductions for variable-length vectors by creating one vector
6343 result for each scalar result, with the elements associated
6344 with other scalar results stubbed out with the neutral value.
6345 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6346 Require IFN_VEC_SHL_INSERT for double reductions on
6347 variable-length vectors, or SLP reductions that have
6348 a neutral value. Require can_duplicate_and_interleave_p
6349 support for variable-length unchained SLP reductions if there
6350 is no neutral value, such as for MIN/MAX reductions. Also require
6351 the number of vector elements to be a multiple of the number of
6352 SLP statements when doing variable-length unchained SLP reductions.
6353 Update call to vect_create_epilog_for_reduction.
6354 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6355 and remove initial values.
6356 (duplicate_and_interleave): Make public.
6357 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6358 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6361 Alan Hayward <alan.hayward@arm.com>
6362 David Sherwood <david.sherwood@arm.com>
6364 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6365 (can_duplicate_and_interleave_p): New function.
6366 (vect_get_and_check_slp_defs): Take the vector of statements
6367 rather than just the current one. Remove excess parentheses.
6368 Restriction rejectinon of vect_constant_def and vect_external_def
6369 for variable-length vectors to boolean types, or types for which
6370 can_duplicate_and_interleave_p is false.
6371 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6372 (duplicate_and_interleave): New function.
6373 (vect_get_constant_vectors): Use gimple_build_vector for
6374 constant-length vectors and suitable variable-length constant
6375 vectors. Use duplicate_and_interleave for other variable-length
6376 vectors. Don't defer the update when inserting new statements.
6378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6379 Alan Hayward <alan.hayward@arm.com>
6380 David Sherwood <david.sherwood@arm.com>
6382 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6383 min_profitable_iters doesn't go negative.
6385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6386 Alan Hayward <alan.hayward@arm.com>
6387 David Sherwood <david.sherwood@arm.com>
6389 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6390 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6391 * optabs.def (vec_mask_load_lanes_optab): New optab.
6392 (vec_mask_store_lanes_optab): Likewise.
6393 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6394 (MASK_STORE_LANES): Likewise.
6395 * internal-fn.c (mask_load_lanes_direct): New macro.
6396 (mask_store_lanes_direct): Likewise.
6397 (expand_mask_load_optab_fn): Handle masked operations.
6398 (expand_mask_load_lanes_optab_fn): New macro.
6399 (expand_mask_store_optab_fn): Handle masked operations.
6400 (expand_mask_store_lanes_optab_fn): New macro.
6401 (direct_mask_load_lanes_optab_supported_p): Likewise.
6402 (direct_mask_store_lanes_optab_supported_p): Likewise.
6403 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6405 (vect_load_lanes_supported): Likewise.
6406 * tree-vect-data-refs.c (strip_conversion): New function.
6407 (can_group_stmts_p): Likewise.
6408 (vect_analyze_data_ref_accesses): Use it instead of checking
6409 for a pair of assignments.
6410 (vect_store_lanes_supported): Take a masked_p parameter.
6411 (vect_load_lanes_supported): Likewise.
6412 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6413 vect_store_lanes_supported and vect_load_lanes_supported.
6414 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6415 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6416 parameter. Don't allow gaps for masked accesses.
6417 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6418 and vect_load_lanes_supported.
6419 (get_load_store_type): Take a masked_p parameter and update
6420 call to get_group_load_store_type.
6421 (vectorizable_store): Update call to get_load_store_type.
6422 Handle IFN_MASK_STORE_LANES.
6423 (vectorizable_load): Update call to get_load_store_type.
6424 Handle IFN_MASK_LOAD_LANES.
6426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6427 Alan Hayward <alan.hayward@arm.com>
6428 David Sherwood <david.sherwood@arm.com>
6430 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6432 * config/aarch64/aarch64-protos.h
6433 (aarch64_sve_struct_memory_operand_p): Declare.
6434 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6435 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6436 (VPRED, vpred): Handle SVE structure modes.
6437 * config/aarch64/constraints.md (Utx): New constraint.
6438 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6439 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6440 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6441 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6442 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6443 structure modes. Split into pieces after RA.
6444 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6445 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6447 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6448 SVE structure modes.
6449 (aarch64_classify_address): Likewise.
6450 (sizetochar): Move earlier in file.
6451 (aarch64_print_operand): Handle SVE register lists.
6452 (aarch64_array_mode): New function.
6453 (aarch64_sve_struct_memory_operand_p): Likewise.
6454 (TARGET_ARRAY_MODE): Redefine.
6456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6457 Alan Hayward <alan.hayward@arm.com>
6458 David Sherwood <david.sherwood@arm.com>
6460 * target.def (array_mode): New target hook.
6461 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6462 * doc/tm.texi: Regenerate.
6463 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6464 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6465 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6467 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6471 Alan Hayward <alan.hayward@arm.com>
6472 David Sherwood <david.sherwood@arm.com>
6474 * fold-const.c (fold_binary_loc): Check the argument types
6475 rather than the result type when testing for a vector operation.
6477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6479 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6480 * doc/tm.texi: Regenerate.
6482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6483 Alan Hayward <alan.hayward@arm.com>
6484 David Sherwood <david.sherwood@arm.com>
6486 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6487 (sve): Document new AArch64 extension.
6488 * doc/md.texi (w): Extend the description of the AArch64
6489 constraint to include SVE vectors.
6490 (Upl, Upa): Document new AArch64 predicate constraints.
6491 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6493 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6494 (msve-vector-bits=): New option.
6495 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6496 SVE when these are disabled.
6497 (sve): New extension.
6498 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6499 modes. Adjust their number of units based on aarch64_sve_vg.
6500 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6501 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6502 aarch64_addr_query_type.
6503 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6504 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6505 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6506 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6507 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6508 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6509 (aarch64_simd_imm_zero_p): Delete.
6510 (aarch64_check_zero_based_sve_index_immediate): Declare.
6511 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6512 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6513 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6514 (aarch64_sve_float_mul_immediate_p): Likewise.
6515 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6517 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6518 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6519 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6520 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6521 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6522 (aarch64_regmode_natural_size): Likewise.
6523 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6524 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6526 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6527 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6528 for VG and the SVE predicate registers.
6529 (V_ALIASES): Add a "z"-prefixed alias.
6530 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6531 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6532 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6533 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6534 (REG_CLASS_NAMES): Add entries for them.
6535 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6536 and the predicate registers.
6537 (aarch64_sve_vg): Declare.
6538 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6539 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6540 (REGMODE_NATURAL_SIZE): Define.
6541 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6543 * config/aarch64/aarch64.c: Include cfgrtl.h.
6544 (simd_immediate_info): Add a constructor for series vectors,
6545 and an associated step field.
6546 (aarch64_sve_vg): New variable.
6547 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6548 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6549 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6550 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6551 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6552 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6553 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6554 (aarch64_get_mask_mode): New functions.
6555 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6556 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6557 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6558 predicate modes and predicate registers. Explicitly restrict
6559 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6560 to store a vector mode if it is recognized by
6561 aarch64_classify_vector_mode.
6562 (aarch64_regmode_natural_size): New function.
6563 (aarch64_hard_regno_caller_save_mode): Return the original mode
6565 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6566 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6567 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6568 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6570 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6571 does not overlap dest if the function is frame-related. Handle
6573 (aarch64_split_add_offset): New function.
6574 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6575 them aarch64_add_offset.
6576 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6577 and update call to aarch64_sub_sp.
6578 (aarch64_add_cfa_expression): New function.
6579 (aarch64_expand_prologue): Pass extra temporary registers to the
6580 functions above. Handle the case in which we need to emit new
6581 DW_CFA_expressions for registers that were originally saved
6582 relative to the stack pointer, but now have to be expressed
6583 relative to the frame pointer.
6584 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6586 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6587 IP0 and IP1 values for SVE frames.
6588 (aarch64_expand_vec_series): New function.
6589 (aarch64_expand_sve_widened_duplicate): Likewise.
6590 (aarch64_expand_sve_const_vector): Likewise.
6591 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6592 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6593 into the register, rather than emitting a SET directly.
6594 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6595 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6596 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6597 (offset_9bit_signed_scaled_p): New functions.
6598 (aarch64_replicate_bitmask_imm): New function.
6599 (aarch64_bitmask_imm): Use it.
6600 (aarch64_cannot_force_const_mem): Reject expressions involving
6601 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6602 (aarch64_classify_index): Handle SVE indices, by requiring
6603 a plain register index with a scale that matches the element size.
6604 (aarch64_classify_address): Handle SVE addresses. Assert that
6605 the mode of the address is VOIDmode or an integer mode.
6606 Update call to aarch64_classify_symbol.
6607 (aarch64_classify_symbolic_expression): Update call to
6608 aarch64_classify_symbol.
6609 (aarch64_const_vec_all_in_range_p): New function.
6610 (aarch64_print_vector_float_operand): Likewise.
6611 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6612 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6613 and the FP immediates 1.0 and 0.5.
6614 (aarch64_print_address_internal): Handle SVE addresses.
6615 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6616 (aarch64_regno_regclass): Handle predicate registers.
6617 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6619 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6620 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6621 (aarch64_convert_sve_vector_bits): New function.
6622 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6623 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6625 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6626 Handle SVE vector and predicate modes. Accept VL-based constants
6627 that need only one temporary register, and VL offsets that require
6628 no temporary registers.
6629 (aarch64_conditional_register_usage): Mark the predicate registers
6630 as fixed if SVE isn't available.
6631 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6632 Return true for SVE vector and predicate modes.
6633 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6634 rather than an unsigned int. Handle SVE modes.
6635 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6637 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6639 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6640 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6641 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6642 (aarch64_sve_float_mul_immediate_p): New functions.
6643 (aarch64_sve_valid_immediate): New function.
6644 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6645 Explicitly reject structure modes. Check for INDEX constants.
6646 Handle PTRUE and PFALSE constants.
6647 (aarch64_check_zero_based_sve_index_immediate): New function.
6648 (aarch64_simd_imm_zero_p): Delete.
6649 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6650 vector modes. Accept constants in the range of CNT[BHWD].
6651 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6652 ask for an Advanced SIMD mode.
6653 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6654 (aarch64_simd_vector_alignment): Handle SVE predicates.
6655 (aarch64_vectorize_preferred_vector_alignment): New function.
6656 (aarch64_simd_vector_alignment_reachable): Use it instead of
6658 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6659 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6661 (MAX_VECT_LEN): Delete.
6662 (expand_vec_perm_d): Add a vec_flags field.
6663 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6664 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6665 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6667 (aarch64_evpc_rev): Rename to...
6668 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6669 (aarch64_evpc_rev_global): New function.
6670 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6671 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6673 (aarch64_evpc_sve_tbl): New function.
6674 (aarch64_expand_vec_perm_const_1): Update after rename of
6675 aarch64_evpc_rev. Handle SVE permutes too, trying
6676 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6677 than aarch64_evpc_tbl.
6678 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6679 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6680 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6681 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6682 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6683 (aarch64_expand_sve_vcond): New functions.
6684 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6685 of aarch64_vector_mode_p.
6686 (aarch64_dwarf_poly_indeterminate_value): New function.
6687 (aarch64_compute_pressure_classes): Likewise.
6688 (aarch64_can_change_mode_class): Likewise.
6689 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6690 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6691 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6692 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6693 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6694 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6695 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6696 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6698 (Dn, Dl, Dr): Accept const as well as const_vector.
6699 (Dz): Likewise. Compare against CONST0_RTX.
6700 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6701 of "vector" where appropriate.
6702 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6703 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6704 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6705 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6706 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6707 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6708 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6709 (v_int_equiv): Extend to SVE modes.
6710 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6712 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6713 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6714 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6715 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6716 (SVE_COND_FP_CMP): New int iterators.
6717 (perm_hilo): Handle the new unpack unspecs.
6718 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6720 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6721 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6722 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6723 (aarch64_equality_operator, aarch64_constant_vector_operand)
6724 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6725 (aarch64_sve_nonimmediate_operand): Likewise.
6726 (aarch64_sve_general_operand): Likewise.
6727 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6728 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6729 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6730 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6731 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6732 (aarch64_sve_float_arith_immediate): Likewise.
6733 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6734 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6735 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6736 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6737 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6738 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6739 (aarch64_sve_float_arith_operand): Likewise.
6740 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6741 (aarch64_sve_float_mul_operand): Likewise.
6742 (aarch64_sve_vec_perm_operand): Likewise.
6743 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6744 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6745 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6746 as well as const_vector.
6747 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6748 in file. Use CONST0_RTX and CONSTM1_RTX.
6749 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6750 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6751 Use aarch64_simd_imm_zero.
6752 * config/aarch64/aarch64-sve.md: New file.
6753 * config/aarch64/aarch64.md: Include it.
6754 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6755 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6756 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6757 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6758 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6759 (sve): New attribute.
