1 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/85257
4 * fold-const.c (native_encode_vector): If not all elts could fit
5 and off is -1, return 0 rather than offset.
6 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9 adjust buffer in native_interpret_expr call.
11 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13 * config/nds32/constants.md (unspec_volatile_element): Add cache
15 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17 * config/nds32/nds32.c (nds32_cctl_names): New.
18 (nds32_print_operand): Handle cache control register names.
19 * config/nds32/nds32.h (nds32_builtins): New enum values.
20 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22 * config/nds32/nds32.md (type): Add mmu.
23 * config/nds32/pipelines.md (simple_insn): Add mmu.
25 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
27 * config/nds32/nds32.md (type): Remove call.
28 * config/nds32/pipelines.md (simple_insn): Likewise.
30 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
32 * config/nds32/constants.md (unspec_volatile_element): Add
33 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
34 UNSPEC_VOLATILE_FMFCFG.
35 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
36 description for fmfcfg and fmfcsr.
37 (bdesc_1arg): Add fmtcsr.
38 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
39 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
40 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
41 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
42 unspec_fmfcfg): New patterns.
43 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
44 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
45 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
46 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
47 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
48 __nds32__fmfcfg): Define.
50 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
52 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
53 intrinsic register names.
54 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
55 intrinsic register enum values and macros.
57 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
59 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
60 for load/store addressing form.
61 (nds32_print_operand_address): Likewise.
63 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
66 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
67 based on LABEL_REF. Remove useless assertion.
68 (pic_address_needs_scratch): Fix formatting.
69 (sparc_legitimize_pic_address): Minor tweaks.
70 (sparc_delegitimize_address): Adjust assertion accordingly.
71 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
72 into symbolic_operand.
73 (movsi_high_pic_label_ref): Likewise.
74 (movsi_lo_sum_pic_label_ref): Likewise.
75 (movdi_pic_label_ref): Likewise.
76 (movdi_high_pic_label_ref): Likewise.
77 (movdi_lo_sum_pic_label_ref): Likewise.
79 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
81 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
82 custom LIB_SPEC setup.
84 2018-04-06 Ruslan Bukin <br@bsdpad.com>
85 Kito Cheng <kito.cheng@gmail.com>
87 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
88 * config/riscv/freebsd.h: New.
90 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
92 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
93 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
96 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
97 Kito Cheng <kito.cheng@gmail.com>
99 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
100 nds32_output_call, nds32_symbol_binds_local_p): New functions.
101 * config/nds32/nds32-protos.h (nds32_output_call,
102 nds32_output_return): Declare.
103 * config/nds32/nds32.md: Refine all the call and return patterns.
105 2018-04-06 Jakub Jelinek <jakub@redhat.com>
108 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
109 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
111 PR rtl-optimization/84872
112 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
113 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
116 2018-04-06 Tamar Christina <tamar.christina@arm.com>
118 * expr.c (copy_blkmode_to_reg): Revert 254862.
119 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
121 2018-04-06 Richard Biener <rguenther@suse.de>
124 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
125 after seeing a component reference with an adjacent field. Treat
126 refs to arrays at struct end of external decls similar to
127 refs to unconstrained commons.
129 2018-04-06 Jakub Jelinek <jakub@redhat.com>
132 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
133 look through SAVE_EXPRs with non-side-effects argument. Adjust
135 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
138 2018-04-06 Richard Biener <rguenther@suse.de>
141 * alias.c (find_base_term): New wrapper around find_base_term
142 unwinding CSELIB_VAL_PTR changes.
143 (find_base_term): Do not restore CSELIB_VAL_PTR during the
146 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
148 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
150 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
151 constant definitions.
152 ("nop"): lr 0,0 -> nopr r0
153 ("nop_lr0", "nop_lr1"): New insn definitions.
155 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
157 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
158 NDS32_V3PUSH_AVAILABLE_P macro.
160 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
161 Chung-Ju Wu <jasonwucj@gmail.com>
163 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
164 (nds32*-*-*): Add float and fpu_config into supported_defaults.
165 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
166 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
167 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
168 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
169 * config/nds32/constraints.md: New constraints and checking for hard
171 * config/nds32/iterators.md: New mode iterator and attribute for hard
173 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
175 * config/nds32/nds32-fpu.md: New file.
176 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
177 deal with hard float code generation.
178 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
180 (abi_type, float_reg_number): New enum type.
181 * config/nds32/nds32-predicates.c: New predicates for hard float.
182 * config/nds32/nds32-protos.h: Declare functions for hard float.
183 * config/nds32/nds32.c: Implementation for hard float configuration.
184 * config/nds32/nds32.h: Definitions for hard float configuration.
185 * config/nds32/nds32.md: Include hard float machine description and
186 modify patterns for hard float configuration.
187 * config/nds32/nds32.opt: New options for hard float configuration.
188 * config/nds32/predicates.md: New predicates for hard float
191 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
193 * common/config/nds32/nds32-common.c
194 (nds32_option_optimization_table): Enable -mreleax-hint by default.
196 2018-04-05 Jakub Jelinek <jakub@redhat.com>
199 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
200 CONSTRUCTOR_ELT (ctor, ...)->value.
202 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
205 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
207 2018-04-05 Tom de Vries <tom@codesourcery.com>
210 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
213 2018-04-05 Shiva Chen <shiva0217@gmail.com>
214 Kito Cheng <kito.cheng@gmail.com>
216 * config/nds32/constraints.md (U33): Fine-tune checking condition.
217 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
218 * config/nds32/nds32.h (nds32_16bit_address_type): Add
219 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
221 2018-04-05 Shiva Chen <shiva0217@gmail.com>
222 Kito Cheng <kito.cheng@gmail.com>
224 * config/nds32/constraints.md (Ufe): New memory constraint.
225 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
226 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
227 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
229 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
230 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
232 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
234 * config/nds32/nds32.md: Use optimize_size in the condition for
235 alu-shift instructions.
237 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
239 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
241 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
243 * config/nds32/nds32.md (negsi2): Refine pattern.
245 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
246 Chung-Ju Wu <jasonwucj@gmail.com>
248 * config/nds32/iterators.md (shift_rotate): New code iterator.
249 (shift): New code attribute.
250 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
251 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
252 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
253 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
255 (andsi3, *andsi3): Ditto.
256 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
257 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
258 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
259 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
260 nds32_ior_operand, nds32_xor_operand): New predicates.
262 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
264 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
265 (addsi3, subsi3): ... this.
267 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
269 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
271 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
273 * config/nds32/nds32.md: Adjust indention.
275 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
277 * config/nds32/nds32.md (feature): New attribute.
279 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
281 * config/nds32/nds32.md (subtype): New attribute.
283 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
286 * config/arm/arm-builtins.c (arm_expand_builtin): Change
287 expansion to perform a bitwise AND of the argument followed by a
288 boolean negation of the result.
290 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
292 PR rtl-optimization/84878
293 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
294 the basic block. Assert the use reference is not artificial and that
295 it has an associated insn.
297 2018-04-04 Michael Matz <matz@suse.de>
299 * builtins.c (compute_objsize): Pass correct operand
300 to array_at_struct_end_p.
302 2018-04-04 Richard Biener <rguenther@suse.de>
305 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
306 from contexts for DINFO_LEVEL_TERSE and below.
308 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
310 * config/nds32/nds32-doubleword.md (move_<mode>): Require
311 resiter_operand condition.
312 * config/nds32/nds32.md (*move<mode>): Ditto.
314 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
315 Monk Chiang <sh.chiang04@gmail.com>
317 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
319 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
321 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
323 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
324 Kito Cheng <kito.cheng@gmail.com>
326 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
327 nds32_cond_code_str, output_cond_branch,
328 output_cond_branch_compare_zero, nds32_expand_cbranch,
329 nds32_expand_cstore, nds32_expand_movcc,
330 nds32_output_cbranchsi4_equality_zero,
331 nds32_output_cbranchsi4_equality_reg,
332 nds32_output_cbranchsi4_equality_reg_or_const_int,
333 nds32_output_cbranchsi4_greater_less_zero: New functions.
334 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
335 nds32_expand_cstore, nds32_expand_movcc,
336 nds32_output_cbranchsi4_equality_zero,
337 nds32_output_cbranchsi4_equality_reg,
338 nds32_output_cbranchsi4_equality_reg_or_const_int,
339 nds32_output_cbranchsi4_greater_less_zero): Declare.
340 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
341 nds32_rimm11s_operand): New predicates.
342 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
343 * config/nds32/nds32.md: Rewrite all the branch and conditional move
346 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
348 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
349 * config/nds32/nds32.md: Ditto.
350 * config/nds32/pipelines.md: Ditto.
352 2018-04-04 Richard Biener <rguenther@suse.de>
354 PR tree-optimization/85168
355 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
356 propagating abnormals.
358 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
360 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
362 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
363 Kito Cheng <kito.cheng@gmail.com>
365 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
366 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
367 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
368 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
369 * config/nds32/nds32.md (sibcall_internal): New.
370 (sibcall_register): Remove.
371 (sibcall_immediate): Remove.
372 (sibcall_value_internal): New.
373 (sibcall_value_register): Remove.
374 (sibcall_value_immediate): Remove.
375 * config/nds32/predicates.md (nds32_general_register_operand): New.
376 (nds32_call_address_operand): New.
378 2018-04-03 Jakub Jelinek <jakub@redhat.com>
380 PR rtl-optimization/85167
381 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
382 bb_defs if *split_p, instead preinitialize it to NULL.
384 PR tree-optimization/85156
385 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
386 evaluating the argument multiple times.
388 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
390 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
392 (_mm_cvtpd_ps): Likewise.
393 (_mm_cvttpd_epi32): Likewise.
394 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
395 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
396 vector, pixel, and bool following altivec.h include.
398 2018-04-03 Martin Sebor <msebor@redhat.com>
400 * doc/extend.texi (Common Function Attributes): Clarify.
401 (const attribute): Likewise.
402 (pure attribute): Likewise.
404 2018-04-03 Jakub Jelinek <jakub@redhat.com>
407 * config/i386/i386.c (ix86_expand_vector_set): Use
408 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
410 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
412 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
413 instructions when changing rounding bits to preserve precision bits
414 in the x87 control word.
416 2018-04-03 Martin Liska <mliska@suse.cz>
418 PR tree-optimization/82491
419 * rtl.h (strip_offset_and_add): Replace += suboffset with
420 poly_uint64 () + suboffset.
422 2018-03-29 Martin Liska <mliska@suse.cz>
423 Martin Jambor <mjambor@suse.cz>
426 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
427 param_type is not an integral or pointer type.
