1 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5 based on LABEL_REF. Remove useless assertion.
6 (pic_address_needs_scratch): Fix formatting.
7 (sparc_legitimize_pic_address): Minor tweaks.
8 (sparc_delegitimize_address): Adjust assertion accordingly.
9 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10 into symbolic_operand.
11 (movsi_high_pic_label_ref): Likewise.
12 (movsi_lo_sum_pic_label_ref): Likewise.
13 (movdi_pic_label_ref): Likewise.
14 (movdi_high_pic_label_ref): Likewise.
15 (movdi_lo_sum_pic_label_ref): Likewise.
17 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
19 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20 custom LIB_SPEC setup.
22 2018-04-06 Ruslan Bukin <br@bsdpad.com>
23 Kito Cheng <kito.cheng@gmail.com>
25 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
26 * config/riscv/freebsd.h: New.
28 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
30 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
31 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
34 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
35 Kito Cheng <kito.cheng@gmail.com>
37 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
38 nds32_output_call, nds32_symbol_binds_local_p): New functions.
39 * config/nds32/nds32-protos.h (nds32_output_call,
40 nds32_output_return): Declare.
41 * config/nds32/nds32.md: Refine all the call and return patterns.
43 2018-04-06 Jakub Jelinek <jakub@redhat.com>
46 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
47 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
49 PR rtl-optimization/84872
50 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
51 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
54 2018-04-06 Tamar Christina <tamar.christina@arm.com>
56 * expr.c (copy_blkmode_to_reg): Revert 254862.
57 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
59 2018-04-06 Richard Biener <rguenther@suse.de>
62 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
63 after seeing a component reference with an adjacent field. Treat
64 refs to arrays at struct end of external decls similar to
65 refs to unconstrained commons.
67 2018-04-06 Jakub Jelinek <jakub@redhat.com>
70 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
71 look through SAVE_EXPRs with non-side-effects argument. Adjust
73 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
76 2018-04-06 Richard Biener <rguenther@suse.de>
79 * alias.c (find_base_term): New wrapper around find_base_term
80 unwinding CSELIB_VAL_PTR changes.
81 (find_base_term): Do not restore CSELIB_VAL_PTR during the
84 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
86 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
88 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
90 ("nop"): lr 0,0 -> nopr r0
91 ("nop_lr0", "nop_lr1"): New insn definitions.
93 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
95 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
96 NDS32_V3PUSH_AVAILABLE_P macro.
98 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
99 Chung-Ju Wu <jasonwucj@gmail.com>
101 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
102 (nds32*-*-*): Add float and fpu_config into supported_defaults.
103 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
104 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
105 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
106 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
107 * config/nds32/constraints.md: New constraints and checking for hard
109 * config/nds32/iterators.md: New mode iterator and attribute for hard
111 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
113 * config/nds32/nds32-fpu.md: New file.
114 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
115 deal with hard float code generation.
116 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
118 (abi_type, float_reg_number): New enum type.
119 * config/nds32/nds32-predicates.c: New predicates for hard float.
120 * config/nds32/nds32-protos.h: Declare functions for hard float.
121 * config/nds32/nds32.c: Implementation for hard float configuration.
122 * config/nds32/nds32.h: Definitions for hard float configuration.
123 * config/nds32/nds32.md: Include hard float machine description and
124 modify patterns for hard float configuration.
125 * config/nds32/nds32.opt: New options for hard float configuration.
126 * config/nds32/predicates.md: New predicates for hard float
129 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
131 * common/config/nds32/nds32-common.c
132 (nds32_option_optimization_table): Enable -mreleax-hint by default.
134 2018-04-05 Jakub Jelinek <jakub@redhat.com>
137 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
138 CONSTRUCTOR_ELT (ctor, ...)->value.
140 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
143 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
145 2018-04-05 Tom de Vries <tom@codesourcery.com>
148 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
151 2018-04-05 Shiva Chen <shiva0217@gmail.com>
152 Kito Cheng <kito.cheng@gmail.com>
154 * config/nds32/constraints.md (U33): Fine-tune checking condition.
155 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
156 * config/nds32/nds32.h (nds32_16bit_address_type): Add
157 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
159 2018-04-05 Shiva Chen <shiva0217@gmail.com>
160 Kito Cheng <kito.cheng@gmail.com>
162 * config/nds32/constraints.md (Ufe): New memory constraint.
163 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
164 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
165 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
167 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
168 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
170 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
172 * config/nds32/nds32.md: Use optimize_size in the condition for
173 alu-shift instructions.
175 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
177 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
179 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
181 * config/nds32/nds32.md (negsi2): Refine pattern.
183 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
184 Chung-Ju Wu <jasonwucj@gmail.com>
186 * config/nds32/iterators.md (shift_rotate): New code iterator.
187 (shift): New code attribute.
188 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
189 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
190 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
191 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
193 (andsi3, *andsi3): Ditto.
194 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
195 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
196 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
197 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
198 nds32_ior_operand, nds32_xor_operand): New predicates.
200 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
202 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
203 (addsi3, subsi3): ... this.
205 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
207 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
209 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
211 * config/nds32/nds32.md: Adjust indention.
213 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
215 * config/nds32/nds32.md (feature): New attribute.
217 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
219 * config/nds32/nds32.md (subtype): New attribute.
221 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
224 * config/arm/arm-builtins.c (arm_expand_builtin): Change
225 expansion to perform a bitwise AND of the argument followed by a
226 boolean negation of the result.
228 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
230 PR rtl-optimization/84878
231 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
232 the basic block. Assert the use reference is not artificial and that
233 it has an associated insn.
235 2018-04-04 Michael Matz <matz@suse.de>
237 * builtins.c (compute_objsize): Pass correct operand
238 to array_at_struct_end_p.
240 2018-04-04 Richard Biener <rguenther@suse.de>
243 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
244 from contexts for DINFO_LEVEL_TERSE and below.
246 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
248 * config/nds32/nds32-doubleword.md (move_<mode>): Require
249 resiter_operand condition.
250 * config/nds32/nds32.md (*move<mode>): Ditto.
252 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
253 Monk Chiang <sh.chiang04@gmail.com>
255 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
257 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
259 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
261 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
262 Kito Cheng <kito.cheng@gmail.com>
264 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
265 nds32_cond_code_str, output_cond_branch,
266 output_cond_branch_compare_zero, nds32_expand_cbranch,
267 nds32_expand_cstore, nds32_expand_movcc,
268 nds32_output_cbranchsi4_equality_zero,
269 nds32_output_cbranchsi4_equality_reg,
270 nds32_output_cbranchsi4_equality_reg_or_const_int,
271 nds32_output_cbranchsi4_greater_less_zero: New functions.
272 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
273 nds32_expand_cstore, nds32_expand_movcc,
274 nds32_output_cbranchsi4_equality_zero,
275 nds32_output_cbranchsi4_equality_reg,
276 nds32_output_cbranchsi4_equality_reg_or_const_int,
277 nds32_output_cbranchsi4_greater_less_zero): Declare.
278 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
279 nds32_rimm11s_operand): New predicates.
280 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
281 * config/nds32/nds32.md: Rewrite all the branch and conditional move
284 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
286 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
287 * config/nds32/nds32.md: Ditto.
288 * config/nds32/pipelines.md: Ditto.
290 2018-04-04 Richard Biener <rguenther@suse.de>
292 PR tree-optimization/85168
293 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
294 propagating abnormals.
296 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
298 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
300 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
301 Kito Cheng <kito.cheng@gmail.com>
303 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
304 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
305 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
306 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
307 * config/nds32/nds32.md (sibcall_internal): New.
308 (sibcall_register): Remove.
309 (sibcall_immediate): Remove.
310 (sibcall_value_internal): New.
311 (sibcall_value_register): Remove.
312 (sibcall_value_immediate): Remove.
313 * config/nds32/predicates.md (nds32_general_register_operand): New.
314 (nds32_call_address_operand): New.
316 2018-04-03 Jakub Jelinek <jakub@redhat.com>
318 PR rtl-optimization/85167
319 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
320 bb_defs if *split_p, instead preinitialize it to NULL.
322 PR tree-optimization/85156
323 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
324 evaluating the argument multiple times.
326 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
328 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
330 (_mm_cvtpd_ps): Likewise.
331 (_mm_cvttpd_epi32): Likewise.
332 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
333 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
334 vector, pixel, and bool following altivec.h include.
336 2018-04-03 Martin Sebor <msebor@redhat.com>
338 * doc/extend.texi (Common Function Attributes): Clarify.
339 (const attribute): Likewise.
340 (pure attribute): Likewise.
342 2018-04-03 Jakub Jelinek <jakub@redhat.com>
345 * config/i386/i386.c (ix86_expand_vector_set): Use
346 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
348 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
350 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
351 instructions when changing rounding bits to preserve precision bits
352 in the x87 control word.
354 2018-04-03 Martin Liska <mliska@suse.cz>
356 PR tree-optimization/82491
357 * rtl.h (strip_offset_and_add): Replace += suboffset with
358 poly_uint64 () + suboffset.
360 2018-03-29 Martin Liska <mliska@suse.cz>
361 Martin Jambor <mjambor@suse.cz>
364 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
365 param_type is not an integral or pointer type.
367 2018-04-03 Richard Biener <rguenther@suse.de>
369 * sese.h (recompute_all_dominators): Remove.
371 2018-04-02 Martin Sebor <msebor@redhat.com>
373 * doc/invoke.texi (-Wrestrict): Fix typos.
375 2018-04-02 Jim Wilson <jimw@sifive.com>
377 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
378 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
379 (<optab>di3, <optab>si3_extend): Likewise.
380 (<optab>si3_mask, <optab>si3_mask_1): New.
381 (<optab>di3_mask, <optab>di3_mask_1): New.
382 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
383 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
384 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
386 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
388 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
391 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
393 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
394 (nds32_canonicalize_comparison): New function.
396 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
397 Kito Cheng <kito.cheng@gmail.com>
398 Kuan-Lin Chen <kuanlinchentw@gmail.com>
400 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
401 * config/nds32/constants.md (unspec_volatile_element): Add
402 UNSPEC_VOLATILE_RELAX_GROUP.
403 * config/nds32/nds32-relax-opt.c: New file.
404 * config/nds32/nds32-predicates.c
405 (nds32_symbol_load_store_p): New function.
406 * config/nds32/nds32-protos.h
407 (nds32_symbol_load_store_p): Declare function.
408 (make_pass_nds32_relax_opt): Declare new rtl pass function.
409 * config/nds32/nds32.c
410 (nds32_register_pass): New function to register pass.
411 (nds32_register_passes): New function to register passes.
412 * config/nds32/nds32.md (relax_group): New pattern.
413 * config/nds32/nds32.opt (mrelax-hint): New option.
414 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
416 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
418 * config/nds32/t-nds32: Modify files dependency.
420 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
422 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
423 (PROFILE_HOOK): Define its implementation.
425 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
427 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
428 type and 32-bit size.
430 2018-04-01 Jakub Jelinek <jakub@redhat.com>
433 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
434 (V_128_256): New mode iterator.
435 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
436 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
437 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
439 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
440 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
442 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
445 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
446 NaN inputs correctly.
448 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
451 * config/rs6000/vsx.md (??r): New mode attribute.
452 (*vsx_mov<mode>_64bit): Use it.
453 (*vsx_mov<mode>_32bit): Likewise.
455 2018-03-30 Martin Sebor <msebor@redhat.com>
457 PR tree-optimization/84818
458 * builtins.c (check_access): Use warning_n.
460 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
463 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
465 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
468 2018-03-30 Julia Koval <julia.koval@intel.com>
471 * x86-tune.def (movx, partial_reg_dependency): Enable for
474 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
477 * lra-constraints.c (process_alt_operands): Move setting
478 this_alternative_matches below.
480 2018-03-29 Martin Liska <mliska@suse.cz>
483 * doc/invoke.texi: Document how LTO works with debug info.
484 Describe auto-load support of binutils. Mention 'x86-64'
485 as valid option value of -march option.
487 2018-03-29 Jakub Jelinek <jakub@redhat.com>
489 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
492 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
493 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
494 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
497 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
500 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
501 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
502 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
503 for RS6000_BTM_POWERPC64.
504 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
505 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
506 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
511 2018-03-28 Carl Love <cel@us.ibm.com>
514 2017-09-27 Carl Love <cel@us.ibm.com>
516 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
517 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
518 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
521 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
523 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
524 instead of __vector bool.
525 (_mm_max_pu8): Likewise.
526 (_mm_min_pi16): Likewise.
528 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
531 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
535 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
537 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
538 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
539 (div_extend): Likewise.
540 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
542 (__builtin_divweuo): Likewise.
543 (__builtin_divdeo): Likewise.
544 (__builtin_divdeuo): Likewise.
546 2018-03-28 Jakub Jelinek <jakub@redhat.com>
549 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
550 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
552 PR tree-optimization/82004
553 * gimple-match-head.c (optimize_pow_to_exp): New function.
554 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
555 Don't fold to exp if optimize_pow_to_exp is false.
557 2018-03-28 Martin Liska <mliska@suse.cz>
560 * calls.c (initialize_argument_information): Fix trailing space.
561 * common.opt: Fix typo and provide better explanation for
562 -fsanitize-coverage option.
563 * config/i386/i386.opt: Fix typo.
565 2018-03-28 Jakub Jelinek <jakub@redhat.com>
566 Martin Liska <mliska@suse.cz>
569 * gimplify.c (asan_poison_variable): Don't do the check for
570 gimplify_omp_ctxp here.
571 (gimplify_decl_expr): Do it here.
572 (gimplify_target_expr): Likewise.
574 2018-03-28 Martin Liska <mliska@suse.cz>
577 * config/i386/i386.c (ix86_function_arg_advance): Do not call
578 chkp_type_bounds_count if MPX is not enabled.
580 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
582 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
584 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
587 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
588 function to create the function decl for complex long double
589 multiply and divide for -mabi=ieeelongdouble.
590 (init_float128_ieee): Call it.
592 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
595 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
596 -fcf-protection=branch -mibt.
597 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
599 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
602 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
604 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
607 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
608 extern array declarations.
610 2018-03-27 Richard Biener <rguenther@suse.de>
613 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
614 explicit single_use checks.
616 2018-03-27 Richard Biener <rguenther@suse.de>
618 PR tree-optimization/85082
619 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
622 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
624 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
625 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
626 Turn on fasynchronous-unwind-tables and funwind-tables.
628 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
631 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
632 (*bmi_blsr_<mode>_ccz): Ditto.
634 2018-03-26 Tom de Vries <tom@codesourcery.com>
636 PR tree-optimization/85063
637 * omp-general.c (offloading_function_p): New function. Factor out
639 * omp-offload.c (pass_omp_target_link::gate): ... here.
640 * omp-general.h (offloading_function_p): Declare.
641 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
642 with attribute omp declare target for offloading functions.
644 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
646 PR tree-optimization/84005
647 * tree-data-ref.h (get_base_for_alignment): Declare.
648 * tree-data-ref.c (get_base_for_alignment_1): New function.
649 (get_base_for_alignment): Likewise.
650 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
651 get_base_for_alignment to find a suitable base object, instead
652 of always using drb->base_address.
654 2018-03-23 Jakub Jelinek <jakub@redhat.com>
657 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
658 known size by default.
660 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
663 * lra-constraints.c (process_alt_operands): Don't match BLKmode
664 and non BLKmode operands.
666 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
669 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
672 2018-03-23 Richard Biener <rguenther@suse.de>
675 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
676 we are going to emit early debug for LTO.
678 2018-03-23 Jakub Jelinek <jakub@redhat.com>
681 * function.c (match_asm_constraints_1): Don't optimize if input
682 doesn't satisfy general_operand predicate for output's mode.
685 * alias.c (write_dependence_p): Don't require for x_canonicalized
686 non-VOIDmode if x has VOIDmode.
689 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
690 just don't try to optimize it rather than assert it never happens.
692 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
694 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
695 macro expansions for definition of ST_INTERNAL_<mode> and
696 LD_INTERNAL_<mode> builtins.
697 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
699 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
701 (altivec_expand_st_builtin): Likewise.
702 (altivec_expand_builtin): Remove calls to deleted functions.
703 (rs6000_address_for_altivec): Delete this function.
704 * config/rs6000/vector.md: Remove expands for
705 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
707 2018-03-22 Sudakshina Das <sudi.das@arm.com>
710 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
711 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
712 re-computing once computed.
713 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
714 (arm_init_machine_status): Initialize
715 machine->static_chain_stack_bytes.
717 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
720 * doc/extend.texi: Add four new prototypes for vec_ld.
721 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
722 definitions for more logical presentation.
723 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
724 entries for V1TI variants of __builtin_altivec_ld builtin.
725 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
726 handling of V1TI variant of LVX icode pattern.
727 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
728 (rs6000_gimple_fold_builtin): Likewise.
729 (altivec_init_builtins): Add code to define
730 __builtin_altivec_lvx_v1ti function.
732 2018-03-22 Jakub Jelinek <jakub@redhat.com>
735 * function.c (match_asm_constraints_1): Don't do the optimization
736 if input isn't a REG, SUBREG, MEM or constant.
738 2018-03-22 Tom de Vries <tom@codesourcery.com>
740 PR tree-optimization/84956
741 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
742 bb_has_abnormal_pred.
744 2018-03-22 Jakub Jelinek <jakub@redhat.com>
747 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
748 DECL_INITIAL (decl) to decl at the end.
749 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
752 2018-03-21 Joseph Myers <joseph@codesourcery.com>
754 * doc/extend.texi (__builtin_tgmath): Document when complex
755 integer types are treated as _Complex _Float64.
757 2018-03-21 Tom de Vries <tom@codesourcery.com>
759 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
761 2018-03-21 Jakub Jelinek <jakub@redhat.com>
763 PR tree-optimization/84960
764 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
765 if it is ENTRY block, move them into single succ of ENTRY in that case.
767 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
769 PR tree-optimization/84811
770 * poly-int.h (poly_span_traits): Remove the T3 parameter and
771 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
772 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
773 (known_subrange_p): Update accordingly. Cast each value involved
774 in the size comparison, rather than casting the result of the
777 2018-03-21 Jakub Jelinek <jakub@redhat.com>
779 PR tree-optimization/84982
780 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
781 by flipping the least significant bit rather than all bits from
782 bitpos to bitpos + bitsize - 1.
784 2018-03-21 Nathan Sidwell <nathan@acm.org>
786 * doc/extend.texi (Deprecated Features): Remove mention of
787 long-deleted deprecations.
789 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
792 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
793 * configure: Regenerate.
795 2018-03-21 Tom de Vries <tom@codesourcery.com>
797 PR tree-optimization/83126
798 * tree-parloops.c (num_phis): New function.
799 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
801 2018-03-21 Nathan Sidwell <nathan@acm.org>
803 * doc/extend.texi (Deprecated Features): Update deprecated flags,
804 mention anon-struct/union members and trailing attributes.
806 2018-03-21 Bin Cheng <bin.cheng@arm.com>
808 PR tree-optimization/84969
809 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
810 builtin memset partitions if they set different rhs values.
812 2018-03-21 Jakub Jelinek <jakub@redhat.com>
814 PR rtl-optimization/84989
815 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
816 VEC_DUPLICATE with scalar result mode.
818 2018-03-21 Martin Liska <mliska@suse.cz>
821 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
822 not intended return statement.
824 2018-03-21 Martin Liska <mliska@suse.cz>
827 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
828 (chkp_find_bound_slots_1): Limit number of iterations.
830 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
833 * Minor grammar fixes for x86 options.
835 2018-03-20 Jakub Jelinek <jakub@redhat.com>
838 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
839 holding REG_CFA_RESTORE notes, instead turn them into a USE.
841 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
844 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
845 (altivec_lvx_<mode>_1op): Likewise.
846 (altivec_stvx_<mode>_2op): Likewise.
847 (altivec_stvx_<mode>_1op): Likewise.
848 (altivec_lvx_<VM2:mode>): New define_expand.
849 (altivec_stvx_<VM2:mode>): Likewise.
850 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
851 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
852 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
853 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
854 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
855 (rs6000_gen_lvx): Likewise.
856 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
857 (altivec_expand_stv_builtin): Likewise.
858 (altivec_expand_builtin): Likewise.
859 * config/rs6000/vector.md: Likewise.
861 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
864 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
867 2018-03-20 Richard Biener <rguenther@suse.de>
870 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
871 sign-conversions as zero, fall back to standard scalar_stmt
874 2018-03-20 Martin Liska <mliska@suse.cz>
877 * predict.c (rebuild_frequencies): Handle case when we have
878 PROFILE_ABSENT, but flag_guess_branch_prob is false.
880 2018-03-20 Jakub Jelinek <jakub@redhat.com>
883 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
884 flag_section_anchors.
885 * varasm.c (use_blocks_for_decl_p): Remove hack for
889 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
891 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
892 be created, use lowpart_subreg of operands[0] rather than operands[0]
894 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
895 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
896 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
897 and n constraint instead of aarch64_shift_imm_di and Usd.
898 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
899 (*aarch64_<optab>_reg_minus<mode>3): ... this.
901 2018-03-20 Sudakshina Das <sudi.das@arm.com>
904 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
905 to favor GPR over NEON registers.
906 (<shift>di3_neon): Likewise.
908 2018-03-20 Tom de Vries <tom@codesourcery.com>
911 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
912 (nvptx_process_pars): Emit bar.sync asap and alap.
914 2018-03-20 Tom de Vries <tom@codesourcery.com>
917 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
918 seen_label if seen_label is already set.
920 2018-03-20 Jakub Jelinek <jakub@redhat.com>
923 * config/i386/i386.c (fold_builtin_cpu): For features above 31
924 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
925 Use 1U instead of 1. Formatting fixes.
928 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
929 instead of TREE_TYPE (s1) for the return value.
931 2018-03-19 Jakub Jelinek <jakub@redhat.com>
933 PR tree-optimization/84946
934 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
935 bitsize + bitsize in poly_uint64 rather than poly_int64.
938 * dwarf2asm.c: Include fold-const.c.
939 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
940 of decl rather than decl itself.
942 PR rtl-optimization/84643
943 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
945 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
948 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
949 calling assemble_variable.
951 2018-03-19 Sudakshina Das <sudi.das@arm.com>
954 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
955 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
957 2018-03-19 Jim Wilson <jimw@sifive.com>
960 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
961 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
962 (riscv_first_stack_step): Likewise.
963 (riscv_option_override): Use STACK_BOUNDARY instead of
965 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
967 (BIGGEST_ALIGNMENT): Set to 128.
968 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
969 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
972 2018-03-19 Richard Biener <rguenther@suse.de>
974 PR tree-optimization/84933
975 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
976 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
978 2018-03-19 Richard Biener <rguenther@suse.de>
980 PR tree-optimization/84859
981 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
982 (cond_if_else_store_replacement): Perform sinking operation on
983 single-store BBs regardless of MAX_STORES_TO_SINK setting.
984 Generalize what a BB with a single eligible store is.
986 2018-03-19 Richard Biener <rguenther@suse.de>
988 PR tree-optimization/84929
989 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
990 chrec_is_positive against non-chrec arg.
992 2018-03-19 Tamar Christina <tamar.christina@arm.com>
995 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
997 2018-03-18 Martin Liska <mliska@suse.cz>
999 PR rtl-optimization/84635
1000 * regrename.c (build_def_use): Use matches_mode only when
1003 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1005 PR tree-optimization/84913
1006 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1007 vectorize chains of COND_EXPRs.
1009 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1011 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1013 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1015 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1017 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1019 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1021 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1022 Kito Cheng <kito.cheng@gmail.com>
1024 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1025 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1026 (nds32_adjust_reg_alloc_order): New function.
1027 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1029 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1031 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1032 nds32_print_operand, nds32_print_operand_address): Use
1033 HOST_WIDE_INT_PRINT_DEC instead.
1035 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1037 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1039 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1042 * config/i386/i386.c (initial_ix86_tune_features,
1043 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1045 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1046 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1047 rather than 1u << ix86_tune. Formatting fix.
1048 (ix86_option_override_internal): Change ix86_arch_mask from
1049 unsigned int to unsigned HOST_WIDE_INT, initialize to
1050 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1051 (ix86_function_specific_restore): Likewise.
1053 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1056 * postreload.c (reload_combine_recognize_pattern): Perform
1057 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1058 truncate_int_for_mode the result for the destination's mode.
1061 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1062 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1064 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1067 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1070 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1071 regno_allocno_class_array and sorted_pseudos.
1072 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1073 insns where regno is used.
1075 2018-03-16 Martin Liska <mliska@suse.cz>
1078 * multiple_target.c (create_dispatcher_calls): Redirect
1079 reference in the symbol table.
1081 2018-03-16 Martin Liska <mliska@suse.cz>
1084 * multiple_target.c (create_dispatcher_calls): Redirect also
1087 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1091 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1092 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1093 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1095 2018-03-16 Julia Koval <julia.koval@intel.com>
1097 * doc/invoke.texi (Skylake Server): Add CLWB.
1098 Cannonlake): Remove CLWB.
1100 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1102 PR tree-optimization/84841
1103 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1105 (FLOAT_ONE_CONST_TYPE): Define.
1106 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1107 (sort_by_operand_rank): Put entries with higher constant_type last
1108 rather than first to match comments.
1110 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1112 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1115 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1118 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1119 check and use address of target always.
1121 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1124 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1125 (indirect_thunk_bnd_needed): Likewise.
1126 (indirect_thunks_used): Likewise.
1127 (indirect_thunks_bnd_used): Likewise.
1128 (indirect_return_needed): New.
1129 (indirect_return_bnd_needed): Likewise.
1130 (output_indirect_thunk_function): Add a bool argument for
1132 (output_indirect_thunk_function): Don't generate alias for
1133 function return thunk.
1134 (ix86_code_end): Call output_indirect_thunk_function to generate
1135 function return thunks.
1136 (ix86_output_function_return): Set indirect_return_bnd_needed
1137 and indirect_return_needed instead of indirect_thunk_bnd_needed
1138 and indirect_thunk_needed.
1140 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1142 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1143 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1144 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1146 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1147 Paul Hua <paul.hua.gm@gmail.com>
1150 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1152 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1154 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1155 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1158 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1161 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1162 instead of GET_MODE_SIZE when comparing Units.
1164 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1167 * varasm.c (hash_section): Return an unchangeble hash value
1168 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1169 Return !aarch64_can_use_per_function_literal_pools_p ().
1171 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1174 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1175 rather than address of cmode as last argument to prepare_cmp_insn.
1177 2018-03-15 Julia Koval <julia.koval@intel.com>
1179 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1180 F_AVX512VNNI, F_AVX512BITALG): New.
1182 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1185 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1186 insn for floating-point loads and stores.
1188 2018-03-14 Carl Love <cel@us.ibm.com>
1190 * config/rs6000/rs6000-c.c: Add macro definitions for
1191 ALTIVEC_BUILTIN_VEC_PERMXOR.
1192 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1193 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1194 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1195 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1197 * config/doc/extend.texi: Add prototypes for vec_permxor.
1199 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1202 * diagnostic-show-locus.c (class layout_point): Convert m_line
1203 from int to linenum_type.
1204 (line_span::comparator): Use linenum "compare" function when
1205 comparing line numbers.
1206 (test_line_span): New function.
1207 (layout_range::contains_point): Convert param "row" from int to
1209 (layout_range::intersects_line_p): Likewise.
1210 (layout::will_show_line_p): Likewise.
1211 (layout::print_source_line): Likewise.
1212 (layout::should_print_annotation_line_p): Likewise.
1213 (layout::print_annotation_line): Likewise.
1214 (layout::print_leading_fixits): Likewise.
1215 (layout::annotation_line_showed_range_p): Likewise.
1216 (struct line_corrections): Likewise for field m_row.
1217 (line_corrections::line_corrections): Likewise for param "row".
1218 (layout::print_trailing_fixits): Likewise.
1219 (layout::get_state_at_point): Likewise.
1220 (layout::get_x_bound_for_row): Likewise.
1221 (layout::print_line): Likewise.
1222 (diagnostic_show_locus): Likewise for locals "last_line" and
1224 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1225 * input.c (selftest::test_linenum_comparisons): New function.
1226 (selftest::input_c_tests): Call it.
1227 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1228 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1229 * selftest.h (ASSERT_GT): New macro.
1230 (ASSERT_GT_AT): New macro.
1231 (ASSERT_LT): New macro.
1232 (ASSERT_LT_AT): New macro.
1234 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1236 PR rtl-optimization/84780
1237 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1239 2018-03-14 Martin Liska <mliska@suse.cz>
1241 * tree.c (record_node_allocation_statistics): Use
1242 get_stats_node_kind.
1243 (get_stats_node_kind): New function extracted from
1244 record_node_allocation_statistics.
1245 (free_node): Use get_stats_node_kind.
1247 2018-03-14 Richard Biener <rguenther@suse.de>
1249 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1250 that the value-set of ANTIC_IN doesn't grow.
1253 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1255 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1256 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1258 2018-03-14 Julia Koval <julia.koval@intel.com>
1260 * config.gcc (icelake-client, icelake-server): New.
1262 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1263 (initial_ix86_arch_features): Ditto.
1264 (PTA_SKYLAKE): Add SGX.
1265 (PTA_ICELAKE): Remove.
1266 (PTA_ICELAKE_CLIENT): New.
1267 (PTA_ICELAKE_SERVER): New.
1268 (ix86_option_override_internal): Split up icelake on icelake client and
1270 (get_builtin_code_for_version): Ditto.
1271 (fold_builtin_cpu): Ditto.
1272 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1273 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1274 * config/i386/i386.h (processor_type): Ditto.
1275 * doc/invoke.texi: Ditto.
1277 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1280 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1281 INTEGER_CST offset, add it together with bitpos / 8 and
1282 sign extend based on POINTER_SIZE.
1286 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1289 * config/i386/constraints.md (Yc): New register constraint.
1290 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1291 Use Yc constraint for alternative 2 of operand 0. Remove
1292 preferred_for_speed attribute.
1294 2018-03-14 Richard Biener <rguenther@suse.de>
1296 PR tree-optimization/84830
1297 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1298 with the old one to avoid oscillations.
1300 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1303 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1305 (assign_by_spills): Return a flag of reload assignment failure.
1306 Do not process the reload assignment failures. Do not spill other
1307 reload pseudos if they has the same reg class. Update n if
1309 (lra_assign): Add a return arg. Set up from the result of
1310 assign_by_spills call.
1311 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1312 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1313 usage_insns if it is not NULL.
1314 (spill_hard_reg_in_range): New function.
1315 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1316 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1317 function prototypes.
1318 (lra_assign): Change prototype.
1319 * lra.c (lra): Add code to deal with fails by splitting hard reg
1322 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1324 * config/riscv/riscv.opt (mrelax): New option.
1325 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1326 "norelax" when riscv_mrelax is disabled.
1327 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1329 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1332 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1333 reassociation for int modes.
1335 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1337 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1338 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1340 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1341 * config/aarch64/aarch64-sve.md
1342 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1343 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1344 (*extend<mode><Vwide>2): Rename to...
1345 (aarch64_sve_extend<mode><Vwide>2): ...this.
1346 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1347 renaming the old pattern to...
1348 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1350 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1351 define_expand, renaming the old pattern to...
1352 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1353 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1354 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1355 account when deciding which SVE instruction the optab should use.
1356 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1358 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1360 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1361 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1362 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1363 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1364 (tlsdesc_small_<mode>): Turn a define_expand and use
1365 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1366 (tlsdesc_small_advsimd_<mode>): ...this.
1367 (tlsdesc_small_sve_<mode>): New pattern.
1369 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1371 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1372 (UNSPEC_UMUL_HIGHPART): New constants.
1373 (MUL_HIGHPART): New int iteraor.
1374 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1375 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1377 (*<su>mul<mode>3_highpart): New define_insn.
1379 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1382 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1385 2018-03-13 Martin Liska <mliska@suse.cz>
1388 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1390 (sem_item_optimizer::~sem_item_optimizer): Release it.
1391 (sem_item_optimizer::merge_classes): Register variable aliases.
1392 (sem_item_optimizer::fixup_pt_set): New function.
1393 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1394 * ipa-icf.h: Declare new variables and functions.
1396 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1399 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1400 integer_pow2p@2 and test integer_pow2p in condition.
1401 (A < 0 ? C : 0): Similarly for @1.
1404 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1405 characters starting at p contain '\0' character, don't look beyond
1409 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1410 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1413 * reg-stack.c (change_stack): Change update_end var from int to
1414 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1415 also call set_block_for_insn on the newly added insns and rescan.
1418 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1419 on the last operand.
1422 * tree.c (stabilize_reference_1): Return save_expr (e) for
1423 STATEMENT_LIST even if it doesn't have side-effects.
1425 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1427 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1429 2018-03-12 Renlin Li <renlin.li@arm.com>
1431 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1432 aarch64_output_scalar_simd_mov_immediate.
1434 2018-03-12 Martin Sebor <msebor@redhat.com>
1436 PR tree-optimization/83456
1437 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1438 for perfectly overlapping calls to memcpy.
1439 (gimple_fold_builtin_memory_chk): Same.
1440 (gimple_fold_builtin_strcpy): Handle no-warning.
1441 (gimple_fold_builtin_stxcpy_chk): Same.
1442 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1444 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1446 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1447 parameter. Use it for SFmode.
1448 (rs6000_function_arg_advance_1): Adjust.
1449 (rs6000_function_arg): Adjust.
1450 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1452 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1454 PR rtl-optimization/84169
1455 PR rtl-optimization/84780
1456 * combine.c (can_combine_p): Check for a 2-insn combination whether
1457 the destination register is used between the two insns, too.
1459 2018-03-12 Richard Biener <rguenther@suse.de>
1461 PR tree-optimization/84803
1462 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1463 for refs DR analysis didn't process.