6760 (enabled): Disable instructions with the sve attribute unless
6762 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6763 aarch64_expand_mov_immediate.
6764 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6765 CNT[BHSD] immediates.
6766 (movti): Split CONST_POLY_INT moves into two halves.
6767 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6768 Split additions that need a temporary here if the destination
6769 is the stack pointer.
6770 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6771 (*add<mode>3_poly_1): New instruction.
6772 (set_clobber_cc): New expander.
6774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6776 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6777 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6778 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6779 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6780 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6781 Change innermode from fixed_mode_size to machine_mode.
6782 (simplify_subreg): Update call accordingly. Handle a constant-sized
6783 subreg of a variable-length CONST_VECTOR.
6785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6786 Alan Hayward <alan.hayward@arm.com>
6787 David Sherwood <david.sherwood@arm.com>
6789 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6790 (add_offset_to_base): New function, split out from...
6791 (create_mem_ref): ...here. When handling a scale other than 1,
6792 check first whether the address is valid without the offset.
6793 Add it into the base if so, leaving the index and scale as-is.
6795 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6798 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6799 fold_for_warn before checking if arg2 is INTEGER_CST.
6801 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6803 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6804 (store_multiple_operation): Delete.
6805 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6806 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6807 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6808 guarded by TARGET_STRING.
6809 (rs6000_output_load_multiple): Delete.
6810 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6811 OPTION_MASK_STRING / TARGET_STRING handling.
6812 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6813 (const rs6000_opt_masks) <"string">: Change mask to 0.
6814 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6815 (MASK_STRING): Delete.
6816 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6818 (load_multiple): Delete.
6825 (store_multiple): Delete.
6832 (movmemsi_8reg): Delete.
6833 (corresponding unnamed define_insn): Delete.
6834 (movmemsi_6reg): Delete.
6835 (corresponding unnamed define_insn): Delete.
6836 (movmemsi_4reg): Delete.
6837 (corresponding unnamed define_insn): Delete.
6838 (movmemsi_2reg): Delete.
6839 (corresponding unnamed define_insn): Delete.
6840 (movmemsi_1reg): Delete.
6841 (corresponding unnamed define_insn): Delete.
6842 * config/rs6000/rs6000.opt (mno-string): New.
6843 (mstring): Replace by deprecation warning stub.
6844 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6846 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6848 * regrename.c (regrename_do_replace): If replacing the same
6849 reg multiple times, try to reuse last created gen_raw_REG.
6852 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6853 main to workaround a bug in GDB.
6855 2018-01-12 Tom de Vries <tom@codesourcery.com>
6858 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6860 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6862 PR rtl-optimization/80481
6863 * ira-color.c (get_cap_member): New function.
6864 (allocnos_conflict_by_live_ranges_p): Use it.
6865 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6866 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6868 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6871 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6872 (*saddl_se_1): Ditto.
6874 (*ssubl_se_1): Ditto.
6876 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6878 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6879 rather than wi::to_widest for DR_INITs.
6880 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6881 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6882 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6884 (vect_analyze_group_access_1): Note that here.
6886 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6888 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6889 polynomial type sizes.
6891 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6893 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6894 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6895 (gimple_add_tmp_var): Likewise.
6897 2018-01-12 Martin Liska <mliska@suse.cz>
6899 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6900 (gimple_alloc_sizes): Likewise.
6901 (dump_gimple_statistics): Use PRIu64 in printf format.
6902 * gimple.h: Change uint64_t to int.
6904 2018-01-12 Martin Liska <mliska@suse.cz>
6906 * tree-core.h: Use uint64_t instead of int.
6907 * tree.c (tree_node_counts): Likewise.
6908 (tree_node_sizes): Likewise.
6909 (dump_tree_statistics): Use PRIu64 in printf format.
6911 2018-01-12 Martin Liska <mliska@suse.cz>
6913 * Makefile.in: As qsort_chk is implemented in vec.c, add
6914 vec.o to linkage of gencfn-macros.
6915 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6916 passing the info to record_node_allocation_statistics.
6917 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6919 * ggc-common.c (struct ggc_usage): Add operator== and use
6920 it in operator< and compare function.
6921 * mem-stats.h (struct mem_usage): Likewise.
6922 * vec.c (struct vec_usage): Remove operator< and compare
6923 function. Can be simply inherited.
6925 2018-01-12 Martin Jambor <mjambor@suse.cz>
6928 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6929 * tree-ssa-math-opts.c: Include domwalk.h.
6930 (convert_mult_to_fma_1): New function.
6931 (fma_transformation_info): New type.
6932 (fma_deferring_state): Likewise.
6933 (cancel_fma_deferring): New function.
6934 (result_of_phi): Likewise.
6935 (last_fma_candidate_feeds_initial_phi): Likewise.
6936 (convert_mult_to_fma): Added deferring logic, split actual
6937 transformation to convert_mult_to_fma_1.
6938 (math_opts_dom_walker): New type.
6939 (math_opts_dom_walker::after_dom_children): New method, body moved
6940 here from pass_optimize_widening_mul::execute, added deferring logic
6942 (pass_optimize_widening_mul::execute): Moved most of code to
6943 math_opts_dom_walker::after_dom_children.
6944 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6945 * config/i386/i386.c (ix86_option_override_internal): Added
6946 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6948 2018-01-12 Richard Biener <rguenther@suse.de>
6951 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6952 inline instance vars.
6954 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6957 * config/rx/rx.c (rx_is_restricted_memory_address):
6960 2018-01-12 Richard Biener <rguenther@suse.de>
6962 PR tree-optimization/80846
6963 * target.def (split_reduction): New target hook.
6964 * targhooks.c (default_split_reduction): New function.
6965 * targhooks.h (default_split_reduction): Declare.
6966 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6967 target requests first reduce vectors by combining low and high
6969 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6970 (get_vectype_for_scalar_type_and_size): Export.
6971 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6972 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6973 * doc/tm.texi: Regenerate.
6974 * config/i386/i386.c (ix86_split_reduction): Implement
6975 TARGET_VECTORIZE_SPLIT_REDUCTION.
6977 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6980 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6981 in PIC mode except for TARGET_VXWORKS_RTP.
6982 * config/sparc/sparc.c: Include cfgrtl.h.
6983 (TARGET_INIT_PIC_REG): Define.
6984 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6985 (sparc_pic_register_p): New predicate.
6986 (sparc_legitimate_address_p): Use it.
6987 (sparc_legitimize_pic_address): Likewise.
6988 (sparc_delegitimize_address): Likewise.
6989 (sparc_mode_dependent_address_p): Likewise.
6990 (gen_load_pcrel_sym): Remove 4th parameter.
6991 (load_got_register): Adjust call to above. Remove obsolete stuff.
6992 (sparc_expand_prologue): Do not call load_got_register here.
6993 (sparc_flat_expand_prologue): Likewise.
6994 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6995 (sparc_use_pseudo_pic_reg): New function.
6996 (sparc_init_pic_reg): Likewise.
6997 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6998 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7000 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7002 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7003 Add item for branch_cost.
7005 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7007 PR rtl-optimization/83565
7008 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7009 not extend the result to a larger mode for rotate operations.
7010 (num_sign_bit_copies1): Likewise.
7012 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7015 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7017 Use values-Xc.o for -pedantic.
7018 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7020 2018-01-12 Martin Liska <mliska@suse.cz>
7023 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7025 (possible_polymorphic_call_targets): Use it.
7026 (ipa_devirt): Likewise.
7028 2018-01-12 Martin Liska <mliska@suse.cz>
7030 * profile-count.h (enum profile_quality): Use 0 as invalid
7031 enum value of profile_quality.
7033 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7035 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7036 -mext-string options.
7038 2018-01-12 Richard Biener <rguenther@suse.de>
7040 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7041 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7042 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7044 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7046 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7048 * configure.ac (--with-long-double-format): Add support for the
7049 configuration option to change the default long double format on
7051 * config.gcc (powerpc*-linux*-*): Likewise.
7052 * configure: Regenerate.
7053 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7054 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7055 used without modification.
7057 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7059 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7060 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7061 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7062 MISC_BUILTIN_SPEC_BARRIER.
7063 (rs6000_init_builtins): Likewise.
7064 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7066 (speculation_barrier): New define_insn.
7067 * doc/extend.texi: Document __builtin_speculation_barrier.
7069 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7072 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7073 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7074 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7076 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7077 integral modes instead of "ss" and "sd".
7078 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7079 vectors with 32-bit and 64-bit elements.
7080 (vecdupssescalarmodesuffix): New mode attribute.
7081 (vec_dup<mode>): Use it.
7083 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7086 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7087 frame if argument is passed on stack.
7089 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7092 * ree.c (combine_reaching_defs): Optimize also
7093 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7094 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7096 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7099 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7101 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7104 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7105 after they are computed.
7107 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7109 PR tree-optimization/83695
7110 * gimple-loop-linterchange.cc
7111 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7112 reset cached scev information after interchange.
7113 (pass_linterchange::execute): Remove call to scev_reset_htab.
7115 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7117 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7118 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7119 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7120 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7121 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7122 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7123 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7124 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7125 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7126 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7127 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7128 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7129 (V_lane_reg): Likewise.
7130 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7132 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7133 (vfmal_lane_low<mode>_intrinsic,
7134 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7135 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7136 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7137 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7138 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7139 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7141 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7143 * config/arm/arm-cpus.in (fp16fml): New feature.
7144 (ALL_SIMD): Add fp16fml.
7145 (armv8.2-a): Add fp16fml as an option.
7146 (armv8.3-a): Likewise.
7147 (armv8.4-a): Add fp16fml as part of fp16.
7148 * config/arm/arm.h (TARGET_FP16FML): Define.
7149 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7151 * config/arm/arm-modes.def (V2HF): Define.
7152 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7153 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7154 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7155 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7156 vfmsl_low, vfmsl_high): New set of builtins.
7157 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7158 (vfml_op): New code attribute.
7159 (VFMLHALVES): New int iterator.
7160 (VFML, VFMLSEL): New mode attributes.
7161 (V_reg): Define mapping for V2HF.
7162 (V_hi, V_lo): New mode attributes.
7163 (VF_constraint): Likewise.
7164 (vfml_half, vfml_half_selector): New int attributes.
7165 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7167 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7168 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7170 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7171 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7172 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7173 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7175 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7176 Document new effective target and option set.
7178 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7180 * config/arm/arm-cpus.in (armv8_4): New feature.
7181 (ARMv8_4a): New fgroup.
7182 (armv8.4-a): New arch.
7183 * config/arm/arm-tables.opt: Regenerate.
7184 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7185 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7186 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7187 Add matching rules for -march=armv8.4-a and extensions.
7188 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7190 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7193 * config/rx/rx.md (BW): New mode attribute.
7194 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7196 2018-01-11 Richard Biener <rguenther@suse.de>
7198 PR tree-optimization/83435
7199 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7200 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7201 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7203 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7204 Alan Hayward <alan.hayward@arm.com>
7205 David Sherwood <david.sherwood@arm.com>
7207 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7209 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7210 (aarch64_print_address_internal): Use it to check for a zero offset.
7212 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7213 Alan Hayward <alan.hayward@arm.com>
7214 David Sherwood <david.sherwood@arm.com>
7216 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7217 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7218 Return a poly_int64 rather than a HOST_WIDE_INT.
7219 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7220 rather than a HOST_WIDE_INT.
7221 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7222 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7223 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7224 final_offset from HOST_WIDE_INT to poly_int64.
7225 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7226 to_constant when getting the number of units in an Advanced SIMD
7228 (aarch64_builtin_vectorized_function): Check for a constant number
7230 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7232 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7233 attribute instead of GET_MODE_NUNITS.
7234 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7235 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7236 GET_MODE_SIZE for fixed-size registers.
7237 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7238 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7239 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7240 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7241 (aarch64_print_operand, aarch64_print_address_internal)
7242 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7243 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7244 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7245 Handle polynomial GET_MODE_SIZE.
7246 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7247 wider than SImode without modification.
7248 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7249 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7250 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7251 passing and returning SVE modes.
7252 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7253 rather than GEN_INT.
7254 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7255 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7256 (aarch64_allocate_and_probe_stack_space): Likewise.
7257 (aarch64_layout_frame): Cope with polynomial offsets.
7258 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7259 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7261 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7262 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7263 poly_int64 rather than a HOST_WIDE_INT.
7264 (aarch64_get_separate_components, aarch64_process_components)
7265 (aarch64_expand_prologue, aarch64_expand_epilogue)
7266 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7267 (aarch64_anchor_offset): New function, split out from...
7268 (aarch64_legitimize_address): ...here.
7269 (aarch64_builtin_vectorization_cost): Handle polynomial
7270 TYPE_VECTOR_SUBPARTS.
7271 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7273 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7274 number of elements from the PARALLEL rather than the mode.
7275 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7276 rather than GET_MODE_BITSIZE.