429 2018-04-03 Richard Biener <rguenther@suse.de>
431 * sese.h (recompute_all_dominators): Remove.
433 2018-04-02 Martin Sebor <msebor@redhat.com>
435 * doc/invoke.texi (-Wrestrict): Fix typos.
437 2018-04-02 Jim Wilson <jimw@sifive.com>
439 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
440 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
441 (<optab>di3, <optab>si3_extend): Likewise.
442 (<optab>si3_mask, <optab>si3_mask_1): New.
443 (<optab>di3_mask, <optab>di3_mask_1): New.
444 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
445 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
446 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
448 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
450 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
453 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
455 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
456 (nds32_canonicalize_comparison): New function.
458 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
459 Kito Cheng <kito.cheng@gmail.com>
460 Kuan-Lin Chen <kuanlinchentw@gmail.com>
462 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
463 * config/nds32/constants.md (unspec_volatile_element): Add
464 UNSPEC_VOLATILE_RELAX_GROUP.
465 * config/nds32/nds32-relax-opt.c: New file.
466 * config/nds32/nds32-predicates.c
467 (nds32_symbol_load_store_p): New function.
468 * config/nds32/nds32-protos.h
469 (nds32_symbol_load_store_p): Declare function.
470 (make_pass_nds32_relax_opt): Declare new rtl pass function.
471 * config/nds32/nds32.c
472 (nds32_register_pass): New function to register pass.
473 (nds32_register_passes): New function to register passes.
474 * config/nds32/nds32.md (relax_group): New pattern.
475 * config/nds32/nds32.opt (mrelax-hint): New option.
476 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
478 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
480 * config/nds32/t-nds32: Modify files dependency.
482 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
484 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
485 (PROFILE_HOOK): Define its implementation.
487 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
489 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
490 type and 32-bit size.
492 2018-04-01 Jakub Jelinek <jakub@redhat.com>
495 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
496 (V_128_256): New mode iterator.
497 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
498 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
499 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
501 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
502 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
504 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
507 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
508 NaN inputs correctly.
510 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
513 * config/rs6000/vsx.md (??r): New mode attribute.
514 (*vsx_mov<mode>_64bit): Use it.
515 (*vsx_mov<mode>_32bit): Likewise.
517 2018-03-30 Martin Sebor <msebor@redhat.com>
519 PR tree-optimization/84818
520 * builtins.c (check_access): Use warning_n.
522 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
525 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
527 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
530 2018-03-30 Julia Koval <julia.koval@intel.com>
533 * x86-tune.def (movx, partial_reg_dependency): Enable for
536 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
539 * lra-constraints.c (process_alt_operands): Move setting
540 this_alternative_matches below.
542 2018-03-29 Martin Liska <mliska@suse.cz>
545 * doc/invoke.texi: Document how LTO works with debug info.
546 Describe auto-load support of binutils. Mention 'x86-64'
547 as valid option value of -march option.
549 2018-03-29 Jakub Jelinek <jakub@redhat.com>
551 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
554 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
555 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
556 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
559 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
562 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
563 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
564 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
565 for RS6000_BTM_POWERPC64.
566 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
567 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
568 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
573 2018-03-28 Carl Love <cel@us.ibm.com>
576 2017-09-27 Carl Love <cel@us.ibm.com>
578 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
579 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
580 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
583 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
585 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
586 instead of __vector bool.
587 (_mm_max_pu8): Likewise.
588 (_mm_min_pi16): Likewise.
590 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
593 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
597 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
599 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
600 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
601 (div_extend): Likewise.
602 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
604 (__builtin_divweuo): Likewise.
605 (__builtin_divdeo): Likewise.
606 (__builtin_divdeuo): Likewise.
608 2018-03-28 Jakub Jelinek <jakub@redhat.com>
611 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
612 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
614 PR tree-optimization/82004
615 * gimple-match-head.c (optimize_pow_to_exp): New function.
616 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
617 Don't fold to exp if optimize_pow_to_exp is false.
619 2018-03-28 Martin Liska <mliska@suse.cz>
622 * calls.c (initialize_argument_information): Fix trailing space.
623 * common.opt: Fix typo and provide better explanation for
624 -fsanitize-coverage option.
625 * config/i386/i386.opt: Fix typo.
627 2018-03-28 Jakub Jelinek <jakub@redhat.com>
628 Martin Liska <mliska@suse.cz>
631 * gimplify.c (asan_poison_variable): Don't do the check for
632 gimplify_omp_ctxp here.
633 (gimplify_decl_expr): Do it here.
634 (gimplify_target_expr): Likewise.
636 2018-03-28 Martin Liska <mliska@suse.cz>
639 * config/i386/i386.c (ix86_function_arg_advance): Do not call
640 chkp_type_bounds_count if MPX is not enabled.
642 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
644 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
646 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
649 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
650 function to create the function decl for complex long double
651 multiply and divide for -mabi=ieeelongdouble.
652 (init_float128_ieee): Call it.
654 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
657 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
658 -fcf-protection=branch -mibt.
659 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
661 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
664 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
666 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
669 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
670 extern array declarations.
672 2018-03-27 Richard Biener <rguenther@suse.de>
675 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
676 explicit single_use checks.
678 2018-03-27 Richard Biener <rguenther@suse.de>
680 PR tree-optimization/85082
681 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
684 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
686 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
687 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
688 Turn on fasynchronous-unwind-tables and funwind-tables.
690 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
693 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
694 (*bmi_blsr_<mode>_ccz): Ditto.
696 2018-03-26 Tom de Vries <tom@codesourcery.com>
698 PR tree-optimization/85063
699 * omp-general.c (offloading_function_p): New function. Factor out
701 * omp-offload.c (pass_omp_target_link::gate): ... here.
702 * omp-general.h (offloading_function_p): Declare.
703 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
704 with attribute omp declare target for offloading functions.
706 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
708 PR tree-optimization/84005
709 * tree-data-ref.h (get_base_for_alignment): Declare.
710 * tree-data-ref.c (get_base_for_alignment_1): New function.
711 (get_base_for_alignment): Likewise.
712 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
713 get_base_for_alignment to find a suitable base object, instead
714 of always using drb->base_address.
716 2018-03-23 Jakub Jelinek <jakub@redhat.com>
719 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
720 known size by default.
722 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
725 * lra-constraints.c (process_alt_operands): Don't match BLKmode
726 and non BLKmode operands.
728 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
731 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
734 2018-03-23 Richard Biener <rguenther@suse.de>
737 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
738 we are going to emit early debug for LTO.
740 2018-03-23 Jakub Jelinek <jakub@redhat.com>
743 * function.c (match_asm_constraints_1): Don't optimize if input
744 doesn't satisfy general_operand predicate for output's mode.
747 * alias.c (write_dependence_p): Don't require for x_canonicalized
748 non-VOIDmode if x has VOIDmode.
751 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
752 just don't try to optimize it rather than assert it never happens.
754 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
756 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
757 macro expansions for definition of ST_INTERNAL_<mode> and
758 LD_INTERNAL_<mode> builtins.
759 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
761 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
763 (altivec_expand_st_builtin): Likewise.
764 (altivec_expand_builtin): Remove calls to deleted functions.
765 (rs6000_address_for_altivec): Delete this function.
766 * config/rs6000/vector.md: Remove expands for
767 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
769 2018-03-22 Sudakshina Das <sudi.das@arm.com>
772 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
773 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
774 re-computing once computed.
775 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
776 (arm_init_machine_status): Initialize
777 machine->static_chain_stack_bytes.
779 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
782 * doc/extend.texi: Add four new prototypes for vec_ld.
783 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
784 definitions for more logical presentation.
785 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
786 entries for V1TI variants of __builtin_altivec_ld builtin.
787 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
788 handling of V1TI variant of LVX icode pattern.
789 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
790 (rs6000_gimple_fold_builtin): Likewise.
791 (altivec_init_builtins): Add code to define
792 __builtin_altivec_lvx_v1ti function.
794 2018-03-22 Jakub Jelinek <jakub@redhat.com>
797 * function.c (match_asm_constraints_1): Don't do the optimization
798 if input isn't a REG, SUBREG, MEM or constant.
800 2018-03-22 Tom de Vries <tom@codesourcery.com>
802 PR tree-optimization/84956
803 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
804 bb_has_abnormal_pred.
806 2018-03-22 Jakub Jelinek <jakub@redhat.com>
809 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
810 DECL_INITIAL (decl) to decl at the end.
811 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
814 2018-03-21 Joseph Myers <joseph@codesourcery.com>
816 * doc/extend.texi (__builtin_tgmath): Document when complex
817 integer types are treated as _Complex _Float64.
819 2018-03-21 Tom de Vries <tom@codesourcery.com>
821 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
823 2018-03-21 Jakub Jelinek <jakub@redhat.com>
825 PR tree-optimization/84960
826 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
827 if it is ENTRY block, move them into single succ of ENTRY in that case.
829 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
831 PR tree-optimization/84811
832 * poly-int.h (poly_span_traits): Remove the T3 parameter and
833 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
834 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
835 (known_subrange_p): Update accordingly. Cast each value involved
836 in the size comparison, rather than casting the result of the
839 2018-03-21 Jakub Jelinek <jakub@redhat.com>
841 PR tree-optimization/84982
842 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
843 by flipping the least significant bit rather than all bits from
844 bitpos to bitpos + bitsize - 1.
846 2018-03-21 Nathan Sidwell <nathan@acm.org>
848 * doc/extend.texi (Deprecated Features): Remove mention of
849 long-deleted deprecations.
851 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
854 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
855 * configure: Regenerate.
857 2018-03-21 Tom de Vries <tom@codesourcery.com>
859 PR tree-optimization/83126
860 * tree-parloops.c (num_phis): New function.
861 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
863 2018-03-21 Nathan Sidwell <nathan@acm.org>
865 * doc/extend.texi (Deprecated Features): Update deprecated flags,
866 mention anon-struct/union members and trailing attributes.
868 2018-03-21 Bin Cheng <bin.cheng@arm.com>
870 PR tree-optimization/84969
871 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
872 builtin memset partitions if they set different rhs values.
874 2018-03-21 Jakub Jelinek <jakub@redhat.com>
876 PR rtl-optimization/84989
877 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
878 VEC_DUPLICATE with scalar result mode.
880 2018-03-21 Martin Liska <mliska@suse.cz>
883 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
884 not intended return statement.
886 2018-03-21 Martin Liska <mliska@suse.cz>
889 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
890 (chkp_find_bound_slots_1): Limit number of iterations.
892 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
895 * Minor grammar fixes for x86 options.
897 2018-03-20 Jakub Jelinek <jakub@redhat.com>
900 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
901 holding REG_CFA_RESTORE notes, instead turn them into a USE.