1465 2018-03-12 Richard Biener <rguenther@suse.de>
1467 PR tree-optimization/84777
1468 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1469 force-vectorize loops ignore whether we are optimizing for size.
1471 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1473 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1474 (TARGET_MD_ASM_ADJUST): Define.
1476 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1477 Kito Cheng <kito.cheng@gmail.com>
1478 Chung-Ju Wu <jasonwucj@gmail.com>
1480 * config/nds32/nds32.c (nds32_compute_stack_frame,
1481 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1482 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1483 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1484 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1485 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1486 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1487 * config/nds32/nds32.md (prologue, epilogue): Use macro
1488 NDS32_V3PUSH_AVAILABLE_P to do checking.
1490 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1493 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1494 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1495 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1496 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1497 addition of most attributes on !orig_type_die or the attribute not
1498 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1500 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1501 Chung-Ju Wu <jasonwucj@gmail.com>
1503 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1505 * config/nds32/nds32.opt (mvh): New option.
1507 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1508 Chung-Ju Wu <jasonwucj@gmail.com>
1510 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1512 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1513 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1516 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1517 Chung-Ju Wu <jasonwucj@gmail.com>
1519 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1521 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1522 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1524 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1525 Kito Cheng <kito.cheng@gmail.com>
1526 Chung-Ju Wu <jasonwucj@gmail.com>
1528 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1529 UNSPEC_FFMISM and UNSPEC_FLMISM.
1530 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1531 for ffb, ffmism and flmism.
1532 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1533 (unspec_ffmism): Ditto.
1534 (unspec_flmism): Ditto.
1535 (nds32_expand_builtin_impl): Check if string extension is available.
1536 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1537 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1539 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1542 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1545 * lra-assigns.c (assign_by_spills): Return a flag of reload
1546 assignment failure. Do not process the reload assignment
1547 failures. Do not spill other reload pseudos if they has the same
1549 (lra_assign): Add a return arg. Set up from the result of
1550 assign_by_spills call.
1551 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1552 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1553 usage_insns if it is not NULL.
1554 (spill_hard_reg_in_range): New function.
1555 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1556 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1557 function prototypes.
1558 (lra_assign): Change prototype.
1559 * lra.c (lra): Add code to deal with fails by splitting hard reg
1562 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1565 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1567 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1570 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1571 (dw_val_node): Add val_symbolic_view.
1572 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1573 (symview_upper_bound): New.
1574 (new_line_info_table): Initialize symviews_since_reset.
1575 (dwarf2out_source_line): Count symviews_since_reset and set
1576 symview_upper_bound.
1577 (dw_val_equal_p): Handle symview.
1578 (add_AT_symview): New.
1579 (print_dw_val): Handle symview.
1580 (attr_checksum, attr_checksum_ordered): Likewise.
1581 (same_dw_val_p, size_of_die): Likewise.
1582 (value_format, output_die): Likewise.
1583 (add_high_low_attributes): Use add_AT_symview for entry_view.
1584 (dwarf2out_finish): Reset symview_upper_bound, clear
1587 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1590 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1591 Add strict argument and use it.
1592 (rs6000_split_multireg_move): Update for new strict argument.
1593 (mem_operand_gpr): Disallow all non-offsettable addresses.
1594 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1596 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1599 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1600 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1601 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1604 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1605 decl, use remap_type if we want to use the type.
1607 2018-03-09 Martin Sebor <msebor@redhat.com>
1609 PR tree-optimization/84526
1610 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1612 (builtin_access::generic_overlap): Be prepared to handle non-array
1615 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1617 PR rtl-optimization/84682
1618 * lra-constraints.c (process_address_1): Check is_address flag
1619 for address constraints.
1620 (process_alt_operands): Likewise.
1621 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1622 preprocess_constraints.
1623 * recog.h (preprocess_constraints): Add oploc parameter.
1625 * recog.c (preprocess_constraints): Test address_operand for
1626 CT_ADDRESS constraints.
1628 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1631 * lra-assigns.c (assign_by_spills): Return a flag of reload
1632 assignment failure. Do not process the reload assignment
1633 failures. Do not spill other reload pseudos if they has the same
1635 (lra_assign): Add a return arg. Set up from the result of
1636 assign_by_spills call.
1637 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1638 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1639 usage_insns if it is not NULL.
1640 (spill_hard_reg_in_range): New function.
1641 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1642 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1643 function prototypes.
1644 (lra_assign): Change prototype.
1645 * lra.c (lra): Add code to deal with fails by splitting hard reg
1648 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1651 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1652 Accept complain bool parameter. Only emit errors if it is true.
1653 (arm_parse_cpu_option_name): Likewise.
1654 (arm_target_thumb_only): Adjust callers of the above.
1655 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1656 prototype to take a default true bool parameter.
1657 (arm_parse_arch_option_name): Likewise.
1659 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1660 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1664 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1665 * configure: Regenerate.
1666 * configure.ac ("linker --version-script option"): New.
1667 ("linker soname option"): New.
1669 2018-03-09 Richard Biener <rguenther@suse.de>
1671 PR tree-optimization/84775
1672 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1673 immediate uses of predicate stmts and mark them modified.
1676 PR tree-optimization/84178
1677 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1679 (version_loop_for_if_conversion): Delay update_ssa call.
1680 (tree_if_conversion): Delay update_ssa until after predicate
1683 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1686 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1687 when the function accesses prior frames.
1689 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1692 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1693 gen_llsym, otherwise call maybe_gen_llsym.
1696 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1697 has ',' character inside of it.
1699 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1702 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1703 as clobbering CC_REGNUM.
1705 2018-03-08 Richard Biener <rguenther@suse.de>
1708 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1709 (follow_copies_to_constant): Do not follow SSA names registered
1712 2018-03-08 Richard Biener <rguenther@suse.de>
1714 PR tree-optimization/84178
1715 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1717 (version_loop_for_if_conversion): Delay update_ssa call.
1718 (tree_if_conversion): Delay update_ssa until after predicate
1721 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1723 PR tree-optimization/84178
1724 * tree-if-conv.c (release_bb_predicate): Remove the
1725 the assertion that the stmts have NULL use_ops.
1726 Discard the statements, asserting that they haven't
1727 yet been added to a BB.
1729 2018-03-08 Richard Biener <rguenther@suse.de>
1731 PR tree-optimization/84746
1732 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1733 (phi_translate): Pass in destination ANTIC_OUT set.
1734 (phi_translate_1): Likewise. For a simplified result lookup
1735 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1736 (phi_translate_set): Adjust.
1737 (do_pre_regular_insertion): Likewise.
1738 (do_pre_partial_partial_insertion): Likewise.
1740 2018-03-08 Martin Liska <mliska@suse.cz>
1742 PR gcov-profile/84735
1743 * doc/gcov.texi: Document usage of profile files.
1744 * gcov-io.h: Document changes in the format.
1746 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1750 * dwarf2out.c (struct dw_line_info_table): Update comments for
1752 (FORCE_RESET_NEXT_VIEW): New.
1753 (FORCE_RESETTING_VIEW_P): New.
1754 (RESETTING_VIEW_P): Check for -1 too.
1755 (ZERO_VIEW_P): Likewise.
1756 (new_line_info_table): Force-reset next view.
1757 (dwarf2out_begin_function): Likewise.
1758 (dwarf2out_source_line): Simplify zero_view_p initialization.
1759 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1760 view directly. Omit view when omitting .loc at line 0.
1762 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1764 PR tree-optimization/84740
1765 * tree-switch-conversion.c (process_switch): Call build_constructors
1766 only if info.phi_count is non-zero.
1768 PR tree-optimization/84739
1769 * tree-tailcall.c (find_tail_calls): Check call arguments against
1770 DECL_ARGUMENTS (current_function_decl) rather than
1771 DECL_ARGUMENTS (func) when checking for tail recursion.
1773 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1775 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1776 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1777 Volker Reichelt's entry and add entries for people that perform
1778 GCC fuzzy testing and report numerous bugs.
1780 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1783 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1784 readonly data in sdata, if that is disabled.
1785 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1786 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1787 -mreadonly-in-sdata option.
1789 2018-03-07 Martin Sebor <msebor@redhat.com>
1791 PR tree-optimization/84468
1792 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1793 basic block when looking for nul assignment.
1795 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1798 * except.h (output_function_exception_table): Adjust prototype.
1799 * except.c (output_function_exception_table): Remove FNNAME parameter
1800 and add SECTION parameter. Ouput one part of the table at a time.
1801 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1802 the first part of the exception table and emit unwind directives.
1803 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1804 (i386_pe_seh_cold_init): Likewise.
1805 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1806 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1807 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1808 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1809 * config/i386/winnt.c: Include except.h.
1810 (struct seh_frame_state): Add reg_offset, after_prologue and
1811 in_cold_section fields.
1812 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1813 (i386_pe_seh_cold_init): New function.
1814 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1815 to seh->in_cold_section.
1816 (seh_emit_push): Record the offset of the push.
1817 (seh_emit_save): Record the offet of the save.
1818 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1819 Test seh->after_prologue to disregard the epilogue.
1820 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1821 (i386_pe_end_cold_function): New function.
1823 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1826 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1827 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1830 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1832 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1833 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1836 * multiple_target.c: Include tree-inline.h and intl.h.
1837 (expand_target_clones): Diagnose and fail if node->definition and
1838 !tree_versionable_function_p (node->decl).
1840 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1842 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1844 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1845 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1846 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1848 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1851 * combine.c (try_combine): Use reg_or_subregno instead of handling
1852 just paradoxical SUBREGs and REGs.
1854 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1856 * config/arc/arc.c (arc_finalize_pic): Remove function.
1857 (arc_must_save_register): We use single base PIC register, remove
1858 checks to save/restore the PIC register.
1859 (arc_expand_prologue): Likewise.
1860 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1862 (arc_verify_short): Likewise.
1863 (arc_attr_type): Likewise.
1864 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1865 (walk_stores): Likewise.
1866 (arc_address_cost): Make it static.
1867 (arc_verify_short): Likewise.
1868 (branch_dest): Likewise.
1869 (arc_attr_type): Likewise.
1870 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1871 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1872 (arc_final_prescan_insn): Remove inserting the nops due to
1873 hardware hazards. It is done in reorg step.
1874 (insn_length_variant_t): Remove.
1875 (insn_length_parameters_t): Likewise.
1876 (arc_insn_length_parameters): Likewise.
1877 (arc_get_insn_variants): Likewise.
1878 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1880 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1883 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1886 PR tree-optimization/84687
1887 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1889 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1891 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1893 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1894 Rename to ppc_speculation_barrier.
1895 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1896 __builtin_ppc_speculation_barrier.
1898 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1901 * combine.c (combine_simplify_rtx): Don't try to simplify if
1902 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1905 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1907 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1908 to 32 bytes when compiling for POWER9.
1910 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1913 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1914 regparm >= 3 with no arg reg available also for calls with
1915 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1918 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1920 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1922 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1925 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1927 2018-03-05 Richard Biener <rguenther@suse.de>
1929 PR tree-optimization/84486
1930 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1931 When inserting a __builtin_assume_aligned call set the LHS
1932 SSA name alignment info accordingly.
1934 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1936 PR tree-optimization/84114
1937 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1938 Avoid reassociation of FLOAT_MODE addition.
1940 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1942 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1943 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1944 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1945 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1946 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1947 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1948 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1950 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1951 __builtin_ia32_wbinvd): New builtins.
1952 (SPECIAL_ARGS2): New.
1953 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1954 (SPECIAL_ARGS2): New.
1955 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1956 (ix86_valid_target_attribute_inner_p): Ditto.
1957 (ix86_init_mmx_sse_builtins): Add special_args2.
1958 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1959 TARGET_WBNOINVD_P): New.
1960 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1961 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1962 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1963 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1964 * config/i386/pconfigintrin.h: New file.
1965 * config/i386/wbnoinvdintrin.h: Ditto.
1966 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1967 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1969 2018-03-05 Richard Biener <rguenther@suse.de>
1971 PR tree-optimization/84670
1972 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1974 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1975 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1976 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1977 doesn't grow if all successors (recursively) were visited at least
1980 2018-03-05 Richard Biener <rguenther@suse.de>
1982 PR tree-optimization/84650
1983 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1984 if executed in the loop pipeline.
1986 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1988 * doc/configfiles.texi (Configuration Files): Move info about
1989 conditionalizing $target-protos.h to...
1990 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1991 differs from $target-protos.h.
1993 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1994 Chung-Ju Wu <jasonwucj@gmail.com>
1996 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1997 * config/nds32/nds32-multiple.md (setmemsi): Define.
1998 * config/nds32/nds32-memory-manipulation.c
1999 (nds32_gen_dup_4_byte_to_word_value): New.
2000 (emit_setmem_word_loop): New.
2001 (emit_setmem_byte_loop): New.
2002 (nds32_expand_setmem_loop): New.
2003 (nds32_expand_setmem_loop_v3m): New.
2004 (nds32_expand_setmem_unroll): New.
2005 (nds32_expand_setmem): New.
2007 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2008 Chung-Ju Wu <jasonwucj@gmail.com>
2010 * config/nds32/nds32-memory-manipulation.c
2011 (nds32_emit_load_store): New.
2012 (nds32_emit_post_inc_load_store): New.
2013 (nds32_emit_mem_move): New.
2014 (nds32_emit_mem_move_block): New.
2015 (nds32_expand_movmemsi_loop_unknown_size): New.
2016 (nds32_expand_movmemsi_loop_known_size): New.
2017 (nds32_expand_movmemsi_loop): New.
2018 (nds32_expand_movmemsi_unroll): New.
2019 (nds32_expand_movmemqi): Rename ...
2020 (nds32_expand_movmemsi): ... to this.
2021 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2022 (movmemsi): ... to this.
2023 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2024 (nds32_expand_movmemsi): ... to this.
2026 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2027 Monk Chiang <sh.chiang04@gmail.com>
2028 Chung-Ju Wu <jasonwucj@gmail.com>
2030 * config/nds32/nds32-protos.h
2031 (nds32_expand_load_multiple): New arguments.
2032 (nds32_expand_store_multiple): Ditto.
2033 (nds32_valid_multiple_load_store): Rename ...
2034 (nds32_valid_multiple_load_store_p): ... to this.
2035 * config/nds32/nds32-memory-manipulation.c
2036 (nds32_expand_load_multiple): Refine implementation.
2037 (nds32_expand_store_multiple): Ditto.
2038 * config/nds32/nds32-multiple.md
2039 (load_multiple): Update nds32_expand_load_multiple interface.
2040 (store_multiple): Update nds32_expand_store_multiple interface.
2041 * config/nds32/nds32-predicates.c
2042 (nds32_valid_multiple_load_store): Rename ...
2043 (nds32_valid_multiple_load_store_p): ... to this and refine
2045 * config/nds32/predicates.md
2046 (nds32_load_multiple_and_update_address_operation): New predicate.
2047 (nds32_store_multiple_and_update_address_operation): New predicate.
2049 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2050 Chung-Ju Wu <jasonwucj@gmail.com>
2052 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2053 (combo): New attribute.
2054 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2056 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2058 * config/nds32/nds32.opt: Change -mcmodel= default value.
2060 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2061 Monk Chiang <sh.chiang04@gmail.com>
2062 Chung-Ju Wu <jasonwucj@gmail.com>
2064 * config/nds32/constants.md (unspec_element): New enum.
2065 * config/nds32/constraints.md (Umw): New constraint.
2066 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2067 * config/nds32/nds32-intrinsic.md: Likewise.
2068 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2069 (nds32_valid_smw_lwm_base_p): New.
2070 (nds32_output_smw_single_word): New.
2071 (nds32_output_lmw_single_word): New.
2072 (nds32_expand_unaligned_load): New.
2073 (nds32_expand_unaligned_store): New.
2074 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2075 (nds32_output_smw_single_word): Declare.
2076 (nds32_output_lmw_single_word): Declare.
2077 (nds32_expand_unaligned_load): Declare.
2078 (nds32_expand_unaligned_store): Declare.
2079 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2080 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2081 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2082 NDS32_BUILTIN_UASTORE_DW.
2083 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2086 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2087 Kito Cheng <kito.cheng@gmail.com>
2088 Chung-Ju Wu <jasonwucj@gmail.com>
2090 * config/nds32/nds32-intrinsic.c
2091 (nds32_expand_builtin_null_ftype_reg): Delete.
2092 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2093 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2094 (nds32_read_argument): New.
2095 (nds32_legitimize_target): Ditto.
2096 (nds32_legitimize_argument): Ditto.
2097 (nds32_check_constant_argument): Ditto.
2098 (nds32_expand_unop_builtin): Ditto.
2099 (nds32_expand_unopimm_builtin): Ditto.
2100 (nds32_expand_binop_builtin): Ditto.
2101 (nds32_builtin_decl_impl): Ditto.
2102 (builtin_description): Ditto.
2103 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2104 (nds32_init_builtins_impl): Ditto.
2105 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2106 (nds32_builtin_decl): New.
2107 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2108 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2110 2018-03-02 Jeff Law <law@redhat.com>
2112 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2113 (redundant_insn, fill_simple_delay_slots): Likewise.
2114 (fill_slots_from_thread): Likewise.
2115 * resource.c (mark_referenced_resources): Likewise.
2116 (mark_set_resources, find_dead_or_set_registers): Likewise.
2118 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2120 * substring-locations.h (format_warning_va): Formatting fix for
2122 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2124 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2125 * substring-locations.c: Include intl.h.
2126 (format_warning_va): Turned into small wrapper around
2127 format_warning_n_va, renamed to ...
2128 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2129 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2131 (format_warning_at_substring_n): New function.
2132 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2133 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2134 format_warning_at_substring with just a shorter name instead of
2135 const function pointer.
2136 (fmtwarn_n): New function.
2137 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2138 appropriate, get rid of all the fmtstr temporaries, move conditionals
2139 with G_() wrapped string literals directly into fmtwarn arguments,
2140 cast dir.len to (int), formatting fixes.
2142 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2144 * doc/invoke.texi: Remove "Cilk Plus" references.
2146 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2147 Richard Biener <rguenther@suse.de>
2150 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2151 for error or warning attributes if CALL_FROM_THUNK_P is set.
2154 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2157 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2158 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2161 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2162 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2164 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2165 instead of -1U in last predictors element's probability member.
2167 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2170 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2171 arguments if they are comparable.
2173 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2175 PR tree-optimization/84634
2176 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2177 masks and masked_loop_p with a single loop_masks, making sure it's
2178 null for bb vectorization.
2180 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2182 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2183 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2184 loop->force_vectorize to check whether there is no alias.
2186 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2189 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2191 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2193 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2194 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2195 instead of a loop around prev_real_insn.
2196 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2200 * config/i386/i386.c (ix86_print_operand): Use conditional
2201 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2204 2018-03-02 Richard Biener <rguenther@suse.de>
2206 PR tree-optimization/84427
2207 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2208 (bitmap_set_subtract_values): Rewrite to handle multiple
2211 (prune_clobbered_mems): Likewise.
2212 (phi_translate): Take edge instead of pred/phiblock.
2213 (phi_translate_1): Likewise.
2214 (phi_translate_set): Likewise. Insert all translated
2215 exprs for a value into the set, keeping possibly multiple
2216 expressions per value.
2217 (compute_antic_aux): Adjust for phi_translate changes.
2218 When intersecting union the expressions and prune those
2219 not in the final value set, keeping possibly multiple
2220 expressions per value. Do not use value-insertion
2221 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2222 all expressions. Add verification that the value-sets
2223 only shrink during iteration.
2224 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2225 (do_pre_regular_insertion): Likewise.
2226 (do_pre_partial_partial_insertion): Likewise.
2228 2018-03-02 Richard Biener <rguenther@suse.de>
2231 * config/darwin.c (saved_debug_info_level): New static global.
2232 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2233 (darwin_asm_lto_end): Restore debug info generation settings.
2235 2018-03-01 Martin Liska <mliska@suse.cz>
2238 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2241 2018-03-01 Richard Biener <rguenther@suse.de>
2244 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2245 type annotation with LTO when debug was disabled at compile-time.
2247 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2249 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2251 (mips_final_postscan_insn): Likewise.
2253 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2255 PR rtl-optimization/84528
2256 * alias.c (init_alias_target): Add commentary.
2257 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2258 a unique base value if the frame pointer is not eliminated
2259 to the stack pointer.
2261 2018-03-01 Tom de Vries <tom@codesourcery.com>
2263 PR rtl-optimization/83327
2264 * lra-int.h (hard_regs_spilled_into): Declare.
2265 * lra.c (hard_regs_spilled_into): Define.
2266 (init_reg_info): Init hard_regs_spilled_into.
2267 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2268 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2269 (process_bb_lives): Handle hard_regs_spilled_into.
2270 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2271 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2273 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2275 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2276 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2277 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2278 * config/rs6000/aix72.h: New file.
2280 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2282 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2283 instead of warning_at with conditional singular and plural messages
2287 * stor-layout.c (update_alignment_for_field): For
2288 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2289 && !DECL_PACKED (field), do the alignment update, just use
2290 only desired_align instead of MAX (type_align, desired_align)
2292 (place_field): Don't do known_align < desired_align handling
2293 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2294 is non-NULL, instead do it after rli->prev_field handling and
2295 only if not within a bitfield word. For DECL_PACKED (field)
2296 use type_align of BITS_PER_UNIT.
2298 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2300 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2301 superfluous parentheses and trailing spaces.
2303 2018-02-28 Richard Biener <rguenther@suse.de>
2305 PR tree-optimization/84584
2306 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2307 SCoPs with fake exit edge.
2309 2018-02-28 Martin Liska <mliska@suse.cz>
2312 * timevar.c (timer::print): Fix format to properly print 100%
2315 2018-02-28 Richard Biener <rguenther@suse.de>
2318 * genmatch.c (capture_info::walk_match): Do not mark
2319 captured expressions without operands as expr_p given
2320 they act more like predicates and should be subject to
2321 "lost tail" side-effect preserving.
2323 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2325 PR rtl-optimization/81611
2326 * auto-inc-dec.c (attempt_change): Move dead note from
2327 mem_insn if it's the next use of regno
2328 (find_address): Take address use of reg holding
2329 non-incremented value. Add parm to limit search to the named
2331 (merge_in_block): Attempt to use a mem insn that is the next
2332 use of the original regno.
2334 2018-02-27 Martin Sebor <msebor@redhat.com>
2337 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2338 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2340 2018-02-27 Martin Sebor <msebor@redhat.com>
2342 PR translation/84207
2343 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2344 n argument to unsigned HOST_WIDE_INT.
2345 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2346 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2347 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2348 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2350 2018-02-27 Richard Biener <rguenther@suse.de>
2352 PR tree-optimization/84512
2353 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2354 Do not use the estimate returned from record_stmt_cost for
2355 the scalar iteration cost but sum properly using add_stmt_cost.
2357 2018-02-27 Richard Biener <rguenther@suse.de>
2359 PR tree-optimization/84466
2360 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2361 Adjust last change to less strictly validate use operands.
2363 2018-02-27 Martin Liska <mliska@suse.cz>
2365 PR gcov-profile/84548
2366 * gcov.c (process_file): Allow partial overlap and consider it
2367 also as group functions.
2368 (output_lines): Properly calculate range of lines for a group.
2370 2018-02-27 Martin Liska <mliska@suse.cz>
2372 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2373 'ggc' suffixes. Change first column width.
2374 (timer::print): Fix formatting of the column.
2376 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2378 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2379 preserve inline entry blocks for the sake of debug inline
2380 entry point markers alone.
2381 (remove_unused_locals): Suggest in comments a better place to
2382 force the preservation of inline entry blocks that are
2383 otherwise unused, but do not preserve them.
2385 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2387 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2389 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2392 * config/i386/constraints.md (Bs): Replace
2393 ix86_indirect_branch_register with
2394 TARGET_INDIRECT_BRANCH_REGISTER.
2396 * config/i386/i386.md (indirect_jump): Likewise.
2397 (tablejump): Likewise.
2398 (*sibcall_memory): Likewise.
2399 (*sibcall_value_memory): Likewise.
2400 Peepholes of indirect call and jump via memory: Likewise.
2401 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2402 (*sibcall_value_GOT_32): Likewise.
2403 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2404 (GOT_memory_operand): Likewise.
2405 (call_insn_operand): Likewise.
2406 (sibcall_insn_operand): Likewise.
2407 (GOT32_symbol_operand): Likewise.
2408 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2410 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2412 PR rtl-optimization/83496
2413 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2414 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2415 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2416 redundant insn, if any.
2417 (relax_delay_slots): Likewise.
2418 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2420 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2422 PR tree-optimization/83965
2423 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2424 that grouped statements are part of a reduction chain. Return
2425 true if the statement is not marked as a reduction itself but
2427 (vect_recog_dot_prod_pattern): Don't check whether the statement
2428 is part of a group here.
2429 (vect_recog_sad_pattern): Likewise.
2430 (vect_recog_widen_sum_pattern): Likewise.
2432 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2435 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2437 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2440 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2442 (ix86_output_indirect_function_return): New prototype.
2443 (ix86_split_simple_return_pop_internal): Likewise.
2444 * config/i386/i386.c (indirect_return_via_cx): New.
2445 (indirect_return_via_cx_bnd): Likewise.
2446 (indirect_thunk_name): Handle return va CX_REG.
2447 (output_indirect_thunk_function): Create alias for
2448 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2449 (ix86_output_indirect_jmp): Remove the bool argument.
2450 (ix86_output_indirect_function_return): New function.
2451 (ix86_split_simple_return_pop_internal): Likewise.
2452 * config/i386/i386.md (*indirect_jump): Don't pass false
2453 to ix86_output_indirect_jmp.
2454 (*tablejump_1): Likewise.
2455 (simple_return_pop_internal): Change it to define_insn_and_split.
2456 Call ix86_split_simple_return_pop_internal to split it for
2458 (simple_return_indirect_internal): Call
2459 ix86_output_indirect_function_return instead of
2460 ix86_output_indirect_jmp.
2462 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2465 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2466 memset and value initialization afterwards.
2468 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2470 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2472 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2475 * common/config/aarch64/aarch64-common.c
2476 (aarch_option_optimization_table[]): Switch
2477 off fomit-frame-pointer
2479 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2480 Chung-Ju Wu <jasonwucj@gmail.com>
2482 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2484 (store_multiple): Ditto.
2486 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2488 * config.gcc: Add --with-cpu support for nds32 target.
2489 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2490 * config/nds32/nds32.opt: Add -mcpu= option.
2492 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2494 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2495 isel=yes): Warn for these deprecated options.
2497 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2499 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2500 ISA_2_5_MASKS_EMBEDDED.
2502 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2504 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2505 p->max as pointers rather than using iterative_hash_expr.
2507 2018-02-23 Carl Love <cel@us.ibm.com>
2509 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2510 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2512 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2513 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2514 P8V_BUILTIN_VEC_VUNSIGNED2.
2516 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2519 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2520 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2522 * lra-constraints.c (curr_insn_transform): Set up
2523 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2525 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2526 * lra-eliminations.c (spill_pseudos): Ditto.
2527 (process_insn_for_elimination): Ditto.
2528 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2529 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2530 LRA_NON_CLOBBERED_ALT.
2532 2018-02-22 Martin Sebor <msebor@redhat.com>
2534 PR tree-optimization/84480
2535 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2536 to maybe_diag_stxncpy_trunc. Call it.
2537 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2538 from gimple_fold_builtin_strcpy. Print inlining stack.
2539 (handle_builtin_stxncpy): Print inlining stack.
2540 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2542 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2545 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2546 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2547 and -fcheck-pointer-bounds are used together.
2548 (indirect_thunk_prefix): New enum.
2549 (indirect_thunk_need_prefix): New function.
2550 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2551 "_nt" instead of "_bnd" for NOTRACK prefix.
2552 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2553 (output_indirect_thunk_function): Likewise.
2555 (ix86_code_end): Update output_indirect_thunk_function calls.
2556 (ix86_output_indirect_branch_via_reg): Replace
2557 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2558 (ix86_output_indirect_branch_via_push): Likewise.
2559 (ix86_output_function_return): Likewise.
2560 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2561 incompatible with -fcf-protection=branch and
2562 -fcheck-pointer-bounds.
2564 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2567 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2568 Change gcc_assert call to output_operand_lossage.
2570 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2572 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2574 2018-02-22 DJ Delorie <dj@redhat.com>
2575 Sebastian Perta <sebastian.perta@renesas.com>
2576 Oleg Endo <olegendo@gcc.gnu.org>
2578 * config/rx/rx.c (rx_rtx_costs): New function.
2579 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2581 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2583 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2585 2018-02-22 Martin Liska <mliska@suse.cz>
2588 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2589 Add "native" as a possible value.
2591 2018-02-22 Martin Liska <mliska@suse.cz>
2594 * config/i386/i386.c (ix86_option_override_internal):
2595 Add "native" as a possible value for -march and -mtune.
2597 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2600 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2601 to all type variants.
2603 PR tree-optimization/84503
2604 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2605 width as info->bitpos + info->bitsize - start.
2606 (merged_store_group::merge_overlapping): Simplify width computation.
2607 (check_no_overlap): New function.
2608 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2609 start + width and last_order of the group, fail if check_no_overlap
2611 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2612 to group if check_no_overlap fails.
2614 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2616 * config/rs6000/altivec.md: Delete contraint arguments to
2617 define_expand, define_split, and define_peephole2, and in
2618 define_insn_and_split if always unused.
2619 * config/rs6000/darwin.md: Ditto.
2620 * config/rs6000/dfp.md: Ditto.
2621 * config/rs6000/rs6000.md: Ditto.
2622 * config/rs6000/sync.md: Ditto.
2623 * config/rs6000/vector.md: Ditto.
2624 * config/rs6000/vsx.md: Ditto.
2626 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2628 * config/rs6000/altivec.md: Write output control strings as braced
2629 blocks instead of double-quoted strings.
2630 * config/rs6000/darwin.md: Ditto.
2631 * config/rs6000/rs6000.md: Ditto.
2632 * config/rs6000/vector.md: Ditto.
2633 * config/rs6000/vsx.md: Ditto.
2635 2018-02-21 Jason Merrill <jason@redhat.com>
2637 PR c++/84314 - ICE with templates and fastcall attribute.
2638 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2640 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2642 * ipa-cp.c (determine_versionability): Fix comment typos.
2644 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2647 * ipa-cp.c (determine_versionability): Do not version functions caling
2650 2018-02-21 Martin Liska <mliska@suse.cz>
2653 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2654 Add "native" as a possible value.
2655 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2656 the macro when native cpu detection is available.
2658 2018-02-21 Martin Liska <mliska@suse.cz>
2661 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2662 Add "native" as a possible value.
2663 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2664 when native cpu detection is available.
2666 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2667 Martin Sebor <msebor@redhat.com>
2669 PR tree-optimization/84478
2670 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2672 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2673 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2674 support which is conservatively correct, for 2 only stay conservative
2675 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2676 argument to the 2 argument get_range_strlen, adjust 6 arg
2677 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2679 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2680 (gimple_fold_builtin_strlen): Pass true as last argument to
2683 2018-02-20 Martin Sebor <msebor@redhat.com>
2686 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2687 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2688 (builtin_memref::builtin_memref): Factor out parts into
2689 set_base_and_offset and call it.
2691 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2694 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2695 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2696 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2697 search at the associated MODE_INT.
2699 2018-02-20 Jeff Law <law@redhat.com>
2702 PR tree-optimization/81592
2704 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2705 for range data rather than using global data.
2706 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2707 range data rather than using global data.
2708 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2709 pass it to children as needed.
2710 (struct directive::fmtresult): Similarly.
2711 (struct directive::set_width): Similarly.
2712 (struct directive::set_precision): Similarly.
2713 (format_integer, format_directive, parse_directive): Similarly.
2714 (format_none): Accept unnamed vr_values parameter.
2715 (format_percent, format_floating, format_character): Similarly.
2716 (format_string, format_plain): Similarly.
2717 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2718 the EVRP range analyzer for range data rather than using global data.
2719 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2720 gimple-ssa-evrp-analyze.h
2721 (class sprintf_dom_walker): Add after_dom_children member function.
2722 Add evrp_range_analyzer member.
2723 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2724 range analyzer as needed.
2725 (sprintf_dom_walker::after_dom_children): New member function.
2726 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2728 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2729 (evrp_range_analyzer::pop_to_marker): Likewise.
2731 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2733 PR tree-optimization/84419
2734 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2735 with the required type if its current type is compatible but
2738 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2741 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2742 after vectorization.
2744 2018-02-20 Martin Liska <mliska@suse.cz>
2747 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2748 possible values if we don't have a hint.
2750 2018-02-20 Martin Liska <mliska@suse.cz>
2754 * final.c (shorten_branches): Build align_tab array with one
2756 * opts.c (finish_options): Add alignment option limit check.
2757 (MAX_CODE_ALIGN): Likewise.
2758 (MAX_CODE_ALIGN_VALUE): Likewise.
2759 * doc/invoke.texi: Document maximum allowed option value for
2760 all -falign-* options.
2762 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2765 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2766 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2767 * var-tracking.c (emit_note_insn_var_location): Remove all references
2768 to NOTE_INSN_CALL_ARG_LOCATION.
2769 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2770 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2771 Use copy_rtx_if_shared.
2772 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2773 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2774 (dwarf2out_var_location): Remove handling of
2775 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2777 * final.c (final_scan_insn): Remove all references to
2778 NOTE_INSN_CALL_ARG_LOCATION.
2779 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2780 before dumping final insns.
2781 * except.c (emit_note_eh_region_end): Remove all references to
2782 NOTE_INSN_CALL_ARG_LOCATION.
2783 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2784 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2785 * config/arc/arc.c (hwloop_optimize): Likewise.
2786 * config/arm/arm.c (create_fix_barrier): Likewise.