7277 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7278 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7279 (aarch64_expand_vec_perm_const_1): Handle polynomial
7280 d->perm.length () and d->perm elements.
7281 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7282 Apply to_constant to d->perm elements.
7283 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7284 polynomial CONST_VECTOR_NUNITS.
7285 (aarch64_move_pointer): Take amount as a poly_int64 rather
7287 (aarch64_progress_pointer): Avoid temporary variable.
7288 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7289 the mode attribute instead of GET_MODE.
7291 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7292 Alan Hayward <alan.hayward@arm.com>
7293 David Sherwood <david.sherwood@arm.com>
7295 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7296 x exists before using it.
7297 (aarch64_add_constant_internal): Rename to...
7298 (aarch64_add_offset_1): ...this. Replace regnum with separate
7299 src and dest rtxes. Handle the case in which they're different,
7300 including when the offset is zero. Replace scratchreg with an rtx.
7301 Use 2 additions if there is no spare register into which we can
7302 move a 16-bit constant.
7303 (aarch64_add_constant): Delete.
7304 (aarch64_add_offset): Replace reg with separate src and dest
7305 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7306 Use aarch64_add_offset_1.
7307 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7308 an rtx rather than an int. Take the delta as a poly_int64
7309 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7310 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7311 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7312 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7313 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7315 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7316 aarch64_add_constant.
7318 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7320 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7321 Use scalar_float_mode.
7323 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7325 * config/aarch64/aarch64-simd.md
7326 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7327 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7328 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7329 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7330 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7331 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7332 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7333 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7334 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7335 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7337 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7340 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7341 targ_options->x_arm_arch_string is non NULL.
7343 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7345 * config/aarch64/aarch64.h
7346 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7348 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7351 * expmed.c (emit_store_flag_force): Swap if const op0
7352 and change VOIDmode to mode of op0.
7354 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7356 PR rtl-optimization/83761
7357 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7358 than bytes to mode_for_size.
7360 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7363 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7364 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7367 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7370 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7371 when in layout mode.
7372 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7373 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7376 2018-01-10 Michael Collison <michael.collison@arm.com>
7378 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7379 * config/aarch64/aarch64-option-extension.def: Add
7380 AARCH64_OPT_EXTENSION of 'fp16fml'.
7381 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7382 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7383 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7384 * config/aarch64/constraints.md (Ui7): New constraint.
7385 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7386 (VFMLA_SEL_W): Ditto.
7389 (VFMLA16_LOW): New int iterator.
7390 (VFMLA16_HIGH): Ditto.
7391 (UNSPEC_FMLAL): New unspec.
7392 (UNSPEC_FMLSL): Ditto.
7393 (UNSPEC_FMLAL2): Ditto.
7394 (UNSPEC_FMLSL2): Ditto.
7395 (f16mac): New code attribute.
7396 * config/aarch64/aarch64-simd-builtins.def
7397 (aarch64_fmlal_lowv2sf): Ditto.
7398 (aarch64_fmlsl_lowv2sf): Ditto.
7399 (aarch64_fmlalq_lowv4sf): Ditto.
7400 (aarch64_fmlslq_lowv4sf): Ditto.
7401 (aarch64_fmlal_highv2sf): Ditto.
7402 (aarch64_fmlsl_highv2sf): Ditto.
7403 (aarch64_fmlalq_highv4sf): Ditto.
7404 (aarch64_fmlslq_highv4sf): Ditto.
7405 (aarch64_fmlal_lane_lowv2sf): Ditto.
7406 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7407 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7408 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7409 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7410 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7411 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7412 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7413 (aarch64_fmlal_lane_highv2sf): Ditto.
7414 (aarch64_fmlsl_lane_highv2sf): Ditto.
7415 (aarch64_fmlal_laneq_highv2sf): Ditto.
7416 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7417 (aarch64_fmlalq_lane_highv4sf): Ditto.
7418 (aarch64_fmlsl_lane_highv4sf): Ditto.
7419 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7420 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7421 * config/aarch64/aarch64-simd.md:
7422 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7423 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7424 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7425 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7426 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7427 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7428 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7429 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7430 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7431 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7432 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7433 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7434 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7435 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7436 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7437 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7438 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7439 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7440 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7441 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7442 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7443 (vfmlsl_low_u32): Ditto.
7444 (vfmlalq_low_u32): Ditto.
7445 (vfmlslq_low_u32): Ditto.
7446 (vfmlal_high_u32): Ditto.
7447 (vfmlsl_high_u32): Ditto.
7448 (vfmlalq_high_u32): Ditto.
7449 (vfmlslq_high_u32): Ditto.
7450 (vfmlal_lane_low_u32): Ditto.
7451 (vfmlsl_lane_low_u32): Ditto.
7452 (vfmlal_laneq_low_u32): Ditto.
7453 (vfmlsl_laneq_low_u32): Ditto.
7454 (vfmlalq_lane_low_u32): Ditto.
7455 (vfmlslq_lane_low_u32): Ditto.
7456 (vfmlalq_laneq_low_u32): Ditto.
7457 (vfmlslq_laneq_low_u32): Ditto.
7458 (vfmlal_lane_high_u32): Ditto.
7459 (vfmlsl_lane_high_u32): Ditto.
7460 (vfmlal_laneq_high_u32): Ditto.
7461 (vfmlsl_laneq_high_u32): Ditto.
7462 (vfmlalq_lane_high_u32): Ditto.
7463 (vfmlslq_lane_high_u32): Ditto.
7464 (vfmlalq_laneq_high_u32): Ditto.
7465 (vfmlslq_laneq_high_u32): Ditto.
7466 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7467 (AARCH64_FL_FOR_ARCH8_4): New.
7468 (AARCH64_ISA_F16FML): New ISA flag.
7469 (TARGET_F16FML): New feature flag for fp16fml.
7470 (doc/invoke.texi): Document new fp16fml option.
7472 2018-01-10 Michael Collison <michael.collison@arm.com>
7474 * config/aarch64/aarch64-builtins.c:
7475 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7476 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7477 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7478 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7479 (AARCH64_ISA_SHA3): New ISA flag.
7480 (TARGET_SHA3): New feature flag for sha3.
7481 * config/aarch64/iterators.md (sha512_op): New int attribute.
7482 (CRYPTO_SHA512): New int iterator.
7483 (UNSPEC_SHA512H): New unspec.
7484 (UNSPEC_SHA512H2): Ditto.
7485 (UNSPEC_SHA512SU0): Ditto.
7486 (UNSPEC_SHA512SU1): Ditto.
7487 * config/aarch64/aarch64-simd-builtins.def
7488 (aarch64_crypto_sha512hqv2di): New builtin.
7489 (aarch64_crypto_sha512h2qv2di): Ditto.
7490 (aarch64_crypto_sha512su0qv2di): Ditto.
7491 (aarch64_crypto_sha512su1qv2di): Ditto.
7492 (aarch64_eor3qv8hi): Ditto.
7493 (aarch64_rax1qv2di): Ditto.
7494 (aarch64_xarqv2di): Ditto.
7495 (aarch64_bcaxqv8hi): Ditto.
7496 * config/aarch64/aarch64-simd.md:
7497 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7498 (aarch64_crypto_sha512su0qv2di): Ditto.
7499 (aarch64_crypto_sha512su1qv2di): Ditto.
7500 (aarch64_eor3qv8hi): Ditto.
7501 (aarch64_rax1qv2di): Ditto.
7502 (aarch64_xarqv2di): Ditto.
7503 (aarch64_bcaxqv8hi): Ditto.
7504 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7505 (vsha512h2q_u64): Ditto.
7506 (vsha512su0q_u64): Ditto.
7507 (vsha512su1q_u64): Ditto.
7508 (veor3q_u16): Ditto.
7509 (vrax1q_u64): Ditto.
7511 (vbcaxq_u16): Ditto.
7512 * config/arm/types.md (crypto_sha512): New type attribute.
7513 (crypto_sha3): Ditto.
7514 (doc/invoke.texi): Document new sha3 option.
7516 2018-01-10 Michael Collison <michael.collison@arm.com>
7518 * config/aarch64/aarch64-builtins.c:
7519 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7520 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7521 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7522 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7523 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7524 (AARCH64_ISA_SM4): New ISA flag.
7525 (TARGET_SM4): New feature flag for sm4.
7526 * config/aarch64/aarch64-simd-builtins.def
7527 (aarch64_sm3ss1qv4si): Ditto.
7528 (aarch64_sm3tt1aq4si): Ditto.
7529 (aarch64_sm3tt1bq4si): Ditto.
7530 (aarch64_sm3tt2aq4si): Ditto.
7531 (aarch64_sm3tt2bq4si): Ditto.
7532 (aarch64_sm3partw1qv4si): Ditto.
7533 (aarch64_sm3partw2qv4si): Ditto.
7534 (aarch64_sm4eqv4si): Ditto.
7535 (aarch64_sm4ekeyqv4si): Ditto.
7536 * config/aarch64/aarch64-simd.md:
7537 (aarch64_sm3ss1qv4si): Ditto.
7538 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7539 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7540 (aarch64_sm4eqv4si): Ditto.
7541 (aarch64_sm4ekeyqv4si): Ditto.
7542 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7543 (sm3part_op): Ditto.
7544 (CRYPTO_SM3TT): Ditto.
7545 (CRYPTO_SM3PART): Ditto.
7546 (UNSPEC_SM3SS1): New unspec.
7547 (UNSPEC_SM3TT1A): Ditto.
7548 (UNSPEC_SM3TT1B): Ditto.
7549 (UNSPEC_SM3TT2A): Ditto.
7550 (UNSPEC_SM3TT2B): Ditto.
7551 (UNSPEC_SM3PARTW1): Ditto.
7552 (UNSPEC_SM3PARTW2): Ditto.
7553 (UNSPEC_SM4E): Ditto.
7554 (UNSPEC_SM4EKEY): Ditto.
7555 * config/aarch64/constraints.md (Ui2): New constraint.
7556 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7557 * config/arm/types.md (crypto_sm3): New type attribute.
7558 (crypto_sm4): Ditto.
7559 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7560 (vsm3tt1aq_u32): Ditto.
7561 (vsm3tt1bq_u32): Ditto.
7562 (vsm3tt2aq_u32): Ditto.
7563 (vsm3tt2bq_u32): Ditto.
7564 (vsm3partw1q_u32): Ditto.
7565 (vsm3partw2q_u32): Ditto.
7566 (vsm4eq_u32): Ditto.
7567 (vsm4ekeyq_u32): Ditto.
7568 (doc/invoke.texi): Document new sm4 option.
7570 2018-01-10 Michael Collison <michael.collison@arm.com>
7572 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7573 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7574 (AARCH64_FL_FOR_ARCH8_4): New.
7575 (AARCH64_FL_V8_4): New flag.
7576 (doc/invoke.texi): Document new armv8.4-a option.
7578 2018-01-10 Michael Collison <michael.collison@arm.com>
7580 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7581 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7582 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7583 * config/aarch64/aarch64-option-extension.def: Add
7584 AARCH64_OPT_EXTENSION of 'sha2'.
7585 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7586 (crypto): Disable sha2 and aes if crypto disabled.
7587 (crypto): Enable aes and sha2 if enabled.
7588 (simd): Disable sha2 and aes if simd disabled.
7589 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7591 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7592 (TARGET_SHA2): New feature flag for sha2.
7593 (TARGET_AES): New feature flag for aes.
7594 * config/aarch64/aarch64-simd.md:
7595 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7596 conditional on TARGET_AES.
7597 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7598 (aarch64_crypto_sha1hsi): Make pattern conditional
7600 (aarch64_crypto_sha1hv4si): Ditto.
7601 (aarch64_be_crypto_sha1hv4si): Ditto.
7602 (aarch64_crypto_sha1su1v4si): Ditto.
7603 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7604 (aarch64_crypto_sha1su0v4si): Ditto.
7605 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7606 (aarch64_crypto_sha256su0v4si): Ditto.
7607 (aarch64_crypto_sha256su1v4si): Ditto.
7608 (doc/invoke.texi): Document new aes and sha2 options.
7610 2018-01-10 Martin Sebor <msebor@redhat.com>
7612 PR tree-optimization/83781
7613 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7616 2018-01-11 Martin Sebor <msebor@gmail.com>
7617 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7619 PR tree-optimization/83501
7620 PR tree-optimization/81703
7622 * tree-ssa-strlen.c (get_string_cst): Rename...
7623 (get_string_len): ...to this. Handle global constants.
7624 (handle_char_store): Adjust.
7626 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7627 Jim Wilson <jimw@sifive.com>
7629 * config/riscv/riscv-protos.h (riscv_output_return): New.
7630 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7631 (riscv_attribute_table, riscv_output_return),
7632 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7633 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7634 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7636 (riscv_expand_prologue): Add early return for naked function.
7637 (riscv_expand_epilogue): Likewise.