903 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
906 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
907 (altivec_lvx_<mode>_1op): Likewise.
908 (altivec_stvx_<mode>_2op): Likewise.
909 (altivec_stvx_<mode>_1op): Likewise.
910 (altivec_lvx_<VM2:mode>): New define_expand.
911 (altivec_stvx_<VM2:mode>): Likewise.
912 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
913 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
914 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
915 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
916 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
917 (rs6000_gen_lvx): Likewise.
918 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
919 (altivec_expand_stv_builtin): Likewise.
920 (altivec_expand_builtin): Likewise.
921 * config/rs6000/vector.md: Likewise.
923 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
926 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
929 2018-03-20 Richard Biener <rguenther@suse.de>
932 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
933 sign-conversions as zero, fall back to standard scalar_stmt
936 2018-03-20 Martin Liska <mliska@suse.cz>
939 * predict.c (rebuild_frequencies): Handle case when we have
940 PROFILE_ABSENT, but flag_guess_branch_prob is false.
942 2018-03-20 Jakub Jelinek <jakub@redhat.com>
945 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
946 flag_section_anchors.
947 * varasm.c (use_blocks_for_decl_p): Remove hack for
951 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
953 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
954 be created, use lowpart_subreg of operands[0] rather than operands[0]
956 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
957 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
958 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
959 and n constraint instead of aarch64_shift_imm_di and Usd.
960 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
961 (*aarch64_<optab>_reg_minus<mode>3): ... this.
963 2018-03-20 Sudakshina Das <sudi.das@arm.com>
966 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
967 to favor GPR over NEON registers.
968 (<shift>di3_neon): Likewise.
970 2018-03-20 Tom de Vries <tom@codesourcery.com>
973 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
974 (nvptx_process_pars): Emit bar.sync asap and alap.
976 2018-03-20 Tom de Vries <tom@codesourcery.com>
979 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
980 seen_label if seen_label is already set.
982 2018-03-20 Jakub Jelinek <jakub@redhat.com>
985 * config/i386/i386.c (fold_builtin_cpu): For features above 31
986 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
987 Use 1U instead of 1. Formatting fixes.
990 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
991 instead of TREE_TYPE (s1) for the return value.
993 2018-03-19 Jakub Jelinek <jakub@redhat.com>
995 PR tree-optimization/84946
996 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
997 bitsize + bitsize in poly_uint64 rather than poly_int64.
1000 * dwarf2asm.c: Include fold-const.c.
1001 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1002 of decl rather than decl itself.
1004 PR rtl-optimization/84643
1005 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1007 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1010 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1011 calling assemble_variable.
1013 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1016 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1017 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1019 2018-03-19 Jim Wilson <jimw@sifive.com>
1022 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1023 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1024 (riscv_first_stack_step): Likewise.
1025 (riscv_option_override): Use STACK_BOUNDARY instead of
1027 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1029 (BIGGEST_ALIGNMENT): Set to 128.
1030 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1031 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1034 2018-03-19 Richard Biener <rguenther@suse.de>
1036 PR tree-optimization/84933
1037 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1038 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1040 2018-03-19 Richard Biener <rguenther@suse.de>
1042 PR tree-optimization/84859
1043 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1044 (cond_if_else_store_replacement): Perform sinking operation on
1045 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1046 Generalize what a BB with a single eligible store is.
1048 2018-03-19 Richard Biener <rguenther@suse.de>
1050 PR tree-optimization/84929
1051 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1052 chrec_is_positive against non-chrec arg.
1054 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1057 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1059 2018-03-18 Martin Liska <mliska@suse.cz>
1061 PR rtl-optimization/84635
1062 * regrename.c (build_def_use): Use matches_mode only when
1065 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1067 PR tree-optimization/84913
1068 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1069 vectorize chains of COND_EXPRs.
1071 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1073 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1075 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1077 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1079 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1081 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1083 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1084 Kito Cheng <kito.cheng@gmail.com>
1086 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1087 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1088 (nds32_adjust_reg_alloc_order): New function.
1089 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1091 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1093 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1094 nds32_print_operand, nds32_print_operand_address): Use
1095 HOST_WIDE_INT_PRINT_DEC instead.
1097 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1101 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1104 * config/i386/i386.c (initial_ix86_tune_features,
1105 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1107 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1108 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1109 rather than 1u << ix86_tune. Formatting fix.
1110 (ix86_option_override_internal): Change ix86_arch_mask from
1111 unsigned int to unsigned HOST_WIDE_INT, initialize to
1112 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1113 (ix86_function_specific_restore): Likewise.
1115 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1118 * postreload.c (reload_combine_recognize_pattern): Perform
1119 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1120 truncate_int_for_mode the result for the destination's mode.
1123 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1124 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1126 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1129 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1132 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1133 regno_allocno_class_array and sorted_pseudos.
1134 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1135 insns where regno is used.
1137 2018-03-16 Martin Liska <mliska@suse.cz>
1140 * multiple_target.c (create_dispatcher_calls): Redirect
1141 reference in the symbol table.
1143 2018-03-16 Martin Liska <mliska@suse.cz>
1146 * multiple_target.c (create_dispatcher_calls): Redirect also
1149 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1153 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1154 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1155 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1157 2018-03-16 Julia Koval <julia.koval@intel.com>
1159 * doc/invoke.texi (Skylake Server): Add CLWB.
1160 Cannonlake): Remove CLWB.
1162 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1164 PR tree-optimization/84841
1165 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1167 (FLOAT_ONE_CONST_TYPE): Define.
1168 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1169 (sort_by_operand_rank): Put entries with higher constant_type last
1170 rather than first to match comments.
1172 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1174 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1177 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1180 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1181 check and use address of target always.
1183 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1186 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1187 (indirect_thunk_bnd_needed): Likewise.
1188 (indirect_thunks_used): Likewise.
1189 (indirect_thunks_bnd_used): Likewise.
1190 (indirect_return_needed): New.
1191 (indirect_return_bnd_needed): Likewise.
1192 (output_indirect_thunk_function): Add a bool argument for
1194 (output_indirect_thunk_function): Don't generate alias for
1195 function return thunk.
1196 (ix86_code_end): Call output_indirect_thunk_function to generate
1197 function return thunks.
1198 (ix86_output_function_return): Set indirect_return_bnd_needed
1199 and indirect_return_needed instead of indirect_thunk_bnd_needed
1200 and indirect_thunk_needed.
1202 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1204 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1205 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1206 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1208 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1209 Paul Hua <paul.hua.gm@gmail.com>
1212 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1214 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1216 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1217 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1220 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1223 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1224 instead of GET_MODE_SIZE when comparing Units.
1226 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1229 * varasm.c (hash_section): Return an unchangeble hash value
1230 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1231 Return !aarch64_can_use_per_function_literal_pools_p ().
1233 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1236 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1237 rather than address of cmode as last argument to prepare_cmp_insn.
1239 2018-03-15 Julia Koval <julia.koval@intel.com>
1241 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1242 F_AVX512VNNI, F_AVX512BITALG): New.
1244 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1247 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1248 insn for floating-point loads and stores.
1250 2018-03-14 Carl Love <cel@us.ibm.com>
1252 * config/rs6000/rs6000-c.c: Add macro definitions for
1253 ALTIVEC_BUILTIN_VEC_PERMXOR.
1254 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1255 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1256 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1257 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1259 * config/doc/extend.texi: Add prototypes for vec_permxor.
1261 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1264 * diagnostic-show-locus.c (class layout_point): Convert m_line
1265 from int to linenum_type.
1266 (line_span::comparator): Use linenum "compare" function when
1267 comparing line numbers.
1268 (test_line_span): New function.
1269 (layout_range::contains_point): Convert param "row" from int to
1271 (layout_range::intersects_line_p): Likewise.
1272 (layout::will_show_line_p): Likewise.
1273 (layout::print_source_line): Likewise.
1274 (layout::should_print_annotation_line_p): Likewise.
1275 (layout::print_annotation_line): Likewise.
1276 (layout::print_leading_fixits): Likewise.
1277 (layout::annotation_line_showed_range_p): Likewise.
1278 (struct line_corrections): Likewise for field m_row.
1279 (line_corrections::line_corrections): Likewise for param "row".
1280 (layout::print_trailing_fixits): Likewise.
1281 (layout::get_state_at_point): Likewise.
1282 (layout::get_x_bound_for_row): Likewise.
1283 (layout::print_line): Likewise.
1284 (diagnostic_show_locus): Likewise for locals "last_line" and
1286 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1287 * input.c (selftest::test_linenum_comparisons): New function.
1288 (selftest::input_c_tests): Call it.
1289 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1290 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1291 * selftest.h (ASSERT_GT): New macro.
1292 (ASSERT_GT_AT): New macro.
1293 (ASSERT_LT): New macro.
1294 (ASSERT_LT_AT): New macro.
1296 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1298 PR rtl-optimization/84780
1299 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1301 2018-03-14 Martin Liska <mliska@suse.cz>
1303 * tree.c (record_node_allocation_statistics): Use
1304 get_stats_node_kind.
1305 (get_stats_node_kind): New function extracted from
1306 record_node_allocation_statistics.
1307 (free_node): Use get_stats_node_kind.
1309 2018-03-14 Richard Biener <rguenther@suse.de>
1311 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1312 that the value-set of ANTIC_IN doesn't grow.
1315 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1317 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1318 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1320 2018-03-14 Julia Koval <julia.koval@intel.com>
1322 * config.gcc (icelake-client, icelake-server): New.
1324 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1325 (initial_ix86_arch_features): Ditto.
1326 (PTA_SKYLAKE): Add SGX.
1327 (PTA_ICELAKE): Remove.
1328 (PTA_ICELAKE_CLIENT): New.
1329 (PTA_ICELAKE_SERVER): New.
1330 (ix86_option_override_internal): Split up icelake on icelake client and
1332 (get_builtin_code_for_version): Ditto.
1333 (fold_builtin_cpu): Ditto.
1334 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1335 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1336 * config/i386/i386.h (processor_type): Ditto.
1337 * doc/invoke.texi: Ditto.
1339 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1342 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1343 INTEGER_CST offset, add it together with bitpos / 8 and
1344 sign extend based on POINTER_SIZE.
1348 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1351 * config/i386/constraints.md (Yc): New register constraint.
1352 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1353 Use Yc constraint for alternative 2 of operand 0. Remove
1354 preferred_for_speed attribute.
1356 2018-03-14 Richard Biener <rguenther@suse.de>
1358 PR tree-optimization/84830
1359 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1360 with the old one to avoid oscillations.
1362 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1365 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1367 (assign_by_spills): Return a flag of reload assignment failure.