2787 * config/s390/s390.c (s390_chunkify_start): Likewise.
2788 * config/sh/sh.c (find_barrier): Likewise.
2789 * config/i386/i386.c (rest_of_insert_endbranch,
2790 ix86_seh_fixup_eh_fallthru): Likewise.
2791 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2792 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2793 * config/frv/frv.c (frv_function_prologue): Likewise.
2794 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2796 (note_outside_basic_block_p): Remove all references to
2797 NOTE_INSN_CALL_ARG_LOCATION.
2798 * gengtype.c (adjust_field_rtx_def): Likewise.
2799 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2801 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2802 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2805 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2808 PR tree-optimization/84452
2809 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2810 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2813 2018-02-19 Martin Liska <mliska@suse.cz>
2816 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2818 2018-02-19 Martin Liska <mliska@suse.cz>
2819 Richard Sandiford <richard.sandiford@linaro.org>
2821 PR tree-optimization/82491
2822 * gimple-fold.c (get_base_constructor): Make earlier bail out
2825 2018-02-19 Carl Love <cel@us.ibm.com>
2827 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2828 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2830 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2831 P8V_BUILTIN_VEC_NEG.
2833 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2835 * config/rl78/rl78.md (movdf): New define expand.
2837 2018-02-19 Martin Liska <mliska@suse.cz>
2840 * doc/invoke.texi: Fix typo.
2841 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2843 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2845 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2846 handle rs6000_single_float and rs6000_double_float specially for
2849 2018-02-16 Jeff Law <law@redhat.com>
2851 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2854 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2857 * ipa-inline.c (inline_small_functions): Fix a typo.
2859 2018-02-16 Nathan Sidwell <nathan@acm.org>
2861 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2863 2018-02-16 Carl Love <cel@us.ibm.com>
2865 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2866 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2867 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2868 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2869 expansion to P8V_BUILTIN_VEC_FLOAT2.
2871 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2873 PR rtl-optimization/70023
2874 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2875 src_regno into account.
2877 2018-02-16 Carl Love <cel@us.ibm.com>
2879 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2880 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2881 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2882 * config/rs6000/rs6000.c: Remove case statements for
2883 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2884 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2885 and P9V_BUILTIN_VEC_VINSERT4B.
2886 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2887 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2888 * config/rs6000/vsx.md:
2889 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2892 2018-02-16 Carl Love <cel@us.ibm.com>
2894 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2896 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2898 * config/rs6000/rs6000-c.c: Add the definitions for
2899 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2900 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2901 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2902 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2903 definition for insert4b and define insn *insert3b_internal.
2904 * doc/extend.texi: Add documentation for vec_extract4b.
2906 2018-02-16 Nathan Sidwell <nathan@acm.org>
2908 * doc/extend.texi (Backwards Compatibility): Mention friend
2909 injection. Note for-scope is deprecated.
2910 * doc/invoke.texi (-ffriend-injection): Deprecate.
2912 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2914 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2915 that moved to I2, also allow destinations that are a paradoxical
2916 subreg (instead of a normal reg).
2918 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2921 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2924 2018-02-16 Richard Biener <rguenther@suse.de>
2926 PR tree-optimization/84037
2927 PR tree-optimization/84016
2929 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2930 Adjust vec_construct for the fact we need additional higher latency
2931 128bit inserts for AVX256 and AVX512 vector builds.
2932 (ix86_add_stmt_cost): Scale vector construction cost for
2935 2018-02-16 Richard Biener <rguenther@suse.de>
2937 PR tree-optimization/84417
2938 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2939 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2940 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2942 2018-02-16 Martin Liska <mliska@suse.cz>
2945 * internal-fn.def (ASAN_CHECK): Set proper flags.
2946 (ASAN_MARK): Likewise.
2948 2018-02-16 Julia Koval <julia.koval@intel.com>
2950 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2951 from PTA_CANNONLAKE.
2953 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2956 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2957 Use ++iter rather than iter++ for std::list iterators.
2958 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2959 defer deleting them until all nodes in the forest are processed. Do
2960 free even leaf nodes. Change to_process into auto_vec.
2963 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2964 * vec.h (vec_default_construct): Use memset instead of placement new
2965 if BROKEN_VALUE_INITIALIZATION is defined.
2966 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2967 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2970 PR rtl-optimization/83723
2971 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2972 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2973 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2975 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2977 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2979 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2981 PR rtl-optimization/81443
2982 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2983 from inner REGs to paradoxical SUBREGs.
2985 2018-02-16 Richard Biener <rguenther@suse.de>
2987 PR tree-optimization/84399
2988 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2989 For operands we can analyze at their definition make sure we can
2990 analyze them at each use as well.
2992 2018-02-16 Richard Biener <rguenther@suse.de>
2994 PR tree-optimization/84190
2995 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2996 volatile accesses if the decl isn't volatile.
2998 2018-02-15 Jason Merrill <jason@redhat.com>
3000 PR c++/84314 - ICE with templates and fastcall attribute.
3001 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3002 TYPE_CANONICAL on an existing type.
3004 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3006 PR tree-optimization/84383
3007 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3008 dstoff nor call operand_equal_p if dstbase is NULL.
3010 PR tree-optimization/84334
3011 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3012 also a CONSTANT_CLASS_P, punt.
3014 2018-02-14 Jim Wilson <jimw@sifive.com>
3016 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3017 first SMALL_OPERAND check. New local min_second_step. Move assert
3018 to where locals are set. Add TARGET_RVC support.
3019 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3021 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3023 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3025 2018-02-14 Martin Sebor <msebor@redhat.com>
3027 PR tree-optimization/83698
3028 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3029 arrays constrain the offset range to their bounds.
3030 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3031 (builtin_access::overlap): Avoid setting the size of overlap if it's
3033 (maybe_diag_overlap): Also consider arrays when deciding what values
3034 of offsets to include in diagnostics.
3036 2018-02-14 Martin Sebor <msebor@redhat.com>
3039 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3040 that correspond to the kind of a declaration.
3042 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3045 * config/pa/pa.md: Load address of PIC label using the linkage table
3046 if the label is nonlocal.
3048 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3050 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3051 warning message if user requests -maltivec=be.
3052 * doc/invoke.texi: Document deprecation of -maltivec=be.
3054 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3057 * config/rs6000/rs6000-c.c: Update definitions for
3058 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3059 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3061 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3064 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3065 add _get_ssp intrinsics. Remove argument from
3066 __builtin_ia32_rdssp[d|q].
3067 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3068 * config/i386/i386-builtin.def: Remove argument from
3069 __builtin_ia32_rdssp[d|q].
3070 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3071 ix86_expand_special_args_builtin for _rdssp[d|q].
3072 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3073 Clear register before usage.
3074 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3075 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3077 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3079 PR tree-optimization/84357
3080 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3081 operand 1 of an ARRAY_REF too.
3083 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3086 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3087 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3089 (set_of_reg): New struct.
3090 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3091 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3092 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3094 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3095 Split into bitclr, bitset, bitinvert patterns if appropriate.
3096 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3097 use rx_fuse_in_memory_bitop.
3098 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3099 to named insn, correct maximum insn length.
3101 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3104 * machmode.def: Define a complex mode for PARTIAL_INT.
3105 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3107 * doc/rtl.texi: Document CSPImode.
3108 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3110 (msp430_hard_regno_nregs_with_padding): Likewise.
3112 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3115 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3117 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3119 PR rtl-optimization/84169
3120 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3121 we generated a parallel as new i3 and we split that to new i2 and i3
3122 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3123 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3124 those to i2, not i1. Partially rewrite this scan code.
3126 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3129 * stor-layout.c (place_field): For variable length fields, adjust
3130 offset_align afterwards not just based on the field's alignment,
3131 but also on the size.
3134 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3135 of exps and logs in the use_exp2 case.
3137 2018-02-13 Jeff Law <law@redhat.com>
3139 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3142 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3145 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3149 * common.opt (gas-loc-support, gas-locview-support): New.
3150 (ginline-points, ginternal-reset-location-views): New.
3151 * doc/invoke.texi: Document them. Use @itemx where intended.
3152 (gvariable-location-views): Adjust.
3153 * target.def (reset_location_view): New.
3154 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3155 (TARGET_RESET_LOCATION_VIEW): New.
3156 * doc/tm.texi: Rebuilt.
3157 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3158 (dwarf2out_default_as_locview_support): New.
3159 (output_asm_line_debug_info): Use option variables.
3160 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3161 (output_loc_list): Likewise.
3162 (add_high_low_attributes): Check option variables.
3163 Don't output entry view attribute in strict mode.
3164 (gen_inlined_subroutine_die): Check option variables.
3165 (dwarf2out_inline_entry): Likewise.
3166 (init_sections_and_labels): Likewise.
3167 (dwarf2out_early_finish): Likewise.
3168 (maybe_reset_location_view): New, from...
3169 (dwarf2out_var_location): ... here. Call it.
3170 * debug.h (dwarf2out_default_as_loc_support): Declare.
3171 (dwarf2out_default_as_locview_support): Declare.
3172 * hooks.c (hook_int_rtx_insn_0): New.
3173 * hooks.h (hook_int_rtx_insn_0): Declare.
3174 * toplev.c (process_options): Take -gas-loc-support and
3175 -gas-locview-support from dwarf2out. Enable
3176 -gvariable-location-views by default only with locview
3177 assembler support. Enable -ginternal-reset-location-views by
3178 default only if the target defines the corresponding hook.
3179 Enable -ginline-points by default if location views are
3180 enabled; force it disabled if statement frontiers are
3182 * tree-inline.c (expand_call_inline): Check option variables.
3183 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3185 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3187 PR tree-optimization/84321
3188 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3189 handling. Also check whether the anti-range contains any values
3190 that satisfy the mask; switch to a VR_RANGE if not.
3192 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3195 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3197 2018-02-13 Martin Jambor <mjambor@suse.cz>
3200 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3201 of call statements, also set location of a load to a temporary.
3203 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3205 * config/rl78/rl78.c (add_vector_labels): New function.
3206 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3207 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3208 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3209 which checks that no arguments are passed.
3210 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3211 * doc/extend.texi: Documentation for the new attribute.
3213 2018-02-13 Andreas Schwab <schwab@suse.de>
3215 * config/riscv/linux.h (CPP_SPEC): Define.
3217 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3220 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3221 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3222 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3223 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3224 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3225 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3226 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3227 temporarily for AES and PCLMUL builtins.
3229 PR tree-optimization/84339
3230 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3231 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3235 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3236 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3237 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3239 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3241 * omp-simd-clone.h: New file.
3242 * omp-simd-clone.c: Include omp-simd-clone.h.
3243 (expand_simd_clones): No longer static.
3244 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3245 cgraph.h and omp-simd-clone.h.
3246 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3247 (vect_recog_widen_shift_pattern): Formatting fix.
3248 (vect_pattern_recog_1): Don't check optab for calls.
3251 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3252 operands[2] into a REG before using gen_lowpart on it.
3254 2018-02-12 Jeff Law <law@redhat.com>
3257 * config/sh/sh.c (find_barrier): Consider a sibling call
3260 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3261 successfully back substituting a reg.
3263 2018-02-12 Richard Biener <rguenther@suse.de>
3265 PR tree-optimization/84037
3266 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3267 parameter, move visited init to caller.
3268 (vect_slp_analyze_operations): Separate cost from validity
3269 check, initialize visited once for all instances.
3270 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3272 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3274 (vect_model_promotion_demotion_cost): Likewise.
3275 (vectorizable_bswap): Guard cost modeling with !slp_node
3276 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3278 (vectorizable_call): Likewise.
3279 (vectorizable_conversion): Likewise.
3280 (vectorizable_assignment): Likewise.
3281 (vectorizable_shift): Likewise.
3282 (vectorizable_operation): Likewise.
3283 (vectorizable_store): Likewise.
3284 (vectorizable_load): Likewise.
3285 (vectorizable_condition): Likewise.
3286 (vectorizable_comparison): Likewise.
3288 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3291 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3292 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3293 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3295 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3298 * params.def (inline-min-speedup): Increase from 8 to 15.
3299 (max-inline-insns-auto): Decrease from 40 to 30.
3300 * ipa-split.c (consider_split): Add some buffer for function to
3301 be considered inlining candidate.
3302 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3305 2018-02-12 Richard Biener <rguenther@suse.de>
3307 PR tree-optimization/84037
3308 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3309 matched stmts if we cannot swap the non-matched ones.
3311 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3313 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3314 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3315 _mm_maskz_scalef_round_ss): New intrinsics.
3316 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3317 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3318 __builtin_ia32_scalefss_round): Remove.
3319 (__builtin_ia32_scalefsd_mask_round,
3320 __builtin_ia32_scalefss_mask_round): New intrinsics.
3321 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3322 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3323 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3324 "<round_constraint>")): Changed to ...
3325 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3326 "<round_scalar_constraint>")): ... this.
3327 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3328 %0, %1, %2<round_op3>}"): Changed to ...
3329 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3330 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3331 %2<round_scalar_mask_op3>}"): ... this.
3332 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3334 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3336 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3337 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3338 (_mm_maskz_sqrt_round_ss): New intrinsics.
3339 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3340 (__builtin_ia32_sqrtsd_mask_round)
3341 (__builtin_ia32_sqrtss_mask_round): New builtins.
3342 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3343 (__builtin_ia32_sqrtss_round): Remove.
3344 (__builtin_ia32_sqrtsd_mask_round)
3345 (__builtin_ia32_sqrtss_mask_round): New builtins.
3346 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3347 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3348 ((match_operand:VF_128 1 "vector_operand"
3349 "xBm,<round_constraint>")): Changed to ...
3350 ((match_operand:VF_128 1 "vector_operand"
3351 "xBm,<round_scalar_constraint>")): ... this.
3352 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3353 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3354 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3355 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3356 %<iptr>1<round_scalar_mask_op3>}): ... this.
3357 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3358 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3360 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3363 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3364 Cast vec_cmpeq result to correct type.
3365 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3366 Cast vec_cmpgt result to correct type.
3368 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3370 * final.c (final_scan_insn_1): Renamed from...
3371 (final_scan_insn): ... this. New wrapper, to recover
3372 seen from the outermost call in recursive ones.
3373 * config/sparc/sparc.c (output_return): Drop seen from call.
3374 (output_sibcall): Likewise.
3375 * config/visium/visium.c (output_branch): Likewise.
3377 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3379 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3382 2018-02-10 Alan Modra <amodra@gmail.com>
3385 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3386 Specify LR as an input.
3388 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3391 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3392 remove_member_access_dummy_vars): New functions.
3393 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3394 lower_omp_1, execute_lower_omp): Use them.
3396 PR rtl-optimization/84308
3397 * shrink-wrap.c (spread_components): Release todo vector.
3399 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3401 PR rtl-optimization/57193
3402 * ira-color.c (struct allocno_color_data): Add member
3403 conflict_allocno_hard_prefs.
3404 (update_conflict_allocno_hard_prefs): New.
3405 (bucket_allocno_compare_func): Add a preference based on
3406 conflict_allocno_hard_prefs.
3407 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3408 (color_allocnos): Remove a dead code. Initiate
3409 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3411 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3414 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3415 constraint from =wa to wa. Avoid a subreg on the output operand,
3416 instead use a pseudo and subreg it in a move.
3417 (p9_xxbrd_<mode>): Changed to ...
3418 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3419 (p9_xxbrd_v2df): New expander.
3420 (p9_xxbrw_<mode>): Changed to ...
3421 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3422 (p9_xxbrw_v4sf): New expander.
3424 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3426 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3428 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3431 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3432 multiply in 32-bit mode.
3433 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3434 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3437 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3439 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3440 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3441 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3442 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3444 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3447 * dwarf2out.c (is_trivial_indirect_ref): New function.
3448 (dwarf2out_late_global_decl): Do not generate a location
3449 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3450 and that are not defined in the current unit.
3452 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3454 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3455 instead of a libcall for UNORDERED.
3457 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3460 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3461 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3463 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3466 * config/s390/s390.c (s390_set_current_function): Invoke
3467 s390_indirect_branch_settings also if fndecl didn't change.
3469 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3471 * config/rs6000/rs6000.md (blockage): Set length to zero.
3473 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3475 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3477 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3480 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3481 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3485 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3486 PARALLEL incoming that failed vt_get_decl_and_offset check.
3489 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3490 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3492 (get_variable_section): For decls in named .bss* sections pass true as
3493 second argument to bss_initializer_p.
3495 2018-02-09 Marek Polacek <polacek@redhat.com>
3496 Jakub Jelinek <jakub@redhat.com>
3499 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3500 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3501 Sync some changes from cxx_fold_indirect_ref.
3503 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3505 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3507 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3508 (BLOCK_INLINE_ENTRY_LABEL): New.
3509 (dwarf2out_var_location): Disregard inline entry markers.
3510 (inline_entry_data): New struct.
3511 (inline_entry_data_hasher): New hashtable type.
3512 (inline_entry_data_hasher::hash): New.
3513 (inline_entry_data_hasher::equal): New.
3514 (inline_entry_data_table): New variable.
3515 (add_high_low_attributes): Add DW_AT_entry_pc and
3516 DW_AT_GNU_entry_view attributes if a pending entry is found
3517 in inline_entry_data_table. Add old entry_pc attribute only
3518 if debug nonbinding markers are disabled.
3519 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3520 markers are enabled.
3521 (block_within_block_p, dwarf2out_inline_entry): New.
3522 (dwarf2out_finish): Check that no entries remained in
3523 inline_entry_data_table.
3524 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3525 (final_scan_insn, notice_source_line): Likewise.
3526 (rest_of_clean_state): Skip inline entry markers.
3527 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3529 * gimple.c (gimple_build_debug_inline_entry): New.
3530 * gimple.h (enum gimple_debug_subcode): Add
3531 GIMPLE_DEBUG_INLINE_ENTRY.
3532 (gimple_build_debug_inline_entry): Declare.
3533 (gimple_debug_inline_entry_p): New.
3534 (gimple_debug_nonbind_marker_p): Adjust.
3535 * insn-notes.def (INLINE_ENTRY): New.
3536 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3537 inline entry marker notes.
3538 (print_insn): Likewise.
3539 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3540 (INSN_DEBUG_MARKER_KIND): Likewise.
3541 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3542 * tree-inline.c (expand_call_inline): Build and insert
3543 debug_inline_entry stmt.
3544 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3545 inline entry blocks early, if nonbind markers are enabled.
3546 (dump_scope_block): Dump fragment info.
3547 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3548 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3549 (gimple_build_debug_inline_entry): New.
3550 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3551 Enable/disable inline entry points too.
3552 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3553 (DEBUG_INSN): Describe inline entry markers.
3555 * common.opt (gvariable-location-views): New.
3556 (gvariable-location-views=incompat5): New.
3557 * config.in: Rebuilt.
3558 * configure: Rebuilt.
3559 * configure.ac: Test assembler for view support.
3560 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3561 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3562 * dwarf2out.c (var_loc_view): New typedef.
3563 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3564 (dwarf2out_locviews_in_attribute): New.
3565 (dwarf2out_locviews_in_loclist): New.
3566 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3567 (enum dw_line_info_opcode): Add LI_adv_address.
3568 (struct dw_line_info_table): Add view.
3569 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3570 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3571 (zero_view_p): New variable.
3572 (ZERO_VIEW_P): New macro.
3573 (output_asm_line_debug_info): New.
3574 (struct var_loc_node): Add view.
3575 (add_AT_view_list, AT_loc_list): New.
3576 (add_var_loc_to_decl): Add view param. Test it against last.
3577 (new_loc_list): Add view params. Record them.
3578 (AT_loc_list_ptr): Handle loc and view lists.
3579 (view_list_to_loc_list_val_node): New.
3580 (print_dw_val): Handle dw_val_class_view_list.
3581 (size_of_die): Likewise.
3582 (value_format): Likewise.
3583 (loc_list_has_views): New.
3584 (gen_llsym): Set vl_symbol too.
3585 (maybe_gen_llsym, skip_loc_list_entry): New.
3586 (dwarf2out_maybe_output_loclist_view_pair): New.
3587 (output_loc_list): Output view list or entries too.
3588 (output_view_list_offset): New.
3589 (output_die): Handle dw_val_class_view_list.
3590 (output_dwarf_version): New.
3591 (output_compilation_unit_header): Use it.
3592 (output_skeleton_debug_sections): Likewise.
3593 (output_rnglists, output_line_info): Likewise.
3594 (output_pubnames, output_aranges): Update version comments.
3595 (output_one_line_info_table): Output view numbers in asm comments.
3596 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3597 Call maybe_gen_llsym.
3598 (loc_list_from_tree_1): Adjust.
3599 (add_AT_location_description): Create view list attribute if
3600 needed, check it's absent otherwise.
3601 (convert_cfa_to_fb_loc_list): Adjust.
3602 (maybe_emit_file): Call output_asm_line_debug_info for test.
3603 (dwarf2out_var_location): Reset views as needed. Precompute
3604 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3605 attribute. Set view.
3606 (new_line_info_table): Reset next view.
3607 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3608 (dwarf2out_source_line): Likewise. Output view resets and labels to
3609 the assembler, or select appropriate line info opcodes.
3610 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3611 (optimize_string_length): Catch it. Adjust.
3612 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3613 dw_val_class_view_list, and remove it if no longer needed.
3614 (hash_loc_list): Hash view numbers.
3615 (loc_list_hasher::equal): Compare them.
3616 (optimize_location_lists): Check whether a view list symbol is
3617 needed, and whether the locview attribute is present, and
3618 whether they match. Remove the locview attribute if no longer
3620 (index_location_lists): Call skip_loc_list_entry for test.
3621 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3622 Use output_dwarf_version.
3623 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3624 (struct dw_val_node): Add val_view_list.
3625 * final.c (SEEN_NEXT_VIEW): New.
3626 (set_next_view_needed): New.
3627 (clear_next_view_needed): New.
3628 (maybe_output_next_view): New.
3629 (final_start_function): Rename to...
3630 (final_start_function_1): ... this. Take pointer to FIRST,
3631 add SEEN parameter. Emit param bindings in the initial view.
3632 (final_start_function): Reintroduce SEEN-less interface.
3633 (final): Rename to...
3634 (final_1): ... this. Take SEEN parameter. Output final pending
3635 next view at the end.
3636 (final): Reintroduce seen-less interface.
3637 (final_scan_insn): Output pending next view before switching
3638 sections or ending a block. Mark the next view as needed when
3639 outputting variable locations. Notify debug backend of section
3640 changes, and of location view changes.
3641 (rest_of_handle_final): Adjust.
3642 * toplev.c (process_options): Autodetect value for debug variable
3643 location views option. Warn on incompat5 without -gdwarf-5.
3644 * doc/invoke.texi (gvariable-location-views): New.
3645 (gvariable-location-views=incompat5): New.
3646 (gno-variable-location-views): New.
3648 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3650 PR tree-optimization/84136
3651 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3652 that the result of find_edge is non-NULL.
3654 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3657 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3658 storing integer register in SImode. Fix cost of 256 and 512
3659 byte aligned SSE register store.
3661 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3663 * config/i386/i386.c (ix86_multiplication_cost): Fix
3664 multiplication cost for TARGET_AVX512DQ.
3666 2018-02-08 Marek Polacek <polacek@redhat.com>
3668 PR tree-optimization/84238
3669 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3672 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3674 PR tree-optimization/84265
3675 * tree-vect-stmts.c (vectorizable_store): Don't treat
3676 VMAT_CONTIGUOUS accesses as grouped.
3677 (vectorizable_load): Likewise.
3679 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3681 PR tree-optimization/81635
3682 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3683 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3684 (test_round_for_mask): New functions.
3685 (wide_int_cc_tests): Call test_round_for_mask.
3686 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3687 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3688 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3689 range returned by get_range_info.
3691 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3694 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3695 * symtab.c: Include builtins.h
3696 (symtab_node::output_to_lto_symbol_table_p): Move here
3697 from lto-streamer-out.c:output_symbol_p.
3698 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3699 (output_symbol_p): Move all logic to symtab.c
3700 (produce_symtab): Update.
3702 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3704 * config/s390/s390-opts.h (enum indirect_branch): Define.
3705 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3706 (s390_indirect_branch_via_thunk)
3707 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3708 (enum s390_indirect_branch_type): Define.
3709 * config/s390/s390.c (struct s390_frame_layout, struct
3710 machine_function): Remove.
3711 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3712 (indirect_branch_table_label_no, indirect_branch_table_name):
3714 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3715 (enum s390_indirect_branch_option): Define.
3716 (s390_return_addr_from_memory): New function.
3717 (s390_handle_string_attribute): New function.
3718 (s390_attribute_table): Add new attribute handler.
3719 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3720 (s390_indirect_branch_via_thunk): New function.
3721 (s390_indirect_branch_via_inline_thunk): New function.
3722 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3723 sibling call optimization for non z10 compiles.
3724 (s390_emit_call): Force indirect branch target to be a single
3725 register. Add r1 clobber for non-z10 compiles.
3726 (s390_emit_epilogue): Emit return jump via return_use expander.
3727 (s390_reorg): Handle JUMP_INSNs as execute targets.
3728 (s390_option_override_internal): Perform validity checks for the
3729 new command line options.
3730 (s390_indirect_branch_attrvalue): New function.
3731 (s390_indirect_branch_settings): New function.
3732 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3733 (s390_output_indirect_thunk_function): New function.
3734 (s390_code_end): Implement target hook.
3735 (s390_case_values_threshold): Implement target hook.
3736 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3738 * config/s390/s390.h (struct s390_frame_layout)
3739 (struct machine_function): Move here from s390.c.
3740 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3741 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3742 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3743 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3744 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3745 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3746 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3747 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3748 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3749 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3750 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3751 (mnemonic attribute): Add values which aren't recognized
3753 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3754 pattern for branch conversion. Fix mnemonic attribute.
3755 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3756 indirect branch via thunk if requested.
3757 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3758 ("*indirect_jump"): Disable for branch conversion using out of
3760 ("indirect_jump_via_thunk<mode>_z10")
3761 ("indirect_jump_via_thunk<mode>")
3762 ("indirect_jump_via_inlinethunk<mode>_z10")
3763 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3764 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3765 ("casesi_jump_via_inlinethunk<mode>_z10")
3766 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3767 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3768 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3769 ("*indirect2_jump"): Disable for branch conversion.
3770 ("casesi_jump"): Turn into expander and expand patterns for branch
3772 ("return_use"): New expander.
3773 ("*return"): Emit return via thunk and rename it to ...
3774 ("*return<mode>"): ... this one.
3775 * config/s390/s390.opt: Add new options and and enum for the
3778 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3780 * lra-constraints.c (match_reload): Unconditionally use
3781 gen_lowpart_SUBREG, rather than selecting between that
3782 and equivalent gen_rtx_SUBREG code.
3784 2018-02-08 Richard Biener <rguenther@suse.de>
3786 PR tree-optimization/84233
3787 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3788 changed flag instead of boguously re-using phi_inserted.
3790 2018-02-08 Martin Jambor <mjambor@suse.cz>
3792 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3793 static local variables.
3795 2018-02-08 Richard Biener <rguenther@suse.de>
3797 PR tree-optimization/84278
3798 * tree-vect-stmts.c (vectorizable_store): When looking for
3799 smaller vector types to perform grouped strided loads/stores
3800 make sure the mode is supported by the target.
3801 (vectorizable_load): Likewise.
3803 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3805 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3806 Increase LDP/STP opportunities by adding adjacent callee-saves.
3808 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3810 PR rtl-optimization/84068
3811 PR rtl-optimization/83459
3812 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3814 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3816 PR tree-optimization/84224
3817 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3818 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3821 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3824 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3825 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3826 count, remove one USE.
3828 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3830 * doc/install.texi (Configuration): Document the
3831 --with-long-double-format={ibm,ieee} PowerPC configuration
3835 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3836 Convert from define_expand to be define_insn_and_split. Rework
3837 float/double/_Float128 conversions to QI/HI/SImode to work with
3838 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3839 conversions to QI/HImode types did a store and then a load to
3840 truncate the value. For conversions to VSX registers, don't split
3841 the insn, instead emit the code directly. Use the code iterator
3842 any_fix to combine signed and unsigned conversions.
3843 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3844 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3845 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3846 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3847 (fix_<mode>di2_hw): Likewise.
3848 (fixuns_<mode>di2_hw): Likewise.
3849 (fix_<mode>si2_hw): Likewise.
3850 (fixuns_<mode>si2_hw): Likewise.
3851 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3852 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3853 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3854 fix<uns>_trunc<SFDF:mode>si2_p8.
3855 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3857 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3858 (fix<uns>_<mode>_mem): Likewise.
3859 (fctiw<u>z_<mode>_mem): Likewise.
3860 (fix<uns>_<mode>_mem): Likewise.
3861 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3862 the register allocator from doing a direct move to the GPRs to do
3863 a store, and instead use the ISA 3.0 store byte/half-word from
3864 vector register instruction. For IEEE 128-bit floating point,
3865 also optimize stores of 32-bit ints.
3866 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3868 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3870 * genextract.c (push_pathstr_operand): New function to support
3872 (walk_rtx): Call push_pathstr_operand.
3873 (print_path): Support [a-zA-Z].
3875 2018-02-07 Richard Biener <rguenther@suse.de>
3877 PR tree-optimization/84037
3878 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3879 (cse_and_gimplify_to_preheader): Declare.
3880 (vect_get_place_in_interleaving_chain): Likewise.
3881 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3883 (_loop_vec_info::~_loop_vec_info): Delete it.
3884 (cse_and_gimplify_to_preheader): New function.
3885 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3886 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3887 (vectorizable_load): Likewise. For grouped stores always base
3888 the IV on the first element.
3889 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3890 condition before gimplifying.
3892 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3894 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3895 *DIV_EXPR and *MOD_EXPR.
3897 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3900 * config/i386/i386.c (ix86_option_override_internal): Mask out
3901 the CF_SET bit when checking -fcf-protection.
3903 2018-02-07 Tom de Vries <tom@codesourcery.com>
3906 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3909 2018-02-07 Richard Biener <rguenther@suse.de>
3911 PR tree-optimization/84204
3912 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3915 PR tree-optimization/84205
3916 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3917 special-case isl_ast_op_zdiv_r.
3919 PR tree-optimization/84223
3920 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3921 Only add conditions from within the region.
3922 (gather_bbs::after_dom_children): Adjust.
3924 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3927 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3928 * config/avr/avr.md: Only post-reload split REG-REG moves if
3929 either register is GENERAL_REG_P.
3931 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3933 PR tree-optimization/84235
3934 * tree-ssa-scopedtables.c
3935 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3936 if the subtraction is performed in floating point type where NaNs are
3937 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3938 build 1. Formatting fix.
3940 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3943 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3944 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3945 and skip it regardless of bb boundaries. Use CALL_P macro,
3946 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3947 unnecessarily, formatting fix.
3949 2018-02-06 Michael Collison <michael.collison@arm.com>
3951 * config/arm/thumb2.md:
3952 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3953 (*thumb_mov_notscc): Ditto.
3955 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3958 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3959 unsigned_fix, not "s".
3961 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3963 * configure.ac (gcc_fn_eh_frame_ro): New function.
3964 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3965 correct .eh_frame permissions.
3966 * configure: Regenerate.
3968 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3970 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3973 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3975 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3976 Display warning message for -mno-speculate-indirect-jumps.
3978 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3980 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3982 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3984 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3986 PR tree-optimization/84225
3987 * tree-eh.c (find_trapping_overflow): Only call
3988 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3990 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3993 * config/i386/i386.c: Reimplement the check of possible options
3994 -mibt/-mshstk conbination. Change error messages.
3995 * doc/invoke.texi: Fix a typo: remove extra '='.
3997 2018-02-06 Marek Polacek <polacek@redhat.com>
3999 PR tree-optimization/84228
4000 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4002 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4005 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4006 emitted arch directives.
4007 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4008 __ARM_FEATURE_COPROC before changing architectures.
4010 2018-02-06 Richard Biener <rguenther@suse.de>
4012 * config/i386/i386.c (print_reg): Fix typo.
4013 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4015 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4017 * configure: Regenerate.
4019 2018-02-05 Martin Sebor <msebor@redhat.com>
4021 PR tree-optimization/83369
4022 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4025 2018-02-05 Martin Liska <mliska@suse.cz>
4027 * doc/invoke.texi: Cherry-pick upstream r323995.
4029 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4031 * ira.c (ira_init_register_move_cost): Adjust comment.
4033 2018-02-05 Martin Liska <mliska@suse.cz>
4035 PR gcov-profile/84137
4036 * doc/gcov.texi: Fix typo in documentation.
4038 2018-02-05 Martin Liska <mliska@suse.cz>
4040 PR gcov-profile/83879
4041 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4042 using dlopen functionality.
4044 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4046 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4047 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4048 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4049 _mm_maskz_range_round_ss): New intrinsics.
4050 (__builtin_ia32_rangesd128_round)
4051 (__builtin_ia32_rangess128_round): Remove.
4052 (__builtin_ia32_rangesd128_mask_round,
4053 __builtin_ia32_rangess128_mask_round): New builtins.
4054 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4055 __builtin_ia32_rangess128_round): Remove.
4056 (__builtin_ia32_rangesd128_mask_round,
4057 __builtin_ia32_rangess128_mask_round): New builtins.
4058 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4059 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4060 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4061 "<round_saeonly_constraint>")): Changed to ...
4062 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4063 "<round_saeonly_scalar_constraint>")): ... this.
4064 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4065 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4066 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4067 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4068 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4070 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4072 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4074 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4075 Remove all values except native, 8540 and 8548.
4077 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4079 * config/i386/i386.c (ix86_output_function_return): Pass
4080 INVALID_REGNUM, instead of -1, as invalid register number to
4081 indirect_thunk_name and output_indirect_thunk.