7638 (riscv_function_ok_for_sibcall): Return false for naked function.
7639 (riscv_set_current_function): New.
7640 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7641 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7642 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7643 * doc/extend.texi (RISC-V Function Attributes): New.
7645 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7647 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7648 check for 128-bit long double before checking TCmode.
7649 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7650 128-bit long doubles before checking TFmode or TCmode.
7651 (FLOAT128_IBM_P): Likewise.
7653 2018-01-10 Martin Sebor <msebor@redhat.com>
7655 PR tree-optimization/83671
7656 * builtins.c (c_strlen): Unconditionally return zero for the empty
7658 Use -Warray-bounds for warnings.
7659 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7660 for non-constant array indices with COMPONENT_REF, arrays of
7661 arrays, and pointers to arrays.
7662 (gimple_fold_builtin_strlen): Determine and set length range for
7663 non-constant character arrays.
7665 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7668 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7671 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7673 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7675 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7678 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7679 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7680 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7681 indexed_or_indirect_operand predicate.
7682 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7683 (*vsx_le_perm_load_v8hi): Likewise.
7684 (*vsx_le_perm_load_v16qi): Likewise.
7685 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7686 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7687 (*vsx_le_perm_store_v8hi): Likewise.
7688 (*vsx_le_perm_store_v16qi): Likewise.
7689 (eight unnamed splitters): Likewise.
7691 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7693 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7694 * config/rs6000/emmintrin.h: Likewise.
7695 * config/rs6000/mmintrin.h: Likewise.
7696 * config/rs6000/xmmintrin.h: Likewise.
7698 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7701 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7703 * tree.c (tree_nop_conversion): Return true for location wrapper
7705 (maybe_wrap_with_location): New function.
7706 (selftest::check_strip_nops): New function.
7707 (selftest::test_location_wrappers): New function.
7708 (selftest::tree_c_tests): Call it.
7709 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7710 (maybe_wrap_with_location): New decl.
7711 (EXPR_LOCATION_WRAPPER_P): New macro.
7712 (location_wrapper_p): New inline function.
7713 (tree_strip_any_location_wrapper): New inline function.
7715 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7718 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7719 stack_realign_offset for the largest alignment of stack slot
7721 (ix86_find_max_used_stack_alignment): New function.
7722 (ix86_finalize_stack_frame_flags): Use it. Set
7723 max_used_stack_alignment if we don't realign stack.
7724 * config/i386/i386.h (machine_function): Add
7725 max_used_stack_alignment.
7727 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7729 * config/arm/arm.opt (-mbranch-cost): New option.
7730 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7733 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7736 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7737 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7739 2018-01-10 Richard Biener <rguenther@suse.de>
7742 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7743 early out so it also covers the case where we have a non-NULL
7746 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7748 PR tree-optimization/83753
7749 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7750 for non-strided grouped accesses if the number of elements is 1.
7752 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7755 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7756 * i386.h (TARGET_USE_GATHER): Define.
7757 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7759 2018-01-10 Martin Liska <mliska@suse.cz>
7762 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7763 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7765 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7766 CLEANUP_NO_PARTITIONING is not set.
7768 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7770 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7771 for vectors, as a partial revert of r254296.
7772 * rtl.h (const_vec_p): Delete.
7773 (const_vec_duplicate_p): Don't test for vector CONSTs.
7774 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7775 * expmed.c (make_tree): Likewise.
7778 * common.md (E, F): Use CONSTANT_P instead of checking for
7780 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7781 checking for CONST_VECTOR.
7783 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7786 * predict.c (force_edge_cold): Handle in more sane way edges
7789 2018-01-09 Carl Love <cel@us.ibm.com>
7791 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7793 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7794 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7795 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7796 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7797 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7798 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7799 * config/rs6000/rs6000-protos.h: Add extern defition for
7800 rs6000_generate_float2_double_code.
7801 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7803 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7804 (float2_v2df): Add define_expand.
7806 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7809 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7810 op_mode in the force_to_mode call.
7812 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7814 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7815 instead of checking each element individually.
7816 (aarch64_evpc_uzp): Likewise.
7817 (aarch64_evpc_zip): Likewise.
7818 (aarch64_evpc_ext): Likewise.
7819 (aarch64_evpc_rev): Likewise.
7820 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7821 instead of checking each element individually. Return true without
7823 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7824 whether all selected elements come from the same input, instead of
7825 checking each element individually. Remove calls to gen_rtx_REG,
7826 start_sequence and end_sequence and instead assert that no rtl is
7829 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7831 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7832 order of HIGH and CONST checks.
7834 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7836 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7837 if the destination isn't an SSA_NAME.
7839 2018-01-09 Richard Biener <rguenther@suse.de>
7841 PR tree-optimization/83668
7842 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7844 (canonicalize_loop_form): ... here, renamed from ...
7845 (canonicalize_loop_closed_ssa_form): ... this and amended to
7846 swap successor edges for loop exit blocks to make us use
7847 the RPO order we need for initial schedule generation.
7849 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7851 PR tree-optimization/64811
7852 * match.pd: When optimizing comparisons with Inf, avoid
7853 introducing or losing exceptions from comparisons with NaN.
7855 2018-01-09 Martin Liska <mliska@suse.cz>
7858 * asan.c (shadow_mem_size): Add gcc_assert.
7860 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7862 Don't save registers in main().
7865 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7866 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7867 * config/avr/avr.c (avr_set_current_function): Don't error if
7868 naked, OS_task or OS_main are specified at the same time.
7869 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7871 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7873 * common/config/avr/avr-common.c (avr_option_optimization_table):
7874 Switch on -mmain-is-OS_task for optimizing compilations.
7876 2018-01-09 Richard Biener <rguenther@suse.de>
7878 PR tree-optimization/83572
7879 * graphite.c: Include cfganal.h.
7880 (graphite_transform_loops): Connect infinite loops to exit
7881 and remove fake edges at the end.
7883 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7885 * ipa-inline.c (edge_badness): Revert accidental checkin.
7887 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7890 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7891 symbols; not inline clones.
7893 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7896 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7897 hard registers. Formatting fixes.
7899 PR preprocessor/83722
7900 * gcc.c (try_generate_repro): Pass
7901 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7902 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7905 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7906 Kito Cheng <kito.cheng@gmail.com>
7908 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7909 (riscv_leaf_function_p): Delete.
7910 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7912 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7914 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7916 (do_ifelse): New function.
7917 (do_isel): New function.
7918 (do_sub3): New function.
7919 (do_add3): New function.
7920 (do_load_mask_compare): New function.
7921 (do_overlap_load_compare): New function.
7922 (expand_compare_loop): New function.
7923 (expand_block_compare): Call expand_compare_loop() when appropriate.
7924 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7926 (-mblock-compare-inline-loop-limit): New option.
7928 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7931 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7932 Reverse order of second and third operands in first alternative.
7933 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7934 of first and second elements in UNSPEC_VPERMR vector.
7935 (altivec_expand_vec_perm_le): Likewise.
7937 2018-01-08 Jeff Law <law@redhat.com>
7939 PR rtl-optimizatin/81308
7940 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7941 (process_switch): If group_case_labels makes a change, then set
7943 (pass_convert_switch::execute): If a switch is converted, then
7944 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7946 PR rtl-optimization/81308
7947 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7950 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7952 PR target/83663 - Revert r255946
7953 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7954 generation for cases where splatting a value is not useful.
7955 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7956 across a vec_duplicate and a paradoxical subreg forming a vector
7957 mode to a vec_concat.
7959 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7961 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7962 -march=armv8.3-a variants.
7963 * config/arm/t-multilib: Likewise.
7964 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7966 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7968 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7970 (cceq_ior_compare_complement): Give it a name so I can use it, and
7971 change boolean_or_operator predicate to boolean_operator so it can
7972 be used to generate a crand.
7973 (eqne): New code iterator.
7974 (bd/bd_neg): New code_attrs.
7975 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7976 a single define_insn.
7977 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7978 decrement (bdnzt/bdnzf/bdzt/bdzf).
7979 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7980 with the new names of the branch decrement patterns, and added the
7981 names of the branch decrement conditional patterns.
7983 2018-01-08 Richard Biener <rguenther@suse.de>
7985 PR tree-optimization/83563
7986 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7989 2018-01-08 Richard Biener <rguenther@suse.de>
7992 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7994 2018-01-08 Richard Biener <rguenther@suse.de>
7996 PR tree-optimization/83685
7997 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7998 references to abnormals.
8000 2018-01-08 Richard Biener <rguenther@suse.de>
8003 * dwarf2out.c (output_indirect_strings): Handle empty
8004 skeleton_debug_str_hash.
8005 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8007 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8009 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8010 (emit_store_direct): Likewise.
8011 (arc_trampoline_adjust_address): Likewise.
8012 (arc_asm_trampoline_template): New function.
8013 (arc_initialize_trampoline): Use asm_trampoline_template.
8014 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8015 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8016 * config/arc/arc.md (flush_icache): Delete pattern.
8018 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8020 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8021 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8024 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8027 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8028 by not USED_FOR_TARGET.
8029 (make_pass_resolve_sw_modes): Likewise.
8031 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8033 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8036 2018-01-08 Richard Biener <rguenther@suse.de>
8039 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8041 2018-01-08 Richard Biener <rguenther@suse.de>
8044 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8046 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8049 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8050 basic blocks with a small number of successors.
8051 (convert_control_dep_chain_into_preds): Improve handling of
8053 (dump_predicates): Split apart into...
8054 (dump_pred_chain): ...here...
8055 (dump_pred_info): ...and here.
8056 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8057 (can_chain_union_be_invalidated_p): Improve check for invalidation
8059 (uninit_uses_cannot_happen): Avoid unnecessary if
8060 convert_control_dep_chain_into_preds yielded nothing.
8062 2018-01-06 Martin Sebor <msebor@redhat.com>
8064 PR tree-optimization/83640
8065 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8066 subtracting negative offset from size.
8067 (builtin_access::overlap): Adjust offset bounds of the access to fall
8068 within the size of the object if possible.
8070 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8072 PR rtl-optimization/83699
8073 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8074 extract_bit_field_as_subreg to cases in which the extracted
8075 value is also a vector.
8077 * lra-constraints.c (process_alt_operands): Test for the equivalence
8078 substitutions when detecting a possible reload cycle.
8080 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8083 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8084 by default if flag_selective_schedling{,2}. Formatting fixes.
8086 PR rtl-optimization/83682
8087 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8088 if it has non-VECTOR_MODE element mode.
8089 (vec_duplicate_p): Likewise.
8092 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8093 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8095 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8098 * config/i386/i386-builtin.def
8099 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8100 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8101 Require also OPTION_MASK_ISA_AVX512F in addition to
8102 OPTION_MASK_ISA_GFNI.
8103 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8104 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8105 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8106 to OPTION_MASK_ISA_GFNI.
8107 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8108 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8109 OPTION_MASK_ISA_AVX512BW.
8110 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8111 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8112 addition to OPTION_MASK_ISA_GFNI.
8113 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8114 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8115 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8116 to OPTION_MASK_ISA_GFNI.
8117 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8118 a requirement for all ISAs rather than any of them with a few
8120 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8122 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8123 bitmasks to be enabled with 3 exceptions, instead of requiring any
8124 enabled ISA with lots of exceptions.
8125 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8126 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8127 Change avx512bw in isa attribute to avx512f.
8128 * config/i386/sgxintrin.h: Add license boilerplate.
8129 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8130 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8131 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8132 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8134 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8135 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8136 temporarily sse2 rather than sse if not enabled already.
8139 * config/i386/sse.md (VI248_VLBW): Rename to ...
8140 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8141 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8142 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8143 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8144 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8145 mode iterator instead of VI248_VLBW.
8147 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8149 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8150 (record_modified): Skip clobbers; add debug output.
8151 (param_change_prob): Use sreal frequencies.
8153 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8155 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8156 punt for user-aligned variables.
8158 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8160 * tree-chrec.c (chrec_contains_symbols): Return true for
8163 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8166 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8167 of (x|y) == x for BICS pattern.
8169 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8171 PR tree-optimization/83605
8172 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8173 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8176 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8178 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8179 * config/epiphany/rtems.h: New file.
8181 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8182 Uros Bizjak <ubizjak@gmail.com>
8185 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8186 QIreg_operand instead of register_operand predicate.
8187 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8188 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8189 comments instead of -fmitigate[-_]rop.
8191 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8194 * cgraphunit.c (symbol_table::compile): Switch to text_section
8195 before calling assembly_start debug hook.
8196 * run-rtl-passes.c (run_rtl_passes): Likewise.
8199 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8201 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8202 range_int_cst_p rather than !symbolic_range_p before calling
8203 extract_range_from_multiplicative_op_1.
8205 2018-01-04 Jeff Law <law@redhat.com>
8207 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8208 redundant test in assertion.