1368 Do not process the reload assignment failures. Do not spill other
1369 reload pseudos if they has the same reg class. Update n if
1371 (lra_assign): Add a return arg. Set up from the result of
1372 assign_by_spills call.
1373 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1374 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1375 usage_insns if it is not NULL.
1376 (spill_hard_reg_in_range): New function.
1377 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1378 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1379 function prototypes.
1380 (lra_assign): Change prototype.
1381 * lra.c (lra): Add code to deal with fails by splitting hard reg
1384 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1386 * config/riscv/riscv.opt (mrelax): New option.
1387 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1388 "norelax" when riscv_mrelax is disabled.
1389 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1391 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1394 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1395 reassociation for int modes.
1397 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1399 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1400 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1402 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1403 * config/aarch64/aarch64-sve.md
1404 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1405 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1406 (*extend<mode><Vwide>2): Rename to...
1407 (aarch64_sve_extend<mode><Vwide>2): ...this.
1408 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1409 renaming the old pattern to...
1410 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1412 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1413 define_expand, renaming the old pattern to...
1414 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1415 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1416 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1417 account when deciding which SVE instruction the optab should use.
1418 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1420 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1422 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1423 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1424 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1425 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1426 (tlsdesc_small_<mode>): Turn a define_expand and use
1427 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1428 (tlsdesc_small_advsimd_<mode>): ...this.
1429 (tlsdesc_small_sve_<mode>): New pattern.
1431 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1433 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1434 (UNSPEC_UMUL_HIGHPART): New constants.
1435 (MUL_HIGHPART): New int iteraor.
1436 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1437 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1439 (*<su>mul<mode>3_highpart): New define_insn.
1441 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1444 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1447 2018-03-13 Martin Liska <mliska@suse.cz>
1450 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1452 (sem_item_optimizer::~sem_item_optimizer): Release it.
1453 (sem_item_optimizer::merge_classes): Register variable aliases.
1454 (sem_item_optimizer::fixup_pt_set): New function.
1455 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1456 * ipa-icf.h: Declare new variables and functions.
1458 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1461 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1462 integer_pow2p@2 and test integer_pow2p in condition.
1463 (A < 0 ? C : 0): Similarly for @1.
1466 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1467 characters starting at p contain '\0' character, don't look beyond
1471 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1472 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1475 * reg-stack.c (change_stack): Change update_end var from int to
1476 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1477 also call set_block_for_insn on the newly added insns and rescan.
1480 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1481 on the last operand.
1484 * tree.c (stabilize_reference_1): Return save_expr (e) for
1485 STATEMENT_LIST even if it doesn't have side-effects.
1487 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1489 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1491 2018-03-12 Renlin Li <renlin.li@arm.com>
1493 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1494 aarch64_output_scalar_simd_mov_immediate.
1496 2018-03-12 Martin Sebor <msebor@redhat.com>
1498 PR tree-optimization/83456
1499 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1500 for perfectly overlapping calls to memcpy.
1501 (gimple_fold_builtin_memory_chk): Same.
1502 (gimple_fold_builtin_strcpy): Handle no-warning.
1503 (gimple_fold_builtin_stxcpy_chk): Same.
1504 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1506 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1508 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1509 parameter. Use it for SFmode.
1510 (rs6000_function_arg_advance_1): Adjust.
1511 (rs6000_function_arg): Adjust.
1512 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1514 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1516 PR rtl-optimization/84169
1517 PR rtl-optimization/84780
1518 * combine.c (can_combine_p): Check for a 2-insn combination whether
1519 the destination register is used between the two insns, too.
1521 2018-03-12 Richard Biener <rguenther@suse.de>
1523 PR tree-optimization/84803
1524 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1525 for refs DR analysis didn't process.
1527 2018-03-12 Richard Biener <rguenther@suse.de>
1529 PR tree-optimization/84777
1530 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1531 force-vectorize loops ignore whether we are optimizing for size.
1533 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1535 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1536 (TARGET_MD_ASM_ADJUST): Define.
1538 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1539 Kito Cheng <kito.cheng@gmail.com>
1540 Chung-Ju Wu <jasonwucj@gmail.com>
1542 * config/nds32/nds32.c (nds32_compute_stack_frame,
1543 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1544 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1545 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1546 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1547 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1548 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1549 * config/nds32/nds32.md (prologue, epilogue): Use macro
1550 NDS32_V3PUSH_AVAILABLE_P to do checking.
1552 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1555 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1556 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1557 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1558 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1559 addition of most attributes on !orig_type_die or the attribute not
1560 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1562 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1563 Chung-Ju Wu <jasonwucj@gmail.com>
1565 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1567 * config/nds32/nds32.opt (mvh): New option.
1569 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1570 Chung-Ju Wu <jasonwucj@gmail.com>
1572 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1574 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1575 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1578 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1579 Chung-Ju Wu <jasonwucj@gmail.com>
1581 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1583 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1584 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1586 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1587 Kito Cheng <kito.cheng@gmail.com>
1588 Chung-Ju Wu <jasonwucj@gmail.com>
1590 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1591 UNSPEC_FFMISM and UNSPEC_FLMISM.
1592 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1593 for ffb, ffmism and flmism.
1594 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1595 (unspec_ffmism): Ditto.
1596 (unspec_flmism): Ditto.
1597 (nds32_expand_builtin_impl): Check if string extension is available.
1598 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1599 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1601 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1604 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1607 * lra-assigns.c (assign_by_spills): Return a flag of reload
1608 assignment failure. Do not process the reload assignment
1609 failures. Do not spill other reload pseudos if they has the same
1611 (lra_assign): Add a return arg. Set up from the result of
1612 assign_by_spills call.
1613 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1614 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1615 usage_insns if it is not NULL.
1616 (spill_hard_reg_in_range): New function.
1617 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1618 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1619 function prototypes.
1620 (lra_assign): Change prototype.
1621 * lra.c (lra): Add code to deal with fails by splitting hard reg
1624 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1627 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1629 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1632 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1633 (dw_val_node): Add val_symbolic_view.
1634 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1635 (symview_upper_bound): New.
1636 (new_line_info_table): Initialize symviews_since_reset.
1637 (dwarf2out_source_line): Count symviews_since_reset and set
1638 symview_upper_bound.
1639 (dw_val_equal_p): Handle symview.
1640 (add_AT_symview): New.
1641 (print_dw_val): Handle symview.
1642 (attr_checksum, attr_checksum_ordered): Likewise.
1643 (same_dw_val_p, size_of_die): Likewise.
1644 (value_format, output_die): Likewise.
1645 (add_high_low_attributes): Use add_AT_symview for entry_view.
1646 (dwarf2out_finish): Reset symview_upper_bound, clear
1649 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1652 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1653 Add strict argument and use it.
1654 (rs6000_split_multireg_move): Update for new strict argument.
1655 (mem_operand_gpr): Disallow all non-offsettable addresses.
1656 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1658 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1661 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1662 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1663 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1666 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1667 decl, use remap_type if we want to use the type.
1669 2018-03-09 Martin Sebor <msebor@redhat.com>
1671 PR tree-optimization/84526
1672 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1674 (builtin_access::generic_overlap): Be prepared to handle non-array
1677 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1679 PR rtl-optimization/84682
1680 * lra-constraints.c (process_address_1): Check is_address flag
1681 for address constraints.
1682 (process_alt_operands): Likewise.
1683 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1684 preprocess_constraints.
1685 * recog.h (preprocess_constraints): Add oploc parameter.
1687 * recog.c (preprocess_constraints): Test address_operand for
1688 CT_ADDRESS constraints.
1690 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1693 * lra-assigns.c (assign_by_spills): Return a flag of reload
1694 assignment failure. Do not process the reload assignment
1695 failures. Do not spill other reload pseudos if they has the same
1697 (lra_assign): Add a return arg. Set up from the result of
1698 assign_by_spills call.
1699 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1700 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1701 usage_insns if it is not NULL.
1702 (spill_hard_reg_in_range): New function.
1703 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1704 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1705 function prototypes.
1706 (lra_assign): Change prototype.
1707 * lra.c (lra): Add code to deal with fails by splitting hard reg
1710 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1713 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1714 Accept complain bool parameter. Only emit errors if it is true.
1715 (arm_parse_cpu_option_name): Likewise.
1716 (arm_target_thumb_only): Adjust callers of the above.
1717 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1718 prototype to take a default true bool parameter.
1719 (arm_parse_arch_option_name): Likewise.
1721 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1722 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1726 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1727 * configure: Regenerate.
1728 * configure.ac ("linker --version-script option"): New.
1729 ("linker soname option"): New.
1731 2018-03-09 Richard Biener <rguenther@suse.de>
1733 PR tree-optimization/84775
1734 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1735 immediate uses of predicate stmts and mark them modified.
1738 PR tree-optimization/84178
1739 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1741 (version_loop_for_if_conversion): Delay update_ssa call.
1742 (tree_if_conversion): Delay update_ssa until after predicate
1745 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1748 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1749 when the function accesses prior frames.
1751 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1754 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1755 gen_llsym, otherwise call maybe_gen_llsym.
1758 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1759 has ',' character inside of it.
1761 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1764 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1765 as clobbering CC_REGNUM.
1767 2018-03-08 Richard Biener <rguenther@suse.de>
1770 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1771 (follow_copies_to_constant): Do not follow SSA names registered
1774 2018-03-08 Richard Biener <rguenther@suse.de>
1776 PR tree-optimization/84178
1777 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1779 (version_loop_for_if_conversion): Delay update_ssa call.
1780 (tree_if_conversion): Delay update_ssa until after predicate
1783 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1785 PR tree-optimization/84178
1786 * tree-if-conv.c (release_bb_predicate): Remove the
1787 the assertion that the stmts have NULL use_ops.
1788 Discard the statements, asserting that they haven't
1789 yet been added to a BB.
1791 2018-03-08 Richard Biener <rguenther@suse.de>
1793 PR tree-optimization/84746
1794 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1795 (phi_translate): Pass in destination ANTIC_OUT set.
1796 (phi_translate_1): Likewise. For a simplified result lookup
1797 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1798 (phi_translate_set): Adjust.
1799 (do_pre_regular_insertion): Likewise.
1800 (do_pre_partial_partial_insertion): Likewise.
1802 2018-03-08 Martin Liska <mliska@suse.cz>
1804 PR gcov-profile/84735
1805 * doc/gcov.texi: Document usage of profile files.
1806 * gcov-io.h: Document changes in the format.
1808 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1812 * dwarf2out.c (struct dw_line_info_table): Update comments for
1814 (FORCE_RESET_NEXT_VIEW): New.
1815 (FORCE_RESETTING_VIEW_P): New.