4083 2018-02-02 Julia Koval <julia.koval@intel.com>
4085 * config.gcc: Add -march=icelake.
4086 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4087 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4088 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4089 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4090 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4091 (processor_target_table): Add icelake.
4092 (ix86_option_override_internal): Handle new PTAs.
4093 (get_builtin_code_for_version): Handle icelake.
4094 (M_INTEL_COREI7_ICELAKE): New.
4095 (fold_builtin_cpu): Handle icelake.
4096 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4097 * doc/invoke.texi: Add -march=icelake.
4099 2018-02-02 Julia Koval <julia.koval@intel.com>
4101 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4102 to wide_int_bitmask.
4103 * wide-int-bitmask.h: New.
4105 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4108 * config/i386/i386.md: Replace Pmode with word_mode in
4109 builtin_setjmp_setup and builtin_longjmp to support x32.
4111 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4115 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4117 (rs6000_supported_cpu_names): New static variable.
4118 (linux_cpu_translation_table): Likewise.
4119 (elf_platform) <cpu>: Define new static variable and use it.
4120 Translate kernel AT_PLATFORM name to canonical name if needed.
4121 Error if platform name is unknown.
4123 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4126 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4128 2018-02-01 Jeff Law <law@redhat.com>
4131 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4132 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4133 the scratch if RELEASE_VIA_POP is false.
4134 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4135 If we have to save a temporary register, decrement SIZE appropriately.
4136 Pass new arguments to release_scratch_register_on_entry.
4137 (ix86_adjust_stack_and_probe): Likewise.
4138 (ix86_emit_probe_stack_range): Pass new arguments to
4139 release_scratch_register_on_entry.
4141 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4143 PR rtl-optimization/84157
4144 * combine.c (change_zero_ext): Use REG_P predicate in
4145 front of HARD_REGISTER_P predicate.
4147 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4149 * config/avr/avr.c (avr_option_override): Move disabling of
4150 -fdelete-null-pointer-checks to...
4151 * common/config/avr/avr-common.c (avr_option_optimization_table):
4154 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4156 PR tree-optimization/81635
4157 * tree-data-ref.c (split_constant_offset_1): For types that
4158 wrap on overflow, try to use range info to prove that wrapping
4161 2018-02-01 Renlin Li <renlin.li@arm.com>
4164 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4166 (aarch64_register_move_cost): Likewise.
4167 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4169 (REG_CLASS_NAMES): Likewise.
4170 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4171 TAILCALL_ADDR_REGS. Remove IP registers.
4172 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4174 2018-02-01 Richard Biener <rguenther@suse.de>
4176 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4177 for specifying RPO order and allow NULL for that.
4178 * domwalk.c (dom_walker::dom_walker): Likewise.
4179 (dom_walker::walk): Handle NULL RPO order.
4180 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4182 (rewrite_update_dom_walker): Likewise.
4183 (mark_def_dom_walker): Likewise.
4185 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4187 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4188 (aarch64_maybe_expand_sve_subreg_move): Declare.
4189 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4190 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4192 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4193 that are semantically a reverse operation.
4194 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4195 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4196 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4198 (aarch64_can_change_mode_class): For big-endian, forbid changes
4199 between two SVE modes if they have different element sizes.
4201 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4203 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4204 the TImode handling for big-endian targets.
4206 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4208 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4209 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4211 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4212 Remove BSWAP handing for big-endian targets and use the form of
4213 LD1RQ appropariate for the mode.
4215 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4217 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4218 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4221 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4224 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4225 check for operands that need to go through aarch64_sve_reload_be.
4227 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4229 PR tree-optimization/81661
4230 PR tree-optimization/84117
4231 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4232 * tree-eh.c: Include gimplify.h.
4233 (find_trapping_overflow, replace_trapping_overflow,
4234 rewrite_to_non_trapping_overflow): New functions.
4235 * tree-vect-loop.c: Include tree-eh.h.
4236 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4237 * tree-data-ref.c: Include tree-eh.h.
4238 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4240 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4242 PR rtl-optimization/84123
4243 * combine.c (change_zero_ext): Check if hard register satisfies
4244 can_change_dest_mode before calling gen_lowpart_SUBREG.
4246 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4249 * ira.c (ira_init_register_move_cost): Remove assert.
4251 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4253 PR rtl-optimization/84071
4254 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4255 * doc/tm.texi: Regenerate.
4257 2018-01-31 Richard Biener <rguenther@suse.de>
4259 PR tree-optimization/84132
4260 * tree-data-ref.c (analyze_miv_subscript): Properly
4261 check whether evolution_function_is_affine_multivariate_p
4262 before calling gcd_of_steps_may_divide_p.
4264 2018-01-31 Julia Koval <julia.koval@intel.com>
4267 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4268 * config/i386/i386.md (rdpid_rex64) New.
4269 (rdpid): Make 32bit only.
4271 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4274 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4275 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4277 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4280 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4282 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4284 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4286 PR rtl-optimization/84071
4287 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4288 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4290 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4292 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4293 (arc_attribute_table): Add 'aux' attribute.
4294 (arc_in_small_data_p): Consider aux like variables.
4295 (arc_is_aux_reg_p): New function.
4296 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4297 (arc_get_aux_arg): New function.
4298 (prepare_move_operands): Handle aux-register access.
4299 (arc_handle_aux_attribute): New function.
4300 * doc/extend.texi (ARC Variable attributes): Add subsection.
4302 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4304 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4305 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4306 (arc_attribute_table): Add 'uncached' attribute.
4307 (arc_print_operand): Print '.di' flag for uncached memory
4309 (arc_in_small_data_p): Do not consider for small data the uncached
4311 (arc_is_uncached_mem_p): New function.
4312 * config/arc/predicates.md (compact_store_memory_operand): Check
4313 for uncached memory accesses.
4314 (nonvol_nonimm_operand): Likewise.
4315 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4317 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4320 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4321 falign-loops=): Add Optimization flag.
4323 2018-01-30 Jeff Law <law@redhat.com>
4326 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4327 INT_REGISTERS_SAVED. Check it prior to calling
4328 get_scratch_register_on_entry.
4329 (ix86_adjust_stack_and_probe): Similarly.
4330 (ix86_emit_probe_stack_range): Similarly.
4331 (ix86_expand_prologue): Corresponding changes.
4333 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4336 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4337 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4339 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4342 * lra-constraints.c (curr_insn_transform): Process AND in the
4345 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4347 PR rtl-optimization/83986
4348 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4349 dependence against last_pending_memory_flush in addition to
4352 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4354 PR tree-optimization/81611
4355 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4358 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4361 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4364 2018-01-30 Richard Biener <rguenther@suse.de>
4365 Jakub Jelinek <jakub@redhat.com>
4367 PR tree-optimization/84111
4368 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4369 inner loops added during recursion, as they don't have up-to-date
4372 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4375 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4376 (can_inline_edge_by_limits_p): ... here.
4377 (can_early_inline_edge_p, check_callers,
4378 update_caller_keys, update_callee_keys, recursive_inlining,
4379 add_new_edges_to_heap, speculation_useful_p,
4380 inline_small_functions,
4381 inline_small_functions, flatten_function,
4382 inline_to_all_callers_1): Update.
4384 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4386 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4389 2018-01-30 Richard Biener <rguenther@suse.de>
4391 PR tree-optimization/83008
4392 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4393 invariant and constant vector uses in stmts when they need
4396 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4399 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4400 * configure: Regenerate.
4402 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4404 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4406 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4407 Use gen_rtx_REG rather than gen_lowpart.
4409 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4411 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4412 rather than 0 when creating partial subregs.
4414 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4416 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4419 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4422 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4423 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4424 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4425 flags. This restores the settings used before the 2017-07-24.
4426 Turning off pre increment/decrement/modify allows IVOPTS to
4427 optimize DF/SF loops where the index is an int.
4429 2018-01-29 Richard Biener <rguenther@suse.de>
4430 Kelvin Nilsen <kelvin@gcc.gnu.org>
4433 * tree-vect-stmts.c (vectorizable_call): Don't call
4434 targetm.vectorize_builtin_md_vectorized_function if callee is
4437 2018-01-22 Carl Love <cel@us.ibm.com>
4439 * doc/extend.tex: Fix typo in second arg in
4440 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4442 2018-01-29 Richard Biener <rguenther@suse.de>
4444 PR tree-optimization/84086
4445 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4446 (flush_ssaname_freelist): When SSA names were released reset
4447 the SCEV hash table.
4449 2018-01-29 Richard Biener <rguenther@suse.de>
4451 PR tree-optimization/84057
4452 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4453 removed paths when removing edges.
4455 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4457 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4458 -mfunction-return=@var{choice}.
4460 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4463 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4465 (layout::print_source_line): Likewise.
4466 (test_get_line_width_without_trailing_whitespace): Add test cases.
4468 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4471 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4474 2018-01-26 Jim Wilson <jimw@sifive.com>
4476 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4478 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4481 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4483 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4484 and CMP + SUB-immediate -> SUBS.
4486 2018-01-26 Martin Sebor <msebor@redhat.com>
4488 PR tree-optimization/83896
4489 * tree-ssa-strlen.c (get_string_len): Rename...
4490 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4491 Avoid assuming length is constant.
4492 (handle_char_store): Use HOST_WIDE_INT for string length.
4494 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4497 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4498 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4500 2018-01-26 Richard Biener <rguenther@suse.de>
4502 PR rtl-optimization/84003
4503 * dse.c (record_store): Only record redundant stores when
4504 the earlier store aliases at least all accesses the later one does.
4506 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4508 PR rtl-optimization/83985
4509 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4510 REG_CFA_RESTORE insns.
4511 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4512 REG_CFA_RESTORE insns here.
4515 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4516 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4518 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4520 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4521 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4522 (arc_init): Likewise.
4523 (arc_override_options): Likewise.
4524 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4526 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4528 * config/arc/arc.h (TARGET_DBNZ): Define.
4529 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4530 properly set the tune attribute.
4531 (dbnz): Use TARGET_DBNZ guard.
4532 * config/arc/arc.opt (mtune): Add core3 option.
4534 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4536 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4537 recognize new pic like addresses.
4538 (arc_delegitimize_address): Clean up.
4540 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4542 * config/arc/arc-arches.def: Option mrf16 valid for all
4544 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4545 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4546 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4547 * config/arc/arc-tables.opt: Regenerate.
4548 * config/arc/arc.c (arc_conditional_register_usage): Handle
4549 reduced register file case.
4550 (arc_file_start): Set must have build attributes.
4551 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4553 * config/arc/arc.opt (mrf16): Add new option.
4554 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4555 * config/arc/genmultilib.awk: Handle new mrf16 option.
4556 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4557 * config/arc/t-multilib: Regenerate.
4558 * doc/invoke.texi (ARC Options): Document mrf16 option.
4560 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4562 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4563 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4564 (arc_attribute_table): Add 'secure_call' attribute.
4565 (arc_print_operand): Print secure call operand.
4566 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4568 (arc_is_secure_call_p): New function. * config/arc/arc.md
4569 (call_i): Add support for sjli instruction.
4570 (call_value_i): Likewise.
4571 * config/arc/constraints.md (Csc): New constraint.
4573 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4574 John Eric Martin <John.Martin@emmicro-us.com>
4576 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4577 * config/arc/arc.c (_arc_jli_section): New struct.
4578 (arc_jli_section): New type.
4579 (rc_jli_sections): New static variable.
4580 (arc_handle_jli_attribute): New function.
4581 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4582 (arc_file_end): New function.
4583 (TARGET_ASM_FILE_END): Define.
4584 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4585 (arc_add_jli_section): New function.
4586 (jli_call_scan): Likewise.
4587 (arc_reorg): Call jli_call_scan.
4588 (arc_output_addsi): Remove 'S' from printing asm operand.
4589 (arc_is_jli_call_p): New function.
4590 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4592 (movhi_insn): Likewise.
4593 (movsi_insn): Likewise.
4594 (movsi_set_cc_insn): Likewise.
4595 (loadqi_update): Likewise.
4596 (load_zeroextendqisi_update): Likewise.
4597 (load_signextendqisi_update): Likewise.
4598 (loadhi_update): Likewise.
4599 (load_zeroextendhisi_update): Likewise.
4600 (load_signextendhisi_update): Likewise.
4601 (loadsi_update): Likewise.
4602 (loadsf_update): Likewise.
4603 (movsicc_insn): Likewise.
4604 (bset_insn): Likewise.
4605 (bxor_insn): Likewise.
4606 (bclr_insn): Likewise.
4607 (bmsk_insn): Likewise.
4608 (bicsi3_insn): Likewise.
4609 (cmpsi_cc_c_insn): Likewise.
4610 (movsi_ne): Likewise.
4611 (movsi_cond_exec): Likewise.
4612 (clrsbsi2): Likewise.
4623 (call_i): Remove 'S' asm letter, add jli instruction.
4624 (call_value_i): Likewise.
4625 * config/arc/arc.op (mjli-always): New option.
4626 * config/arc/constraints.md (Cji): New constraint.
4627 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4629 (subsf3_fpx): Likewise.
4630 (mulsf3_fpx): Likewise.
4631 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4633 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4635 * doc/invoke.texi (ARC): Document mjli-always option.
4637 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4639 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4640 avoid addition with 0 and use incw and decw where possible.
4642 2018-01-26 Richard Biener <rguenther@suse.de>
4644 PR tree-optimization/81082
4645 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4646 association if it requires casting to unsigned.
4647 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4648 from fold_plusminus_mult_expr to catch important cases late when
4649 range info is available.
4651 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4653 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4654 * configure.ac (hidden_linkonce): New test.
4655 * configure: Regenerate.
4656 * config.in: Regenerate.
4658 2018-01-26 Julia Koval <julia.koval@intel.com>
4660 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4661 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4662 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4663 _mm_mask_bitshuffle_epi64_mask): Fix type.
4664 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4665 USI_FTYPE_V4DI_V4DI_USI): Remove.
4666 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4667 __builtin_ia32_vpshufbitqmb256_mask,
4668 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4669 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4670 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4672 2018-01-26 Alan Modra <amodra@gmail.com>
4675 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4676 UNSPEC_VBPERMQ. Sort other unspecs.
4678 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4680 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4682 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4685 * predict.c (drop_profile): Do not push/pop cfun; update also
4687 (handle_missing_profiles): Fix logic looking for zero profiles.
4689 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4692 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4693 on functions with #pragma omp declare simd or functions with simd
4695 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4696 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4697 Remove trailing \n from warning_at calls.
4699 2018-01-25 Tom de Vries <tom@codesourcery.com>
4702 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4703 for neutered workers.
4705 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4708 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4709 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4711 2018-01-24 Jeff Law <law@redhat.com>
4714 * i386.c (get_probe_interval): Move to earlier point.
4715 (ix86_compute_frame_layout): If -fstack-clash-protection and
4716 the frame is larger than the probe interval, then use pushes
4717 to save registers rather than reg->mem moves.
4718 (ix86_expand_prologue): Remove conditional for int_registers_saved
4721 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4724 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4725 min/max for never referenced object.
4727 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4730 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4732 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4733 attributes from DECL_ATTRIBUTES (decl) without affecting
4734 DECL_ATTRIBUTES (current_function_decl).
4735 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4736 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4738 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4740 PR tree-optimization/83979
4741 * fold-const.c (fold_comparison): Use constant_boolean_node
4742 instead of boolean_{true,false}_node.
4744 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4746 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4749 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4751 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4752 Simplify the clause that sets the length attribute.
4753 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4754 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4755 clause that sets the length attribute.
4756 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4758 2018-01-24 Tom de Vries <tom@codesourcery.com>
4761 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4762 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4763 Add strict parameter.
4764 (prevent_branch_around_nothing): Insert dummy insn between branch to
4765 label and label with no ptx insn inbetween.
4766 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4768 2018-01-24 Tom de Vries <tom@codesourcery.com>
4771 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4772 for neutered threads in warp.
4773 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4775 2018-01-24 Richard Biener <rguenther@suse.de>
4777 PR tree-optimization/83176
4778 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4781 2018-01-24 Richard Biener <rguenther@suse.de>
4783 PR tree-optimization/82819
4784 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4785 code generating pluses that are no-ops in the target precision.
4787 2018-01-24 Richard Biener <rguenther@suse.de>
4790 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4792 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4794 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4795 to merge probabilities.
4796 * predict.c (probably_never_executed): Also mark as cold functions
4797 with global 0 profile and guessed local profile.
4798 * profile-count.c (profile_probability::combine_with_count): New
4800 * profile-count.h (profile_probability::operator*,
4801 profile_probability::operator*=, profile_probability::operator/,
4802 profile_probability::operator/=): Reduce precision to adjusted
4803 and set value to guessed on contradictory divisions.
4804 (profile_probability::combine_with_freq): Remove.
4805 (profile_probability::combine_wiht_count): Declare.
4806 (profile_count::force_nonzero):: Set to adjusted.
4807 (profile_count::probability_in):: Set quality to adjusted.
4808 * tree-ssa-tail-merge.c (replace_block_by): Use
4811 2018-01-23 Andrew Waterman <andrew@sifive.com>
4812 Jim Wilson <jimw@sifive.com>
4814 * config/riscv/riscv.c (riscv_stack_boundary): New.
4815 (riscv_option_override): Set riscv_stack_boundary. Handle
4816 riscv_preferred_stack_boundary_arg.
4817 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4818 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4819 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4820 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4821 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4822 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4824 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4827 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4828 of struct ix86_frame.
4829 (ix86_expand_epilogue): Likewise. Add a local variable for
4830 the reg_save_offset field in struct ix86_frame.
4832 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4834 PR tree-optimization/82604
4835 * tree-loop-distribution.c (enum partition_kind): New enum item
4836 PKIND_PARTIAL_MEMSET.
4837 (partition_builtin_p): Support above new enum item.
4838 (generate_code_for_partition): Ditto.
4839 (compute_access_range): Differentiate cases that equality can be
4840 proven at all loops, the innermost loops or no loops.
4841 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4842 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4843 (finalize_partitions, distribute_loop): Don't fuse partition of
4844 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4845 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4848 2018-01-23 Martin Liska <mliska@suse.cz>
4850 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4851 order to ignore the predictor.
4852 (PRED_POLYMORPHIC_CALL): Likewise.
4853 (PRED_RECURSIVE_CALL): Likewise.
4855 2018-01-23 Martin Liska <mliska@suse.cz>
4857 * tree-profile.c (tree_profiling): Print function header to
4858 aware reader which function we are working on.
4859 * value-prof.c (gimple_find_values_to_profile): Do not print
4860 not interesting value histograms.
4862 2018-01-23 Martin Liska <mliska@suse.cz>
4864 * profile-count.h (enum profile_quality): Add
4865 profile_uninitialized as the first value. Do not number values
4866 as they are zero based.
4867 (profile_count::verify): Update sanity check.
4868 (profile_probability::verify): Likewise.
4870 2018-01-23 Nathan Sidwell <nathan@acm.org>
4872 * doc/invoke.texi (ffor-scope): Deprecate.
4874 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4876 PR tree-optimization/83510
4877 * domwalk.c (set_all_edges_as_executable): New function.
4878 (dom_walker::dom_walker): Convert bool param
4879 "skip_unreachable_blocks" to enum reachability. Move setup of
4880 edge flags to set_all_edges_as_executable and only do it when
4881 reachability is REACHABLE_BLOCKS.
4882 * domwalk.h (enum dom_walker::reachability): New enum.
4883 (dom_walker::dom_walker): Convert bool param
4884 "skip_unreachable_blocks" to enum reachability.
4885 (set_all_edges_as_executable): New decl.
4886 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4887 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4889 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4890 but converting true to REACHABLE_BLOCKS.
4891 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4893 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4894 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4895 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4897 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4898 if check_all_array_refs will be called.
4900 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4902 * tree.c (selftest::test_location_wrappers): Add more test
4905 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4907 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4908 (selftest::test_bit_in_range): Likewise.
4910 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4913 * doc/sourcebuild.texi (vect_float): Say that the selector
4914 only describes the situation when -funsafe-math-optimizations is on.
4915 (vect_float_strict): Document.
4917 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4919 PR tree-optimization/83965
4920 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4921 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4922 instead of checking only for a reduction.
4923 (vect_recog_widen_sum_pattern): Likewise.
4925 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4927 * predict.c (probably_never_executed): Only use precise profile info.
4928 (compute_function_frequency): Skip after inlining hack since we now
4929 have quality checking.
4931 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4933 * profile-count.h (profile_probability::very_unlikely,
4934 profile_probability::unlikely, profile_probability::even): Set
4935 precision to guessed.
4937 2018-01-23 Richard Biener <rguenther@suse.de>
4939 PR tree-optimization/83963
4940 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4941 Properly terminate dominator walk when crossing the exit edge not
4942 when visiting its source block.
4944 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4947 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4948 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4950 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4952 PR tree-optimization/83957
4953 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4954 semicolon after for body surrounded by braces.
4956 PR tree-optimization/83081
4957 * profile-count.h (profile_probability::split): New method.
4958 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4959 Use profile_probability::split.
4960 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4961 when splitting a single conditional jump into 2.
4963 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4965 PR tree-optimization/69452
4966 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4969 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4971 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4972 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4973 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4975 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4977 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4978 * config/rl78/rl78.md (movdi): New define_expand.
4979 * config/rl78/rl78.c (rl78_split_movdi): New function.
4981 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4984 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4986 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4987 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4988 128-bit to produce an UNSPEC move to get the double word with the
4989 signbit and then a shift directly to do signbit.
4990 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4991 implementation with a new version that just does either a direct
4992 move or a regular move. Move memory interface to separate insns.
4993 Move insns so they are next to the expander.
4994 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4995 with signbit move. Split big and little endian case.
4996 (signbit<mode>2_dm_mem_le): Likewise.
4997 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4998 (signbit<mode>2_dm2): Likewise.
5000 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5002 * config/rl78/rl78.md (anddi3): New define_expand.
5004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5006 * config/rl78/rl78.md (umindi3): New define_expand.
5008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5010 * config/rl78/rl78.md (smindi3): New define_expand.
5012 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5014 * config/rl78/rl78.md (smaxdi3): New define_expand.
5016 2018-01-22 Carl Love <cel@us.ibm.com>
5018 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5019 LVX_V1TI): Add macro expansion.
5020 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5021 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5022 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5023 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5024 Change check to determine if the instruction is a byte reversing
5025 entry. Fix typo in comment.
5026 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5027 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5028 Add def_builtin calls for new builtins.
5029 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5030 Add define_insn expansion.
5032 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5034 * config/rl78/rl78.md (umaxdi3): New define_expand.
5036 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5038 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5039 for non-QImode registers.
5041 2018-01-22 Richard Biener <rguenther@suse.de>
5043 PR tree-optimization/83963
5044 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5045 including the loop exit block.
5046 (scop_detection::merge_sese): Likewise.
5047 (scop_detection::add_scop): Do it here instead.
5049 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5051 * doc/sourcebuild.texi (arm_softfloat): Document.
5053 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5056 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5057 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5058 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5060 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5061 David Edelsohn <dje.gcc@gmail.com>
5064 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5065 Change "crset eq" to "crset 2".
5066 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5067 (*call_indirect_aix<mode>_nospec): Likewise.
5068 (*call_value_indirect_aix<mode>_nospec): Likewise.
5069 (*call_indirect_elfv2<mode>_nospec): Likewise.
5070 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5071 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5072 change assembly output from . to $.
5073 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5074 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5075 (*tablejump<mode>_internal1_nospec): Likewise.
5077 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5080 * config/sh/sh_optimize_sett_clrt.cc:
5081 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5083 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5085 PR tree-optimization/83940
5086 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5087 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5088 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5089 use it to pass back the definition type.
5090 (vect_check_store_rhs): Likewise rhs_dt_out.
5091 (vect_build_gather_load_calls): Add a mask_dt argument and use
5092 it instead of a call to vect_is_simple_use.
5093 (vectorizable_store): Update calls to vect_check_load_store_mask
5094 and vect_check_store_rhs. Use the dt returned by the latter instead
5095 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5096 instead of calls to vect_is_simple_use. Pass the scalar rather
5097 than the vector operand to vect_is_simple_use when handling
5098 second and subsequent copies of an rhs value.
5099 (vectorizable_load): Update calls to vect_check_load_store_mask
5100 and vect_build_gather_load_calls. Use the cached mask_dt and
5101 gs_info.offset_dt instead of calls to vect_is_simple_use.
5103 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5106 * tree-emutls.c: Include gimplify.h.
5107 (lower_emutls_2): New function.
5108 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5109 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5110 it before further processing.
5113 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5114 UINTVAL (trueop1) instead of INTVAL (op1).
5116 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5120 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5121 INCOMING_FRAME_SP_OFFSET if not defined.
5122 (scan_trace): Add ENTRY argument. If true and
5123 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5124 emit a note to adjust the CFA offset.
5125 (create_cfi_notes): Adjust scan_trace callers.
5126 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5127 INCOMING_FRAME_SP_OFFSET in the CIE.
5128 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5129 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5131 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5132 * doc/tm.texi: Regenerated.
5134 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5136 PR rtl-optimization/83147
5137 * lra-constraints.c (remove_inheritance_pseudos): Use
5138 lra_substitute_pseudo_within_insn.
5140 2018-01-19 Tom de Vries <tom@codesourcery.com>
5141 Cesar Philippidis <cesar@codesourcery.com>
5144 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5146 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5149 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5150 spaces for function labels.
5152 2018-01-19 Martin Liska <mliska@suse.cz>
5154 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5155 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5156 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5157 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5158 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5159 (PRED_CONST_RETURN): Change from 69 to 65.
5160 (PRED_NULL_RETURN): Change from 91 to 71.
5161 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5162 (PRED_LOOP_GUARD): Change from 66 to 73.
5164 2018-01-19 Martin Liska <mliska@suse.cz>
5166 * predict.c (predict_insn_def): Add new assert.
5167 (struct branch_predictor): Change type to signed integer.
5168 (test_prediction_value_range): Amend test to cover
5170 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5171 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5172 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5173 (PRED_LOOP_IV_COMPARE): Likewise.
5174 * predict.h (PROB_UNINITIALIZED): Define new constant.
5176 2018-01-19 Martin Liska <mliska@suse.cz>
5178 * predict.c (dump_prediction): Add new format for
5179 analyze_brprob.py script which is enabled with -details
5181 * profile-count.h (precise_p): New function.
5183 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5185 PR tree-optimization/83922
5186 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5187 there are no statements that need masking.
5188 (vect_active_double_reduction_p): New function.
5189 (vect_analyze_loop_operations): Use it when handling phis that
5190 are not in the loop header.
5192 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5194 PR tree-optimization/83914
5195 * tree-vect-loop.c (vectorizable_induction): Don't convert
5196 init_expr or apply the peeling adjustment for inductions
5197 that are nested within the vectorized loop.
5199 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5201 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5204 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5208 * function.h (gimplify_parameters): Add gimple_seq * argument.
5209 * function.c: Include gimple.h and options.h.
5210 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5211 for the added local temporaries if needed.
5212 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5213 if there are any parameter cleanups, wrap whole body into a
5214 try/finally with the cleanups.
5216 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5219 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5220 Use GET_MODE_CLASS for scalar floating point.
5222 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5226 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5227 Fix call of call_cgraph_insertion_hooks.
5229 2018-01-18 Martin Sebor <msebor@redhat.com>
5231 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5233 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5236 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5239 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5242 * common.opt: (-ffile-prefix-map): New option.
5243 * opts.c (common_handle_option): Defer it.
5244 * opts-global.c (handle_common_deferred_options): Handle it.
5245 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5246 * file-prefix-map.h: New file.
5247 (remap_debug_filename, add_debug_prefix_map): ...here.
5248 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5249 * final.c (debug_prefix_map, add_debug_prefix_map
5250 remap_debug_filename): Move to...
5251 * file-prefix-map.c: New file.
5252 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5253 generalize, get rid of alloca(), use strrchr() instead of strchr().
5254 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5255 Implement in terms of add_prefix_map().
5256 (remap_macro_filename, remap_debug_filename): Implement in term of
5258 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5259 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5260 * dbxout.c: Include file-prefix-map.h.
5261 * varasm.c: Likewise.
5262 * vmsdbgout.c: Likewise.
5263 * xcoffout.c: Likewise.
5264 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5265 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5266 * doc/invoke.texi (-ffile-prefix-map): Document.
5267 (-fdebug-prefix-map): Update description.
5269 2018-01-18 Martin Liska <mliska@suse.cz>
5271 * config/i386/i386.c (indirect_thunk_name): Document that also
5273 (output_indirect_thunk): Document why both instructions
5274 (pause and lfence) are generated.
5276 2018-01-18 Richard Biener <rguenther@suse.de>
5278 PR tree-optimization/83887
5279 * graphite-scop-detection.c
5280 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5281 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5282 (scop_detection::merge_sese): Re-implement with a flood-fill
5283 algorithm that properly finds a SESE region if it exists.
5285 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5288 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5289 pointer_diff optimizations use view_convert instead of convert.
5291 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5293 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5294 Generate different code for -mno-speculate-indirect-jumps.
5295 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5296 (*call_indirect_aix<mode>): Disable for
5297 -mno-speculate-indirect-jumps.
5298 (*call_indirect_aix<mode>_nospec): New define_insn.
5299 (*call_value_indirect_aix<mode>): Disable for
5300 -mno-speculate-indirect-jumps.
5301 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5302 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5303 -mno-speculate-indirect-jumps.
5304 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5306 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5308 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5309 long double type, set the flags for noting the default long double
5310 type, even if we don't pass or return a long double type.
5312 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5315 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5318 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5320 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5321 support for merge[hl].
5322 (fold_mergehl_helper): New helper function.
5323 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5324 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5325 (altivec_vmrglw_direct): Add xxmrglw insn.
5327 2018-01-17 Andrew Waterman <andrew@sifive.com>
5329 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5330 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5332 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5335 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5336 call the lto_location_cache before reading the
5337 DECL_SOURCE_LOCATION of the types.
5339 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5340 Richard Sandiford <richard.sandiford@linaro.org>
5342 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5343 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5344 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5345 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5346 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5348 * config/aarch64/constraints.md (aarch64_movti_operand):
5350 * config/aarch64/predicates.md (Uti): Add new constraint.
5352 2018-01-17 Carl Love <cel@us.ibm.com>
5354 * config/rs6000/vsx.md (define_expand xl_len_r,
5355 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5356 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5358 (define_expand, define_insn): Move the shift left from the
5359 define_insn to the define_expand for lxvl and stxvl instructions.
5360 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5361 and XL_LEN_R definitions to PURE.
5363 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5365 * config/i386/i386.c (indirect_thunk_name): Declare regno
5366 as unsigned int. Compare regno with INVALID_REGNUM.
5367 (output_indirect_thunk): Ditto.
5368 (output_indirect_thunk_function): Ditto.
5369 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5370 in the call to output_indirect_thunk_function.
5372 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5375 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5376 rather than the size of inner_type to determine the stack slot size
5377 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5379 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5382 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5385 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5387 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5388 endian Linux systems to optionally enable multilibs for selecting
5389 the long double type if the user configured an explicit type.
5390 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5391 have no long double multilibs if not defined.
5392 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5393 warn if the user used -mabi={ieee,ibm}longdouble and we built
5394 multilibs for long double.
5395 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5396 appropriate multilib option.
5397 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5399 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5400 for building long double multilibs.
5401 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5403 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5405 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5408 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5410 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5413 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5416 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5419 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5421 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5422 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5424 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5426 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5427 different rtl trees depending on TARGET_64BIT.
5428 (rs6000_gen_lvx): Likewise.
5430 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5432 * config/visium/visium.md (nop): Tweak comment.
5433 (hazard_nop): Likewise.
5435 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5437 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5438 -mspeculate-indirect-jumps.
5439 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5440 for -mno-speculate-indirect-jumps.
5441 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5442 (*call_value_indirect_elfv2<mode>): Disable for
5443 -mno-speculate-indirect-jumps.
5444 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5445 (indirect_jump): Emit different RTL for
5446 -mno-speculate-indirect-jumps.
5447 (*indirect_jump<mode>): Disable for
5448 -mno-speculate-indirect-jumps.
5449 (*indirect_jump<mode>_nospec): New define_insn.
5450 (tablejump): Emit different RTL for
5451 -mno-speculate-indirect-jumps.
5452 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5453 (tablejumpsi_nospec): New define_expand.
5454 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5455 (tablejumpdi_nospec): New define_expand.
5456 (*tablejump<mode>_internal1): Disable for
5457 -mno-speculate-indirect-jumps.
5458 (*tablejump<mode>_internal1_nospec): New define_insn.
5459 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5462 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5464 * caller-save.c (insert_save): Drop unnecessary parameter. All
5467 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5468 Richard Biener <rguenth@suse.de>
5471 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5472 return early, inline manually is_gimple_sizepos. Make sure if we
5473 call gimplify_expr we don't end up with a gimple constant.
5474 * tree.c (variably_modified_type_p): Don't return true for
5475 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5476 * gimplify.h (is_gimple_sizepos): Remove.
5478 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5480 PR tree-optimization/83857
5481 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5482 vectorizable_live_operation for pure SLP statements.
5483 (vectorizable_live_operation): Handle PHIs.
5485 2018-01-16 Richard Biener <rguenther@suse.de>
5487 PR tree-optimization/83867
5488 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5489 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5491 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5494 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5495 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5496 If off is not INTEGER_CST, issue a may not be aligned warning
5497 rather than isn't aligned. Use isn%'t rather than isn't.
5498 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5500 <case MULT_EXPR>: Improve the case when bottom and one of the
5501 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5502 operand, in that case check if the other operand is multiple of
5503 bottom divided by the INTEGER_CST operand.
5505 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5508 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5509 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5510 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5511 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5512 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5513 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5514 (pa_function_arg_size): New function.
5516 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5518 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5519 in a separate statement.