8210 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8212 * doc/rtl.texi: Document machine_mode wrapper classes.
8214 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8216 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8219 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8221 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8222 the VEC_PERM_EXPR fold to fail.
8224 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8227 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8228 to switched_sections.
8230 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8233 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8236 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8239 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8240 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8242 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8245 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8246 is BLKmode and bitpos not zero or mode change is needed.
8248 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8251 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8254 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8257 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8258 instead of MULT rtx. Update all corresponding splitters.
8260 (*ssub<modesuffix>): Ditto.
8262 (*cmp_sadd_di): Update split patterns.
8263 (*cmp_sadd_si): Ditto.
8264 (*cmp_sadd_sidi): Ditto.
8265 (*cmp_ssub_di): Ditto.
8266 (*cmp_ssub_si): Ditto.
8267 (*cmp_ssub_sidi): Ditto.
8268 * config/alpha/predicates.md (const23_operand): New predicate.
8269 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8270 Look for ASHIFT, not MULT inner operand.
8271 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8273 2018-01-04 Martin Liska <mliska@suse.cz>
8275 PR gcov-profile/83669
8276 * gcov.c (output_intermediate_file): Add version to intermediate
8278 * doc/gcov.texi: Document new field 'version' in intermediate
8279 file format. Fix location of '-k' option of gcov command.
8281 2018-01-04 Martin Liska <mliska@suse.cz>
8284 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8286 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8288 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8290 2018-01-03 Martin Sebor <msebor@redhat.com>
8292 PR tree-optimization/83655
8293 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8294 checking calls with invalid arguments.
8296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8298 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8299 (vectorizable_mask_load_store): Delete.
8300 (vectorizable_call): Return false for masked loads and stores.
8301 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8302 instead of gimple_assign_rhs1.
8303 (vectorizable_load): Handle IFN_MASK_LOAD.
8304 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8308 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8310 (vectorizable_mask_load_store): ...here.
8311 (vectorizable_load): ...and here.
8313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8315 * tree-vect-stmts.c (vect_build_all_ones_mask)
8316 (vect_build_zero_merge_argument): New functions, split out from...
8317 (vectorizable_load): ...here.
8319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8321 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8323 (vectorizable_mask_load_store): ...here.
8324 (vectorizable_store): ...and here.
8326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8328 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8330 (vectorizable_mask_load_store): ...here.
8332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8334 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8335 (vect_model_store_cost): Take a vec_load_store_type instead of a
8337 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8338 (vect_model_store_cost): Take a vec_load_store_type instead of a
8340 (vectorizable_mask_load_store): Update accordingly.
8341 (vectorizable_store): Likewise.
8342 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8346 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8347 IFN_MASK_LOAD calls here rather than...
8348 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8351 Alan Hayward <alan.hayward@arm.com>
8352 David Sherwood <david.sherwood@arm.com>
8354 * expmed.c (extract_bit_field_1): For vector extracts,
8355 fall back to extract_bit_field_as_subreg if vec_extract
8358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8359 Alan Hayward <alan.hayward@arm.com>
8360 David Sherwood <david.sherwood@arm.com>
8362 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8363 they are variable or constant sized.
8364 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8365 slots for constant-sized data.
8367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8368 Alan Hayward <alan.hayward@arm.com>
8369 David Sherwood <david.sherwood@arm.com>
8371 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8372 handling COND_EXPRs with boolean comparisons, try to find a better
8373 basis for the mask type than the boolean itself.
8375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8377 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8378 is calculated and how it can be overridden.
8379 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8380 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8382 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8386 Alan Hayward <alan.hayward@arm.com>
8387 David Sherwood <david.sherwood@arm.com>
8389 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8390 Remove the mode argument.
8391 (aarch64_simd_valid_immediate): Remove the mode and inverse
8393 * config/aarch64/iterators.md (bitsize): New iterator.
8394 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8395 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8396 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8397 aarch64_simd_valid_immediate.
8398 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8399 (aarch64_reg_or_bic_imm): Likewise.
8400 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8401 with an insn_type enum and msl with a modifier_type enum.
8402 Replace element_width with a scalar_mode. Change the shift
8403 to unsigned int. Add constructors for scalar_float_mode and
8404 scalar_int_mode elements.
8405 (aarch64_vect_float_const_representable_p): Delete.
8406 (aarch64_can_const_movi_rtx_p)
8407 (aarch64_simd_scalar_immediate_valid_for_move)
8408 (aarch64_simd_make_constant): Update call to
8409 aarch64_simd_valid_immediate.
8410 (aarch64_advsimd_valid_immediate_hs): New function.
8411 (aarch64_advsimd_valid_immediate): Likewise.
8412 (aarch64_simd_valid_immediate): Remove mode and inverse
8413 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8414 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8415 and aarch64_float_const_representable_p on the result.
8416 (aarch64_output_simd_mov_immediate): Remove mode argument.
8417 Update call to aarch64_simd_valid_immediate and use of
8418 simd_immediate_info.
8419 (aarch64_output_scalar_simd_mov_immediate): Update call
8422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8423 Alan Hayward <alan.hayward@arm.com>
8424 David Sherwood <david.sherwood@arm.com>
8426 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8427 (mode_nunits): Likewise CONST_MODE_NUNITS.
8428 * machmode.def (ADJUST_NUNITS): Document.
8429 * genmodes.c (mode_data::need_nunits_adj): New field.
8430 (blank_mode): Update accordingly.
8431 (adj_nunits): New variable.
8432 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8434 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8435 listed in adj_nunits.
8436 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8437 listed in adj_nunits. Don't emit case statements for such modes.
8438 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8439 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8440 nothing if adj_nunits is nonnull.
8441 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8442 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8443 (emit_mode_fbit): Update use of print_maybe_const_decl.
8444 (emit_move_size): Likewise. Treat the array as non-const
8446 (emit_mode_adjustments): Handle adj_nunits.
8448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8450 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8451 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8452 (VECTOR_MODES): Use it.
8453 (make_vector_modes): Take the prefix as an argument.
8455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8456 Alan Hayward <alan.hayward@arm.com>
8457 David Sherwood <david.sherwood@arm.com>
8459 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8460 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8461 for MODE_VECTOR_BOOL.
8462 * machmode.def (VECTOR_BOOL_MODE): Document.
8463 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8464 (make_vector_bool_mode): New function.
8465 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8467 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8468 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8470 * stor-layout.c (int_mode_for_mode): Likewise.
8471 * tree.c (build_vector_type_for_mode): Likewise.
8472 * varasm.c (output_constant_pool_2): Likewise.
8473 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8474 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8475 for MODE_VECTOR_BOOL.
8476 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8477 of mode class checks.
8478 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8479 instead of a list of mode class checks.
8480 (expand_vector_scalar_condition): Likewise.
8481 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8484 Alan Hayward <alan.hayward@arm.com>
8485 David Sherwood <david.sherwood@arm.com>
8487 * machmode.h (mode_size): Change from unsigned short to
8489 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8490 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8491 or if measurement_type is not polynomial.
8492 (fixed_size_mode::includes_p): Check for constant-sized modes.
8493 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8494 return a poly_uint16 rather than an unsigned short.
8495 (emit_mode_size): Change the type of mode_size from unsigned short
8496 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8497 (emit_mode_adjustments): Cope with polynomial vector sizes.
8498 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8500 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8502 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8503 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8504 * caller-save.c (setup_save_areas): Likewise.
8505 (replace_reg_with_saved_mem): Likewise.
8506 * calls.c (emit_library_call_value_1): Likewise.
8507 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8508 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8509 (gen_lowpart_for_combine): Likewise.
8510 * convert.c (convert_to_integer_1): Likewise.
8511 * cse.c (equiv_constant, cse_insn): Likewise.
8512 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8513 (cselib_subst_to_values): Likewise.
8514 * dce.c (word_dce_process_block): Likewise.
8515 * df-problems.c (df_word_lr_mark_ref): Likewise.
8516 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8517 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8518 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8519 (rtl_for_decl_location): Likewise.
8520 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8521 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8522 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8523 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8524 (expand_expr_real_1): Likewise.
8525 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8526 (pad_below): Likewise.
8527 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8528 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8529 * ira.c (get_subreg_tracking_sizes): Likewise.
8530 * ira-build.c (ira_create_allocno_objects): Likewise.
8531 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8532 (ira_sort_regnos_for_alter_reg): Likewise.
8533 * ira-costs.c (record_operand_costs): Likewise.
8534 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8535 (resolve_simple_move): Likewise.
8536 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8537 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8538 (lra_constraints): Likewise.
8539 (CONST_POOL_OK_P): Reject variable-sized modes.
8540 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8541 (add_pseudo_to_slot, lra_spill): Likewise.
8542 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8543 * optabs-query.c (get_best_extraction_insn): Likewise.
8544 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8545 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8546 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8547 * recog.c (offsettable_address_addr_space_p): Likewise.
8548 * regcprop.c (maybe_mode_change): Likewise.
8549 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8550 * regrename.c (build_def_use): Likewise.
8551 * regstat.c (dump_reg_info): Likewise.
8552 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8553 (find_reloads, find_reloads_subreg_address): Likewise.
8554 * reload1.c (eliminate_regs_1): Likewise.
8555 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8556 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8557 (simplify_binary_operation_1, simplify_subreg): Likewise.
8558 * targhooks.c (default_function_arg_padding): Likewise.
8559 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8560 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8561 (verify_gimple_assign_ternary): Likewise.
8562 * tree-inline.c (estimate_move_cost): Likewise.
8563 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8564 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8565 (get_address_cost_ainc): Likewise.
8566 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8567 (vect_supportable_dr_alignment): Likewise.
8568 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8569 (vectorizable_reduction): Likewise.
8570 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8571 (vectorizable_operation, vectorizable_load): Likewise.
8572 * tree.c (build_same_sized_truth_vector_type): Likewise.
8573 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8574 * var-tracking.c (emit_note_insn_var_location): Likewise.
8575 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8576 (ADDR_VEC_ALIGN): Likewise.
8578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8579 Alan Hayward <alan.hayward@arm.com>
8580 David Sherwood <david.sherwood@arm.com>
8582 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8584 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8585 or if measurement_type is polynomial.
8586 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8587 * combine.c (make_extraction): Likewise.
8588 * dse.c (find_shift_sequence): Likewise.
8589 * dwarf2out.c (mem_loc_descriptor): Likewise.
8590 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8591 (extract_bit_field, extract_low_bits): Likewise.
8592 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8593 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8594 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8595 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8596 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8597 * reload.c (find_reloads): Likewise.
8598 * reload1.c (alter_reg): Likewise.
8599 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8600 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8601 * tree-if-conv.c (predicate_mem_writes): Likewise.
8602 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8603 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8604 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8605 * valtrack.c (dead_debug_insert_temp): Likewise.
8606 * varasm.c (mergeable_constant_section): Likewise.
8607 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8610 Alan Hayward <alan.hayward@arm.com>
8611 David Sherwood <david.sherwood@arm.com>
8613 * expr.c (expand_assignment): Cope with polynomial mode sizes
8614 when assigning to a CONCAT.
8616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8617 Alan Hayward <alan.hayward@arm.com>
8618 David Sherwood <david.sherwood@arm.com>
8620 * machmode.h (mode_precision): Change from unsigned short to
8622 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8624 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8625 or if measurement_type is not polynomial.
8626 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8627 in which the mode is already known to be a scalar_int_mode.
8628 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8629 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8631 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8632 for GET_MODE_PRECISION.
8633 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8634 for GET_MODE_PRECISION.
8635 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8637 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8638 (expand_field_assignment, make_extraction): Likewise.
8639 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8640 (get_last_value): Likewise.
8641 * convert.c (convert_to_integer_1): Likewise.
8642 * cse.c (cse_insn): Likewise.
8643 * expr.c (expand_expr_real_1): Likewise.
8644 * lra-constraints.c (simplify_operand_subreg): Likewise.
8645 * optabs-query.c (can_atomic_load_p): Likewise.
8646 * optabs.c (expand_atomic_load): Likewise.
8647 (expand_atomic_store): Likewise.
8648 * ree.c (combine_reaching_defs): Likewise.
8649 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8650 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8651 * tree.h (type_has_mode_precision_p): Likewise.
8652 * ubsan.c (instrument_si_overflow): Likewise.
8654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8655 Alan Hayward <alan.hayward@arm.com>
8656 David Sherwood <david.sherwood@arm.com>
8658 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8659 polynomial numbers of units.
8660 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8661 (valid_vector_subparts_p): New function.
8662 (build_vector_type): Remove temporary shim and take the number
8663 of units as a poly_uint64 rather than an int.
8664 (build_opaque_vector_type): Take the number of units as a
8665 poly_uint64 rather than an int.