1816 (RESETTING_VIEW_P): Check for -1 too.
1817 (ZERO_VIEW_P): Likewise.
1818 (new_line_info_table): Force-reset next view.
1819 (dwarf2out_begin_function): Likewise.
1820 (dwarf2out_source_line): Simplify zero_view_p initialization.
1821 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1822 view directly. Omit view when omitting .loc at line 0.
1824 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1826 PR tree-optimization/84740
1827 * tree-switch-conversion.c (process_switch): Call build_constructors
1828 only if info.phi_count is non-zero.
1830 PR tree-optimization/84739
1831 * tree-tailcall.c (find_tail_calls): Check call arguments against
1832 DECL_ARGUMENTS (current_function_decl) rather than
1833 DECL_ARGUMENTS (func) when checking for tail recursion.
1835 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1837 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1838 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1839 Volker Reichelt's entry and add entries for people that perform
1840 GCC fuzzy testing and report numerous bugs.
1842 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1845 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1846 readonly data in sdata, if that is disabled.
1847 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1848 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1849 -mreadonly-in-sdata option.
1851 2018-03-07 Martin Sebor <msebor@redhat.com>
1853 PR tree-optimization/84468
1854 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1855 basic block when looking for nul assignment.
1857 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1860 * except.h (output_function_exception_table): Adjust prototype.
1861 * except.c (output_function_exception_table): Remove FNNAME parameter
1862 and add SECTION parameter. Ouput one part of the table at a time.
1863 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1864 the first part of the exception table and emit unwind directives.
1865 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1866 (i386_pe_seh_cold_init): Likewise.
1867 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1868 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1869 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1870 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1871 * config/i386/winnt.c: Include except.h.
1872 (struct seh_frame_state): Add reg_offset, after_prologue and
1873 in_cold_section fields.
1874 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1875 (i386_pe_seh_cold_init): New function.
1876 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1877 to seh->in_cold_section.
1878 (seh_emit_push): Record the offset of the push.
1879 (seh_emit_save): Record the offet of the save.
1880 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1881 Test seh->after_prologue to disregard the epilogue.
1882 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1883 (i386_pe_end_cold_function): New function.
1885 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1888 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1889 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1892 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1894 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1895 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1898 * multiple_target.c: Include tree-inline.h and intl.h.
1899 (expand_target_clones): Diagnose and fail if node->definition and
1900 !tree_versionable_function_p (node->decl).
1902 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1904 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1906 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1907 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1908 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1910 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1913 * combine.c (try_combine): Use reg_or_subregno instead of handling
1914 just paradoxical SUBREGs and REGs.
1916 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1918 * config/arc/arc.c (arc_finalize_pic): Remove function.
1919 (arc_must_save_register): We use single base PIC register, remove
1920 checks to save/restore the PIC register.
1921 (arc_expand_prologue): Likewise.
1922 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1924 (arc_verify_short): Likewise.
1925 (arc_attr_type): Likewise.
1926 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1927 (walk_stores): Likewise.
1928 (arc_address_cost): Make it static.
1929 (arc_verify_short): Likewise.
1930 (branch_dest): Likewise.
1931 (arc_attr_type): Likewise.
1932 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1933 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1934 (arc_final_prescan_insn): Remove inserting the nops due to
1935 hardware hazards. It is done in reorg step.
1936 (insn_length_variant_t): Remove.
1937 (insn_length_parameters_t): Likewise.
1938 (arc_insn_length_parameters): Likewise.
1939 (arc_get_insn_variants): Likewise.
1940 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1942 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1945 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1948 PR tree-optimization/84687
1949 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1951 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1953 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1955 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1956 Rename to ppc_speculation_barrier.
1957 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1958 __builtin_ppc_speculation_barrier.
1960 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1963 * combine.c (combine_simplify_rtx): Don't try to simplify if
1964 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1967 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1969 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1970 to 32 bytes when compiling for POWER9.
1972 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1975 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1976 regparm >= 3 with no arg reg available also for calls with
1977 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1980 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1982 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1984 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1987 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1989 2018-03-05 Richard Biener <rguenther@suse.de>
1991 PR tree-optimization/84486
1992 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1993 When inserting a __builtin_assume_aligned call set the LHS
1994 SSA name alignment info accordingly.
1996 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1998 PR tree-optimization/84114
1999 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2000 Avoid reassociation of FLOAT_MODE addition.
2002 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2004 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2005 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2006 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2007 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2008 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2009 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2010 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2012 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2013 __builtin_ia32_wbinvd): New builtins.
2014 (SPECIAL_ARGS2): New.
2015 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2016 (SPECIAL_ARGS2): New.
2017 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2018 (ix86_valid_target_attribute_inner_p): Ditto.
2019 (ix86_init_mmx_sse_builtins): Add special_args2.
2020 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2021 TARGET_WBNOINVD_P): New.
2022 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2023 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2024 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2025 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2026 * config/i386/pconfigintrin.h: New file.
2027 * config/i386/wbnoinvdintrin.h: Ditto.
2028 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2029 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2031 2018-03-05 Richard Biener <rguenther@suse.de>
2033 PR tree-optimization/84670
2034 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2036 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2037 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2038 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2039 doesn't grow if all successors (recursively) were visited at least
2042 2018-03-05 Richard Biener <rguenther@suse.de>
2044 PR tree-optimization/84650
2045 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2046 if executed in the loop pipeline.
2048 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2050 * doc/configfiles.texi (Configuration Files): Move info about
2051 conditionalizing $target-protos.h to...
2052 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2053 differs from $target-protos.h.
2055 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2056 Chung-Ju Wu <jasonwucj@gmail.com>
2058 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2059 * config/nds32/nds32-multiple.md (setmemsi): Define.
2060 * config/nds32/nds32-memory-manipulation.c
2061 (nds32_gen_dup_4_byte_to_word_value): New.
2062 (emit_setmem_word_loop): New.
2063 (emit_setmem_byte_loop): New.
2064 (nds32_expand_setmem_loop): New.
2065 (nds32_expand_setmem_loop_v3m): New.
2066 (nds32_expand_setmem_unroll): New.
2067 (nds32_expand_setmem): New.
2069 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2070 Chung-Ju Wu <jasonwucj@gmail.com>
2072 * config/nds32/nds32-memory-manipulation.c
2073 (nds32_emit_load_store): New.
2074 (nds32_emit_post_inc_load_store): New.
2075 (nds32_emit_mem_move): New.
2076 (nds32_emit_mem_move_block): New.
2077 (nds32_expand_movmemsi_loop_unknown_size): New.
2078 (nds32_expand_movmemsi_loop_known_size): New.
2079 (nds32_expand_movmemsi_loop): New.
2080 (nds32_expand_movmemsi_unroll): New.
2081 (nds32_expand_movmemqi): Rename ...
2082 (nds32_expand_movmemsi): ... to this.
2083 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2084 (movmemsi): ... to this.
2085 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2086 (nds32_expand_movmemsi): ... to this.
2088 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2089 Monk Chiang <sh.chiang04@gmail.com>
2090 Chung-Ju Wu <jasonwucj@gmail.com>
2092 * config/nds32/nds32-protos.h
2093 (nds32_expand_load_multiple): New arguments.
2094 (nds32_expand_store_multiple): Ditto.
2095 (nds32_valid_multiple_load_store): Rename ...
2096 (nds32_valid_multiple_load_store_p): ... to this.
2097 * config/nds32/nds32-memory-manipulation.c
2098 (nds32_expand_load_multiple): Refine implementation.
2099 (nds32_expand_store_multiple): Ditto.
2100 * config/nds32/nds32-multiple.md
2101 (load_multiple): Update nds32_expand_load_multiple interface.
2102 (store_multiple): Update nds32_expand_store_multiple interface.
2103 * config/nds32/nds32-predicates.c
2104 (nds32_valid_multiple_load_store): Rename ...
2105 (nds32_valid_multiple_load_store_p): ... to this and refine
2107 * config/nds32/predicates.md
2108 (nds32_load_multiple_and_update_address_operation): New predicate.
2109 (nds32_store_multiple_and_update_address_operation): New predicate.
2111 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2112 Chung-Ju Wu <jasonwucj@gmail.com>
2114 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2115 (combo): New attribute.
2116 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2118 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2120 * config/nds32/nds32.opt: Change -mcmodel= default value.
2122 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2123 Monk Chiang <sh.chiang04@gmail.com>
2124 Chung-Ju Wu <jasonwucj@gmail.com>
2126 * config/nds32/constants.md (unspec_element): New enum.
2127 * config/nds32/constraints.md (Umw): New constraint.
2128 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2129 * config/nds32/nds32-intrinsic.md: Likewise.
2130 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2131 (nds32_valid_smw_lwm_base_p): New.
2132 (nds32_output_smw_single_word): New.
2133 (nds32_output_lmw_single_word): New.
2134 (nds32_expand_unaligned_load): New.
2135 (nds32_expand_unaligned_store): New.
2136 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2137 (nds32_output_smw_single_word): Declare.
2138 (nds32_output_lmw_single_word): Declare.
2139 (nds32_expand_unaligned_load): Declare.
2140 (nds32_expand_unaligned_store): Declare.
2141 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2142 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2143 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2144 NDS32_BUILTIN_UASTORE_DW.
2145 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2148 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2149 Kito Cheng <kito.cheng@gmail.com>
2150 Chung-Ju Wu <jasonwucj@gmail.com>
2152 * config/nds32/nds32-intrinsic.c
2153 (nds32_expand_builtin_null_ftype_reg): Delete.
2154 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2155 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2156 (nds32_read_argument): New.
2157 (nds32_legitimize_target): Ditto.
2158 (nds32_legitimize_argument): Ditto.
2159 (nds32_check_constant_argument): Ditto.
2160 (nds32_expand_unop_builtin): Ditto.
2161 (nds32_expand_unopimm_builtin): Ditto.
2162 (nds32_expand_binop_builtin): Ditto.
2163 (nds32_builtin_decl_impl): Ditto.
2164 (builtin_description): Ditto.
2165 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2166 (nds32_init_builtins_impl): Ditto.
2167 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2168 (nds32_builtin_decl): New.
2169 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2170 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2172 2018-03-02 Jeff Law <law@redhat.com>
2174 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2175 (redundant_insn, fill_simple_delay_slots): Likewise.
2176 (fill_slots_from_thread): Likewise.
2177 * resource.c (mark_referenced_resources): Likewise.
2178 (mark_set_resources, find_dead_or_set_registers): Likewise.
2180 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2182 * substring-locations.h (format_warning_va): Formatting fix for
2184 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2186 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2187 * substring-locations.c: Include intl.h.
2188 (format_warning_va): Turned into small wrapper around
2189 format_warning_n_va, renamed to ...