5521 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5523 PR tree-optimization/83847
5524 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5525 group gathers and scatters.
5527 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5529 PR rtl-optimization/86620
5530 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5532 PR rtl-optimization/83213
5533 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5534 to last if both are JUMP_INSNs.
5536 PR tree-optimization/83843
5537 * gimple-ssa-store-merging.c
5538 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5539 store_immediate_info for bswap/nop orig_stores.
5541 2018-01-15 Andrew Waterman <andrew@sifive.com>
5543 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5545 <UDIV>: Increase cost if !TARGET_DIV.
5547 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5549 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5550 (define_attr "cr_logical_3op"): New.
5551 (cceq_ior_compare): Adjust.
5552 (cceq_ior_compare_complement): Adjust.
5553 (*cceq_rev_compare): Adjust.
5554 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5555 (is_cracked_insn): Adjust.
5556 (insn_must_be_first_in_group): Adjust.
5557 * config/rs6000/40x.md: Adjust.
5558 * config/rs6000/440.md: Adjust.
5559 * config/rs6000/476.md: Adjust.
5560 * config/rs6000/601.md: Adjust.
5561 * config/rs6000/603.md: Adjust.
5562 * config/rs6000/6xx.md: Adjust.
5563 * config/rs6000/7450.md: Adjust.
5564 * config/rs6000/7xx.md: Adjust.
5565 * config/rs6000/8540.md: Adjust.
5566 * config/rs6000/cell.md: Adjust.
5567 * config/rs6000/e300c2c3.md: Adjust.
5568 * config/rs6000/e500mc.md: Adjust.
5569 * config/rs6000/e500mc64.md: Adjust.
5570 * config/rs6000/e5500.md: Adjust.
5571 * config/rs6000/e6500.md: Adjust.
5572 * config/rs6000/mpc.md: Adjust.
5573 * config/rs6000/power4.md: Adjust.
5574 * config/rs6000/power5.md: Adjust.
5575 * config/rs6000/power6.md: Adjust.
5576 * config/rs6000/power7.md: Adjust.
5577 * config/rs6000/power8.md: Adjust.
5578 * config/rs6000/power9.md: Adjust.
5579 * config/rs6000/rs64.md: Adjust.
5580 * config/rs6000/titan.md: Adjust.
5582 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5584 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5585 ix86_indirect_branch_register logic.
5587 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5589 * config/i386/constraints.md (Bs): Update
5590 ix86_indirect_branch_register check. Don't check
5591 ix86_indirect_branch_register with GOT_memory_operand.
5593 * config/i386/predicates.md (GOT_memory_operand): Don't check
5594 ix86_indirect_branch_register here.
5595 (GOT32_symbol_operand): Likewise.
5597 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5599 * config/i386/predicates.md (constant_call_address_operand):
5600 Rewrite ix86_indirect_branch_register logic.
5601 (sibcall_insn_operand): Likewise.
5603 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5605 * config/i386/constraints.md (Bs): Replace
5606 ix86_indirect_branch_thunk_register with
5607 ix86_indirect_branch_register.
5609 * config/i386/i386.md (indirect_jump): Likewise.
5610 (tablejump): Likewise.
5611 (*sibcall_memory): Likewise.
5612 (*sibcall_value_memory): Likewise.
5613 Peepholes of indirect call and jump via memory: Likewise.
5614 * config/i386/i386.opt: Likewise.
5615 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5616 (GOT_memory_operand): Likewise.
5617 (call_insn_operand): Likewise.
5618 (sibcall_insn_operand): Likewise.
5619 (GOT32_symbol_operand): Likewise.
5621 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5624 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5625 type rather than type addr's type points to.
5626 (expand_omp_atomic_mutex): Likewise.
5627 (expand_omp_atomic): Likewise.
5629 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5632 * config/i386/i386.c (output_indirect_thunk_function): Use
5633 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5634 for __x86_return_thunk.
5636 2018-01-15 Richard Biener <rguenther@suse.de>
5639 * expmed.c (extract_bit_field_1): Fix typo.
5641 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5644 * config/arm/iterators.md (VF): New mode iterator.
5645 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5646 Remove integer-related logic from pattern.
5647 (neon_vabd<mode>_3): Likewise.
5649 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5652 * common.opt (fstrict-overflow): No longer an alias.
5653 (fwrapv-pointer): New option.
5654 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5655 also for pointer types based on flag_wrapv_pointer.
5656 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5657 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5658 opts->x_flag_wrapv got set.
5659 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5660 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5661 POINTER_TYPE_OVERFLOW_UNDEFINED.
5662 * match.pd: Likewise in address comparison pattern.
5663 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5665 2018-01-15 Richard Biener <rguenther@suse.de>
5668 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5669 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5670 Reset type names to their identifier if their TYPE_DECL doesn't
5671 have linkage (and thus is used for ODR and devirt).
5672 (save_debug_info_for_decl): Remove.
5673 (save_debug_info_for_type): Likewise.
5674 (add_tree_to_fld_list): Adjust.
5675 * tree-pretty-print.c (dump_generic_node): Make dumping of
5676 type names more robust.
5678 2018-01-15 Richard Biener <rguenther@suse.de>
5680 * BASE-VER: Bump to 8.0.1.
5682 2018-01-14 Martin Sebor <msebor@redhat.com>
5685 * builtins.c (check_access): Avoid warning when the no-warning bit
5688 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5690 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5691 * ira-color (allocno_hard_regs_compare): Likewise.
5693 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5696 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5697 Use .pushsection/.popsection.
5699 2018-01-14 Martin Sebor <msebor@redhat.com>
5702 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5704 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5706 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5707 entry from extra_headers.
5708 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5709 extra_headers, make the list bitwise identical to the i?86-*-* one.
5711 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5713 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5714 -mcmodel=large with -mindirect-branch=thunk,
5715 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5716 -mfunction-return=thunk-extern.
5717 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5718 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5719 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5721 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5723 * config/i386/i386.c (print_reg): Print the name of the full
5724 integer register without '%'.
5725 (ix86_print_operand): Handle 'V'.
5726 * doc/extend.texi: Document 'V' modifier.
5728 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5730 * config/i386/constraints.md (Bs): Disallow memory operand for
5731 -mindirect-branch-register.
5733 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5734 (GOT_memory_operand): Likewise.
5735 (call_insn_operand): Likewise.
5736 (sibcall_insn_operand): Likewise.
5737 (GOT32_symbol_operand): Likewise.
5738 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5739 for -mindirect-branch-register.
5740 (tablejump): Likewise.
5741 (*sibcall_memory): Likewise.
5742 (*sibcall_value_memory): Likewise.
5743 Disallow peepholes of indirect call and jump via memory for
5744 -mindirect-branch-register.
5745 (*call_pop): Replace m with Bw.
5746 (*call_value_pop): Likewise.
5747 (*sibcall_pop_memory): Replace m with Bs.
5748 * config/i386/i386.opt (mindirect-branch-register): New option.
5749 * doc/invoke.texi: Document -mindirect-branch-register option.
5751 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5753 * config/i386/i386-protos.h (ix86_output_function_return): New.
5754 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5755 set function_return_type.
5756 (indirect_thunk_name): Add ret_p to indicate thunk for function
5758 (output_indirect_thunk_function): Pass false to
5759 indirect_thunk_name.
5760 (ix86_output_indirect_branch_via_reg): Likewise.
5761 (ix86_output_indirect_branch_via_push): Likewise.
5762 (output_indirect_thunk_function): Create alias for function
5763 return thunk if regno < 0.
5764 (ix86_output_function_return): New function.
5765 (ix86_handle_fndecl_attribute): Handle function_return.
5766 (ix86_attribute_table): Add function_return.
5767 * config/i386/i386.h (machine_function): Add
5768 function_return_type.
5769 * config/i386/i386.md (simple_return_internal): Use
5770 ix86_output_function_return.
5771 (simple_return_internal_long): Likewise.
5772 * config/i386/i386.opt (mfunction-return=): New option.
5773 (indirect_branch): Mention -mfunction-return=.
5774 * doc/extend.texi: Document function_return function attribute.
5775 * doc/invoke.texi: Document -mfunction-return= option.
5777 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5779 * config/i386/i386-opts.h (indirect_branch): New.
5780 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5781 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5782 with local indirect jump when converting indirect call and jump.
5783 (ix86_set_indirect_branch_type): New.
5784 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5785 (indirectlabelno): New.
5786 (indirect_thunk_needed): Likewise.
5787 (indirect_thunk_bnd_needed): Likewise.
5788 (indirect_thunks_used): Likewise.
5789 (indirect_thunks_bnd_used): Likewise.
5790 (INDIRECT_LABEL): Likewise.
5791 (indirect_thunk_name): Likewise.
5792 (output_indirect_thunk): Likewise.
5793 (output_indirect_thunk_function): Likewise.
5794 (ix86_output_indirect_branch_via_reg): Likewise.
5795 (ix86_output_indirect_branch_via_push): Likewise.
5796 (ix86_output_indirect_branch): Likewise.
5797 (ix86_output_indirect_jmp): Likewise.
5798 (ix86_code_end): Call output_indirect_thunk_function if needed.
5799 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5801 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5802 (ix86_attribute_table): Add indirect_branch.
5803 * config/i386/i386.h (machine_function): Add indirect_branch_type
5804 and has_local_indirect_jump.
5805 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5807 (tablejump): Likewise.
5808 (*indirect_jump): Use ix86_output_indirect_jmp.
5809 (*tablejump_1): Likewise.
5810 (simple_return_indirect_internal): Likewise.
5811 * config/i386/i386.opt (mindirect-branch=): New option.
5812 (indirect_branch): New.
5815 (thunk-inline): Likewise.
5816 (thunk-extern): Likewise.
5817 * doc/extend.texi: Document indirect_branch function attribute.
5818 * doc/invoke.texi: Document -mindirect-branch= option.
5820 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5823 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5825 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5827 * ipa-inline.c (want_inline_small_function_p): Return false if
5828 inlining has already failed with CIF_FINAL_ERROR.
5829 (update_caller_keys): Call want_inline_small_function_p before
5831 (update_callee_keys): Likewise.
5833 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5835 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5837 (rs6000_quadword_masked_address_p): Likewise.
5838 (quad_aligned_load_p): Likewise.
5839 (quad_aligned_store_p): Likewise.
5840 (const_load_sequence_p): Add comment to describe the outer-most loop.
5841 (mimic_memory_attributes_and_flags): New function.
5842 (rs6000_gen_stvx): Likewise.
5843 (replace_swapped_aligned_store): Likewise.
5844 (rs6000_gen_lvx): Likewise.
5845 (replace_swapped_aligned_load): Likewise.
5846 (replace_swapped_load_constant): Capitalize argument name in
5847 comment describing this function.
5848 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5849 and stores that access quad-word aligned addresses and replace
5850 with stvx or lvx instructions when appropriate.
5851 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5852 New function prototype.
5853 (rs6000_quadword_masked_address_p): Likewise.
5854 (rs6000_gen_lvx): Likewise.
5855 (rs6000_gen_stvx): Likewise.
5856 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5857 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5858 when memory address is aligned.
5859 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5860 this split to select lvx instruction when memory address is aligned.
5861 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5862 instruction when memory address is aligned.
5863 (*vsx_le_perm_load_v16qi): Likewise.
5864 (four unnamed splitters): Modify to select the stvx instruction
5865 when memory is aligned.
5867 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5869 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5870 which appears in the queue multiple times.
5872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5873 Alan Hayward <alan.hayward@arm.com>
5874 David Sherwood <david.sherwood@arm.com>
5876 * tree-vectorizer.h (vec_lower_bound): New structure.
5877 (_loop_vec_info): Add check_nonzero and lower_bounds.
5878 (LOOP_VINFO_CHECK_NONZERO): New macro.
5879 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5880 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5881 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5882 fields. Make seg_len the distance travelled, not including the
5884 (dr_direction_indicator): Declare.
5885 (dr_zero_step_indicator): Likewise.
5886 (dr_known_forward_stride_p): Likewise.
5887 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5889 (runtime_alias_check_p): Allow runtime alias checks with
5891 (operator ==): Compare access_size and align.
5892 (prune_runtime_alias_test_list): Rework for new distinction between
5893 the access_size and seg_len.
5894 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5896 (get_segment_min_max): New function.
5897 (create_intersect_range_checks): Use it.
5898 (dr_step_indicator): New function.
5899 (dr_direction_indicator): Likewise.
5900 (dr_zero_step_indicator): Likewise.
5901 (dr_known_forward_stride_p): Likewise.
5902 * tree-loop-distribution.c (data_ref_segment_size): Return
5903 DR_STEP * (niters - 1).
5904 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5906 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5907 (vect_preserves_scalar_order_p): New function, split out from...
5908 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5909 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5910 (vect_vfa_access_size): New function.
5911 (vect_vfa_align): Likewise.
5912 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5913 (dump_lower_bound): New function.
5914 (vect_check_lower_bound): Likewise.
5915 (vect_small_gap_p): Likewise.
5916 (vectorizable_with_step_bound_p): Likewise.
5917 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5918 depencies if the vectorization factor is 1. Convert the checks
5919 for nonzero steps into checks on the bounds of DR_STEP. Try using
5920 a bunds check for variable steps if the minimum required step is
5921 relatively small. Update calls to the dr_with_seg_len
5922 constructor and to vect_compile_time_alias.
5923 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5925 (vect_loop_versioning): Call it.
5926 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5928 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5931 Alan Hayward <alan.hayward@arm.com>
5932 David Sherwood <david.sherwood@arm.com>
5934 * doc/sourcebuild.texi (vect_scatter_store): Document.
5935 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5937 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5939 * genopinit.c (main): Add supports_vec_scatter_store and
5940 supports_vec_scatter_store_cached to target_optabs.
5941 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5942 IFN_MASK_SCATTER_STORE.
5943 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5945 * internal-fn.h (internal_store_fn_p): Declare.
5946 (internal_fn_stored_value_index): Likewise.
5947 * internal-fn.c (scatter_store_direct): New macro.
5948 (expand_scatter_store_optab_fn): New function.
5949 (direct_scatter_store_optab_supported_p): New macro.
5950 (internal_store_fn_p): New function.
5951 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5952 IFN_MASK_SCATTER_STORE.
5953 (internal_fn_mask_index): Likewise.
5954 (internal_fn_stored_value_index): New function.
5955 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5957 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5958 * optabs-query.c (supports_vec_scatter_store_p): New function.
5959 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5960 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5961 true for scatter stores.
5962 (vect_gather_scatter_fn_p): Handle scatter stores too.
5963 (vect_check_gather_scatter): Consider using scatter stores if
5964 supports_vec_scatter_store_p.
5965 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5967 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5968 internal_fn_stored_value_index.
5969 (check_load_store_masking): Handle scatter stores too.
5970 (vect_get_store_rhs): Make public.
5971 (vectorizable_call): Use internal_store_fn_p.
5972 (vectorizable_store): Handle scatter store internal functions.
5973 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5974 when deciding whether the end of the group has been reached.
5975 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5976 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5977 (mask_scatter_store<mode>): New insns.
5979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5980 Alan Hayward <alan.hayward@arm.com>
5981 David Sherwood <david.sherwood@arm.com>
5983 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5984 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5985 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5987 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5988 Use vect_truncate_gather_scatter_offset if we can't treat the
5989 operation as a normal gather load or scatter store.
5990 (get_group_load_store_type): Take the gather_scatter_info
5991 as argument. Try using a gather load or scatter store for
5992 single-element groups.
5993 (get_load_store_type): Update calls to get_group_load_store_type
5994 and vect_use_strided_gather_scatters_p.
5996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5997 Alan Hayward <alan.hayward@arm.com>
5998 David Sherwood <david.sherwood@arm.com>
6000 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6001 optional tree argument.
6002 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6004 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6005 but continue to use the current value as a fallback.
6006 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6007 to compare the updates.
6008 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6009 (get_load_store_type): Use it when handling a strided access.
6010 (vect_get_strided_load_store_ops): New function.
6011 (vect_get_data_ptr_increment): Likewise.
6012 (vectorizable_load): Handle strided gather loads. Always pass
6013 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6016 Alan Hayward <alan.hayward@arm.com>
6017 David Sherwood <david.sherwood@arm.com>
6019 * doc/md.texi (gather_load@var{m}): Document.
6020 (mask_gather_load@var{m}): Likewise.
6021 * genopinit.c (main): Add supports_vec_gather_load and
6022 supports_vec_gather_load_cached to target_optabs.
6023 * optabs-tree.c (init_tree_optimization_optabs): Use
6024 ggc_cleared_alloc to allocate target_optabs.
6025 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6026 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6028 * internal-fn.h (internal_load_fn_p): Declare.
6029 (internal_gather_scatter_fn_p): Likewise.
6030 (internal_fn_mask_index): Likewise.
6031 (internal_gather_scatter_fn_supported_p): Likewise.
6032 * internal-fn.c (gather_load_direct): New macro.
6033 (expand_gather_load_optab_fn): New function.
6034 (direct_gather_load_optab_supported_p): New macro.
6035 (direct_internal_fn_optab): New function.
6036 (internal_load_fn_p): Likewise.
6037 (internal_gather_scatter_fn_p): Likewise.
6038 (internal_fn_mask_index): Likewise.
6039 (internal_gather_scatter_fn_supported_p): Likewise.
6040 * optabs-query.c (supports_at_least_one_mode_p): New function.
6041 (supports_vec_gather_load_p): Likewise.
6042 * optabs-query.h (supports_vec_gather_load_p): Declare.
6043 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6044 and memory_type field.
6045 (NUM_PATTERNS): Bump to 15.
6046 * tree-vect-data-refs.c: Include internal-fn.h.
6047 (vect_gather_scatter_fn_p): New function.
6048 (vect_describe_gather_scatter_call): Likewise.
6049 (vect_check_gather_scatter): Try using internal functions for
6050 gather loads. Recognize existing calls to a gather load function.
6051 (vect_analyze_data_refs): Consider using gather loads if
6052 supports_vec_gather_load_p.
6053 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6054 (vect_get_gather_scatter_offset_type): Likewise.
6055 (vect_convert_mask_for_vectype): Likewise.
6056 (vect_add_conversion_to_patterm): Likewise.
6057 (vect_try_gather_scatter_pattern): Likewise.
6058 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6059 (vect_vect_recog_func_ptrs): Add it.
6060 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6061 internal_fn_mask_index and internal_gather_scatter_fn_p.
6062 (check_load_store_masking): Take the gather_scatter_info as an
6063 argument and handle gather loads.
6064 (vect_get_gather_scatter_ops): New function.
6065 (vectorizable_call): Check internal_load_fn_p.
6066 (vectorizable_load): Likewise. Handle gather load internal
6068 (vectorizable_store): Update call to check_load_store_masking.
6069 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6070 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6071 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6072 (aarch64_gather_scale_operand_d): New predicates.
6073 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6074 (mask_gather_load<mode>): New insns.
6076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6077 Alan Hayward <alan.hayward@arm.com>
6078 David Sherwood <david.sherwood@arm.com>
6080 * optabs.def (fold_left_plus_optab): New optab.
6081 * doc/md.texi (fold_left_plus_@var{m}): Document.
6082 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6083 * internal-fn.c (fold_left_direct): Define.
6084 (expand_fold_left_optab_fn): Likewise.
6085 (direct_fold_left_optab_supported_p): Likewise.
6086 * fold-const-call.c (fold_const_fold_left): New function.
6087 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6088 * tree-parloops.c (valid_reduction_p): New function.
6089 (gather_scalar_reductions): Use it.
6090 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6091 (vect_finish_replace_stmt): Declare.
6092 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6093 (needs_fold_left_reduction_p): New function, split out from...
6094 (vect_is_simple_reduction): ...here. Accept reductions that
6095 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6096 (vect_force_simple_reduction): Also store the reduction type in
6097 the assignment's STMT_VINFO_REDUC_TYPE.
6098 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6099 (merge_with_identity): New function.
6100 (vect_expand_fold_left): Likewise.
6101 (vectorize_fold_left_reduction): Likewise.
6102 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6103 scalar phi in place for it. Check for target support and reject
6104 cases that would reassociate the operation. Defer the transform
6105 phase to vectorize_fold_left_reduction.
6106 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6107 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6108 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6112 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6113 call to ifc_temp_var.
6115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6116 Alan Hayward <alan.hayward@arm.com>
6117 David Sherwood <david.sherwood@arm.com>
6119 * target.def (legitimize_address_displacement): Take the original
6120 offset as a poly_int.
6121 * targhooks.h (default_legitimize_address_displacement): Update
6123 * targhooks.c (default_legitimize_address_displacement): Likewise.
6124 * doc/tm.texi: Regenerate.
6125 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6126 as an argument, moving assert of ad->disp == ad->disp_term to...
6127 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6128 Try calling targetm.legitimize_address_displacement before expanding
6129 the address rather than afterwards, and adjust for the new interface.
6130 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6131 Match the new hook interface. Handle SVE addresses.
6132 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6137 * Makefile.in (OBJS): Add early-remat.o.
6138 * target.def (select_early_remat_modes): New hook.
6139 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6140 * doc/tm.texi: Regenerate.
6141 * targhooks.h (default_select_early_remat_modes): Declare.
6142 * targhooks.c (default_select_early_remat_modes): New function.
6143 * timevar.def (TV_EARLY_REMAT): New timevar.
6144 * passes.def (pass_early_remat): New pass.
6145 * tree-pass.h (make_pass_early_remat): Declare.
6146 * early-remat.c: New file.
6147 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6149 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6152 Alan Hayward <alan.hayward@arm.com>
6153 David Sherwood <david.sherwood@arm.com>
6155 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6156 vfm1 with a bound_epilog parameter.
6157 (vect_do_peeling): Update calls accordingly, and move the prologue
6158 call earlier in the function. Treat the base bound_epilog as 0 for
6159 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6160 this base when peeling for gaps.
6161 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6162 with fully-masked loops.
6163 (vect_estimate_min_profitable_iters): Handle the single peeled
6164 iteration in that case.
6166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6167 Alan Hayward <alan.hayward@arm.com>
6168 David Sherwood <david.sherwood@arm.com>
6170 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6171 single-element interleaving even if the size is not a power of 2.
6172 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6173 accesses for single-element interleaving if the group size is
6176 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6177 Alan Hayward <alan.hayward@arm.com>
6178 David Sherwood <david.sherwood@arm.com>
6180 * doc/md.texi (fold_extract_last_@var{m}): Document.
6181 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6182 * optabs.def (fold_extract_last_optab): New optab.
6183 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6184 * internal-fn.c (fold_extract_direct): New macro.
6185 (expand_fold_extract_optab_fn): Likewise.
6186 (direct_fold_extract_optab_supported_p): Likewise.
6187 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6188 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6189 EXTRACT_LAST_REDUCTION.
6190 (get_initial_def_for_reduction): Do not create an initial vector
6191 for EXTRACT_LAST_REDUCTION reductions.
6192 (vectorizable_reduction): Leave the scalar phi in place for
6193 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6194 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6195 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6196 transform phase to vectorizable_condition.
6197 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6199 (vect_finish_stmt_generation): ...here.
6200 (vect_finish_replace_stmt): New function.
6201 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6202 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6204 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6207 Alan Hayward <alan.hayward@arm.com>
6208 David Sherwood <david.sherwood@arm.com>
6210 * doc/md.texi (extract_last_@var{m}): Document.
6211 * optabs.def (extract_last_optab): New optab.
6212 * internal-fn.def (EXTRACT_LAST): New internal function.
6213 * internal-fn.c (cond_unary_direct): New macro.
6214 (expand_cond_unary_optab_fn): Likewise.
6215 (direct_cond_unary_optab_supported_p): Likewise.
6216 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6217 loops using EXTRACT_LAST.
6218 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6219 (extract_last_<mode>): ...this optab.
6220 (vec_extract<mode><Vel>): Update accordingly.
6222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6223 Alan Hayward <alan.hayward@arm.com>
6224 David Sherwood <david.sherwood@arm.com>
6226 * target.def (empty_mask_is_expensive): New hook.
6227 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6228 * doc/tm.texi: Regenerate.
6229 * targhooks.h (default_empty_mask_is_expensive): Declare.
6230 * targhooks.c (default_empty_mask_is_expensive): New function.
6231 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6232 if the target says that empty masks are expensive.
6233 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6235 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6238 Alan Hayward <alan.hayward@arm.com>
6239 David Sherwood <david.sherwood@arm.com>
6241 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6242 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6243 (vect_use_loop_mask_for_alignment_p): New function.
6244 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6245 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6246 niters_skip argument. Make sure that the first niters_skip elements
6247 of the first iteration are inactive.
6248 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6249 Update call to vect_set_loop_masks_directly.
6250 (get_misalign_in_elems): New function, split out from...
6251 (vect_gen_prolog_loop_niters): ...here.
6252 (vect_update_init_of_dr): Take a code argument that specifies whether
6253 the adjustment should be added or subtracted.
6254 (vect_update_init_of_drs): Likewise.
6255 (vect_prepare_for_masked_peels): New function.
6256 (vect_do_peeling): Skip prologue peeling if we're using a mask
6257 instead. Update call to vect_update_inits_of_drs.
6258 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6260 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6261 alignment. Do not include the number of peeled iterations in
6262 the minimum threshold in that case.
6263 (vectorizable_induction): Adjust the start value down by
6264 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6265 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6266 Take the number of skipped iterations into account when calculating
6268 * tree-vect-stmts.c (vect_gen_while_not): New function.
6270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6271 Alan Hayward <alan.hayward@arm.com>
6272 David Sherwood <david.sherwood@arm.com>
6274 * doc/sourcebuild.texi (vect_fully_masked): Document.
6275 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6277 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6279 (vect_analyze_loop_2): ...here. Don't check the vectorization
6280 factor against the number of loop iterations if the loop is
6283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6284 Alan Hayward <alan.hayward@arm.com>
6285 David Sherwood <david.sherwood@arm.com>
6287 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6288 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6289 (dump_groups): Update accordingly.
6290 (iv_use::mem_type): New member variable.
6291 (address_p): New function.
6292 (record_use): Add a mem_type argument and initialize the new
6294 (record_group_use): Add a mem_type argument. Use address_p.
6295 Remove obsolete null checks of base_object. Update call to record_use.
6296 (find_interesting_uses_op): Update call to record_group_use.
6297 (find_interesting_uses_cond): Likewise.
6298 (find_interesting_uses_address): Likewise.
6299 (get_mem_type_for_internal_fn): New function.
6300 (find_address_like_use): Likewise.
6301 (find_interesting_uses_stmt): Try find_address_like_use before
6302 calling find_interesting_uses_op.
6303 (addr_offset_valid_p): Use the iv mem_type field as the type
6304 of the addressed memory.
6305 (add_autoinc_candidates): Likewise.
6306 (get_address_cost): Likewise.
6307 (split_small_address_groups_p): Use address_p.
6308 (split_address_groups): Likewise.
6309 (add_iv_candidate_for_use): Likewise.
6310 (autoinc_possible_for_pair): Likewise.
6311 (rewrite_groups): Likewise.
6312 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6313 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6314 (get_alias_ptr_type_for_ptr_address): New function.
6315 (rewrite_use_address): Rewrite address uses in calls that were
6316 identified by find_address_like_use.
6318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6319 Alan Hayward <alan.hayward@arm.com>
6320 David Sherwood <david.sherwood@arm.com>
6322 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6324 * gimple-expr.h (is_gimple_addressable: Likewise.
6325 * gimple-expr.c (is_gimple_address): Likewise.
6326 * internal-fn.c (expand_call_mem_ref): New function.
6327 (expand_mask_load_optab_fn): Use it.
6328 (expand_mask_store_optab_fn): Likewise.
6330 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6331 Alan Hayward <alan.hayward@arm.com>
6332 David Sherwood <david.sherwood@arm.com>
6334 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6335 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6336 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6337 (cond_umax@var{mode}): Document.
6338 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6339 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6340 (cond_umin_optab, cond_umax_optab): New optabs.
6341 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6342 (COND_IOR, COND_XOR): New internal functions.
6343 * internal-fn.h (get_conditional_internal_fn): Declare.
6344 * internal-fn.c (cond_binary_direct): New macro.
6345 (expand_cond_binary_optab_fn): Likewise.
6346 (direct_cond_binary_optab_supported_p): Likewise.
6347 (get_conditional_internal_fn): New function.
6348 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6349 Cope with reduction statements that are vectorized as calls rather
6351 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6352 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6353 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6354 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6355 (UNSPEC_COND_EOR): New unspecs.
6356 (optab): Add mappings for them.
6357 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6358 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
6365 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6366 * internal-fn.def (WHILE_ULT): New internal function.
6367 * internal-fn.h (direct_internal_fn_supported_p): New override
6368 that takes two types as argument.
6369 * internal-fn.c (while_direct): New macro.
6370 (expand_while_optab_fn): New function.
6371 (convert_optab_supported_p): Likewise.
6372 (direct_while_optab_supported_p): New macro.
6373 * wide-int.h (wi::udiv_ceil): New function.
6374 * tree-vectorizer.h (rgroup_masks): New structure.
6375 (vec_loop_masks): New typedef.
6376 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6378 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6379 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6380 (vect_max_vf): New function.
6381 (slpeel_make_loop_iterate_ntimes): Delete.
6382 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6383 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6384 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6385 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6386 internal-fn.h, stor-layout.h and optabs-query.h.
6387 (vect_set_loop_mask): New function.
6388 (add_preheader_seq): Likewise.
6389 (add_header_seq): Likewise.
6390 (interleave_supported_p): Likewise.
6391 (vect_maybe_permute_loop_masks): Likewise.
6392 (vect_set_loop_masks_directly): Likewise.
6393 (vect_set_loop_condition_masked): Likewise.
6394 (vect_set_loop_condition_unmasked): New function, split out from
6395 slpeel_make_loop_iterate_ntimes.
6396 (slpeel_make_loop_iterate_ntimes): Rename to..
6397 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6398 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6399 (vect_do_peeling): Update call accordingly.
6400 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6402 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6403 mask_compare_type, can_fully_mask_p and fully_masked_p.
6404 (release_vec_loop_masks): New function.
6405 (_loop_vec_info): Use it to free the loop masks.
6406 (can_produce_all_loop_masks_p): New function.
6407 (vect_get_max_nscalars_per_iter): Likewise.
6408 (vect_verify_full_masking): Likewise.
6409 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6410 retries, and free the mask rgroups before retrying. Check loop-wide
6411 reasons for disallowing fully-masked loops. Make the final decision
6412 about whether use a fully-masked loop or not.
6413 (vect_estimate_min_profitable_iters): Do not assume that peeling
6414 for the number of iterations will be needed for fully-masked loops.
6415 (vectorizable_reduction): Disable fully-masked loops.
6416 (vectorizable_live_operation): Likewise.
6417 (vect_halve_mask_nunits): New function.
6418 (vect_double_mask_nunits): Likewise.
6419 (vect_record_loop_mask): Likewise.
6420 (vect_get_loop_mask): Likewise.
6421 (vect_transform_loop): Handle the case in which the final loop
6422 iteration might handle a partial vector. Call vect_set_loop_condition
6423 instead of slpeel_make_loop_iterate_ntimes.
6424 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6425 (check_load_store_masking): New function.
6426 (prepare_load_store_mask): Likewise.
6427 (vectorizable_store): Handle fully-masked loops.
6428 (vectorizable_load): Likewise.
6429 (supportable_widening_operation): Use vect_halve_mask_nunits for
6431 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6432 (vect_gen_while): New function.
6433 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6434 (aarch64_uqdec<mode>): New insn.
6436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6437 Alan Hayward <alan.hayward@arm.com>
6438 David Sherwood <david.sherwood@arm.com>
6440 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6441 (reduc_xor_scal_optab): New optabs.
6442 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6443 (reduc_xor_scal_@var{m}): Document.
6444 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6445 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6447 * fold-const-call.c (fold_const_call): Handle them.
6448 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6449 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6450 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6451 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6452 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6453 (UNSPEC_XORV): New unspecs.
6454 (optab): Add entries for them.
6455 (BITWISEV): New int iterator.
6456 (bit_reduc_op): New int attributes.
6458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6459 Alan Hayward <alan.hayward@arm.com>
6460 David Sherwood <david.sherwood@arm.com>
6462 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6463 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6464 * optabs.def (vec_shl_insert_optab): New optab.
6465 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6466 (duplicate_and_interleave): Likewise.
6467 * tree-vect-loop.c: Include internal-fn.h.
6468 (neutral_op_for_slp_reduction): New function, split out from
6469 get_initial_defs_for_reduction.
6470 (get_initial_def_for_reduction): Handle option 2 for variable-length
6471 vectors by loading the neutral value into a vector and then shifting
6472 the initial value into element 0.
6473 (get_initial_defs_for_reduction): Replace the code argument with
6474 the neutral value calculated by neutral_op_for_slp_reduction.
6475 Use gimple_build_vector for constant-length vectors.
6476 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6477 but the first group_size elements have a neutral value.
6478 Use duplicate_and_interleave otherwise.
6479 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6480 Update call to get_initial_defs_for_reduction. Handle SLP
6481 reductions for variable-length vectors by creating one vector
6482 result for each scalar result, with the elements associated
6483 with other scalar results stubbed out with the neutral value.
6484 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6485 Require IFN_VEC_SHL_INSERT for double reductions on
6486 variable-length vectors, or SLP reductions that have
6487 a neutral value. Require can_duplicate_and_interleave_p
6488 support for variable-length unchained SLP reductions if there
6489 is no neutral value, such as for MIN/MAX reductions. Also require
6490 the number of vector elements to be a multiple of the number of
6491 SLP statements when doing variable-length unchained SLP reductions.
6492 Update call to vect_create_epilog_for_reduction.
6493 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6494 and remove initial values.
6495 (duplicate_and_interleave): Make public.