8666 * tree.c (build_vector_from_ctor): Handle polynomial
8667 TYPE_VECTOR_SUBPARTS.
8668 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8669 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8670 (build_vector_from_val): If the number of units is variable,
8671 use build_vec_duplicate_cst for constant operands and
8672 VEC_DUPLICATE_EXPR otherwise.
8673 (make_vector_type): Remove temporary is_constant ().
8674 (build_vector_type, build_opaque_vector_type): Take the number of
8675 units as a poly_uint64 rather than an int.
8676 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8678 * cfgexpand.c (expand_debug_expr): Likewise.
8679 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8680 (store_constructor, expand_expr_real_1): Likewise.
8681 (const_scalar_mask_from_tree): Likewise.
8682 * fold-const-call.c (fold_const_reduction): Likewise.
8683 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8684 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8685 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8686 (fold_relational_const): Likewise.
8687 (native_interpret_vector): Likewise. Change the size from an
8688 int to an unsigned int.
8689 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8690 TYPE_VECTOR_SUBPARTS.
8691 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8692 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8693 duplicating a non-constant operand into a variable-length vector.
8694 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8695 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8696 * ipa-icf.c (sem_variable::equals): Likewise.
8697 * match.pd: Likewise.
8698 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8699 * print-tree.c (print_node): Likewise.
8700 * stor-layout.c (layout_type): Likewise.
8701 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8702 * tree-cfg.c (verify_gimple_comparison): Likewise.
8703 (verify_gimple_assign_binary): Likewise.
8704 (verify_gimple_assign_ternary): Likewise.
8705 (verify_gimple_assign_single): Likewise.
8706 * tree-pretty-print.c (dump_generic_node): Likewise.
8707 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8708 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8709 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8710 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8711 (vect_shift_permute_load_chain): Likewise.
8712 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8713 (expand_vector_condition, optimize_vector_constructor): Likewise.
8714 (lower_vec_perm, get_compute_type): Likewise.
8715 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8716 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8717 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8718 (vect_recog_mask_conversion_pattern): Likewise.
8719 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8720 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8721 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8722 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8723 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8724 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8725 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8726 (supportable_widening_operation): Likewise.
8727 (supportable_narrowing_operation): Likewise.
8728 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8730 * varasm.c (output_constant): Likewise.
8732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8733 Alan Hayward <alan.hayward@arm.com>
8734 David Sherwood <david.sherwood@arm.com>
8736 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8737 so that both the length == 3 and length != 3 cases set up their
8738 own permute vectors. Add comments explaining why we know the
8739 number of elements is constant.
8740 (vect_permute_load_chain): Likewise.
8742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8743 Alan Hayward <alan.hayward@arm.com>
8744 David Sherwood <david.sherwood@arm.com>
8746 * machmode.h (mode_nunits): Change from unsigned char to
8748 (ONLY_FIXED_SIZE_MODES): New macro.
8749 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8750 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8751 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8753 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8754 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8755 or if measurement_type is not polynomial.
8756 * genmodes.c (ZERO_COEFFS): New macro.
8757 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8759 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8760 Use ZERO_COEFFS when emitting initializers.
8761 * data-streamer.h (bp_pack_poly_value): New function.
8762 (bp_unpack_poly_value): Likewise.
8763 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8764 for GET_MODE_NUNITS.
8765 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8766 for GET_MODE_NUNITS.
8767 * tree.c (make_vector_type): Remove temporary shim and make
8768 the real function take the number of units as a poly_uint64
8770 (build_vector_type_for_mode): Handle polynomial nunits.
8771 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8772 * emit-rtl.c (const_vec_series_p_1): Likewise.
8773 (gen_rtx_CONST_VECTOR): Likewise.
8774 * fold-const.c (test_vec_duplicate_folding): Likewise.
8775 * genrecog.c (validate_pattern): Likewise.
8776 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8777 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8778 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8779 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8780 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8781 * rtlanal.c (subreg_get_info): Likewise.
8782 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8783 (vect_grouped_load_supported): Likewise.
8784 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8785 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8786 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8787 (simplify_const_unary_operation, simplify_binary_operation_1)
8788 (simplify_const_binary_operation, simplify_ternary_operation)
8789 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8790 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8791 instead of CONST_VECTOR_NUNITS.
8792 * varasm.c (output_constant_pool_2): Likewise.
8793 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8794 explicit-encoded elements in the XVEC for variable-length vectors.
8796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8798 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8801 Alan Hayward <alan.hayward@arm.com>
8802 David Sherwood <david.sherwood@arm.com>
8804 * coretypes.h (fixed_size_mode): Declare.
8805 (fixed_size_mode_pod): New typedef.
8806 * builtins.h (target_builtins::x_apply_args_mode)
8807 (target_builtins::x_apply_result_mode): Change type to
8808 fixed_size_mode_pod.
8809 * builtins.c (apply_args_size, apply_result_size, result_vector)
8810 (expand_builtin_apply_args_1, expand_builtin_apply)
8811 (expand_builtin_return): Update accordingly.
8813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8815 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8816 * cselib.c (cselib_hash_rtx): Likewise.
8817 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8818 CONST_VECTOR encoding.
8820 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8821 Jeff Law <law@redhat.com>
8824 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8825 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8826 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8827 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8830 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8831 explicitly probe *sp in a noreturn function if there were any callee
8832 register saves or frame pointer is needed.
8834 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8837 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8838 BLKmode for ternary, binary or unary expressions.
8841 * var-tracking.c (delete_vta_debug_insn): New inline function.
8842 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8843 insns from get_insns () to NULL instead of each bb separately.
8844 Use delete_vta_debug_insn. No longer static.
8845 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8846 delete_vta_debug_insns callers.
8847 * rtl.h (delete_vta_debug_insns): Declare.
8848 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8849 instead of variable_tracking_main.
8851 2018-01-03 Martin Sebor <msebor@redhat.com>
8853 PR tree-optimization/83603
8854 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8855 arguments past the endof the argument list in functions declared
8856 without a prototype.
8857 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8858 Avoid checking when arguments are null.
8860 2018-01-03 Martin Sebor <msebor@redhat.com>
8863 * doc/extend.texi (attribute const): Fix a typo.
8864 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8865 issuing -Wsuggest-attribute for void functions.
8867 2018-01-03 Martin Sebor <msebor@redhat.com>
8869 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8870 offset_int::from instead of wide_int::to_shwi.
8871 (maybe_diag_overlap): Remove assertion.
8872 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8873 * gimple-ssa-sprintf.c (format_directive): Same.
8874 (parse_directive): Same.
8875 (sprintf_dom_walker::compute_format_length): Same.
8876 (try_substitute_return_value): Same.
8878 2018-01-03 Jeff Law <law@redhat.com>
8881 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8882 non-constant residual for zero at runtime and avoid probing in
8883 that case. Reorganize code for trailing problem to mirror handling
8886 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8888 PR tree-optimization/83501
8889 * tree-ssa-strlen.c (get_string_cst): New.
8890 (handle_char_store): Call get_string_cst.
8892 2018-01-03 Martin Liska <mliska@suse.cz>
8894 PR tree-optimization/83593
8895 * tree-ssa-strlen.c: Include tree-cfg.h.
8896 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8897 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8898 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8900 (strlen_dom_walker::before_dom_children): Call
8901 gimple_purge_dead_eh_edges. Dump tranformation with details
8903 (strlen_dom_walker::before_dom_children): Update call by adding
8904 new argument cleanup_eh.
8905 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8907 2018-01-03 Martin Liska <mliska@suse.cz>
8910 * cif-code.def (VARIADIC_THUNK): New enum value.
8911 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8914 2018-01-03 Jan Beulich <jbeulich@suse.com>
8916 * sse.md (mov<mode>_internal): Tighten condition for when to use
8917 vmovdqu<ssescalarsize> for TI and OI modes.
8919 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8921 Update copyright years.
8923 2018-01-03 Martin Liska <mliska@suse.cz>
8926 * ipa-visibility.c (function_and_variable_visibility): Skip
8927 functions with noipa attribure.
8929 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8931 * gcc.c (process_command): Update copyright notice dates.
8932 * gcov-dump.c (print_version): Ditto.
8933 * gcov.c (print_version): Ditto.
8934 * gcov-tool.c (print_version): Ditto.
8935 * gengtype.c (create_file): Ditto.
8936 * doc/cpp.texi: Bump @copying's copyright year.
8937 * doc/cppinternals.texi: Ditto.
8938 * doc/gcc.texi: Ditto.
8939 * doc/gccint.texi: Ditto.
8940 * doc/gcov.texi: Ditto.
8941 * doc/install.texi: Ditto.
8942 * doc/invoke.texi: Ditto.
8944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8946 * vector-builder.h (vector_builder::m_full_nelts): Change from
8947 unsigned int to poly_uint64.
8948 (vector_builder::full_nelts): Update prototype accordingly.
8949 (vector_builder::new_vector): Likewise.
8950 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8951 (vector_builder::operator ==): Likewise.
8952 (vector_builder::finalize): Likewise.
8953 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8954 Take the number of elements as a poly_uint64 rather than an
8956 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8957 from unsigned int to poly_uint64.
8958 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8959 (vec_perm_indices::new_vector): Likewise.
8960 (vec_perm_indices::length): Likewise.
8961 (vec_perm_indices::nelts_per_input): Likewise.
8962 (vec_perm_indices::input_nelts): Likewise.
8963 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8964 number of elements per input as a poly_uint64 rather than an
8965 unsigned int. Use the original encoding for variable-length
8966 vectors, rather than clamping each individual element.
8967 For the second and subsequent elements in each pattern,
8968 clamp the step and base before clamping their sum.
8969 (vec_perm_indices::series_p): Handle polynomial element counts.
8970 (vec_perm_indices::all_in_range_p): Likewise.
8971 (vec_perm_indices_to_tree): Likewise.
8972 (vec_perm_indices_to_rtx): Likewise.
8973 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8974 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8975 (tree_vector_builder::new_binary_operation): Handle polynomial
8976 element counts. Return false if we need to know the number
8977 of elements at compile time.
8978 * fold-const.c (fold_vec_perm): Punt if the number of elements
8979 isn't known at compile time.
8981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8983 * vec-perm-indices.h (vec_perm_builder): Change element type
8984 from HOST_WIDE_INT to poly_int64.
8985 (vec_perm_indices::element_type): Update accordingly.
8986 (vec_perm_indices::clamp): Handle polynomial element_types.
8987 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8988 (vec_perm_indices::all_in_range_p): Likewise.
8989 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8991 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8992 polynomial vec_perm_indices element types.
8993 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8994 * fold-const.c (fold_vec_perm): Likewise.
8995 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8996 * tree-vect-generic.c (lower_vec_perm): Likewise.
8997 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8998 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8999 element type to HOST_WIDE_INT.
9001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9002 Alan Hayward <alan.hayward@arm.com>
9003 David Sherwood <david.sherwood@arm.com>
9005 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9006 rather than an int. Use plus_constant.
9007 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9008 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9011 Alan Hayward <alan.hayward@arm.com>
9012 David Sherwood <david.sherwood@arm.com>
9014 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9015 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
9022 mode sizes. Require a constant size for BLKmode parameters
9023 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9024 forces a parameter to be padded at the lsb end in order to
9025 fill a complete number of words, require the parameter size
9026 to be ordered wrt UNITS_PER_WORD.
9028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9029 Alan Hayward <alan.hayward@arm.com>
9030 David Sherwood <david.sherwood@arm.com>
9032 * reload1.c (spill_stack_slot_width): Change element type
9033 from unsigned int to poly_uint64_pod.
9034 (alter_reg): Treat mode sizes as polynomial.
9036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9037 Alan Hayward <alan.hayward@arm.com>
9038 David Sherwood <david.sherwood@arm.com>
9040 * reload.c (complex_word_subreg_p): New function.
9041 (reload_inner_reg_of_subreg, push_reload): Use it.
9043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9044 Alan Hayward <alan.hayward@arm.com>
9045 David Sherwood <david.sherwood@arm.com>
9047 * lra-constraints.c (process_alt_operands): Reject matched
9048 operands whose sizes aren't ordered.
9049 (match_reload): Refer to this check here.
9051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9052 Alan Hayward <alan.hayward@arm.com>
9053 David Sherwood <david.sherwood@arm.com>
9055 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9056 that the mode size is in the set {1, 2, 4, 8, 16}.
9058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9059 Alan Hayward <alan.hayward@arm.com>
9060 David Sherwood <david.sherwood@arm.com>
9062 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9063 Use plus_constant instead of gen_rtx_PLUS.
9065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9066 Alan Hayward <alan.hayward@arm.com>
9067 David Sherwood <david.sherwood@arm.com>
9069 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9070 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9071 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9072 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9073 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9074 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9075 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9076 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9077 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9078 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9080 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9081 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9082 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9083 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9084 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9085 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9086 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9087 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9088 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9089 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9091 * expr.c (emit_move_resolve_push): Treat the input and result
9092 of PUSH_ROUNDING as a poly_int64.