2190 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2191 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2193 (format_warning_at_substring_n): New function.
2194 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2195 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2196 format_warning_at_substring with just a shorter name instead of
2197 const function pointer.
2198 (fmtwarn_n): New function.
2199 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2200 appropriate, get rid of all the fmtstr temporaries, move conditionals
2201 with G_() wrapped string literals directly into fmtwarn arguments,
2202 cast dir.len to (int), formatting fixes.
2204 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2206 * doc/invoke.texi: Remove "Cilk Plus" references.
2208 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2209 Richard Biener <rguenther@suse.de>
2212 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2213 for error or warning attributes if CALL_FROM_THUNK_P is set.
2216 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2219 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2220 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2223 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2224 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2226 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2227 instead of -1U in last predictors element's probability member.
2229 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2232 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2233 arguments if they are comparable.
2235 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2237 PR tree-optimization/84634
2238 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2239 masks and masked_loop_p with a single loop_masks, making sure it's
2240 null for bb vectorization.
2242 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2244 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2245 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2246 loop->force_vectorize to check whether there is no alias.
2248 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2251 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2253 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2255 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2256 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2257 instead of a loop around prev_real_insn.
2258 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2262 * config/i386/i386.c (ix86_print_operand): Use conditional
2263 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2266 2018-03-02 Richard Biener <rguenther@suse.de>
2268 PR tree-optimization/84427
2269 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2270 (bitmap_set_subtract_values): Rewrite to handle multiple
2273 (prune_clobbered_mems): Likewise.
2274 (phi_translate): Take edge instead of pred/phiblock.
2275 (phi_translate_1): Likewise.
2276 (phi_translate_set): Likewise. Insert all translated
2277 exprs for a value into the set, keeping possibly multiple
2278 expressions per value.
2279 (compute_antic_aux): Adjust for phi_translate changes.
2280 When intersecting union the expressions and prune those
2281 not in the final value set, keeping possibly multiple
2282 expressions per value. Do not use value-insertion
2283 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2284 all expressions. Add verification that the value-sets
2285 only shrink during iteration.
2286 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2287 (do_pre_regular_insertion): Likewise.
2288 (do_pre_partial_partial_insertion): Likewise.
2290 2018-03-02 Richard Biener <rguenther@suse.de>
2293 * config/darwin.c (saved_debug_info_level): New static global.
2294 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2295 (darwin_asm_lto_end): Restore debug info generation settings.
2297 2018-03-01 Martin Liska <mliska@suse.cz>
2300 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2303 2018-03-01 Richard Biener <rguenther@suse.de>
2306 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2307 type annotation with LTO when debug was disabled at compile-time.
2309 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2311 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2313 (mips_final_postscan_insn): Likewise.
2315 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2317 PR rtl-optimization/84528
2318 * alias.c (init_alias_target): Add commentary.
2319 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2320 a unique base value if the frame pointer is not eliminated
2321 to the stack pointer.
2323 2018-03-01 Tom de Vries <tom@codesourcery.com>
2325 PR rtl-optimization/83327
2326 * lra-int.h (hard_regs_spilled_into): Declare.
2327 * lra.c (hard_regs_spilled_into): Define.
2328 (init_reg_info): Init hard_regs_spilled_into.
2329 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2330 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2331 (process_bb_lives): Handle hard_regs_spilled_into.
2332 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2333 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2335 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2337 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2338 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2339 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2340 * config/rs6000/aix72.h: New file.
2342 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2344 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2345 instead of warning_at with conditional singular and plural messages
2349 * stor-layout.c (update_alignment_for_field): For
2350 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2351 && !DECL_PACKED (field), do the alignment update, just use
2352 only desired_align instead of MAX (type_align, desired_align)
2354 (place_field): Don't do known_align < desired_align handling
2355 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2356 is non-NULL, instead do it after rli->prev_field handling and
2357 only if not within a bitfield word. For DECL_PACKED (field)
2358 use type_align of BITS_PER_UNIT.
2360 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2362 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2363 superfluous parentheses and trailing spaces.
2365 2018-02-28 Richard Biener <rguenther@suse.de>
2367 PR tree-optimization/84584
2368 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2369 SCoPs with fake exit edge.
2371 2018-02-28 Martin Liska <mliska@suse.cz>
2374 * timevar.c (timer::print): Fix format to properly print 100%
2377 2018-02-28 Richard Biener <rguenther@suse.de>
2380 * genmatch.c (capture_info::walk_match): Do not mark
2381 captured expressions without operands as expr_p given
2382 they act more like predicates and should be subject to
2383 "lost tail" side-effect preserving.
2385 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2387 PR rtl-optimization/81611
2388 * auto-inc-dec.c (attempt_change): Move dead note from
2389 mem_insn if it's the next use of regno
2390 (find_address): Take address use of reg holding
2391 non-incremented value. Add parm to limit search to the named
2393 (merge_in_block): Attempt to use a mem insn that is the next
2394 use of the original regno.
2396 2018-02-27 Martin Sebor <msebor@redhat.com>
2399 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2400 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2402 2018-02-27 Martin Sebor <msebor@redhat.com>
2404 PR translation/84207
2405 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2406 n argument to unsigned HOST_WIDE_INT.
2407 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2408 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2409 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2410 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2412 2018-02-27 Richard Biener <rguenther@suse.de>
2414 PR tree-optimization/84512
2415 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2416 Do not use the estimate returned from record_stmt_cost for
2417 the scalar iteration cost but sum properly using add_stmt_cost.
2419 2018-02-27 Richard Biener <rguenther@suse.de>
2421 PR tree-optimization/84466
2422 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2423 Adjust last change to less strictly validate use operands.
2425 2018-02-27 Martin Liska <mliska@suse.cz>
2427 PR gcov-profile/84548
2428 * gcov.c (process_file): Allow partial overlap and consider it
2429 also as group functions.
2430 (output_lines): Properly calculate range of lines for a group.
2432 2018-02-27 Martin Liska <mliska@suse.cz>
2434 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2435 'ggc' suffixes. Change first column width.
2436 (timer::print): Fix formatting of the column.
2438 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2440 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2441 preserve inline entry blocks for the sake of debug inline
2442 entry point markers alone.
2443 (remove_unused_locals): Suggest in comments a better place to
2444 force the preservation of inline entry blocks that are
2445 otherwise unused, but do not preserve them.
2447 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2449 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2451 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2454 * config/i386/constraints.md (Bs): Replace
2455 ix86_indirect_branch_register with
2456 TARGET_INDIRECT_BRANCH_REGISTER.
2458 * config/i386/i386.md (indirect_jump): Likewise.
2459 (tablejump): Likewise.
2460 (*sibcall_memory): Likewise.
2461 (*sibcall_value_memory): Likewise.
2462 Peepholes of indirect call and jump via memory: Likewise.
2463 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2464 (*sibcall_value_GOT_32): Likewise.
2465 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2466 (GOT_memory_operand): Likewise.
2467 (call_insn_operand): Likewise.
2468 (sibcall_insn_operand): Likewise.
2469 (GOT32_symbol_operand): Likewise.
2470 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2472 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2474 PR rtl-optimization/83496
2475 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2476 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2477 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2478 redundant insn, if any.
2479 (relax_delay_slots): Likewise.
2480 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2482 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2484 PR tree-optimization/83965
2485 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2486 that grouped statements are part of a reduction chain. Return
2487 true if the statement is not marked as a reduction itself but
2489 (vect_recog_dot_prod_pattern): Don't check whether the statement
2490 is part of a group here.
2491 (vect_recog_sad_pattern): Likewise.
2492 (vect_recog_widen_sum_pattern): Likewise.
2494 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2497 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2499 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2502 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2504 (ix86_output_indirect_function_return): New prototype.
2505 (ix86_split_simple_return_pop_internal): Likewise.
2506 * config/i386/i386.c (indirect_return_via_cx): New.
2507 (indirect_return_via_cx_bnd): Likewise.
2508 (indirect_thunk_name): Handle return va CX_REG.
2509 (output_indirect_thunk_function): Create alias for
2510 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2511 (ix86_output_indirect_jmp): Remove the bool argument.
2512 (ix86_output_indirect_function_return): New function.
2513 (ix86_split_simple_return_pop_internal): Likewise.
2514 * config/i386/i386.md (*indirect_jump): Don't pass false
2515 to ix86_output_indirect_jmp.
2516 (*tablejump_1): Likewise.
2517 (simple_return_pop_internal): Change it to define_insn_and_split.
2518 Call ix86_split_simple_return_pop_internal to split it for
2520 (simple_return_indirect_internal): Call
2521 ix86_output_indirect_function_return instead of
2522 ix86_output_indirect_jmp.
2524 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2527 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2528 memset and value initialization afterwards.
2530 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2532 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2534 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2537 * common/config/aarch64/aarch64-common.c
2538 (aarch_option_optimization_table[]): Switch
2539 off fomit-frame-pointer
2541 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2542 Chung-Ju Wu <jasonwucj@gmail.com>
2544 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2546 (store_multiple): Ditto.
2548 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2550 * config.gcc: Add --with-cpu support for nds32 target.
2551 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2552 * config/nds32/nds32.opt: Add -mcpu= option.
2554 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2556 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2557 isel=yes): Warn for these deprecated options.
2559 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2561 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2562 ISA_2_5_MASKS_EMBEDDED.
2564 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2566 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2567 p->max as pointers rather than using iterative_hash_expr.
2569 2018-02-23 Carl Love <cel@us.ibm.com>
2571 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2572 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2574 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2575 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2576 P8V_BUILTIN_VEC_VUNSIGNED2.
2578 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2581 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2582 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2584 * lra-constraints.c (curr_insn_transform): Set up
2585 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2587 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2588 * lra-eliminations.c (spill_pseudos): Ditto.
2589 (process_insn_for_elimination): Ditto.
2590 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2591 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2592 LRA_NON_CLOBBERED_ALT.
2594 2018-02-22 Martin Sebor <msebor@redhat.com>
2596 PR tree-optimization/84480
2597 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2598 to maybe_diag_stxncpy_trunc. Call it.
2599 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2600 from gimple_fold_builtin_strcpy. Print inlining stack.
2601 (handle_builtin_stxncpy): Print inlining stack.
2602 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2604 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2607 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2608 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2609 and -fcheck-pointer-bounds are used together.
2610 (indirect_thunk_prefix): New enum.
2611 (indirect_thunk_need_prefix): New function.
2612 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2613 "_nt" instead of "_bnd" for NOTRACK prefix.
2614 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2615 (output_indirect_thunk_function): Likewise.
2617 (ix86_code_end): Update output_indirect_thunk_function calls.