6496 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6497 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6500 Alan Hayward <alan.hayward@arm.com>
6501 David Sherwood <david.sherwood@arm.com>
6503 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6504 (can_duplicate_and_interleave_p): New function.
6505 (vect_get_and_check_slp_defs): Take the vector of statements
6506 rather than just the current one. Remove excess parentheses.
6507 Restriction rejectinon of vect_constant_def and vect_external_def
6508 for variable-length vectors to boolean types, or types for which
6509 can_duplicate_and_interleave_p is false.
6510 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6511 (duplicate_and_interleave): New function.
6512 (vect_get_constant_vectors): Use gimple_build_vector for
6513 constant-length vectors and suitable variable-length constant
6514 vectors. Use duplicate_and_interleave for other variable-length
6515 vectors. Don't defer the update when inserting new statements.
6517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6518 Alan Hayward <alan.hayward@arm.com>
6519 David Sherwood <david.sherwood@arm.com>
6521 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6522 min_profitable_iters doesn't go negative.
6524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6525 Alan Hayward <alan.hayward@arm.com>
6526 David Sherwood <david.sherwood@arm.com>
6528 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6529 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6530 * optabs.def (vec_mask_load_lanes_optab): New optab.
6531 (vec_mask_store_lanes_optab): Likewise.
6532 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6533 (MASK_STORE_LANES): Likewise.
6534 * internal-fn.c (mask_load_lanes_direct): New macro.
6535 (mask_store_lanes_direct): Likewise.
6536 (expand_mask_load_optab_fn): Handle masked operations.
6537 (expand_mask_load_lanes_optab_fn): New macro.
6538 (expand_mask_store_optab_fn): Handle masked operations.
6539 (expand_mask_store_lanes_optab_fn): New macro.
6540 (direct_mask_load_lanes_optab_supported_p): Likewise.
6541 (direct_mask_store_lanes_optab_supported_p): Likewise.
6542 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6544 (vect_load_lanes_supported): Likewise.
6545 * tree-vect-data-refs.c (strip_conversion): New function.
6546 (can_group_stmts_p): Likewise.
6547 (vect_analyze_data_ref_accesses): Use it instead of checking
6548 for a pair of assignments.
6549 (vect_store_lanes_supported): Take a masked_p parameter.
6550 (vect_load_lanes_supported): Likewise.
6551 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6552 vect_store_lanes_supported and vect_load_lanes_supported.
6553 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6554 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6555 parameter. Don't allow gaps for masked accesses.
6556 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6557 and vect_load_lanes_supported.
6558 (get_load_store_type): Take a masked_p parameter and update
6559 call to get_group_load_store_type.
6560 (vectorizable_store): Update call to get_load_store_type.
6561 Handle IFN_MASK_STORE_LANES.
6562 (vectorizable_load): Update call to get_load_store_type.
6563 Handle IFN_MASK_LOAD_LANES.
6565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6566 Alan Hayward <alan.hayward@arm.com>
6567 David Sherwood <david.sherwood@arm.com>
6569 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6571 * config/aarch64/aarch64-protos.h
6572 (aarch64_sve_struct_memory_operand_p): Declare.
6573 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6574 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6575 (VPRED, vpred): Handle SVE structure modes.
6576 * config/aarch64/constraints.md (Utx): New constraint.
6577 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6578 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6579 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6580 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6581 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6582 structure modes. Split into pieces after RA.
6583 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6584 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6586 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6587 SVE structure modes.
6588 (aarch64_classify_address): Likewise.
6589 (sizetochar): Move earlier in file.
6590 (aarch64_print_operand): Handle SVE register lists.
6591 (aarch64_array_mode): New function.
6592 (aarch64_sve_struct_memory_operand_p): Likewise.
6593 (TARGET_ARRAY_MODE): Redefine.
6595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6596 Alan Hayward <alan.hayward@arm.com>
6597 David Sherwood <david.sherwood@arm.com>
6599 * target.def (array_mode): New target hook.
6600 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6601 * doc/tm.texi: Regenerate.
6602 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6603 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6604 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6606 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6610 Alan Hayward <alan.hayward@arm.com>
6611 David Sherwood <david.sherwood@arm.com>
6613 * fold-const.c (fold_binary_loc): Check the argument types
6614 rather than the result type when testing for a vector operation.
6616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6618 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6619 * doc/tm.texi: Regenerate.
6621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6622 Alan Hayward <alan.hayward@arm.com>
6623 David Sherwood <david.sherwood@arm.com>
6625 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6626 (sve): Document new AArch64 extension.
6627 * doc/md.texi (w): Extend the description of the AArch64
6628 constraint to include SVE vectors.
6629 (Upl, Upa): Document new AArch64 predicate constraints.
6630 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6632 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6633 (msve-vector-bits=): New option.
6634 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6635 SVE when these are disabled.
6636 (sve): New extension.
6637 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6638 modes. Adjust their number of units based on aarch64_sve_vg.
6639 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6640 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6641 aarch64_addr_query_type.
6642 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6643 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6644 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6645 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6646 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6647 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6648 (aarch64_simd_imm_zero_p): Delete.
6649 (aarch64_check_zero_based_sve_index_immediate): Declare.
6650 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6651 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6652 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6653 (aarch64_sve_float_mul_immediate_p): Likewise.
6654 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6656 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6657 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6658 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6659 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6660 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6661 (aarch64_regmode_natural_size): Likewise.
6662 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6663 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6665 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6666 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6667 for VG and the SVE predicate registers.
6668 (V_ALIASES): Add a "z"-prefixed alias.
6669 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6670 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6671 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6672 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6673 (REG_CLASS_NAMES): Add entries for them.
6674 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6675 and the predicate registers.
6676 (aarch64_sve_vg): Declare.
6677 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6678 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6679 (REGMODE_NATURAL_SIZE): Define.
6680 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6682 * config/aarch64/aarch64.c: Include cfgrtl.h.
6683 (simd_immediate_info): Add a constructor for series vectors,
6684 and an associated step field.
6685 (aarch64_sve_vg): New variable.
6686 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6687 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6688 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6689 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6690 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6691 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6692 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6693 (aarch64_get_mask_mode): New functions.
6694 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6695 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6696 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6697 predicate modes and predicate registers. Explicitly restrict
6698 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6699 to store a vector mode if it is recognized by
6700 aarch64_classify_vector_mode.
6701 (aarch64_regmode_natural_size): New function.
6702 (aarch64_hard_regno_caller_save_mode): Return the original mode
6704 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6705 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6706 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6707 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6709 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6710 does not overlap dest if the function is frame-related. Handle
6712 (aarch64_split_add_offset): New function.
6713 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6714 them aarch64_add_offset.
6715 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6716 and update call to aarch64_sub_sp.
6717 (aarch64_add_cfa_expression): New function.
6718 (aarch64_expand_prologue): Pass extra temporary registers to the
6719 functions above. Handle the case in which we need to emit new
6720 DW_CFA_expressions for registers that were originally saved
6721 relative to the stack pointer, but now have to be expressed
6722 relative to the frame pointer.
6723 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6725 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6726 IP0 and IP1 values for SVE frames.
6727 (aarch64_expand_vec_series): New function.
6728 (aarch64_expand_sve_widened_duplicate): Likewise.
6729 (aarch64_expand_sve_const_vector): Likewise.
6730 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6731 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6732 into the register, rather than emitting a SET directly.
6733 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6734 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6735 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6736 (offset_9bit_signed_scaled_p): New functions.
6737 (aarch64_replicate_bitmask_imm): New function.
6738 (aarch64_bitmask_imm): Use it.
6739 (aarch64_cannot_force_const_mem): Reject expressions involving
6740 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6741 (aarch64_classify_index): Handle SVE indices, by requiring
6742 a plain register index with a scale that matches the element size.
6743 (aarch64_classify_address): Handle SVE addresses. Assert that
6744 the mode of the address is VOIDmode or an integer mode.
6745 Update call to aarch64_classify_symbol.
6746 (aarch64_classify_symbolic_expression): Update call to
6747 aarch64_classify_symbol.
6748 (aarch64_const_vec_all_in_range_p): New function.
6749 (aarch64_print_vector_float_operand): Likewise.
6750 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6751 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6752 and the FP immediates 1.0 and 0.5.
6753 (aarch64_print_address_internal): Handle SVE addresses.
6754 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6755 (aarch64_regno_regclass): Handle predicate registers.
6756 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6758 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6759 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6760 (aarch64_convert_sve_vector_bits): New function.
6761 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6762 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6764 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6765 Handle SVE vector and predicate modes. Accept VL-based constants
6766 that need only one temporary register, and VL offsets that require
6767 no temporary registers.
6768 (aarch64_conditional_register_usage): Mark the predicate registers
6769 as fixed if SVE isn't available.
6770 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6771 Return true for SVE vector and predicate modes.
6772 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6773 rather than an unsigned int. Handle SVE modes.
6774 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6776 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6778 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6779 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6780 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6781 (aarch64_sve_float_mul_immediate_p): New functions.
6782 (aarch64_sve_valid_immediate): New function.
6783 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6784 Explicitly reject structure modes. Check for INDEX constants.
6785 Handle PTRUE and PFALSE constants.
6786 (aarch64_check_zero_based_sve_index_immediate): New function.
6787 (aarch64_simd_imm_zero_p): Delete.
6788 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6789 vector modes. Accept constants in the range of CNT[BHWD].
6790 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6791 ask for an Advanced SIMD mode.
6792 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6793 (aarch64_simd_vector_alignment): Handle SVE predicates.
6794 (aarch64_vectorize_preferred_vector_alignment): New function.
6795 (aarch64_simd_vector_alignment_reachable): Use it instead of
6797 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6798 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6800 (MAX_VECT_LEN): Delete.
6801 (expand_vec_perm_d): Add a vec_flags field.
6802 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6803 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6804 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6806 (aarch64_evpc_rev): Rename to...
6807 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6808 (aarch64_evpc_rev_global): New function.
6809 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6810 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6812 (aarch64_evpc_sve_tbl): New function.
6813 (aarch64_expand_vec_perm_const_1): Update after rename of
6814 aarch64_evpc_rev. Handle SVE permutes too, trying
6815 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6816 than aarch64_evpc_tbl.
6817 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6818 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6819 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6820 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6821 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6822 (aarch64_expand_sve_vcond): New functions.
6823 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6824 of aarch64_vector_mode_p.
6825 (aarch64_dwarf_poly_indeterminate_value): New function.
6826 (aarch64_compute_pressure_classes): Likewise.
6827 (aarch64_can_change_mode_class): Likewise.
6828 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6829 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6830 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6831 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6832 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6833 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6834 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6835 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6837 (Dn, Dl, Dr): Accept const as well as const_vector.
6838 (Dz): Likewise. Compare against CONST0_RTX.
6839 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6840 of "vector" where appropriate.
6841 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6842 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6843 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6844 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6845 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6846 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6847 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6848 (v_int_equiv): Extend to SVE modes.
6849 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6851 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6852 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6853 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6854 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6855 (SVE_COND_FP_CMP): New int iterators.
6856 (perm_hilo): Handle the new unpack unspecs.
6857 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6859 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6860 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6861 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6862 (aarch64_equality_operator, aarch64_constant_vector_operand)
6863 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6864 (aarch64_sve_nonimmediate_operand): Likewise.
6865 (aarch64_sve_general_operand): Likewise.
6866 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6867 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6868 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6869 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6870 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6871 (aarch64_sve_float_arith_immediate): Likewise.
6872 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6873 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6874 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6875 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6876 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6877 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6878 (aarch64_sve_float_arith_operand): Likewise.
6879 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6880 (aarch64_sve_float_mul_operand): Likewise.
6881 (aarch64_sve_vec_perm_operand): Likewise.
6882 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6883 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6884 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6885 as well as const_vector.
6886 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6887 in file. Use CONST0_RTX and CONSTM1_RTX.
6888 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6889 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6890 Use aarch64_simd_imm_zero.
6891 * config/aarch64/aarch64-sve.md: New file.
6892 * config/aarch64/aarch64.md: Include it.
6893 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6894 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6895 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6896 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6897 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6898 (sve): New attribute.
6899 (enabled): Disable instructions with the sve attribute unless
6901 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6902 aarch64_expand_mov_immediate.
6903 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6904 CNT[BHSD] immediates.
6905 (movti): Split CONST_POLY_INT moves into two halves.
6906 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6907 Split additions that need a temporary here if the destination
6908 is the stack pointer.
6909 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6910 (*add<mode>3_poly_1): New instruction.
6911 (set_clobber_cc): New expander.
6913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6915 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6916 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6917 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6918 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6919 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6920 Change innermode from fixed_mode_size to machine_mode.
6921 (simplify_subreg): Update call accordingly. Handle a constant-sized
6922 subreg of a variable-length CONST_VECTOR.
6924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6925 Alan Hayward <alan.hayward@arm.com>
6926 David Sherwood <david.sherwood@arm.com>
6928 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6929 (add_offset_to_base): New function, split out from...
6930 (create_mem_ref): ...here. When handling a scale other than 1,
6931 check first whether the address is valid without the offset.
6932 Add it into the base if so, leaving the index and scale as-is.
6934 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6937 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6938 fold_for_warn before checking if arg2 is INTEGER_CST.
6940 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6942 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6943 (store_multiple_operation): Delete.
6944 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6945 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6946 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6947 guarded by TARGET_STRING.
6948 (rs6000_output_load_multiple): Delete.
6949 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6950 OPTION_MASK_STRING / TARGET_STRING handling.
6951 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6952 (const rs6000_opt_masks) <"string">: Change mask to 0.
6953 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6954 (MASK_STRING): Delete.
6955 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6957 (load_multiple): Delete.
6964 (store_multiple): Delete.
6971 (movmemsi_8reg): Delete.
6972 (corresponding unnamed define_insn): Delete.
6973 (movmemsi_6reg): Delete.
6974 (corresponding unnamed define_insn): Delete.
6975 (movmemsi_4reg): Delete.
6976 (corresponding unnamed define_insn): Delete.
6977 (movmemsi_2reg): Delete.
6978 (corresponding unnamed define_insn): Delete.
6979 (movmemsi_1reg): Delete.
6980 (corresponding unnamed define_insn): Delete.
6981 * config/rs6000/rs6000.opt (mno-string): New.
6982 (mstring): Replace by deprecation warning stub.
6983 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6985 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6987 * regrename.c (regrename_do_replace): If replacing the same
6988 reg multiple times, try to reuse last created gen_raw_REG.
6991 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6992 main to workaround a bug in GDB.
6994 2018-01-12 Tom de Vries <tom@codesourcery.com>
6997 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6999 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7001 PR rtl-optimization/80481
7002 * ira-color.c (get_cap_member): New function.
7003 (allocnos_conflict_by_live_ranges_p): Use it.
7004 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7005 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7007 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7010 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7011 (*saddl_se_1): Ditto.
7013 (*ssubl_se_1): Ditto.
7015 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7017 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7018 rather than wi::to_widest for DR_INITs.
7019 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7020 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7021 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7023 (vect_analyze_group_access_1): Note that here.
7025 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7027 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7028 polynomial type sizes.
7030 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7032 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7033 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7034 (gimple_add_tmp_var): Likewise.
7036 2018-01-12 Martin Liska <mliska@suse.cz>
7038 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7039 (gimple_alloc_sizes): Likewise.
7040 (dump_gimple_statistics): Use PRIu64 in printf format.
7041 * gimple.h: Change uint64_t to int.
7043 2018-01-12 Martin Liska <mliska@suse.cz>
7045 * tree-core.h: Use uint64_t instead of int.
7046 * tree.c (tree_node_counts): Likewise.
7047 (tree_node_sizes): Likewise.
7048 (dump_tree_statistics): Use PRIu64 in printf format.
7050 2018-01-12 Martin Liska <mliska@suse.cz>
7052 * Makefile.in: As qsort_chk is implemented in vec.c, add
7053 vec.o to linkage of gencfn-macros.
7054 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7055 passing the info to record_node_allocation_statistics.
7056 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7058 * ggc-common.c (struct ggc_usage): Add operator== and use
7059 it in operator< and compare function.
7060 * mem-stats.h (struct mem_usage): Likewise.
7061 * vec.c (struct vec_usage): Remove operator< and compare
7062 function. Can be simply inherited.
7064 2018-01-12 Martin Jambor <mjambor@suse.cz>
7067 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7068 * tree-ssa-math-opts.c: Include domwalk.h.
7069 (convert_mult_to_fma_1): New function.
7070 (fma_transformation_info): New type.
7071 (fma_deferring_state): Likewise.
7072 (cancel_fma_deferring): New function.
7073 (result_of_phi): Likewise.
7074 (last_fma_candidate_feeds_initial_phi): Likewise.
7075 (convert_mult_to_fma): Added deferring logic, split actual
7076 transformation to convert_mult_to_fma_1.
7077 (math_opts_dom_walker): New type.
7078 (math_opts_dom_walker::after_dom_children): New method, body moved
7079 here from pass_optimize_widening_mul::execute, added deferring logic
7081 (pass_optimize_widening_mul::execute): Moved most of code to
7082 math_opts_dom_walker::after_dom_children.
7083 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7084 * config/i386/i386.c (ix86_option_override_internal): Added
7085 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7087 2018-01-12 Richard Biener <rguenther@suse.de>
7090 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7091 inline instance vars.
7093 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7096 * config/rx/rx.c (rx_is_restricted_memory_address):
7099 2018-01-12 Richard Biener <rguenther@suse.de>
7101 PR tree-optimization/80846
7102 * target.def (split_reduction): New target hook.
7103 * targhooks.c (default_split_reduction): New function.
7104 * targhooks.h (default_split_reduction): Declare.
7105 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7106 target requests first reduce vectors by combining low and high
7108 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7109 (get_vectype_for_scalar_type_and_size): Export.
7110 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7111 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7112 * doc/tm.texi: Regenerate.
7113 * config/i386/i386.c (ix86_split_reduction): Implement
7114 TARGET_VECTORIZE_SPLIT_REDUCTION.
7116 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7119 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7120 in PIC mode except for TARGET_VXWORKS_RTP.
7121 * config/sparc/sparc.c: Include cfgrtl.h.
7122 (TARGET_INIT_PIC_REG): Define.
7123 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7124 (sparc_pic_register_p): New predicate.
7125 (sparc_legitimate_address_p): Use it.
7126 (sparc_legitimize_pic_address): Likewise.
7127 (sparc_delegitimize_address): Likewise.
7128 (sparc_mode_dependent_address_p): Likewise.
7129 (gen_load_pcrel_sym): Remove 4th parameter.
7130 (load_got_register): Adjust call to above. Remove obsolete stuff.
7131 (sparc_expand_prologue): Do not call load_got_register here.
7132 (sparc_flat_expand_prologue): Likewise.
7133 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7134 (sparc_use_pseudo_pic_reg): New function.
7135 (sparc_init_pic_reg): Likewise.
7136 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7137 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7139 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7141 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7142 Add item for branch_cost.
7144 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7146 PR rtl-optimization/83565
7147 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7148 not extend the result to a larger mode for rotate operations.
7149 (num_sign_bit_copies1): Likewise.
7151 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7154 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7156 Use values-Xc.o for -pedantic.
7157 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7159 2018-01-12 Martin Liska <mliska@suse.cz>
7162 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7164 (possible_polymorphic_call_targets): Use it.
7165 (ipa_devirt): Likewise.
7167 2018-01-12 Martin Liska <mliska@suse.cz>
7169 * profile-count.h (enum profile_quality): Use 0 as invalid
7170 enum value of profile_quality.
7172 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7174 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7175 -mext-string options.
7177 2018-01-12 Richard Biener <rguenther@suse.de>
7179 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7180 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7181 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7183 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7185 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7187 * configure.ac (--with-long-double-format): Add support for the
7188 configuration option to change the default long double format on
7190 * config.gcc (powerpc*-linux*-*): Likewise.
7191 * configure: Regenerate.
7192 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7193 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7194 used without modification.
7196 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7198 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7199 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7200 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7201 MISC_BUILTIN_SPEC_BARRIER.
7202 (rs6000_init_builtins): Likewise.
7203 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7205 (speculation_barrier): New define_insn.
7206 * doc/extend.texi: Document __builtin_speculation_barrier.
7208 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7211 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7212 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7213 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7215 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7216 integral modes instead of "ss" and "sd".
7217 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7218 vectors with 32-bit and 64-bit elements.
7219 (vecdupssescalarmodesuffix): New mode attribute.
7220 (vec_dup<mode>): Use it.
7222 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7225 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7226 frame if argument is passed on stack.
7228 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7231 * ree.c (combine_reaching_defs): Optimize also
7232 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7233 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7235 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7238 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7240 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7243 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7244 after they are computed.
7246 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7248 PR tree-optimization/83695
7249 * gimple-loop-linterchange.cc
7250 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7251 reset cached scev information after interchange.
7252 (pass_linterchange::execute): Remove call to scev_reset_htab.
7254 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7256 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7257 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7258 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7259 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7260 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7261 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7262 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7263 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7264 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7265 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7266 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7267 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7268 (V_lane_reg): Likewise.
7269 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7271 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7272 (vfmal_lane_low<mode>_intrinsic,
7273 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7274 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7275 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7276 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7277 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7278 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7280 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7282 * config/arm/arm-cpus.in (fp16fml): New feature.
7283 (ALL_SIMD): Add fp16fml.
7284 (armv8.2-a): Add fp16fml as an option.
7285 (armv8.3-a): Likewise.
7286 (armv8.4-a): Add fp16fml as part of fp16.
7287 * config/arm/arm.h (TARGET_FP16FML): Define.
7288 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7290 * config/arm/arm-modes.def (V2HF): Define.
7291 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7292 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7293 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7294 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7295 vfmsl_low, vfmsl_high): New set of builtins.
7296 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7297 (vfml_op): New code attribute.
7298 (VFMLHALVES): New int iterator.
7299 (VFML, VFMLSEL): New mode attributes.
7300 (V_reg): Define mapping for V2HF.
7301 (V_hi, V_lo): New mode attributes.
7302 (VF_constraint): Likewise.
7303 (vfml_half, vfml_half_selector): New int attributes.
7304 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7306 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7307 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7309 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7310 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7311 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7312 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7314 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7315 Document new effective target and option set.
7317 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7319 * config/arm/arm-cpus.in (armv8_4): New feature.
7320 (ARMv8_4a): New fgroup.
7321 (armv8.4-a): New arch.
7322 * config/arm/arm-tables.opt: Regenerate.
7323 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7324 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7325 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7326 Add matching rules for -march=armv8.4-a and extensions.
7327 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7329 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7332 * config/rx/rx.md (BW): New mode attribute.
7333 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7335 2018-01-11 Richard Biener <rguenther@suse.de>
7337 PR tree-optimization/83435
7338 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7339 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7340 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7342 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7343 Alan Hayward <alan.hayward@arm.com>
7344 David Sherwood <david.sherwood@arm.com>
7346 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7348 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7349 (aarch64_print_address_internal): Use it to check for a zero offset.
7351 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7352 Alan Hayward <alan.hayward@arm.com>
7353 David Sherwood <david.sherwood@arm.com>
7355 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7356 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7357 Return a poly_int64 rather than a HOST_WIDE_INT.
7358 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7359 rather than a HOST_WIDE_INT.
7360 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7361 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7362 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7363 final_offset from HOST_WIDE_INT to poly_int64.
7364 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7365 to_constant when getting the number of units in an Advanced SIMD
7367 (aarch64_builtin_vectorized_function): Check for a constant number
7369 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7371 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7372 attribute instead of GET_MODE_NUNITS.
7373 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7374 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7375 GET_MODE_SIZE for fixed-size registers.
7376 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7377 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7378 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7379 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7380 (aarch64_print_operand, aarch64_print_address_internal)
7381 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7382 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7383 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7384 Handle polynomial GET_MODE_SIZE.
7385 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7386 wider than SImode without modification.
7387 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7388 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7389 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7390 passing and returning SVE modes.
7391 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7392 rather than GEN_INT.
7393 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7394 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7395 (aarch64_allocate_and_probe_stack_space): Likewise.
7396 (aarch64_layout_frame): Cope with polynomial offsets.
7397 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7398 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7400 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7401 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7402 poly_int64 rather than a HOST_WIDE_INT.
7403 (aarch64_get_separate_components, aarch64_process_components)
7404 (aarch64_expand_prologue, aarch64_expand_epilogue)
7405 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7406 (aarch64_anchor_offset): New function, split out from...
7407 (aarch64_legitimize_address): ...here.
7408 (aarch64_builtin_vectorization_cost): Handle polynomial
7409 TYPE_VECTOR_SUBPARTS.
7410 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7412 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7413 number of elements from the PARALLEL rather than the mode.
7414 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7415 rather than GET_MODE_BITSIZE.
7416 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7417 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7418 (aarch64_expand_vec_perm_const_1): Handle polynomial
7419 d->perm.length () and d->perm elements.
7420 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7421 Apply to_constant to d->perm elements.
7422 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7423 polynomial CONST_VECTOR_NUNITS.
7424 (aarch64_move_pointer): Take amount as a poly_int64 rather
7426 (aarch64_progress_pointer): Avoid temporary variable.
7427 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7428 the mode attribute instead of GET_MODE.
7430 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7431 Alan Hayward <alan.hayward@arm.com>
7432 David Sherwood <david.sherwood@arm.com>
7434 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7435 x exists before using it.
7436 (aarch64_add_constant_internal): Rename to...
7437 (aarch64_add_offset_1): ...this. Replace regnum with separate
7438 src and dest rtxes. Handle the case in which they're different,
7439 including when the offset is zero. Replace scratchreg with an rtx.
7440 Use 2 additions if there is no spare register into which we can
7441 move a 16-bit constant.
7442 (aarch64_add_constant): Delete.
7443 (aarch64_add_offset): Replace reg with separate src and dest
7444 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7445 Use aarch64_add_offset_1.
7446 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7447 an rtx rather than an int. Take the delta as a poly_int64
7448 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7449 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7450 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7451 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7452 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7454 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7455 aarch64_add_constant.
7457 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7459 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7460 Use scalar_float_mode.
7462 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7464 * config/aarch64/aarch64-simd.md
7465 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7466 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7467 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7468 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7469 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7470 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7471 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7472 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7473 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7474 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7476 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7479 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7480 targ_options->x_arm_arch_string is non NULL.
7482 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7484 * config/aarch64/aarch64.h
7485 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7487 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7490 * expmed.c (emit_store_flag_force): Swap if const op0
7491 and change VOIDmode to mode of op0.
7493 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7495 PR rtl-optimization/83761
7496 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7497 than bytes to mode_for_size.
7499 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7502 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7503 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7506 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7509 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7510 when in layout mode.
7511 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7512 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7515 2018-01-10 Michael Collison <michael.collison@arm.com>
7517 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7518 * config/aarch64/aarch64-option-extension.def: Add
7519 AARCH64_OPT_EXTENSION of 'fp16fml'.
7520 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7521 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7522 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7523 * config/aarch64/constraints.md (Ui7): New constraint.
7524 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7525 (VFMLA_SEL_W): Ditto.
7528 (VFMLA16_LOW): New int iterator.
7529 (VFMLA16_HIGH): Ditto.
7530 (UNSPEC_FMLAL): New unspec.
7531 (UNSPEC_FMLSL): Ditto.
7532 (UNSPEC_FMLAL2): Ditto.
7533 (UNSPEC_FMLSL2): Ditto.
7534 (f16mac): New code attribute.
7535 * config/aarch64/aarch64-simd-builtins.def
7536 (aarch64_fmlal_lowv2sf): Ditto.
7537 (aarch64_fmlsl_lowv2sf): Ditto.
7538 (aarch64_fmlalq_lowv4sf): Ditto.
7539 (aarch64_fmlslq_lowv4sf): Ditto.
7540 (aarch64_fmlal_highv2sf): Ditto.
7541 (aarch64_fmlsl_highv2sf): Ditto.
7542 (aarch64_fmlalq_highv4sf): Ditto.
7543 (aarch64_fmlslq_highv4sf): Ditto.
7544 (aarch64_fmlal_lane_lowv2sf): Ditto.
7545 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7546 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7547 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7548 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7549 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7550 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7551 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7552 (aarch64_fmlal_lane_highv2sf): Ditto.
7553 (aarch64_fmlsl_lane_highv2sf): Ditto.
7554 (aarch64_fmlal_laneq_highv2sf): Ditto.
7555 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7556 (aarch64_fmlalq_lane_highv4sf): Ditto.
7557 (aarch64_fmlsl_lane_highv4sf): Ditto.
7558 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7559 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7560 * config/aarch64/aarch64-simd.md:
7561 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7562 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7563 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7564 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7565 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7566 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7567 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7568 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7569 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7570 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7571 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7572 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7573 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7574 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7575 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7576 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7577 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7578 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7579 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7580 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7581 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7582 (vfmlsl_low_u32): Ditto.
7583 (vfmlalq_low_u32): Ditto.
7584 (vfmlslq_low_u32): Ditto.
7585 (vfmlal_high_u32): Ditto.
7586 (vfmlsl_high_u32): Ditto.
7587 (vfmlalq_high_u32): Ditto.
7588 (vfmlslq_high_u32): Ditto.
7589 (vfmlal_lane_low_u32): Ditto.
7590 (vfmlsl_lane_low_u32): Ditto.
7591 (vfmlal_laneq_low_u32): Ditto.
7592 (vfmlsl_laneq_low_u32): Ditto.
7593 (vfmlalq_lane_low_u32): Ditto.
7594 (vfmlslq_lane_low_u32): Ditto.
7595 (vfmlalq_laneq_low_u32): Ditto.
7596 (vfmlslq_laneq_low_u32): Ditto.
7597 (vfmlal_lane_high_u32): Ditto.
7598 (vfmlsl_lane_high_u32): Ditto.
7599 (vfmlal_laneq_high_u32): Ditto.
7600 (vfmlsl_laneq_high_u32): Ditto.
7601 (vfmlalq_lane_high_u32): Ditto.
7602 (vfmlslq_lane_high_u32): Ditto.
7603 (vfmlalq_laneq_high_u32): Ditto.
7604 (vfmlslq_laneq_high_u32): Ditto.
7605 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7606 (AARCH64_FL_FOR_ARCH8_4): New.
7607 (AARCH64_ISA_F16FML): New ISA flag.
7608 (TARGET_F16FML): New feature flag for fp16fml.
7609 (doc/invoke.texi): Document new fp16fml option.
7611 2018-01-10 Michael Collison <michael.collison@arm.com>
7613 * config/aarch64/aarch64-builtins.c:
7614 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7615 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7616 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7617 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7618 (AARCH64_ISA_SHA3): New ISA flag.
7619 (TARGET_SHA3): New feature flag for sha3.
7620 * config/aarch64/iterators.md (sha512_op): New int attribute.
7621 (CRYPTO_SHA512): New int iterator.
7622 (UNSPEC_SHA512H): New unspec.
7623 (UNSPEC_SHA512H2): Ditto.
7624 (UNSPEC_SHA512SU0): Ditto.
7625 (UNSPEC_SHA512SU1): Ditto.
7626 * config/aarch64/aarch64-simd-builtins.def
7627 (aarch64_crypto_sha512hqv2di): New builtin.
7628 (aarch64_crypto_sha512h2qv2di): Ditto.
7629 (aarch64_crypto_sha512su0qv2di): Ditto.
7630 (aarch64_crypto_sha512su1qv2di): Ditto.
7631 (aarch64_eor3qv8hi): Ditto.
7632 (aarch64_rax1qv2di): Ditto.
7633 (aarch64_xarqv2di): Ditto.
7634 (aarch64_bcaxqv8hi): Ditto.
7635 * config/aarch64/aarch64-simd.md:
7636 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7637 (aarch64_crypto_sha512su0qv2di): Ditto.
7638 (aarch64_crypto_sha512su1qv2di): Ditto.
7639 (aarch64_eor3qv8hi): Ditto.
7640 (aarch64_rax1qv2di): Ditto.
7641 (aarch64_xarqv2di): Ditto.
7642 (aarch64_bcaxqv8hi): Ditto.
7643 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7644 (vsha512h2q_u64): Ditto.
7645 (vsha512su0q_u64): Ditto.
7646 (vsha512su1q_u64): Ditto.
7647 (veor3q_u16): Ditto.
7648 (vrax1q_u64): Ditto.
7650 (vbcaxq_u16): Ditto.
7651 * config/arm/types.md (crypto_sha512): New type attribute.
7652 (crypto_sha3): Ditto.
7653 (doc/invoke.texi): Document new sha3 option.
7655 2018-01-10 Michael Collison <michael.collison@arm.com>
7657 * config/aarch64/aarch64-builtins.c:
7658 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7659 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7660 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7661 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7662 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7663 (AARCH64_ISA_SM4): New ISA flag.
7664 (TARGET_SM4): New feature flag for sm4.
7665 * config/aarch64/aarch64-simd-builtins.def
7666 (aarch64_sm3ss1qv4si): Ditto.
7667 (aarch64_sm3tt1aq4si): Ditto.
7668 (aarch64_sm3tt1bq4si): Ditto.
7669 (aarch64_sm3tt2aq4si): Ditto.
7670 (aarch64_sm3tt2bq4si): Ditto.
7671 (aarch64_sm3partw1qv4si): Ditto.
7672 (aarch64_sm3partw2qv4si): Ditto.
7673 (aarch64_sm4eqv4si): Ditto.
7674 (aarch64_sm4ekeyqv4si): Ditto.
7675 * config/aarch64/aarch64-simd.md:
7676 (aarch64_sm3ss1qv4si): Ditto.
7677 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7678 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7679 (aarch64_sm4eqv4si): Ditto.
7680 (aarch64_sm4ekeyqv4si): Ditto.
7681 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7682 (sm3part_op): Ditto.
7683 (CRYPTO_SM3TT): Ditto.