9093 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9094 (emit_push_insn): Likewise.
9095 * lra-eliminations.c (mark_not_eliminable): Likewise.
9096 * recog.c (push_operand): Likewise.
9097 * reload1.c (elimination_effects): Likewise.
9098 * rtlanal.c (nonzero_bits1): Likewise.
9099 * calls.c (store_one_arg): Likewise. Require the padding to be
9100 known at compile time.
9102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9103 Alan Hayward <alan.hayward@arm.com>
9104 David Sherwood <david.sherwood@arm.com>
9106 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9107 Use plus_constant instead of gen_rtx_PLUS.
9109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9110 Alan Hayward <alan.hayward@arm.com>
9111 David Sherwood <david.sherwood@arm.com>
9113 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9117 Alan Hayward <alan.hayward@arm.com>
9118 David Sherwood <david.sherwood@arm.com>
9120 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9121 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9122 via stack temporaries. Treat the mode size as polynomial too.
9124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9125 Alan Hayward <alan.hayward@arm.com>
9126 David Sherwood <david.sherwood@arm.com>
9128 * expr.c (expand_expr_real_2): When handling conversions involving
9129 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9130 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9131 as a poly_uint64 too.
9133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9134 Alan Hayward <alan.hayward@arm.com>
9135 David Sherwood <david.sherwood@arm.com>
9137 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9140 Alan Hayward <alan.hayward@arm.com>
9141 David Sherwood <david.sherwood@arm.com>
9143 * combine.c (can_change_dest_mode): Handle polynomial
9144 REGMODE_NATURAL_SIZE.
9145 * expmed.c (store_bit_field_1): Likewise.
9146 * expr.c (store_constructor): Likewise.
9147 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9148 and polynomial REGMODE_NATURAL_SIZE.
9149 (gen_lowpart_common): Likewise.
9150 * reginfo.c (record_subregs_of_mode): Likewise.
9151 * rtlanal.c (read_modify_subreg_p): Likewise.
9153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9154 Alan Hayward <alan.hayward@arm.com>
9155 David Sherwood <david.sherwood@arm.com>
9157 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9158 numbers of elements.
9160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9161 Alan Hayward <alan.hayward@arm.com>
9162 David Sherwood <david.sherwood@arm.com>
9164 * match.pd: Cope with polynomial numbers of vector elements.
9166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9167 Alan Hayward <alan.hayward@arm.com>
9168 David Sherwood <david.sherwood@arm.com>
9170 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9171 in a POINTER_PLUS_EXPR.
9173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9174 Alan Hayward <alan.hayward@arm.com>
9175 David Sherwood <david.sherwood@arm.com>
9177 * omp-simd-clone.c (simd_clone_subparts): New function.
9178 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9179 (ipa_simd_modify_function_body): Likewise.
9181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9182 Alan Hayward <alan.hayward@arm.com>
9183 David Sherwood <david.sherwood@arm.com>
9185 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9186 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9187 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9188 (expand_vector_condition, vector_element): Likewise.
9189 (subparts_gt): New function.
9190 (get_compute_type): Use subparts_gt.
9191 (count_type_subparts): Delete.
9192 (expand_vector_operations_1): Use subparts_gt instead of
9193 count_type_subparts.
9195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9196 Alan Hayward <alan.hayward@arm.com>
9197 David Sherwood <david.sherwood@arm.com>
9199 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9200 (vect_compile_time_alias): ...this new function. Do the calculation
9201 on poly_ints rather than trees.
9202 (vect_prune_runtime_alias_test_list): Update call accordingly.
9204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9205 Alan Hayward <alan.hayward@arm.com>
9206 David Sherwood <david.sherwood@arm.com>
9208 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9210 (vect_schedule_slp_instance): Likewise.
9212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9213 Alan Hayward <alan.hayward@arm.com>
9214 David Sherwood <david.sherwood@arm.com>
9216 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9217 constant and extern definitions for variable-length vectors.
9218 (vect_get_constant_vectors): Note that the number of units
9219 is known to be constant.
9221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9222 Alan Hayward <alan.hayward@arm.com>
9223 David Sherwood <david.sherwood@arm.com>
9225 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9226 of units as polynomial. Choose between WIDE and NARROW based
9229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9230 Alan Hayward <alan.hayward@arm.com>
9231 David Sherwood <david.sherwood@arm.com>
9233 * tree-vect-stmts.c (simd_clone_subparts): New function.
9234 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9237 Alan Hayward <alan.hayward@arm.com>
9238 David Sherwood <david.sherwood@arm.com>
9240 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9241 vectors as polynomial. Use build_index_vector for
9244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9245 Alan Hayward <alan.hayward@arm.com>
9246 David Sherwood <david.sherwood@arm.com>
9248 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9249 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9250 for variable-length vectors.
9251 (vectorizable_mask_load_store): Treat the number of units as
9252 polynomial, asserting that it is constant if the condition has
9253 already been enforced.
9254 (vectorizable_store, vectorizable_load): Likewise.
9256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9257 Alan Hayward <alan.hayward@arm.com>
9258 David Sherwood <david.sherwood@arm.com>
9260 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9261 of units as polynomial. Punt if we can't tell at compile time
9262 which vector contains the final result.
9264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9265 Alan Hayward <alan.hayward@arm.com>
9266 David Sherwood <david.sherwood@arm.com>
9268 * tree-vect-loop.c (vectorizable_induction): Treat the number
9269 of units as polynomial. Punt on SLP inductions. Use an integer
9270 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9271 cast of such a series for variable-length floating-point
9274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9275 Alan Hayward <alan.hayward@arm.com>
9276 David Sherwood <david.sherwood@arm.com>
9278 * tree.h (build_index_vector): Declare.
9279 * tree.c (build_index_vector): New function.
9280 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9281 of units as polynomial, forcibly converting it to a constant if
9282 vectorizable_reduction has already enforced the condition.
9283 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9284 to create a {1,2,3,...} vector.
9285 (vectorizable_reduction): Treat the number of units as polynomial.
9286 Choose vectype_in based on the largest scalar element size rather
9287 than the smallest number of units. Enforce the restrictions
9290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9291 Alan Hayward <alan.hayward@arm.com>
9292 David Sherwood <david.sherwood@arm.com>
9294 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9295 number of units as polynomial.
9297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9298 Alan Hayward <alan.hayward@arm.com>
9299 David Sherwood <david.sherwood@arm.com>
9301 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9302 * target.def (autovectorize_vector_sizes): Return the vector sizes
9303 by pointer, using vector_sizes rather than a bitmask.
9304 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9305 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9306 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9308 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9309 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9310 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9311 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9312 * omp-general.c (omp_max_vf): Likewise.
9313 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9314 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9315 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9316 * tree-vect-slp.c (vect_slp_bb): Likewise.
9317 * doc/tm.texi: Regenerate.
9318 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9320 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9321 the vector size as a poly_uint64 rather than an unsigned int.
9322 (current_vector_size): Change from an unsigned int to a poly_uint64.
9323 (get_vectype_for_scalar_type): Update accordingly.
9324 * tree.h (build_truth_vector_type): Take the size and number of
9325 units as a poly_uint64 rather than an unsigned int.
9326 (build_vector_type): Add a temporary overload that takes
9327 the number of units as a poly_uint64 rather than an unsigned int.
9328 * tree.c (make_vector_type): Likewise.
9329 (build_truth_vector_type): Take the number of units as a poly_uint64
9330 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
9337 as poly_uint64s rather than unsigned ints.
9338 * targhooks.h (default_get_mask_mode): Update accordingly.
9339 * targhooks.c (default_get_mask_mode): Likewise.
9340 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9341 * doc/tm.texi: Regenerate.
9343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9344 Alan Hayward <alan.hayward@arm.com>
9345 David Sherwood <david.sherwood@arm.com>
9347 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9348 * omp-general.c (omp_max_vf): Likewise.
9349 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9350 (expand_omp_simd): Handle polynomial safelen.
9351 * omp-low.c (omplow_simd_context): Add a default constructor.
9352 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9353 (lower_rec_simd_input_clauses): Update accordingly.
9354 (lower_rec_input_clauses): 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-vectorizer.h (vect_nunits_for_cost): New function.
9361 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9362 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9363 (vect_analyze_slp_cost): Likewise.
9364 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9365 (vect_model_load_cost): Likewise.
9367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9368 Alan Hayward <alan.hayward@arm.com>
9369 David Sherwood <david.sherwood@arm.com>
9371 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9372 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9373 from an unsigned int * to a poly_uint64_pod *.
9374 (calculate_unrolling_factor): New function.
9375 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9378 Alan Hayward <alan.hayward@arm.com>
9379 David Sherwood <david.sherwood@arm.com>
9381 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9382 from an unsigned int to a poly_uint64.
9383 (_loop_vec_info::slp_unrolling_factor): Likewise.
9384 (_loop_vec_info::vectorization_factor): Change from an int
9386 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9387 (vect_get_num_vectors): New function.
9388 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9389 (vect_get_num_copies): Use vect_get_num_vectors.
9390 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9391 to an unsigned int *.
9392 (vect_analyze_data_refs): Change min_vf from an int * to a
9394 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9395 than an unsigned HOST_WIDE_INT.
9396 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9397 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9398 to an unsigned int *.
9399 (vect_analyze_data_ref_dependences): Likewise.
9400 (vect_compute_data_ref_alignment): Handle polynomial vf.
9401 (vect_enhance_data_refs_alignment): Likewise.
9402 (vect_prune_runtime_alias_test_list): Likewise.
9403 (vect_shift_permute_load_chain): Likewise.
9404 (vect_supportable_dr_alignment): Likewise.
9405 (dependence_distance_ge_vf): Take the vectorization factor as a
9406 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9407 (vect_analyze_data_refs): Change min_vf from an int * to a
9409 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9410 vfm1 as a poly_uint64 rather than an int. Make the same change
9411 for the returned bound_scalar.
9412 (vect_gen_vector_loop_niters): Handle polynomial vf.
9413 (vect_do_peeling): Likewise. Update call to
9414 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9415 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9417 * tree-vect-loop.c (vect_determine_vectorization_factor)
9418 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9419 (vect_get_known_peeling_cost): Likewise.
9420 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9421 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9422 (vect_transform_loop): Likewise. Use the lowest possible VF when
9423 updating the upper bounds of the loop.
9424 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9426 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9427 polynomial unroll factors.
9428 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9429 (vect_make_slp_decision): Likewise.
9430 (vect_supported_load_permutation_p): Likewise, and polynomial
9432 (vect_analyze_slp_cost): Handle polynomial vf.
9433 (vect_slp_analyze_node_operations): Likewise.
9434 (vect_slp_analyze_bb_1): Likewise.
9435 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9436 than an unsigned HOST_WIDE_INT.
9437 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9438 (vectorizable_load): Handle polynomial vf.
9439 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9441 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9444 Alan Hayward <alan.hayward@arm.com>
9445 David Sherwood <david.sherwood@arm.com>
9447 * match.pd: Handle bit operations involving three constants
9448 and try to fold one pair.
9450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9452 * tree-vect-loop-manip.c: Include gimple-fold.h.
9453 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9454 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9455 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9456 Add a path that uses a step of VF instead of 1, but disable it
9458 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9459 and niters_no_overflow parameters. Update calls to
9460 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9461 Create a new SSA name if the latter choses to use a ste other
9462 than zero, and return it via niters_vector_mult_vf_var.
9463 * tree-vect-loop.c (vect_transform_loop): Update calls to
9464 vect_do_peeling, vect_gen_vector_loop_niters and
9465 slpeel_make_loop_iterate_ntimes.
9466 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9467 (vect_gen_vector_loop_niters): Update declarations after above changes.
9469 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9471 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9472 128-bit round to integer instructions.
9473 (ceil<mode>2): Likewise.
9474 (btrunc<mode>2): Likewise.
9475 (round<mode>2): Likewise.
9477 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9479 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9480 unaligned VSX load/store on P8/P9.
9481 (expand_block_clear): Allow the use of unaligned VSX
9482 load/store on P8/P9.
9484 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9486 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9488 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9489 swap associated with both a load and a store.
9491 2018-01-02 Andrew Waterman <andrew@sifive.com>
9493 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9494 * config/riscv/riscv.md (clear_cache): Use it.
9496 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9498 * web.c: Remove out-of-date comment.
9500 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9502 * expr.c (fixup_args_size_notes): Check that any existing
9503 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9504 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9505 (emit_single_push_insn): ...here.
9507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9509 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9510 (const_vector_encoded_nelts): New function.
9511 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9512 (const_vector_int_elt, const_vector_elt): Declare.
9513 * emit-rtl.c (const_vector_int_elt_1): New function.
9514 (const_vector_elt): Likewise.
9515 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9516 of CONST_VECTOR_ELT.
9518 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9520 * expr.c: Include rtx-vector-builder.h.
9521 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9522 directly on the tree encoding.