2618 (ix86_output_indirect_branch_via_reg): Replace
2619 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2620 (ix86_output_indirect_branch_via_push): Likewise.
2621 (ix86_output_function_return): Likewise.
2622 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2623 incompatible with -fcf-protection=branch and
2624 -fcheck-pointer-bounds.
2626 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2629 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2630 Change gcc_assert call to output_operand_lossage.
2632 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2634 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2636 2018-02-22 DJ Delorie <dj@redhat.com>
2637 Sebastian Perta <sebastian.perta@renesas.com>
2638 Oleg Endo <olegendo@gcc.gnu.org>
2640 * config/rx/rx.c (rx_rtx_costs): New function.
2641 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2643 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2645 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2647 2018-02-22 Martin Liska <mliska@suse.cz>
2650 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2651 Add "native" as a possible value.
2653 2018-02-22 Martin Liska <mliska@suse.cz>
2656 * config/i386/i386.c (ix86_option_override_internal):
2657 Add "native" as a possible value for -march and -mtune.
2659 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2662 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2663 to all type variants.
2665 PR tree-optimization/84503
2666 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2667 width as info->bitpos + info->bitsize - start.
2668 (merged_store_group::merge_overlapping): Simplify width computation.
2669 (check_no_overlap): New function.
2670 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2671 start + width and last_order of the group, fail if check_no_overlap
2673 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2674 to group if check_no_overlap fails.
2676 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2678 * config/rs6000/altivec.md: Delete contraint arguments to
2679 define_expand, define_split, and define_peephole2, and in
2680 define_insn_and_split if always unused.
2681 * config/rs6000/darwin.md: Ditto.
2682 * config/rs6000/dfp.md: Ditto.
2683 * config/rs6000/rs6000.md: Ditto.
2684 * config/rs6000/sync.md: Ditto.
2685 * config/rs6000/vector.md: Ditto.
2686 * config/rs6000/vsx.md: Ditto.
2688 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2690 * config/rs6000/altivec.md: Write output control strings as braced
2691 blocks instead of double-quoted strings.
2692 * config/rs6000/darwin.md: Ditto.
2693 * config/rs6000/rs6000.md: Ditto.
2694 * config/rs6000/vector.md: Ditto.
2695 * config/rs6000/vsx.md: Ditto.
2697 2018-02-21 Jason Merrill <jason@redhat.com>
2699 PR c++/84314 - ICE with templates and fastcall attribute.
2700 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2702 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2704 * ipa-cp.c (determine_versionability): Fix comment typos.
2706 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2709 * ipa-cp.c (determine_versionability): Do not version functions caling
2712 2018-02-21 Martin Liska <mliska@suse.cz>
2715 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2716 Add "native" as a possible value.
2717 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2718 the macro when native cpu detection is available.
2720 2018-02-21 Martin Liska <mliska@suse.cz>
2723 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2724 Add "native" as a possible value.
2725 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2726 when native cpu detection is available.
2728 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2729 Martin Sebor <msebor@redhat.com>
2731 PR tree-optimization/84478
2732 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2734 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2735 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2736 support which is conservatively correct, for 2 only stay conservative
2737 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2738 argument to the 2 argument get_range_strlen, adjust 6 arg
2739 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2741 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2742 (gimple_fold_builtin_strlen): Pass true as last argument to
2745 2018-02-20 Martin Sebor <msebor@redhat.com>
2748 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2749 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2750 (builtin_memref::builtin_memref): Factor out parts into
2751 set_base_and_offset and call it.
2753 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2756 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2757 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2758 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2759 search at the associated MODE_INT.
2761 2018-02-20 Jeff Law <law@redhat.com>
2764 PR tree-optimization/81592
2766 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2767 for range data rather than using global data.
2768 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2769 range data rather than using global data.
2770 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2771 pass it to children as needed.
2772 (struct directive::fmtresult): Similarly.
2773 (struct directive::set_width): Similarly.
2774 (struct directive::set_precision): Similarly.
2775 (format_integer, format_directive, parse_directive): Similarly.
2776 (format_none): Accept unnamed vr_values parameter.
2777 (format_percent, format_floating, format_character): Similarly.
2778 (format_string, format_plain): Similarly.
2779 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2780 the EVRP range analyzer for range data rather than using global data.
2781 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2782 gimple-ssa-evrp-analyze.h
2783 (class sprintf_dom_walker): Add after_dom_children member function.
2784 Add evrp_range_analyzer member.
2785 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2786 range analyzer as needed.
2787 (sprintf_dom_walker::after_dom_children): New member function.
2788 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2790 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2791 (evrp_range_analyzer::pop_to_marker): Likewise.
2793 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2795 PR tree-optimization/84419
2796 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2797 with the required type if its current type is compatible but
2800 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2803 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2804 after vectorization.
2806 2018-02-20 Martin Liska <mliska@suse.cz>
2809 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2810 possible values if we don't have a hint.
2812 2018-02-20 Martin Liska <mliska@suse.cz>
2816 * final.c (shorten_branches): Build align_tab array with one
2818 * opts.c (finish_options): Add alignment option limit check.
2819 (MAX_CODE_ALIGN): Likewise.
2820 (MAX_CODE_ALIGN_VALUE): Likewise.
2821 * doc/invoke.texi: Document maximum allowed option value for
2822 all -falign-* options.
2824 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2827 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2828 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2829 * var-tracking.c (emit_note_insn_var_location): Remove all references
2830 to NOTE_INSN_CALL_ARG_LOCATION.
2831 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2832 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2833 Use copy_rtx_if_shared.
2834 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2835 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2836 (dwarf2out_var_location): Remove handling of
2837 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2839 * final.c (final_scan_insn): Remove all references to
2840 NOTE_INSN_CALL_ARG_LOCATION.
2841 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2842 before dumping final insns.
2843 * except.c (emit_note_eh_region_end): Remove all references to
2844 NOTE_INSN_CALL_ARG_LOCATION.
2845 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2846 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2847 * config/arc/arc.c (hwloop_optimize): Likewise.
2848 * config/arm/arm.c (create_fix_barrier): Likewise.
2849 * config/s390/s390.c (s390_chunkify_start): Likewise.
2850 * config/sh/sh.c (find_barrier): Likewise.
2851 * config/i386/i386.c (rest_of_insert_endbranch,
2852 ix86_seh_fixup_eh_fallthru): Likewise.
2853 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2854 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2855 * config/frv/frv.c (frv_function_prologue): Likewise.
2856 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2858 (note_outside_basic_block_p): Remove all references to
2859 NOTE_INSN_CALL_ARG_LOCATION.
2860 * gengtype.c (adjust_field_rtx_def): Likewise.
2861 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2863 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2864 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2867 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2870 PR tree-optimization/84452
2871 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2872 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2875 2018-02-19 Martin Liska <mliska@suse.cz>
2878 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2880 2018-02-19 Martin Liska <mliska@suse.cz>
2881 Richard Sandiford <richard.sandiford@linaro.org>
2883 PR tree-optimization/82491
2884 * gimple-fold.c (get_base_constructor): Make earlier bail out
2887 2018-02-19 Carl Love <cel@us.ibm.com>
2889 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2890 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2892 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2893 P8V_BUILTIN_VEC_NEG.
2895 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2897 * config/rl78/rl78.md (movdf): New define expand.
2899 2018-02-19 Martin Liska <mliska@suse.cz>
2902 * doc/invoke.texi: Fix typo.
2903 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2905 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2907 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2908 handle rs6000_single_float and rs6000_double_float specially for
2911 2018-02-16 Jeff Law <law@redhat.com>
2913 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2916 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2919 * ipa-inline.c (inline_small_functions): Fix a typo.
2921 2018-02-16 Nathan Sidwell <nathan@acm.org>
2923 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2925 2018-02-16 Carl Love <cel@us.ibm.com>
2927 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2928 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2929 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2930 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2931 expansion to P8V_BUILTIN_VEC_FLOAT2.
2933 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2935 PR rtl-optimization/70023
2936 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2937 src_regno into account.
2939 2018-02-16 Carl Love <cel@us.ibm.com>
2941 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2942 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2943 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2944 * config/rs6000/rs6000.c: Remove case statements for
2945 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2946 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2947 and P9V_BUILTIN_VEC_VINSERT4B.
2948 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2949 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2950 * config/rs6000/vsx.md:
2951 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2954 2018-02-16 Carl Love <cel@us.ibm.com>
2956 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2958 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2960 * config/rs6000/rs6000-c.c: Add the definitions for
2961 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2962 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2963 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2964 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2965 definition for insert4b and define insn *insert3b_internal.
2966 * doc/extend.texi: Add documentation for vec_extract4b.
2968 2018-02-16 Nathan Sidwell <nathan@acm.org>
2970 * doc/extend.texi (Backwards Compatibility): Mention friend
2971 injection. Note for-scope is deprecated.
2972 * doc/invoke.texi (-ffriend-injection): Deprecate.
2974 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2976 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2977 that moved to I2, also allow destinations that are a paradoxical
2978 subreg (instead of a normal reg).
2980 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2983 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2986 2018-02-16 Richard Biener <rguenther@suse.de>
2988 PR tree-optimization/84037
2989 PR tree-optimization/84016
2991 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2992 Adjust vec_construct for the fact we need additional higher latency
2993 128bit inserts for AVX256 and AVX512 vector builds.
2994 (ix86_add_stmt_cost): Scale vector construction cost for
2997 2018-02-16 Richard Biener <rguenther@suse.de>
2999 PR tree-optimization/84417
3000 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3001 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3002 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3004 2018-02-16 Martin Liska <mliska@suse.cz>
3007 * internal-fn.def (ASAN_CHECK): Set proper flags.
3008 (ASAN_MARK): Likewise.
3010 2018-02-16 Julia Koval <julia.koval@intel.com>
3012 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3013 from PTA_CANNONLAKE.
3015 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3018 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3019 Use ++iter rather than iter++ for std::list iterators.
3020 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3021 defer deleting them until all nodes in the forest are processed. Do
3022 free even leaf nodes. Change to_process into auto_vec.
3025 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3026 * vec.h (vec_default_construct): Use memset instead of placement new
3027 if BROKEN_VALUE_INITIALIZATION is defined.
3028 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3029 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3032 PR rtl-optimization/83723
3033 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3034 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3035 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3037 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3039 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3041 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3043 PR rtl-optimization/81443
3044 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3045 from inner REGs to paradoxical SUBREGs.
3047 2018-02-16 Richard Biener <rguenther@suse.de>
3049 PR tree-optimization/84399
3050 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3051 For operands we can analyze at their definition make sure we can
3052 analyze them at each use as well.