7684 (CRYPTO_SM3PART): Ditto.
7685 (UNSPEC_SM3SS1): New unspec.
7686 (UNSPEC_SM3TT1A): Ditto.
7687 (UNSPEC_SM3TT1B): Ditto.
7688 (UNSPEC_SM3TT2A): Ditto.
7689 (UNSPEC_SM3TT2B): Ditto.
7690 (UNSPEC_SM3PARTW1): Ditto.
7691 (UNSPEC_SM3PARTW2): Ditto.
7692 (UNSPEC_SM4E): Ditto.
7693 (UNSPEC_SM4EKEY): Ditto.
7694 * config/aarch64/constraints.md (Ui2): New constraint.
7695 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7696 * config/arm/types.md (crypto_sm3): New type attribute.
7697 (crypto_sm4): Ditto.
7698 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7699 (vsm3tt1aq_u32): Ditto.
7700 (vsm3tt1bq_u32): Ditto.
7701 (vsm3tt2aq_u32): Ditto.
7702 (vsm3tt2bq_u32): Ditto.
7703 (vsm3partw1q_u32): Ditto.
7704 (vsm3partw2q_u32): Ditto.
7705 (vsm4eq_u32): Ditto.
7706 (vsm4ekeyq_u32): Ditto.
7707 (doc/invoke.texi): Document new sm4 option.
7709 2018-01-10 Michael Collison <michael.collison@arm.com>
7711 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7712 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7713 (AARCH64_FL_FOR_ARCH8_4): New.
7714 (AARCH64_FL_V8_4): New flag.
7715 (doc/invoke.texi): Document new armv8.4-a option.
7717 2018-01-10 Michael Collison <michael.collison@arm.com>
7719 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7720 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7721 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7722 * config/aarch64/aarch64-option-extension.def: Add
7723 AARCH64_OPT_EXTENSION of 'sha2'.
7724 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7725 (crypto): Disable sha2 and aes if crypto disabled.
7726 (crypto): Enable aes and sha2 if enabled.
7727 (simd): Disable sha2 and aes if simd disabled.
7728 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7730 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7731 (TARGET_SHA2): New feature flag for sha2.
7732 (TARGET_AES): New feature flag for aes.
7733 * config/aarch64/aarch64-simd.md:
7734 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7735 conditional on TARGET_AES.
7736 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7737 (aarch64_crypto_sha1hsi): Make pattern conditional
7739 (aarch64_crypto_sha1hv4si): Ditto.
7740 (aarch64_be_crypto_sha1hv4si): Ditto.
7741 (aarch64_crypto_sha1su1v4si): Ditto.
7742 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7743 (aarch64_crypto_sha1su0v4si): Ditto.
7744 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7745 (aarch64_crypto_sha256su0v4si): Ditto.
7746 (aarch64_crypto_sha256su1v4si): Ditto.
7747 (doc/invoke.texi): Document new aes and sha2 options.
7749 2018-01-10 Martin Sebor <msebor@redhat.com>
7751 PR tree-optimization/83781
7752 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7755 2018-01-11 Martin Sebor <msebor@gmail.com>
7756 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7758 PR tree-optimization/83501
7759 PR tree-optimization/81703
7761 * tree-ssa-strlen.c (get_string_cst): Rename...
7762 (get_string_len): ...to this. Handle global constants.
7763 (handle_char_store): Adjust.
7765 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7766 Jim Wilson <jimw@sifive.com>
7768 * config/riscv/riscv-protos.h (riscv_output_return): New.
7769 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7770 (riscv_attribute_table, riscv_output_return),
7771 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7772 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7773 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7775 (riscv_expand_prologue): Add early return for naked function.
7776 (riscv_expand_epilogue): Likewise.
7777 (riscv_function_ok_for_sibcall): Return false for naked function.
7778 (riscv_set_current_function): New.
7779 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7780 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7781 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7782 * doc/extend.texi (RISC-V Function Attributes): New.
7784 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7786 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7787 check for 128-bit long double before checking TCmode.
7788 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7789 128-bit long doubles before checking TFmode or TCmode.
7790 (FLOAT128_IBM_P): Likewise.
7792 2018-01-10 Martin Sebor <msebor@redhat.com>
7794 PR tree-optimization/83671
7795 * builtins.c (c_strlen): Unconditionally return zero for the empty
7797 Use -Warray-bounds for warnings.
7798 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7799 for non-constant array indices with COMPONENT_REF, arrays of
7800 arrays, and pointers to arrays.
7801 (gimple_fold_builtin_strlen): Determine and set length range for
7802 non-constant character arrays.
7804 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7807 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7810 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7812 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7814 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7817 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7818 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7819 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7820 indexed_or_indirect_operand predicate.
7821 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7822 (*vsx_le_perm_load_v8hi): Likewise.
7823 (*vsx_le_perm_load_v16qi): Likewise.
7824 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7825 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7826 (*vsx_le_perm_store_v8hi): Likewise.
7827 (*vsx_le_perm_store_v16qi): Likewise.
7828 (eight unnamed splitters): Likewise.
7830 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7832 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7833 * config/rs6000/emmintrin.h: Likewise.
7834 * config/rs6000/mmintrin.h: Likewise.
7835 * config/rs6000/xmmintrin.h: Likewise.
7837 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7840 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7842 * tree.c (tree_nop_conversion): Return true for location wrapper
7844 (maybe_wrap_with_location): New function.
7845 (selftest::check_strip_nops): New function.
7846 (selftest::test_location_wrappers): New function.
7847 (selftest::tree_c_tests): Call it.
7848 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7849 (maybe_wrap_with_location): New decl.
7850 (EXPR_LOCATION_WRAPPER_P): New macro.
7851 (location_wrapper_p): New inline function.
7852 (tree_strip_any_location_wrapper): New inline function.
7854 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7857 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7858 stack_realign_offset for the largest alignment of stack slot
7860 (ix86_find_max_used_stack_alignment): New function.
7861 (ix86_finalize_stack_frame_flags): Use it. Set
7862 max_used_stack_alignment if we don't realign stack.
7863 * config/i386/i386.h (machine_function): Add
7864 max_used_stack_alignment.
7866 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7868 * config/arm/arm.opt (-mbranch-cost): New option.
7869 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7872 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7875 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7876 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7878 2018-01-10 Richard Biener <rguenther@suse.de>
7881 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7882 early out so it also covers the case where we have a non-NULL
7885 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7887 PR tree-optimization/83753
7888 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7889 for non-strided grouped accesses if the number of elements is 1.
7891 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7894 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7895 * i386.h (TARGET_USE_GATHER): Define.
7896 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7898 2018-01-10 Martin Liska <mliska@suse.cz>
7901 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7902 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7904 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7905 CLEANUP_NO_PARTITIONING is not set.
7907 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7909 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7910 for vectors, as a partial revert of r254296.
7911 * rtl.h (const_vec_p): Delete.
7912 (const_vec_duplicate_p): Don't test for vector CONSTs.
7913 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7914 * expmed.c (make_tree): Likewise.
7917 * common.md (E, F): Use CONSTANT_P instead of checking for
7919 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7920 checking for CONST_VECTOR.
7922 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7925 * predict.c (force_edge_cold): Handle in more sane way edges
7928 2018-01-09 Carl Love <cel@us.ibm.com>
7930 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7932 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7933 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7934 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7935 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7936 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7937 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7938 * config/rs6000/rs6000-protos.h: Add extern defition for
7939 rs6000_generate_float2_double_code.
7940 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7942 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7943 (float2_v2df): Add define_expand.
7945 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7948 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7949 op_mode in the force_to_mode call.
7951 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7953 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7954 instead of checking each element individually.
7955 (aarch64_evpc_uzp): Likewise.
7956 (aarch64_evpc_zip): Likewise.
7957 (aarch64_evpc_ext): Likewise.
7958 (aarch64_evpc_rev): Likewise.
7959 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7960 instead of checking each element individually. Return true without
7962 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7963 whether all selected elements come from the same input, instead of
7964 checking each element individually. Remove calls to gen_rtx_REG,
7965 start_sequence and end_sequence and instead assert that no rtl is
7968 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7970 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7971 order of HIGH and CONST checks.
7973 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7975 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7976 if the destination isn't an SSA_NAME.
7978 2018-01-09 Richard Biener <rguenther@suse.de>
7980 PR tree-optimization/83668
7981 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7983 (canonicalize_loop_form): ... here, renamed from ...
7984 (canonicalize_loop_closed_ssa_form): ... this and amended to
7985 swap successor edges for loop exit blocks to make us use
7986 the RPO order we need for initial schedule generation.
7988 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7990 PR tree-optimization/64811
7991 * match.pd: When optimizing comparisons with Inf, avoid
7992 introducing or losing exceptions from comparisons with NaN.
7994 2018-01-09 Martin Liska <mliska@suse.cz>
7997 * asan.c (shadow_mem_size): Add gcc_assert.
7999 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8001 Don't save registers in main().
8004 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8005 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8006 * config/avr/avr.c (avr_set_current_function): Don't error if
8007 naked, OS_task or OS_main are specified at the same time.
8008 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8010 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8012 * common/config/avr/avr-common.c (avr_option_optimization_table):
8013 Switch on -mmain-is-OS_task for optimizing compilations.
8015 2018-01-09 Richard Biener <rguenther@suse.de>
8017 PR tree-optimization/83572
8018 * graphite.c: Include cfganal.h.
8019 (graphite_transform_loops): Connect infinite loops to exit
8020 and remove fake edges at the end.
8022 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8024 * ipa-inline.c (edge_badness): Revert accidental checkin.
8026 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8029 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8030 symbols; not inline clones.
8032 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8035 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8036 hard registers. Formatting fixes.
8038 PR preprocessor/83722
8039 * gcc.c (try_generate_repro): Pass
8040 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8041 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8044 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8045 Kito Cheng <kito.cheng@gmail.com>
8047 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8048 (riscv_leaf_function_p): Delete.
8049 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8051 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8053 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8055 (do_ifelse): New function.
8056 (do_isel): New function.
8057 (do_sub3): New function.
8058 (do_add3): New function.
8059 (do_load_mask_compare): New function.
8060 (do_overlap_load_compare): New function.
8061 (expand_compare_loop): New function.
8062 (expand_block_compare): Call expand_compare_loop() when appropriate.
8063 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8065 (-mblock-compare-inline-loop-limit): New option.
8067 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8070 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8071 Reverse order of second and third operands in first alternative.
8072 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8073 of first and second elements in UNSPEC_VPERMR vector.
8074 (altivec_expand_vec_perm_le): Likewise.
8076 2018-01-08 Jeff Law <law@redhat.com>
8078 PR rtl-optimizatin/81308
8079 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8080 (process_switch): If group_case_labels makes a change, then set
8082 (pass_convert_switch::execute): If a switch is converted, then
8083 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8085 PR rtl-optimization/81308
8086 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8089 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8091 PR target/83663 - Revert r255946
8092 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8093 generation for cases where splatting a value is not useful.
8094 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8095 across a vec_duplicate and a paradoxical subreg forming a vector
8096 mode to a vec_concat.
8098 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8100 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8101 -march=armv8.3-a variants.
8102 * config/arm/t-multilib: Likewise.
8103 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8105 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8107 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8109 (cceq_ior_compare_complement): Give it a name so I can use it, and
8110 change boolean_or_operator predicate to boolean_operator so it can
8111 be used to generate a crand.
8112 (eqne): New code iterator.
8113 (bd/bd_neg): New code_attrs.
8114 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8115 a single define_insn.
8116 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8117 decrement (bdnzt/bdnzf/bdzt/bdzf).
8118 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8119 with the new names of the branch decrement patterns, and added the
8120 names of the branch decrement conditional patterns.
8122 2018-01-08 Richard Biener <rguenther@suse.de>
8124 PR tree-optimization/83563
8125 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8128 2018-01-08 Richard Biener <rguenther@suse.de>
8131 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8133 2018-01-08 Richard Biener <rguenther@suse.de>
8135 PR tree-optimization/83685
8136 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8137 references to abnormals.
8139 2018-01-08 Richard Biener <rguenther@suse.de>
8142 * dwarf2out.c (output_indirect_strings): Handle empty
8143 skeleton_debug_str_hash.
8144 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8146 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8148 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8149 (emit_store_direct): Likewise.
8150 (arc_trampoline_adjust_address): Likewise.
8151 (arc_asm_trampoline_template): New function.
8152 (arc_initialize_trampoline): Use asm_trampoline_template.
8153 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8154 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8155 * config/arc/arc.md (flush_icache): Delete pattern.
8157 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8159 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8160 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8163 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8166 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8167 by not USED_FOR_TARGET.
8168 (make_pass_resolve_sw_modes): Likewise.
8170 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8172 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8175 2018-01-08 Richard Biener <rguenther@suse.de>
8178 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8180 2018-01-08 Richard Biener <rguenther@suse.de>
8183 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8185 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8188 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8189 basic blocks with a small number of successors.
8190 (convert_control_dep_chain_into_preds): Improve handling of
8192 (dump_predicates): Split apart into...
8193 (dump_pred_chain): ...here...
8194 (dump_pred_info): ...and here.
8195 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8196 (can_chain_union_be_invalidated_p): Improve check for invalidation
8198 (uninit_uses_cannot_happen): Avoid unnecessary if
8199 convert_control_dep_chain_into_preds yielded nothing.
8201 2018-01-06 Martin Sebor <msebor@redhat.com>
8203 PR tree-optimization/83640
8204 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8205 subtracting negative offset from size.
8206 (builtin_access::overlap): Adjust offset bounds of the access to fall
8207 within the size of the object if possible.
8209 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8211 PR rtl-optimization/83699
8212 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8213 extract_bit_field_as_subreg to cases in which the extracted
8214 value is also a vector.
8216 * lra-constraints.c (process_alt_operands): Test for the equivalence
8217 substitutions when detecting a possible reload cycle.
8219 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8222 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8223 by default if flag_selective_schedling{,2}. Formatting fixes.
8225 PR rtl-optimization/83682
8226 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8227 if it has non-VECTOR_MODE element mode.
8228 (vec_duplicate_p): Likewise.
8231 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8232 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8234 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8237 * config/i386/i386-builtin.def
8238 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8239 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8240 Require also OPTION_MASK_ISA_AVX512F in addition to
8241 OPTION_MASK_ISA_GFNI.
8242 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8243 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8244 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8245 to OPTION_MASK_ISA_GFNI.
8246 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8247 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8248 OPTION_MASK_ISA_AVX512BW.
8249 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8250 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8251 addition to OPTION_MASK_ISA_GFNI.
8252 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8253 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8254 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8255 to OPTION_MASK_ISA_GFNI.
8256 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8257 a requirement for all ISAs rather than any of them with a few
8259 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8261 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8262 bitmasks to be enabled with 3 exceptions, instead of requiring any
8263 enabled ISA with lots of exceptions.
8264 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8265 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8266 Change avx512bw in isa attribute to avx512f.
8267 * config/i386/sgxintrin.h: Add license boilerplate.
8268 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8269 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8270 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8271 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8273 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8274 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8275 temporarily sse2 rather than sse if not enabled already.
8278 * config/i386/sse.md (VI248_VLBW): Rename to ...
8279 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8280 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8281 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8282 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8283 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8284 mode iterator instead of VI248_VLBW.
8286 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8288 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8289 (record_modified): Skip clobbers; add debug output.
8290 (param_change_prob): Use sreal frequencies.
8292 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8294 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8295 punt for user-aligned variables.
8297 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8299 * tree-chrec.c (chrec_contains_symbols): Return true for
8302 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8305 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8306 of (x|y) == x for BICS pattern.
8308 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8310 PR tree-optimization/83605
8311 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8312 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8315 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8317 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8318 * config/epiphany/rtems.h: New file.
8320 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8321 Uros Bizjak <ubizjak@gmail.com>
8324 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8325 QIreg_operand instead of register_operand predicate.
8326 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8327 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8328 comments instead of -fmitigate[-_]rop.
8330 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8333 * cgraphunit.c (symbol_table::compile): Switch to text_section
8334 before calling assembly_start debug hook.
8335 * run-rtl-passes.c (run_rtl_passes): Likewise.
8338 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8340 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8341 range_int_cst_p rather than !symbolic_range_p before calling
8342 extract_range_from_multiplicative_op_1.
8344 2018-01-04 Jeff Law <law@redhat.com>
8346 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8347 redundant test in assertion.
8349 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8351 * doc/rtl.texi: Document machine_mode wrapper classes.
8353 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8355 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8358 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8360 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8361 the VEC_PERM_EXPR fold to fail.
8363 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8366 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8367 to switched_sections.
8369 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8372 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8375 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8378 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8379 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8381 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8384 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8385 is BLKmode and bitpos not zero or mode change is needed.
8387 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8390 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8393 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8396 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8397 instead of MULT rtx. Update all corresponding splitters.
8399 (*ssub<modesuffix>): Ditto.
8401 (*cmp_sadd_di): Update split patterns.
8402 (*cmp_sadd_si): Ditto.
8403 (*cmp_sadd_sidi): Ditto.
8404 (*cmp_ssub_di): Ditto.
8405 (*cmp_ssub_si): Ditto.
8406 (*cmp_ssub_sidi): Ditto.
8407 * config/alpha/predicates.md (const23_operand): New predicate.
8408 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8409 Look for ASHIFT, not MULT inner operand.
8410 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8412 2018-01-04 Martin Liska <mliska@suse.cz>
8414 PR gcov-profile/83669
8415 * gcov.c (output_intermediate_file): Add version to intermediate
8417 * doc/gcov.texi: Document new field 'version' in intermediate
8418 file format. Fix location of '-k' option of gcov command.
8420 2018-01-04 Martin Liska <mliska@suse.cz>
8423 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8425 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8427 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8429 2018-01-03 Martin Sebor <msebor@redhat.com>
8431 PR tree-optimization/83655
8432 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8433 checking calls with invalid arguments.
8435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8437 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8438 (vectorizable_mask_load_store): Delete.
8439 (vectorizable_call): Return false for masked loads and stores.
8440 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8441 instead of gimple_assign_rhs1.
8442 (vectorizable_load): Handle IFN_MASK_LOAD.
8443 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8447 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8449 (vectorizable_mask_load_store): ...here.
8450 (vectorizable_load): ...and here.
8452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8454 * tree-vect-stmts.c (vect_build_all_ones_mask)
8455 (vect_build_zero_merge_argument): New functions, split out from...
8456 (vectorizable_load): ...here.
8458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8460 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8462 (vectorizable_mask_load_store): ...here.
8463 (vectorizable_store): ...and here.
8465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8467 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8469 (vectorizable_mask_load_store): ...here.
8471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8473 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8474 (vect_model_store_cost): Take a vec_load_store_type instead of a
8476 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8477 (vect_model_store_cost): Take a vec_load_store_type instead of a
8479 (vectorizable_mask_load_store): Update accordingly.
8480 (vectorizable_store): Likewise.
8481 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8485 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8486 IFN_MASK_LOAD calls here rather than...
8487 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8490 Alan Hayward <alan.hayward@arm.com>
8491 David Sherwood <david.sherwood@arm.com>
8493 * expmed.c (extract_bit_field_1): For vector extracts,
8494 fall back to extract_bit_field_as_subreg if vec_extract
8497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8498 Alan Hayward <alan.hayward@arm.com>
8499 David Sherwood <david.sherwood@arm.com>
8501 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8502 they are variable or constant sized.
8503 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8504 slots for constant-sized data.
8506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8507 Alan Hayward <alan.hayward@arm.com>
8508 David Sherwood <david.sherwood@arm.com>
8510 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8511 handling COND_EXPRs with boolean comparisons, try to find a better
8512 basis for the mask type than the boolean itself.
8514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8516 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8517 is calculated and how it can be overridden.
8518 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8519 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8521 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8525 Alan Hayward <alan.hayward@arm.com>
8526 David Sherwood <david.sherwood@arm.com>
8528 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8529 Remove the mode argument.
8530 (aarch64_simd_valid_immediate): Remove the mode and inverse
8532 * config/aarch64/iterators.md (bitsize): New iterator.
8533 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8534 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8535 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8536 aarch64_simd_valid_immediate.
8537 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8538 (aarch64_reg_or_bic_imm): Likewise.
8539 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8540 with an insn_type enum and msl with a modifier_type enum.
8541 Replace element_width with a scalar_mode. Change the shift
8542 to unsigned int. Add constructors for scalar_float_mode and
8543 scalar_int_mode elements.
8544 (aarch64_vect_float_const_representable_p): Delete.
8545 (aarch64_can_const_movi_rtx_p)
8546 (aarch64_simd_scalar_immediate_valid_for_move)
8547 (aarch64_simd_make_constant): Update call to
8548 aarch64_simd_valid_immediate.
8549 (aarch64_advsimd_valid_immediate_hs): New function.
8550 (aarch64_advsimd_valid_immediate): Likewise.
8551 (aarch64_simd_valid_immediate): Remove mode and inverse
8552 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8553 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8554 and aarch64_float_const_representable_p on the result.
8555 (aarch64_output_simd_mov_immediate): Remove mode argument.
8556 Update call to aarch64_simd_valid_immediate and use of
8557 simd_immediate_info.
8558 (aarch64_output_scalar_simd_mov_immediate): Update call
8561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8562 Alan Hayward <alan.hayward@arm.com>
8563 David Sherwood <david.sherwood@arm.com>
8565 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8566 (mode_nunits): Likewise CONST_MODE_NUNITS.
8567 * machmode.def (ADJUST_NUNITS): Document.
8568 * genmodes.c (mode_data::need_nunits_adj): New field.
8569 (blank_mode): Update accordingly.
8570 (adj_nunits): New variable.
8571 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8573 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8574 listed in adj_nunits.
8575 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8576 listed in adj_nunits. Don't emit case statements for such modes.
8577 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8578 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8579 nothing if adj_nunits is nonnull.
8580 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8581 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8582 (emit_mode_fbit): Update use of print_maybe_const_decl.
8583 (emit_move_size): Likewise. Treat the array as non-const
8585 (emit_mode_adjustments): Handle adj_nunits.
8587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8589 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8590 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8591 (VECTOR_MODES): Use it.
8592 (make_vector_modes): Take the prefix as an argument.
8594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8595 Alan Hayward <alan.hayward@arm.com>
8596 David Sherwood <david.sherwood@arm.com>
8598 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8599 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8600 for MODE_VECTOR_BOOL.
8601 * machmode.def (VECTOR_BOOL_MODE): Document.
8602 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8603 (make_vector_bool_mode): New function.
8604 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8606 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8607 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8609 * stor-layout.c (int_mode_for_mode): Likewise.
8610 * tree.c (build_vector_type_for_mode): Likewise.
8611 * varasm.c (output_constant_pool_2): Likewise.
8612 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8613 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8614 for MODE_VECTOR_BOOL.
8615 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8616 of mode class checks.
8617 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8618 instead of a list of mode class checks.
8619 (expand_vector_scalar_condition): Likewise.
8620 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8623 Alan Hayward <alan.hayward@arm.com>
8624 David Sherwood <david.sherwood@arm.com>
8626 * machmode.h (mode_size): Change from unsigned short to
8628 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8629 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8630 or if measurement_type is not polynomial.
8631 (fixed_size_mode::includes_p): Check for constant-sized modes.
8632 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8633 return a poly_uint16 rather than an unsigned short.
8634 (emit_mode_size): Change the type of mode_size from unsigned short
8635 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8636 (emit_mode_adjustments): Cope with polynomial vector sizes.
8637 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8639 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8641 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8642 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8643 * caller-save.c (setup_save_areas): Likewise.
8644 (replace_reg_with_saved_mem): Likewise.
8645 * calls.c (emit_library_call_value_1): Likewise.
8646 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8647 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8648 (gen_lowpart_for_combine): Likewise.
8649 * convert.c (convert_to_integer_1): Likewise.
8650 * cse.c (equiv_constant, cse_insn): Likewise.
8651 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8652 (cselib_subst_to_values): Likewise.
8653 * dce.c (word_dce_process_block): Likewise.
8654 * df-problems.c (df_word_lr_mark_ref): Likewise.
8655 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8656 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8657 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8658 (rtl_for_decl_location): Likewise.
8659 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8660 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8661 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8662 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8663 (expand_expr_real_1): Likewise.
8664 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8665 (pad_below): Likewise.
8666 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8667 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8668 * ira.c (get_subreg_tracking_sizes): Likewise.
8669 * ira-build.c (ira_create_allocno_objects): Likewise.
8670 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8671 (ira_sort_regnos_for_alter_reg): Likewise.
8672 * ira-costs.c (record_operand_costs): Likewise.
8673 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8674 (resolve_simple_move): Likewise.
8675 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8676 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8677 (lra_constraints): Likewise.
8678 (CONST_POOL_OK_P): Reject variable-sized modes.
8679 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8680 (add_pseudo_to_slot, lra_spill): Likewise.
8681 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8682 * optabs-query.c (get_best_extraction_insn): Likewise.
8683 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8684 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8685 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8686 * recog.c (offsettable_address_addr_space_p): Likewise.
8687 * regcprop.c (maybe_mode_change): Likewise.
8688 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8689 * regrename.c (build_def_use): Likewise.
8690 * regstat.c (dump_reg_info): Likewise.
8691 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8692 (find_reloads, find_reloads_subreg_address): Likewise.
8693 * reload1.c (eliminate_regs_1): Likewise.
8694 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8695 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8696 (simplify_binary_operation_1, simplify_subreg): Likewise.
8697 * targhooks.c (default_function_arg_padding): Likewise.
8698 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8699 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8700 (verify_gimple_assign_ternary): Likewise.
8701 * tree-inline.c (estimate_move_cost): Likewise.
8702 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8703 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8704 (get_address_cost_ainc): Likewise.
8705 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8706 (vect_supportable_dr_alignment): Likewise.
8707 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8708 (vectorizable_reduction): Likewise.
8709 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8710 (vectorizable_operation, vectorizable_load): Likewise.
8711 * tree.c (build_same_sized_truth_vector_type): Likewise.
8712 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8713 * var-tracking.c (emit_note_insn_var_location): Likewise.
8714 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8715 (ADDR_VEC_ALIGN): Likewise.
8717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8718 Alan Hayward <alan.hayward@arm.com>
8719 David Sherwood <david.sherwood@arm.com>
8721 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8723 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8724 or if measurement_type is polynomial.
8725 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8726 * combine.c (make_extraction): Likewise.
8727 * dse.c (find_shift_sequence): Likewise.
8728 * dwarf2out.c (mem_loc_descriptor): Likewise.
8729 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8730 (extract_bit_field, extract_low_bits): Likewise.
8731 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8732 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8733 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8734 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8735 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8736 * reload.c (find_reloads): Likewise.
8737 * reload1.c (alter_reg): Likewise.
8738 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8739 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8740 * tree-if-conv.c (predicate_mem_writes): Likewise.
8741 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8742 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8743 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8744 * valtrack.c (dead_debug_insert_temp): Likewise.
8745 * varasm.c (mergeable_constant_section): Likewise.
8746 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8749 Alan Hayward <alan.hayward@arm.com>
8750 David Sherwood <david.sherwood@arm.com>
8752 * expr.c (expand_assignment): Cope with polynomial mode sizes
8753 when assigning to a CONCAT.
8755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8756 Alan Hayward <alan.hayward@arm.com>
8757 David Sherwood <david.sherwood@arm.com>
8759 * machmode.h (mode_precision): Change from unsigned short to
8761 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8763 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8764 or if measurement_type is not polynomial.
8765 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8766 in which the mode is already known to be a scalar_int_mode.
8767 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8768 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8770 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8771 for GET_MODE_PRECISION.
8772 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8773 for GET_MODE_PRECISION.
8774 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8776 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8777 (expand_field_assignment, make_extraction): Likewise.
8778 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8779 (get_last_value): Likewise.
8780 * convert.c (convert_to_integer_1): Likewise.
8781 * cse.c (cse_insn): Likewise.
8782 * expr.c (expand_expr_real_1): Likewise.
8783 * lra-constraints.c (simplify_operand_subreg): Likewise.
8784 * optabs-query.c (can_atomic_load_p): Likewise.
8785 * optabs.c (expand_atomic_load): Likewise.
8786 (expand_atomic_store): Likewise.
8787 * ree.c (combine_reaching_defs): Likewise.
8788 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8789 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8790 * tree.h (type_has_mode_precision_p): Likewise.
8791 * ubsan.c (instrument_si_overflow): Likewise.
8793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8794 Alan Hayward <alan.hayward@arm.com>
8795 David Sherwood <david.sherwood@arm.com>
8797 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8798 polynomial numbers of units.
8799 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8800 (valid_vector_subparts_p): New function.
8801 (build_vector_type): Remove temporary shim and take the number
8802 of units as a poly_uint64 rather than an int.
8803 (build_opaque_vector_type): Take the number of units as a
8804 poly_uint64 rather than an int.
8805 * tree.c (build_vector_from_ctor): Handle polynomial
8806 TYPE_VECTOR_SUBPARTS.
8807 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8808 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8809 (build_vector_from_val): If the number of units is variable,
8810 use build_vec_duplicate_cst for constant operands and
8811 VEC_DUPLICATE_EXPR otherwise.
8812 (make_vector_type): Remove temporary is_constant ().
8813 (build_vector_type, build_opaque_vector_type): Take the number of
8814 units as a poly_uint64 rather than an int.
8815 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8817 * cfgexpand.c (expand_debug_expr): Likewise.
8818 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8819 (store_constructor, expand_expr_real_1): Likewise.
8820 (const_scalar_mask_from_tree): Likewise.
8821 * fold-const-call.c (fold_const_reduction): Likewise.
8822 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8823 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8824 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8825 (fold_relational_const): Likewise.
8826 (native_interpret_vector): Likewise. Change the size from an
8827 int to an unsigned int.
8828 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8829 TYPE_VECTOR_SUBPARTS.
8830 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8831 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8832 duplicating a non-constant operand into a variable-length vector.
8833 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8834 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8835 * ipa-icf.c (sem_variable::equals): Likewise.
8836 * match.pd: Likewise.
8837 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8838 * print-tree.c (print_node): Likewise.
8839 * stor-layout.c (layout_type): Likewise.
8840 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8841 * tree-cfg.c (verify_gimple_comparison): Likewise.
8842 (verify_gimple_assign_binary): Likewise.
8843 (verify_gimple_assign_ternary): Likewise.
8844 (verify_gimple_assign_single): Likewise.
8845 * tree-pretty-print.c (dump_generic_node): Likewise.
8846 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8847 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8848 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8849 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8850 (vect_shift_permute_load_chain): Likewise.
8851 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8852 (expand_vector_condition, optimize_vector_constructor): Likewise.
8853 (lower_vec_perm, get_compute_type): Likewise.
8854 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8855 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8856 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8857 (vect_recog_mask_conversion_pattern): Likewise.
8858 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8859 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8860 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8861 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8862 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8863 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8864 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8865 (supportable_widening_operation): Likewise.
8866 (supportable_narrowing_operation): Likewise.
8867 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8869 * varasm.c (output_constant): Likewise.
8871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8872 Alan Hayward <alan.hayward@arm.com>
8873 David Sherwood <david.sherwood@arm.com>
8875 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8876 so that both the length == 3 and length != 3 cases set up their
8877 own permute vectors. Add comments explaining why we know the
8878 number of elements is constant.
8879 (vect_permute_load_chain): Likewise.
8881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8882 Alan Hayward <alan.hayward@arm.com>
8883 David Sherwood <david.sherwood@arm.com>
8885 * machmode.h (mode_nunits): Change from unsigned char to
8887 (ONLY_FIXED_SIZE_MODES): New macro.
8888 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8889 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8890 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8892 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8893 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8894 or if measurement_type is not polynomial.
8895 * genmodes.c (ZERO_COEFFS): New macro.
8896 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8898 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8899 Use ZERO_COEFFS when emitting initializers.
8900 * data-streamer.h (bp_pack_poly_value): New function.
8901 (bp_unpack_poly_value): Likewise.
8902 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8903 for GET_MODE_NUNITS.
8904 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8905 for GET_MODE_NUNITS.
8906 * tree.c (make_vector_type): Remove temporary shim and make
8907 the real function take the number of units as a poly_uint64
8909 (build_vector_type_for_mode): Handle polynomial nunits.
8910 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8911 * emit-rtl.c (const_vec_series_p_1): Likewise.
8912 (gen_rtx_CONST_VECTOR): Likewise.
8913 * fold-const.c (test_vec_duplicate_folding): Likewise.
8914 * genrecog.c (validate_pattern): Likewise.
8915 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8916 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8917 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8918 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8919 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8920 * rtlanal.c (subreg_get_info): Likewise.
8921 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8922 (vect_grouped_load_supported): Likewise.
8923 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8924 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8925 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8926 (simplify_const_unary_operation, simplify_binary_operation_1)
8927 (simplify_const_binary_operation, simplify_ternary_operation)
8928 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8929 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8930 instead of CONST_VECTOR_NUNITS.
8931 * varasm.c (output_constant_pool_2): Likewise.
8932 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8933 explicit-encoded elements in the XVEC for variable-length vectors.
8935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8937 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8940 Alan Hayward <alan.hayward@arm.com>
8941 David Sherwood <david.sherwood@arm.com>
8943 * coretypes.h (fixed_size_mode): Declare.
8944 (fixed_size_mode_pod): New typedef.
8945 * builtins.h (target_builtins::x_apply_args_mode)
8946 (target_builtins::x_apply_result_mode): Change type to
8947 fixed_size_mode_pod.
8948 * builtins.c (apply_args_size, apply_result_size, result_vector)
8949 (expand_builtin_apply_args_1, expand_builtin_apply)
8950 (expand_builtin_return): Update accordingly.
8952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8954 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8955 * cselib.c (cselib_hash_rtx): Likewise.
8956 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8957 CONST_VECTOR encoding.