9523 (const_vector_from_tree): Likewise.
9524 * optabs.c: Include rtx-vector-builder.h.
9525 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9526 sequence of "u" values.
9527 * vec-perm-indices.c: Include rtx-vector-builder.h.
9528 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9529 directly on the vec_perm_indices encoding.
9531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9533 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9534 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9535 * rtx-vector-builder.h: New file.
9536 * rtx-vector-builder.c: Likewise.
9537 * rtl.h (rtx_def::u2): Add a const_vector field.
9538 (CONST_VECTOR_NPATTERNS): New macro.
9539 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9540 (CONST_VECTOR_DUPLICATE_P): Likewise.
9541 (CONST_VECTOR_STEPPED_P): Likewise.
9542 (CONST_VECTOR_ENCODED_ELT): Likewise.
9543 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9544 (unwrap_const_vec_duplicate): Likewise.
9545 (const_vec_series_p): Check for a non-duplicated vector encoding.
9546 Say that the function only returns true for integer vectors.
9547 * emit-rtl.c: Include rtx-vector-builder.h.
9548 (gen_const_vec_duplicate_1): Delete.
9549 (gen_const_vector): Call gen_const_vec_duplicate instead of
9550 gen_const_vec_duplicate_1.
9551 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9552 (gen_const_vec_duplicate): Use rtx_vector_builder.
9553 (gen_const_vec_series): Likewise.
9554 (gen_rtx_CONST_VECTOR): Likewise.
9555 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9556 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9557 Build a new vector rather than modifying a CONST_VECTOR in-place.
9558 (handle_special_swappables): Update call accordingly.
9559 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9560 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9561 Build a new vector rather than modifying a CONST_VECTOR in-place.
9562 (handle_special_swappables): Update call accordingly.
9564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9566 * simplify-rtx.c (simplify_const_binary_operation): Use
9567 CONST_VECTOR_ELT instead of XVECEXP.
9569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9571 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9572 the selector elements to be different from the data elements
9573 if the selector is a VECTOR_CST.
9574 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9575 ssizetype for the selector.
9577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9579 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9580 before testing each element individually.
9581 * tree-vect-generic.c (lower_vec_perm): Likewise.
9583 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9585 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9586 * selftest-run-tests.c (selftest::run_tests): Call it.
9587 * vector-builder.h (vector_builder::operator ==): New function.
9588 (vector_builder::operator !=): Likewise.
9589 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9590 (vec_perm_indices::all_from_input_p): New function.
9591 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9592 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9593 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9594 instead of reading the VECTOR_CST directly. Detect whether both
9595 vector inputs are the same before constructing the vec_perm_indices,
9596 and update the number of inputs argument accordingly. Use the
9597 utility functions added above. Only construct sel2 if we need to.
9599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9601 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9602 the broadcast of the low byte.
9603 (expand_mult_highpart): Use an explicit encoding for the permutes.
9604 * optabs-query.c (can_mult_highpart_p): Likewise.
9605 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9606 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9607 (vectorizable_bswap): Likewise.
9608 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9609 explicit encoding for the power-of-2 permutes.
9610 (vect_permute_store_chain): Likewise.
9611 (vect_grouped_load_supported): Likewise.
9612 (vect_permute_load_chain): Likewise.
9614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9616 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9617 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9618 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9619 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9620 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9621 (vect_gen_perm_mask_any): Likewise.
9623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9625 * int-vector-builder.h: New file.
9626 * vec-perm-indices.h: Include int-vector-builder.h.
9627 (vec_perm_indices): Redefine as an int_vector_builder.
9628 (auto_vec_perm_indices): Delete.
9629 (vec_perm_builder): Redefine as a stand-alone class.
9630 (vec_perm_indices::vec_perm_indices): New function.
9631 (vec_perm_indices::clamp): Likewise.
9632 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9633 (vec_perm_indices::new_vector): New function.
9634 (vec_perm_indices::new_expanded_vector): Update for new
9635 vec_perm_indices class.
9636 (vec_perm_indices::rotate_inputs): New function.
9637 (vec_perm_indices::all_in_range_p): Operate directly on the
9638 encoded form, without computing elided elements.
9639 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9640 encoding. Update for new vec_perm_indices class.
9641 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9642 the given vec_perm_builder.
9643 (expand_vec_perm_var): Update vec_perm_builder constructor.
9644 (expand_mult_highpart): Use vec_perm_builder instead of
9645 auto_vec_perm_indices.
9646 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9647 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9648 or double series encoding as appropriate.
9649 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9650 vec_perm_indices instead of auto_vec_perm_indices.
9651 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9652 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9653 (vect_permute_store_chain): Likewise.
9654 (vect_grouped_load_supported): Likewise.
9655 (vect_permute_load_chain): Likewise.
9656 (vect_shift_permute_load_chain): Likewise.
9657 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9658 (vect_transform_slp_perm_load): Likewise.
9659 (vect_schedule_slp_instance): Likewise.
9660 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9661 (vectorizable_mask_load_store): Likewise.
9662 (vectorizable_bswap): Likewise.
9663 (vectorizable_store): Likewise.
9664 (vectorizable_load): Likewise.
9665 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9666 vec_perm_indices instead of auto_vec_perm_indices. Use
9667 tree_to_vec_perm_builder to read the vector from a tree.
9668 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9669 vec_perm_builder instead of a vec_perm_indices.
9670 (have_whole_vector_shift): Use vec_perm_builder and
9671 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9672 truncation to calc_vec_perm_mask_for_shift.
9673 (vect_create_epilog_for_reduction): Likewise.
9674 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9675 from auto_vec_perm_indices to vec_perm_indices.
9676 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9677 instead of changing individual elements.
9678 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9679 the vector in d.perm.
9680 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9681 from auto_vec_perm_indices to vec_perm_indices.
9682 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9683 instead of changing individual elements.
9684 (arm_vectorize_vec_perm_const): Use new_vector to install
9685 the vector in d.perm.
9686 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9687 Update vec_perm_builder constructor.
9688 (rs6000_expand_interleave): Likewise.
9689 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9690 (rs6000_expand_interleave): Likewise.
9692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9694 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9695 to qimode could truncate the indices.
9696 * optabs.c (expand_vec_perm_var): Likewise.
9698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9700 * Makefile.in (OBJS): Add vec-perm-indices.o.
9701 * vec-perm-indices.h: New file.
9702 * vec-perm-indices.c: Likewise.
9703 * target.h (vec_perm_indices): Replace with a forward class
9705 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9706 * optabs.h: Include vec-perm-indices.h.
9707 (expand_vec_perm): Delete.
9708 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9709 (expand_vec_perm_const): Declare.
9710 * target.def (vec_perm_const_ok): Replace with...
9711 (vec_perm_const): ...this new hook.
9712 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9713 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9714 * doc/tm.texi: Regenerate.
9715 * optabs.def (vec_perm_const): Delete.
9716 * doc/md.texi (vec_perm_const): Likewise.
9717 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9718 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9719 expand_vec_perm for constant permutation vectors. Assert that
9720 the mode of variable permutation vectors is the integer equivalent
9721 of the mode that is being permuted.
9722 * optabs-query.h (selector_fits_mode_p): Declare.
9723 * optabs-query.c: Include vec-perm-indices.h.
9724 (selector_fits_mode_p): New function.
9725 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9726 is defined, instead of checking whether the vec_perm_const_optab
9727 exists. Use targetm.vectorize.vec_perm_const instead of
9728 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9729 fit in the vector mode before using a variable permute.
9730 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9731 vec_perm_indices instead of an rtx.
9732 (expand_vec_perm): Replace with...
9733 (expand_vec_perm_const): ...this new function. Take the selector
9734 as a vec_perm_indices rather than an rtx. Also take the mode of
9735 the selector. Update call to shift_amt_for_vec_perm_mask.
9736 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9737 Use vec_perm_indices::new_expanded_vector to expand the original
9738 selector into bytes. Check whether the indices fit in the vector
9739 mode before using a variable permute.
9740 (expand_vec_perm_var): Make global.
9741 (expand_mult_highpart): Use expand_vec_perm_const.
9742 * fold-const.c: Includes vec-perm-indices.h.
9743 * tree-ssa-forwprop.c: Likewise.
9744 * tree-vect-data-refs.c: Likewise.
9745 * tree-vect-generic.c: Likewise.
9746 * tree-vect-loop.c: Likewise.
9747 * tree-vect-slp.c: Likewise.
9748 * tree-vect-stmts.c: Likewise.
9749 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9751 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9752 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9753 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9754 (aarch64_vectorize_vec_perm_const): ...this new function.
9755 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9756 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9757 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9758 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9759 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9760 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9761 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9763 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9764 check for NEON modes.
9765 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9766 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9767 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9768 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9770 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9771 the old VEC_PERM_CONST conditions.
9772 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9773 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9774 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9775 (ia64_vectorize_vec_perm_const_ok): Merge into...
9776 (ia64_vectorize_vec_perm_const): ...this new function.
9777 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9778 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9779 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9780 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9781 * config/mips/mips.c (mips_expand_vec_perm_const)
9782 (mips_vectorize_vec_perm_const_ok): Merge into...
9783 (mips_vectorize_vec_perm_const): ...this new function.
9784 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9785 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9786 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9787 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9788 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9789 (rs6000_expand_vec_perm_const): Delete.
9790 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9792 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9793 (altivec_expand_vec_perm_const_le): Take each operand individually.
9794 Operate on constant selectors rather than rtxes.
9795 (altivec_expand_vec_perm_const): Likewise. Update call to
9796 altivec_expand_vec_perm_const_le.
9797 (rs6000_expand_vec_perm_const): Delete.
9798 (rs6000_vectorize_vec_perm_const_ok): Delete.
9799 (rs6000_vectorize_vec_perm_const): New function.
9800 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9801 an element count and rtx array.
9802 (rs6000_expand_extract_even): Update call accordingly.
9803 (rs6000_expand_interleave): Likewise.
9804 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9805 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9806 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9807 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9808 (rs6000_expand_vec_perm_const): Delete.
9809 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9810 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9811 (altivec_expand_vec_perm_const_le): Take each operand individually.
9812 Operate on constant selectors rather than rtxes.
9813 (altivec_expand_vec_perm_const): Likewise. Update call to
9814 altivec_expand_vec_perm_const_le.
9815 (rs6000_expand_vec_perm_const): Delete.
9816 (rs6000_vectorize_vec_perm_const_ok): Delete.
9817 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9818 reference to the SPE evmerge intructions.
9819 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9820 an element count and rtx array.
9821 (rs6000_expand_extract_even): Update call accordingly.
9822 (rs6000_expand_interleave): Likewise.
9823 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9824 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9826 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9830 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9831 vector mode and that that mode matches the mode of the data
9833 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9834 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9835 directly using expand_vec_perm_1 when forcing selectors into
9837 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9841 * optabs-query.h (can_vec_perm_p): Delete.
9842 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9843 * optabs-query.c (can_vec_perm_p): Split into...
9844 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9845 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9846 particular selector is valid.
9847 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9848 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9849 (vect_grouped_load_supported): Likewise.
9850 (vect_shift_permute_load_chain): Likewise.
9851 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9852 (vect_transform_slp_perm_load): Likewise.
9853 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9854 (vectorizable_bswap): Likewise.
9855 (vect_gen_perm_mask_checked): Likewise.
9856 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9857 implementations of variable permutation vectors into account
9858 when deciding which selector to use.
9859 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9860 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9861 with a false third argument.
9862 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9863 to test whether the constant selector is valid and can_vec_perm_var_p
9864 to test whether a variable selector is valid.
9866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9868 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9869 * optabs-query.c (can_vec_perm_p): Likewise.
9870 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9871 instead of vec_perm_indices.
9872 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9873 (vect_gen_perm_mask_checked): Likewise,
9874 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9875 (vect_gen_perm_mask_checked): Likewise,
9877 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9879 * optabs-query.h (qimode_for_vec_perm): Declare.
9880 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9881 (qimode_for_vec_perm): ...this new function.
9882 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9884 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9886 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9887 does not have a conditional at the top.
9889 2018-01-02 Richard Biener <rguenther@suse.de>
9891 * ipa-inline.c (big_speedup_p): Fix expression.
9893 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9896 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9899 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9903 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9904 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9905 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9906 cond_taken_branch_cost 3->4.
9908 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9910 PR tree-optimization/83581
9911 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9912 TODO_cleanup_cfg if any changes have been made.
9915 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9916 convert_modes if target mode has the right side, but different mode
9920 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9921 last argument when extracting from CONCAT. If either from_real or
9922 from_imag is NULL, use expansion through memory. If result is not
9923 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9924 the parts directly to inner mode, if even that fails, use expansion
9928 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9929 check for bswap in mode rather than HImode and use that in expand_unop
9932 Copyright (C) 2018 Free Software Foundation, Inc.
9934 Copying and distribu