8959 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8960 Jeff Law <law@redhat.com>
8963 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8964 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8965 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8966 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8969 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8970 explicitly probe *sp in a noreturn function if there were any callee
8971 register saves or frame pointer is needed.
8973 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8976 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8977 BLKmode for ternary, binary or unary expressions.
8980 * var-tracking.c (delete_vta_debug_insn): New inline function.
8981 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8982 insns from get_insns () to NULL instead of each bb separately.
8983 Use delete_vta_debug_insn. No longer static.
8984 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8985 delete_vta_debug_insns callers.
8986 * rtl.h (delete_vta_debug_insns): Declare.
8987 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8988 instead of variable_tracking_main.
8990 2018-01-03 Martin Sebor <msebor@redhat.com>
8992 PR tree-optimization/83603
8993 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8994 arguments past the endof the argument list in functions declared
8995 without a prototype.
8996 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8997 Avoid checking when arguments are null.
8999 2018-01-03 Martin Sebor <msebor@redhat.com>
9002 * doc/extend.texi (attribute const): Fix a typo.
9003 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9004 issuing -Wsuggest-attribute for void functions.
9006 2018-01-03 Martin Sebor <msebor@redhat.com>
9008 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9009 offset_int::from instead of wide_int::to_shwi.
9010 (maybe_diag_overlap): Remove assertion.
9011 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9012 * gimple-ssa-sprintf.c (format_directive): Same.
9013 (parse_directive): Same.
9014 (sprintf_dom_walker::compute_format_length): Same.
9015 (try_substitute_return_value): Same.
9017 2018-01-03 Jeff Law <law@redhat.com>
9020 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9021 non-constant residual for zero at runtime and avoid probing in
9022 that case. Reorganize code for trailing problem to mirror handling
9025 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9027 PR tree-optimization/83501
9028 * tree-ssa-strlen.c (get_string_cst): New.
9029 (handle_char_store): Call get_string_cst.
9031 2018-01-03 Martin Liska <mliska@suse.cz>
9033 PR tree-optimization/83593
9034 * tree-ssa-strlen.c: Include tree-cfg.h.
9035 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9036 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9037 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9039 (strlen_dom_walker::before_dom_children): Call
9040 gimple_purge_dead_eh_edges. Dump tranformation with details
9042 (strlen_dom_walker::before_dom_children): Update call by adding
9043 new argument cleanup_eh.
9044 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9046 2018-01-03 Martin Liska <mliska@suse.cz>
9049 * cif-code.def (VARIADIC_THUNK): New enum value.
9050 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9053 2018-01-03 Jan Beulich <jbeulich@suse.com>
9055 * sse.md (mov<mode>_internal): Tighten condition for when to use
9056 vmovdqu<ssescalarsize> for TI and OI modes.
9058 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9060 Update copyright years.
9062 2018-01-03 Martin Liska <mliska@suse.cz>
9065 * ipa-visibility.c (function_and_variable_visibility): Skip
9066 functions with noipa attribure.
9068 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9070 * gcc.c (process_command): Update copyright notice dates.
9071 * gcov-dump.c (print_version): Ditto.
9072 * gcov.c (print_version): Ditto.
9073 * gcov-tool.c (print_version): Ditto.
9074 * gengtype.c (create_file): Ditto.
9075 * doc/cpp.texi: Bump @copying's copyright year.
9076 * doc/cppinternals.texi: Ditto.
9077 * doc/gcc.texi: Ditto.
9078 * doc/gccint.texi: Ditto.
9079 * doc/gcov.texi: Ditto.
9080 * doc/install.texi: Ditto.
9081 * doc/invoke.texi: Ditto.
9083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9085 * vector-builder.h (vector_builder::m_full_nelts): Change from
9086 unsigned int to poly_uint64.
9087 (vector_builder::full_nelts): Update prototype accordingly.
9088 (vector_builder::new_vector): Likewise.
9089 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9090 (vector_builder::operator ==): Likewise.
9091 (vector_builder::finalize): Likewise.
9092 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9093 Take the number of elements as a poly_uint64 rather than an
9095 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9096 from unsigned int to poly_uint64.
9097 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9098 (vec_perm_indices::new_vector): Likewise.
9099 (vec_perm_indices::length): Likewise.
9100 (vec_perm_indices::nelts_per_input): Likewise.
9101 (vec_perm_indices::input_nelts): Likewise.
9102 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9103 number of elements per input as a poly_uint64 rather than an
9104 unsigned int. Use the original encoding for variable-length
9105 vectors, rather than clamping each individual element.
9106 For the second and subsequent elements in each pattern,
9107 clamp the step and base before clamping their sum.
9108 (vec_perm_indices::series_p): Handle polynomial element counts.
9109 (vec_perm_indices::all_in_range_p): Likewise.
9110 (vec_perm_indices_to_tree): Likewise.
9111 (vec_perm_indices_to_rtx): Likewise.
9112 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9113 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9114 (tree_vector_builder::new_binary_operation): Handle polynomial
9115 element counts. Return false if we need to know the number
9116 of elements at compile time.
9117 * fold-const.c (fold_vec_perm): Punt if the number of elements
9118 isn't known at compile time.
9120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9122 * vec-perm-indices.h (vec_perm_builder): Change element type
9123 from HOST_WIDE_INT to poly_int64.
9124 (vec_perm_indices::element_type): Update accordingly.
9125 (vec_perm_indices::clamp): Handle polynomial element_types.
9126 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9127 (vec_perm_indices::all_in_range_p): Likewise.
9128 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9130 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9131 polynomial vec_perm_indices element types.
9132 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9133 * fold-const.c (fold_vec_perm): Likewise.
9134 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9135 * tree-vect-generic.c (lower_vec_perm): Likewise.
9136 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9137 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9138 element type to HOST_WIDE_INT.
9140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9141 Alan Hayward <alan.hayward@arm.com>
9142 David Sherwood <david.sherwood@arm.com>
9144 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9145 rather than an int. Use plus_constant.
9146 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9147 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9150 Alan Hayward <alan.hayward@arm.com>
9151 David Sherwood <david.sherwood@arm.com>
9153 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9154 a HOST_WIDE_INT to a poly_int64.
9156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9157 Alan Hayward <alan.hayward@arm.com>
9158 David Sherwood <david.sherwood@arm.com>
9160 * calls.c (load_register_parameters): Cope with polynomial
9161 mode sizes. Require a constant size for BLKmode parameters
9162 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9163 forces a parameter to be padded at the lsb end in order to
9164 fill a complete number of words, require the parameter size
9165 to be ordered wrt UNITS_PER_WORD.
9167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9168 Alan Hayward <alan.hayward@arm.com>
9169 David Sherwood <david.sherwood@arm.com>
9171 * reload1.c (spill_stack_slot_width): Change element type
9172 from unsigned int to poly_uint64_pod.
9173 (alter_reg): Treat mode sizes as polynomial.
9175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9176 Alan Hayward <alan.hayward@arm.com>
9177 David Sherwood <david.sherwood@arm.com>
9179 * reload.c (complex_word_subreg_p): New function.
9180 (reload_inner_reg_of_subreg, push_reload): Use it.
9182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9183 Alan Hayward <alan.hayward@arm.com>
9184 David Sherwood <david.sherwood@arm.com>
9186 * lra-constraints.c (process_alt_operands): Reject matched
9187 operands whose sizes aren't ordered.
9188 (match_reload): Refer to this check here.
9190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9191 Alan Hayward <alan.hayward@arm.com>
9192 David Sherwood <david.sherwood@arm.com>
9194 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9195 that the mode size is in the set {1, 2, 4, 8, 16}.
9197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9198 Alan Hayward <alan.hayward@arm.com>
9199 David Sherwood <david.sherwood@arm.com>
9201 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9202 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9209 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9210 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9211 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9212 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9213 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9214 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9215 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9216 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9217 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9219 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9220 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9221 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9222 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9223 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9224 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9225 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9226 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9227 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9228 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9230 * expr.c (emit_move_resolve_push): Treat the input and result
9231 of PUSH_ROUNDING as a poly_int64.
9232 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9233 (emit_push_insn): Likewise.
9234 * lra-eliminations.c (mark_not_eliminable): Likewise.
9235 * recog.c (push_operand): Likewise.
9236 * reload1.c (elimination_effects): Likewise.
9237 * rtlanal.c (nonzero_bits1): Likewise.
9238 * calls.c (store_one_arg): Likewise. Require the padding to be
9239 known at compile time.
9241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9242 Alan Hayward <alan.hayward@arm.com>
9243 David Sherwood <david.sherwood@arm.com>
9245 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9246 Use plus_constant instead of gen_rtx_PLUS.
9248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9249 Alan Hayward <alan.hayward@arm.com>
9250 David Sherwood <david.sherwood@arm.com>
9252 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9256 Alan Hayward <alan.hayward@arm.com>
9257 David Sherwood <david.sherwood@arm.com>
9259 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9260 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9261 via stack temporaries. Treat the mode size as polynomial too.
9263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9264 Alan Hayward <alan.hayward@arm.com>
9265 David Sherwood <david.sherwood@arm.com>
9267 * expr.c (expand_expr_real_2): When handling conversions involving
9268 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9269 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9270 as a poly_uint64 too.
9272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9273 Alan Hayward <alan.hayward@arm.com>
9274 David Sherwood <david.sherwood@arm.com>
9276 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9279 Alan Hayward <alan.hayward@arm.com>
9280 David Sherwood <david.sherwood@arm.com>
9282 * combine.c (can_change_dest_mode): Handle polynomial
9283 REGMODE_NATURAL_SIZE.
9284 * expmed.c (store_bit_field_1): Likewise.
9285 * expr.c (store_constructor): Likewise.
9286 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9287 and polynomial REGMODE_NATURAL_SIZE.
9288 (gen_lowpart_common): Likewise.
9289 * reginfo.c (record_subregs_of_mode): Likewise.
9290 * rtlanal.c (read_modify_subreg_p): Likewise.
9292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9293 Alan Hayward <alan.hayward@arm.com>
9294 David Sherwood <david.sherwood@arm.com>
9296 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9297 numbers of elements.
9299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9300 Alan Hayward <alan.hayward@arm.com>
9301 David Sherwood <david.sherwood@arm.com>
9303 * match.pd: Cope with polynomial numbers of vector elements.
9305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9306 Alan Hayward <alan.hayward@arm.com>
9307 David Sherwood <david.sherwood@arm.com>
9309 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9310 in a POINTER_PLUS_EXPR.
9312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9313 Alan Hayward <alan.hayward@arm.com>
9314 David Sherwood <david.sherwood@arm.com>
9316 * omp-simd-clone.c (simd_clone_subparts): New function.
9317 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9318 (ipa_simd_modify_function_body): Likewise.
9320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9321 Alan Hayward <alan.hayward@arm.com>
9322 David Sherwood <david.sherwood@arm.com>
9324 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9325 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9326 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9327 (expand_vector_condition, vector_element): Likewise.
9328 (subparts_gt): New function.
9329 (get_compute_type): Use subparts_gt.
9330 (count_type_subparts): Delete.
9331 (expand_vector_operations_1): Use subparts_gt instead of
9332 count_type_subparts.
9334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9335 Alan Hayward <alan.hayward@arm.com>
9336 David Sherwood <david.sherwood@arm.com>
9338 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9339 (vect_compile_time_alias): ...this new function. Do the calculation
9340 on poly_ints rather than trees.
9341 (vect_prune_runtime_alias_test_list): Update call accordingly.
9343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9344 Alan Hayward <alan.hayward@arm.com>
9345 David Sherwood <david.sherwood@arm.com>
9347 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9349 (vect_schedule_slp_instance): Likewise.
9351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9352 Alan Hayward <alan.hayward@arm.com>
9353 David Sherwood <david.sherwood@arm.com>
9355 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9356 constant and extern definitions for variable-length vectors.
9357 (vect_get_constant_vectors): Note that the number of units
9358 is known to be constant.
9360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9361 Alan Hayward <alan.hayward@arm.com>
9362 David Sherwood <david.sherwood@arm.com>
9364 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9365 of units as polynomial. Choose between WIDE and NARROW based
9368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9369 Alan Hayward <alan.hayward@arm.com>
9370 David Sherwood <david.sherwood@arm.com>
9372 * tree-vect-stmts.c (simd_clone_subparts): New function.
9373 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9376 Alan Hayward <alan.hayward@arm.com>
9377 David Sherwood <david.sherwood@arm.com>
9379 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9380 vectors as polynomial. Use build_index_vector for
9383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9384 Alan Hayward <alan.hayward@arm.com>
9385 David Sherwood <david.sherwood@arm.com>
9387 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9388 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9389 for variable-length vectors.
9390 (vectorizable_mask_load_store): Treat the number of units as
9391 polynomial, asserting that it is constant if the condition has
9392 already been enforced.
9393 (vectorizable_store, vectorizable_load): Likewise.
9395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9396 Alan Hayward <alan.hayward@arm.com>
9397 David Sherwood <david.sherwood@arm.com>
9399 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9400 of units as polynomial. Punt if we can't tell at compile time
9401 which vector contains the final result.
9403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9404 Alan Hayward <alan.hayward@arm.com>
9405 David Sherwood <david.sherwood@arm.com>
9407 * tree-vect-loop.c (vectorizable_induction): Treat the number
9408 of units as polynomial. Punt on SLP inductions. Use an integer
9409 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9410 cast of such a series for variable-length floating-point
9413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9414 Alan Hayward <alan.hayward@arm.com>
9415 David Sherwood <david.sherwood@arm.com>
9417 * tree.h (build_index_vector): Declare.
9418 * tree.c (build_index_vector): New function.
9419 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9420 of units as polynomial, forcibly converting it to a constant if
9421 vectorizable_reduction has already enforced the condition.
9422 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9423 to create a {1,2,3,...} vector.
9424 (vectorizable_reduction): Treat the number of units as polynomial.
9425 Choose vectype_in based on the largest scalar element size rather
9426 than the smallest number of units. Enforce the restrictions
9429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9430 Alan Hayward <alan.hayward@arm.com>
9431 David Sherwood <david.sherwood@arm.com>
9433 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9434 number of units as polynomial.
9436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9437 Alan Hayward <alan.hayward@arm.com>
9438 David Sherwood <david.sherwood@arm.com>
9440 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9441 * target.def (autovectorize_vector_sizes): Return the vector sizes
9442 by pointer, using vector_sizes rather than a bitmask.
9443 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9444 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9445 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9447 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9448 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9449 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9450 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9451 * omp-general.c (omp_max_vf): Likewise.
9452 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9453 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9454 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9455 * tree-vect-slp.c (vect_slp_bb): Likewise.
9456 * doc/tm.texi: Regenerate.
9457 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9459 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9460 the vector size as a poly_uint64 rather than an unsigned int.
9461 (current_vector_size): Change from an unsigned int to a poly_uint64.
9462 (get_vectype_for_scalar_type): Update accordingly.
9463 * tree.h (build_truth_vector_type): Take the size and number of
9464 units as a poly_uint64 rather than an unsigned int.
9465 (build_vector_type): Add a temporary overload that takes
9466 the number of units as a poly_uint64 rather than an unsigned int.
9467 * tree.c (make_vector_type): Likewise.
9468 (build_truth_vector_type): Take the number of units as a poly_uint64
9469 rather than an unsigned int.
9471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9472 Alan Hayward <alan.hayward@arm.com>
9473 David Sherwood <david.sherwood@arm.com>
9475 * target.def (get_mask_mode): Take the number of units and length
9476 as poly_uint64s rather than unsigned ints.
9477 * targhooks.h (default_get_mask_mode): Update accordingly.
9478 * targhooks.c (default_get_mask_mode): Likewise.
9479 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9480 * doc/tm.texi: Regenerate.
9482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9483 Alan Hayward <alan.hayward@arm.com>
9484 David Sherwood <david.sherwood@arm.com>
9486 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9487 * omp-general.c (omp_max_vf): Likewise.
9488 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9489 (expand_omp_simd): Handle polynomial safelen.
9490 * omp-low.c (omplow_simd_context): Add a default constructor.
9491 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9492 (lower_rec_simd_input_clauses): Update accordingly.
9493 (lower_rec_input_clauses): Likewise.
9495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9496 Alan Hayward <alan.hayward@arm.com>
9497 David Sherwood <david.sherwood@arm.com>
9499 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9500 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9501 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9502 (vect_analyze_slp_cost): Likewise.
9503 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9504 (vect_model_load_cost): Likewise.
9506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9507 Alan Hayward <alan.hayward@arm.com>
9508 David Sherwood <david.sherwood@arm.com>
9510 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9511 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9512 from an unsigned int * to a poly_uint64_pod *.
9513 (calculate_unrolling_factor): New function.
9514 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9517 Alan Hayward <alan.hayward@arm.com>
9518 David Sherwood <david.sherwood@arm.com>
9520 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9521 from an unsigned int to a poly_uint64.
9522 (_loop_vec_info::slp_unrolling_factor): Likewise.
9523 (_loop_vec_info::vectorization_factor): Change from an int
9525 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9526 (vect_get_num_vectors): New function.
9527 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9528 (vect_get_num_copies): Use vect_get_num_vectors.
9529 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9530 to an unsigned int *.
9531 (vect_analyze_data_refs): Change min_vf from an int * to a
9533 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9534 than an unsigned HOST_WIDE_INT.
9535 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9536 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9537 to an unsigned int *.
9538 (vect_analyze_data_ref_dependences): Likewise.
9539 (vect_compute_data_ref_alignment): Handle polynomial vf.
9540 (vect_enhance_data_refs_alignment): Likewise.
9541 (vect_prune_runtime_alias_test_list): Likewise.
9542 (vect_shift_permute_load_chain): Likewise.
9543 (vect_supportable_dr_alignment): Likewise.
9544 (dependence_distance_ge_vf): Take the vectorization factor as a
9545 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9546 (vect_analyze_data_refs): Change min_vf from an int * to a
9548 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9549 vfm1 as a poly_uint64 rather than an int. Make the same change
9550 for the returned bound_scalar.
9551 (vect_gen_vector_loop_niters): Handle polynomial vf.
9552 (vect_do_peeling): Likewise. Update call to
9553 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9554 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9556 * tree-vect-loop.c (vect_determine_vectorization_factor)
9557 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9558 (vect_get_known_peeling_cost): Likewise.
9559 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9560 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9561 (vect_transform_loop): Likewise. Use the lowest possible VF when
9562 updating the upper bounds of the loop.
9563 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9565 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9566 polynomial unroll factors.
9567 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9568 (vect_make_slp_decision): Likewise.
9569 (vect_supported_load_permutation_p): Likewise, and polynomial
9571 (vect_analyze_slp_cost): Handle polynomial vf.
9572 (vect_slp_analyze_node_operations): Likewise.
9573 (vect_slp_analyze_bb_1): Likewise.
9574 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9575 than an unsigned HOST_WIDE_INT.
9576 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9577 (vectorizable_load): Handle polynomial vf.
9578 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9580 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9583 Alan Hayward <alan.hayward@arm.com>
9584 David Sherwood <david.sherwood@arm.com>
9586 * match.pd: Handle bit operations involving three constants
9587 and try to fold one pair.
9589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9591 * tree-vect-loop-manip.c: Include gimple-fold.h.
9592 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9593 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9594 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9595 Add a path that uses a step of VF instead of 1, but disable it
9597 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9598 and niters_no_overflow parameters. Update calls to
9599 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9600 Create a new SSA name if the latter choses to use a ste other
9601 than zero, and return it via niters_vector_mult_vf_var.
9602 * tree-vect-loop.c (vect_transform_loop): Update calls to
9603 vect_do_peeling, vect_gen_vector_loop_niters and
9604 slpeel_make_loop_iterate_ntimes.
9605 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9606 (vect_gen_vector_loop_niters): Update declarations after above changes.
9608 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9610 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9611 128-bit round to integer instructions.
9612 (ceil<mode>2): Likewise.
9613 (btrunc<mode>2): Likewise.
9614 (round<mode>2): Likewise.
9616 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9618 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9619 unaligned VSX load/store on P8/P9.
9620 (expand_block_clear): Allow the use of unaligned VSX
9621 load/store on P8/P9.
9623 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9625 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9627 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9628 swap associated with both a load and a store.
9630 2018-01-02 Andrew Waterman <andrew@sifive.com>
9632 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9633 * config/riscv/riscv.md (clear_cache): Use it.
9635 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9637 * web.c: Remove out-of-date comment.
9639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9641 * expr.c (fixup_args_size_notes): Check that any existing
9642 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9643 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9644 (emit_single_push_insn): ...here.
9646 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9648 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9649 (const_vector_encoded_nelts): New function.
9650 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9651 (const_vector_int_elt, const_vector_elt): Declare.
9652 * emit-rtl.c (const_vector_int_elt_1): New function.
9653 (const_vector_elt): Likewise.
9654 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9655 of CONST_VECTOR_ELT.
9657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9659 * expr.c: Include rtx-vector-builder.h.
9660 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9661 directly on the tree encoding.
9662 (const_vector_from_tree): Likewise.
9663 * optabs.c: Include rtx-vector-builder.h.
9664 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9665 sequence of "u" values.
9666 * vec-perm-indices.c: Include rtx-vector-builder.h.
9667 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9668 directly on the vec_perm_indices encoding.
9670 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9672 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9673 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9674 * rtx-vector-builder.h: New file.
9675 * rtx-vector-builder.c: Likewise.
9676 * rtl.h (rtx_def::u2): Add a const_vector field.
9677 (CONST_VECTOR_NPATTERNS): New macro.
9678 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9679 (CONST_VECTOR_DUPLICATE_P): Likewise.
9680 (CONST_VECTOR_STEPPED_P): Likewise.
9681 (CONST_VECTOR_ENCODED_ELT): Likewise.
9682 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9683 (unwrap_const_vec_duplicate): Likewise.
9684 (const_vec_series_p): Check for a non-duplicated vector encoding.
9685 Say that the function only returns true for integer vectors.
9686 * emit-rtl.c: Include rtx-vector-builder.h.
9687 (gen_const_vec_duplicate_1): Delete.
9688 (gen_const_vector): Call gen_const_vec_duplicate instead of
9689 gen_const_vec_duplicate_1.
9690 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9691 (gen_const_vec_duplicate): Use rtx_vector_builder.
9692 (gen_const_vec_series): Likewise.
9693 (gen_rtx_CONST_VECTOR): Likewise.
9694 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9695 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9696 Build a new vector rather than modifying a CONST_VECTOR in-place.
9697 (handle_special_swappables): Update call accordingly.
9698 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9699 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9700 Build a new vector rather than modifying a CONST_VECTOR in-place.
9701 (handle_special_swappables): Update call accordingly.
9703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9705 * simplify-rtx.c (simplify_const_binary_operation): Use
9706 CONST_VECTOR_ELT instead of XVECEXP.
9708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9710 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9711 the selector elements to be different from the data elements
9712 if the selector is a VECTOR_CST.
9713 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9714 ssizetype for the selector.
9716 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9718 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9719 before testing each element individually.
9720 * tree-vect-generic.c (lower_vec_perm): Likewise.
9722 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9724 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9725 * selftest-run-tests.c (selftest::run_tests): Call it.
9726 * vector-builder.h (vector_builder::operator ==): New function.
9727 (vector_builder::operator !=): Likewise.
9728 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9729 (vec_perm_indices::all_from_input_p): New function.
9730 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9731 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9732 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9733 instead of reading the VECTOR_CST directly. Detect whether both
9734 vector inputs are the same before constructing the vec_perm_indices,
9735 and update the number of inputs argument accordingly. Use the
9736 utility functions added above. Only construct sel2 if we need to.
9738 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9740 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9741 the broadcast of the low byte.
9742 (expand_mult_highpart): Use an explicit encoding for the permutes.
9743 * optabs-query.c (can_mult_highpart_p): Likewise.
9744 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9745 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9746 (vectorizable_bswap): Likewise.
9747 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9748 explicit encoding for the power-of-2 permutes.
9749 (vect_permute_store_chain): Likewise.
9750 (vect_grouped_load_supported): Likewise.
9751 (vect_permute_load_chain): Likewise.
9753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9755 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9756 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9757 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9758 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9759 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9760 (vect_gen_perm_mask_any): Likewise.
9762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9764 * int-vector-builder.h: New file.
9765 * vec-perm-indices.h: Include int-vector-builder.h.
9766 (vec_perm_indices): Redefine as an int_vector_builder.
9767 (auto_vec_perm_indices): Delete.
9768 (vec_perm_builder): Redefine as a stand-alone class.
9769 (vec_perm_indices::vec_perm_indices): New function.
9770 (vec_perm_indices::clamp): Likewise.
9771 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9772 (vec_perm_indices::new_vector): New function.
9773 (vec_perm_indices::new_expanded_vector): Update for new
9774 vec_perm_indices class.
9775 (vec_perm_indices::rotate_inputs): New function.
9776 (vec_perm_indices::all_in_range_p): Operate directly on the
9777 encoded form, without computing elided elements.
9778 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9779 encoding. Update for new vec_perm_indices class.
9780 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9781 the given vec_perm_builder.
9782 (expand_vec_perm_var): Update vec_perm_builder constructor.
9783 (expand_mult_highpart): Use vec_perm_builder instead of
9784 auto_vec_perm_indices.
9785 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9786 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9787 or double series encoding as appropriate.
9788 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9789 vec_perm_indices instead of auto_vec_perm_indices.
9790 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9791 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9792 (vect_permute_store_chain): Likewise.
9793 (vect_grouped_load_supported): Likewise.
9794 (vect_permute_load_chain): Likewise.
9795 (vect_shift_permute_load_chain): Likewise.
9796 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9797 (vect_transform_slp_perm_load): Likewise.
9798 (vect_schedule_slp_instance): Likewise.
9799 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9800 (vectorizable_mask_load_store): Likewise.
9801 (vectorizable_bswap): Likewise.
9802 (vectorizable_store): Likewise.
9803 (vectorizable_load): Likewise.
9804 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9805 vec_perm_indices instead of auto_vec_perm_indices. Use
9806 tree_to_vec_perm_builder to read the vector from a tree.
9807 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9808 vec_perm_builder instead of a vec_perm_indices.
9809 (have_whole_vector_shift): Use vec_perm_builder and
9810 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9811 truncation to calc_vec_perm_mask_for_shift.
9812 (vect_create_epilog_for_reduction): Likewise.
9813 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9814 from auto_vec_perm_indices to vec_perm_indices.
9815 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9816 instead of changing individual elements.
9817 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9818 the vector in d.perm.
9819 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9820 from auto_vec_perm_indices to vec_perm_indices.
9821 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9822 instead of changing individual elements.
9823 (arm_vectorize_vec_perm_const): Use new_vector to install
9824 the vector in d.perm.
9825 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9826 Update vec_perm_builder constructor.
9827 (rs6000_expand_interleave): Likewise.
9828 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9829 (rs6000_expand_interleave): Likewise.
9831 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9833 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9834 to qimode could truncate the indices.
9835 * optabs.c (expand_vec_perm_var): Likewise.
9837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9839 * Makefile.in (OBJS): Add vec-perm-indices.o.
9840 * vec-perm-indices.h: New file.
9841 * vec-perm-indices.c: Likewise.
9842 * target.h (vec_perm_indices): Replace with a forward class
9844 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9845 * optabs.h: Include vec-perm-indices.h.
9846 (expand_vec_perm): Delete.
9847 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9848 (expand_vec_perm_const): Declare.
9849 * target.def (vec_perm_const_ok): Replace with...
9850 (vec_perm_const): ...this new hook.
9851 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9852 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9853 * doc/tm.texi: Regenerate.
9854 * optabs.def (vec_perm_const): Delete.
9855 * doc/md.texi (vec_perm_const): Likewise.
9856 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9857 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9858 expand_vec_perm for constant permutation vectors. Assert that
9859 the mode of variable permutation vectors is the integer equivalent
9860 of the mode that is being permuted.
9861 * optabs-query.h (selector_fits_mode_p): Declare.
9862 * optabs-query.c: Include vec-perm-indices.h.
9863 (selector_fits_mode_p): New function.
9864 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9865 is defined, instead of checking whether the vec_perm_const_optab
9866 exists. Use targetm.vectorize.vec_perm_const instead of
9867 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9868 fit in the vector mode before using a variable permute.
9869 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9870 vec_perm_indices instead of an rtx.
9871 (expand_vec_perm): Replace with...
9872 (expand_vec_perm_const): ...this new function. Take the selector
9873 as a vec_perm_indices rather than an rtx. Also take the mode of
9874 the selector. Update call to shift_amt_for_vec_perm_mask.
9875 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9876 Use vec_perm_indices::new_expanded_vector to expand the original
9877 selector into bytes. Check whether the indices fit in the vector
9878 mode before using a variable permute.
9879 (expand_vec_perm_var): Make global.
9880 (expand_mult_highpart): Use expand_vec_perm_const.
9881 * fold-const.c: Includes vec-perm-indices.h.
9882 * tree-ssa-forwprop.c: Likewise.
9883 * tree-vect-data-refs.c: Likewise.
9884 * tree-vect-generic.c: Likewise.
9885 * tree-vect-loop.c: Likewise.
9886 * tree-vect-slp.c: Likewise.
9887 * tree-vect-stmts.c: Likewise.
9888 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9890 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9891 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9892 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9893 (aarch64_vectorize_vec_perm_const): ...this new function.
9894 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9895 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9896 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9897 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9898 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9899 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9900 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9902 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9903 check for NEON modes.
9904 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9905 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9906 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9907 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9909 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9910 the old VEC_PERM_CONST conditions.
9911 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9912 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9913 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9914 (ia64_vectorize_vec_perm_const_ok): Merge into...
9915 (ia64_vectorize_vec_perm_const): ...this new function.
9916 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9917 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9918 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9919 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9920 * config/mips/mips.c (mips_expand_vec_perm_const)
9921 (mips_vectorize_vec_perm_const_ok): Merge into...
9922 (mips_vectorize_vec_perm_const): ...this new function.
9923 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9924 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9925 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9926 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9927 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9928 (rs6000_expand_vec_perm_const): Delete.
9929 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9931 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9932 (altivec_expand_vec_perm_const_le): Take each operand individually.
9933 Operate on constant selectors rather than rtxes.
9934 (altivec_expand_vec_perm_const): Likewise. Update call to
9935 altivec_expand_vec_perm_const_le.
9936 (rs6000_expand_vec_perm_const): Delete.
9937 (rs6000_vectorize_vec_perm_const_ok): Delete.
9938 (rs6000_vectorize_vec_perm_const): New function.
9939 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9940 an element count and rtx array.
9941 (rs6000_expand_extract_even): Update call accordingly.
9942 (rs6000_expand_interleave): Likewise.
9943 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9944 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9945 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9946 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9947 (rs6000_expand_vec_perm_const): Delete.
9948 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9949 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9950 (altivec_expand_vec_perm_const_le): Take each operand individually.
9951 Operate on constant selectors rather than rtxes.
9952 (altivec_expand_vec_perm_const): Likewise. Update call to
9953 altivec_expand_vec_perm_const_le.
9954 (rs6000_expand_vec_perm_const): Delete.
9955 (rs6000_vectorize_vec_perm_const_ok): Delete.
9956 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9957 reference to the SPE evmerge intructions.
9958 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9959 an element count and rtx array.
9960 (rs6000_expand_extract_even): Update call accordingly.
9961 (rs6000_expand_interleave): Likewise.
9962 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9963 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9965 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9967 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9969 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9970 vector mode and that that mode matches the mode of the data
9972 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9973 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9974 directly using expand_vec_perm_1 when forcing selectors into
9976 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9978 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9980 * optabs-query.h (can_vec_perm_p): Delete.
9981 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9982 * optabs-query.c (can_vec_perm_p): Split into...
9983 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9984 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9985 particular selector is valid.
9986 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9987 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9988 (vect_grouped_load_supported): Likewise.
9989 (vect_shift_permute_load_chain): Likewise.
9990 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9991 (vect_transform_slp_perm_load): Likewise.
9992 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9993 (vectorizable_bswap): Likewise.
9994 (vect_gen_perm_mask_checked): Likewise.
9995 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9996 implementations of variable permutation vectors into account
9997 when deciding which selector to use.
9998 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9999 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10000 with a false third argument.
10001 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10002 to test whether the constant selector is valid and can_vec_perm_var_p
10003 to test whether a variable selector is valid.
10005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10007 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10008 * optabs-query.c (can_vec_perm_p): Likewise.
10009 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10010 instead of vec_perm_indices.
10011 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10012 (vect_gen_perm_mask_checked): Likewise,
10013 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10014 (vect_gen_perm_mask_checked): Likewise,
10016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10018 * optabs-query.h (qimode_for_vec_perm): Declare.
10019 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10020 (qimode_for_vec_perm): ...this new function.
10021 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10023 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10025 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10026 does not have a conditional at the top.
10028 2018-01-02 Richard Biener <rguenther@suse.de>
10030 * ipa-inline.c (big_speedup_p): Fix expression.
10032 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10035 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10038 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10042 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10043 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10044 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10045 cond_taken_branch_cost 3->4.
10047 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10049 PR tree-optimization/83581
10050 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10051 TODO_cleanup_cfg if any changes have been made.
10053 PR middle-end/83608
10054 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10055 convert_modes if target mode has the right side, but different mode
10058 PR middle-end/83609
10059 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10060 last argument when extracting from CONCAT. If either from_real or
10061 from_imag is NULL, use expansion through memory. If result is not
10062 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10063 the parts directly to inner mode, if even that fails, use expansion
10066 PR middle-end/83623
10067 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10068 check for bswap in mode rather than HImode and use that in expand_unop
10071 Copyright (C) 2018 Free Software Foundation, Inc.
10073 Copying and distribution of this file, with or without modification,
10074 are permitted in any medium without royalty provided the copyright
10075 notice and this notice are preserved.