3054 2018-02-16 Richard Biener <rguenther@suse.de>
3056 PR tree-optimization/84190
3057 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3058 volatile accesses if the decl isn't volatile.
3060 2018-02-15 Jason Merrill <jason@redhat.com>
3062 PR c++/84314 - ICE with templates and fastcall attribute.
3063 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3064 TYPE_CANONICAL on an existing type.
3066 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3068 PR tree-optimization/84383
3069 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3070 dstoff nor call operand_equal_p if dstbase is NULL.
3072 PR tree-optimization/84334
3073 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3074 also a CONSTANT_CLASS_P, punt.
3076 2018-02-14 Jim Wilson <jimw@sifive.com>
3078 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3079 first SMALL_OPERAND check. New local min_second_step. Move assert
3080 to where locals are set. Add TARGET_RVC support.
3081 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3083 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3085 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3087 2018-02-14 Martin Sebor <msebor@redhat.com>
3089 PR tree-optimization/83698
3090 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3091 arrays constrain the offset range to their bounds.
3092 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3093 (builtin_access::overlap): Avoid setting the size of overlap if it's
3095 (maybe_diag_overlap): Also consider arrays when deciding what values
3096 of offsets to include in diagnostics.
3098 2018-02-14 Martin Sebor <msebor@redhat.com>
3101 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3102 that correspond to the kind of a declaration.
3104 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3107 * config/pa/pa.md: Load address of PIC label using the linkage table
3108 if the label is nonlocal.
3110 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3112 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3113 warning message if user requests -maltivec=be.
3114 * doc/invoke.texi: Document deprecation of -maltivec=be.
3116 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3119 * config/rs6000/rs6000-c.c: Update definitions for
3120 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3121 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3123 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3126 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3127 add _get_ssp intrinsics. Remove argument from
3128 __builtin_ia32_rdssp[d|q].
3129 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3130 * config/i386/i386-builtin.def: Remove argument from
3131 __builtin_ia32_rdssp[d|q].
3132 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3133 ix86_expand_special_args_builtin for _rdssp[d|q].
3134 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3135 Clear register before usage.
3136 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3137 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3139 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3141 PR tree-optimization/84357
3142 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3143 operand 1 of an ARRAY_REF too.
3145 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3148 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3149 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3151 (set_of_reg): New struct.
3152 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3153 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3154 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3156 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3157 Split into bitclr, bitset, bitinvert patterns if appropriate.
3158 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3159 use rx_fuse_in_memory_bitop.
3160 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3161 to named insn, correct maximum insn length.
3163 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3166 * machmode.def: Define a complex mode for PARTIAL_INT.
3167 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3169 * doc/rtl.texi: Document CSPImode.
3170 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3172 (msp430_hard_regno_nregs_with_padding): Likewise.
3174 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3177 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3179 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3181 PR rtl-optimization/84169
3182 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3183 we generated a parallel as new i3 and we split that to new i2 and i3
3184 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3185 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3186 those to i2, not i1. Partially rewrite this scan code.
3188 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3191 * stor-layout.c (place_field): For variable length fields, adjust
3192 offset_align afterwards not just based on the field's alignment,
3193 but also on the size.
3196 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3197 of exps and logs in the use_exp2 case.
3199 2018-02-13 Jeff Law <law@redhat.com>
3201 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3204 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3207 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3211 * common.opt (gas-loc-support, gas-locview-support): New.
3212 (ginline-points, ginternal-reset-location-views): New.
3213 * doc/invoke.texi: Document them. Use @itemx where intended.
3214 (gvariable-location-views): Adjust.
3215 * target.def (reset_location_view): New.
3216 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3217 (TARGET_RESET_LOCATION_VIEW): New.
3218 * doc/tm.texi: Rebuilt.
3219 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3220 (dwarf2out_default_as_locview_support): New.
3221 (output_asm_line_debug_info): Use option variables.
3222 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3223 (output_loc_list): Likewise.
3224 (add_high_low_attributes): Check option variables.
3225 Don't output entry view attribute in strict mode.
3226 (gen_inlined_subroutine_die): Check option variables.
3227 (dwarf2out_inline_entry): Likewise.
3228 (init_sections_and_labels): Likewise.
3229 (dwarf2out_early_finish): Likewise.
3230 (maybe_reset_location_view): New, from...
3231 (dwarf2out_var_location): ... here. Call it.
3232 * debug.h (dwarf2out_default_as_loc_support): Declare.
3233 (dwarf2out_default_as_locview_support): Declare.
3234 * hooks.c (hook_int_rtx_insn_0): New.
3235 * hooks.h (hook_int_rtx_insn_0): Declare.
3236 * toplev.c (process_options): Take -gas-loc-support and
3237 -gas-locview-support from dwarf2out. Enable
3238 -gvariable-location-views by default only with locview
3239 assembler support. Enable -ginternal-reset-location-views by
3240 default only if the target defines the corresponding hook.
3241 Enable -ginline-points by default if location views are
3242 enabled; force it disabled if statement frontiers are
3244 * tree-inline.c (expand_call_inline): Check option variables.
3245 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3247 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3249 PR tree-optimization/84321
3250 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3251 handling. Also check whether the anti-range contains any values
3252 that satisfy the mask; switch to a VR_RANGE if not.
3254 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3257 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3259 2018-02-13 Martin Jambor <mjambor@suse.cz>
3262 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3263 of call statements, also set location of a load to a temporary.
3265 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3267 * config/rl78/rl78.c (add_vector_labels): New function.
3268 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3269 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3270 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3271 which checks that no arguments are passed.
3272 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3273 * doc/extend.texi: Documentation for the new attribute.
3275 2018-02-13 Andreas Schwab <schwab@suse.de>
3277 * config/riscv/linux.h (CPP_SPEC): Define.
3279 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3282 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3283 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3284 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3285 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3286 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3287 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3288 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3289 temporarily for AES and PCLMUL builtins.
3291 PR tree-optimization/84339
3292 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3293 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3297 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3298 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3299 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3301 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3303 * omp-simd-clone.h: New file.
3304 * omp-simd-clone.c: Include omp-simd-clone.h.
3305 (expand_simd_clones): No longer static.
3306 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3307 cgraph.h and omp-simd-clone.h.
3308 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3309 (vect_recog_widen_shift_pattern): Formatting fix.
3310 (vect_pattern_recog_1): Don't check optab for calls.
3313 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3314 operands[2] into a REG before using gen_lowpart on it.
3316 2018-02-12 Jeff Law <law@redhat.com>
3319 * config/sh/sh.c (find_barrier): Consider a sibling call
3322 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3323 successfully back substituting a reg.
3325 2018-02-12 Richard Biener <rguenther@suse.de>
3327 PR tree-optimization/84037
3328 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3329 parameter, move visited init to caller.
3330 (vect_slp_analyze_operations): Separate cost from validity
3331 check, initialize visited once for all instances.
3332 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3334 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3336 (vect_model_promotion_demotion_cost): Likewise.
3337 (vectorizable_bswap): Guard cost modeling with !slp_node
3338 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3340 (vectorizable_call): Likewise.
3341 (vectorizable_conversion): Likewise.
3342 (vectorizable_assignment): Likewise.
3343 (vectorizable_shift): Likewise.
3344 (vectorizable_operation): Likewise.
3345 (vectorizable_store): Likewise.
3346 (vectorizable_load): Likewise.
3347 (vectorizable_condition): Likewise.
3348 (vectorizable_comparison): Likewise.
3350 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3353 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3354 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3355 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3357 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3360 * params.def (inline-min-speedup): Increase from 8 to 15.
3361 (max-inline-insns-auto): Decrease from 40 to 30.
3362 * ipa-split.c (consider_split): Add some buffer for function to
3363 be considered inlining candidate.
3364 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3367 2018-02-12 Richard Biener <rguenther@suse.de>
3369 PR tree-optimization/84037
3370 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3371 matched stmts if we cannot swap the non-matched ones.
3373 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3375 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3376 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3377 _mm_maskz_scalef_round_ss): New intrinsics.
3378 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3379 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3380 __builtin_ia32_scalefss_round): Remove.
3381 (__builtin_ia32_scalefsd_mask_round,
3382 __builtin_ia32_scalefss_mask_round): New intrinsics.
3383 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3384 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3385 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3386 "<round_constraint>")): Changed to ...
3387 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3388 "<round_scalar_constraint>")): ... this.
3389 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3390 %0, %1, %2<round_op3>}"): Changed to ...
3391 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3392 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3393 %2<round_scalar_mask_op3>}"): ... this.
3394 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3396 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3398 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3399 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3400 (_mm_maskz_sqrt_round_ss): New intrinsics.
3401 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3402 (__builtin_ia32_sqrtsd_mask_round)
3403 (__builtin_ia32_sqrtss_mask_round): New builtins.
3404 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3405 (__builtin_ia32_sqrtss_round): Remove.
3406 (__builtin_ia32_sqrtsd_mask_round)
3407 (__builtin_ia32_sqrtss_mask_round): New builtins.
3408 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3409 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3410 ((match_operand:VF_128 1 "vector_operand"
3411 "xBm,<round_constraint>")): Changed to ...
3412 ((match_operand:VF_128 1 "vector_operand"
3413 "xBm,<round_scalar_constraint>")): ... this.
3414 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3415 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3416 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3417 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3418 %<iptr>1<round_scalar_mask_op3>}): ... this.
3419 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3420 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3422 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3425 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3426 Cast vec_cmpeq result to correct type.
3427 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3428 Cast vec_cmpgt result to correct type.
3430 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3432 * final.c (final_scan_insn_1): Renamed from...
3433 (final_scan_insn): ... this. New wrapper, to recover
3434 seen from the outermost call in recursive ones.
3435 * config/sparc/sparc.c (output_return): Drop seen from call.
3436 (output_sibcall): Likewise.
3437 * config/visium/visium.c (output_branch): Likewise.
3439 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3441 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3444 2018-02-10 Alan Modra <amodra@gmail.com>
3447 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3448 Specify LR as an input.
3450 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3453 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3454 remove_member_access_dummy_vars): New functions.
3455 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3456 lower_omp_1, execute_lower_omp): Use them.
3458 PR rtl-optimization/84308
3459 * shrink-wrap.c (spread_components): Release todo vector.
3461 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3463 PR rtl-optimization/57193
3464 * ira-color.c (struct allocno_color_data): Add member
3465 conflict_allocno_hard_prefs.
3466 (update_conflict_allocno_hard_prefs): New.
3467 (bucket_allocno_compare_func): Add a preference based on
3468 conflict_allocno_hard_prefs.
3469 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3470 (color_allocnos): Remove a dead code. Initiate
3471 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3473 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3476 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3477 constraint from =wa to wa. Avoid a subreg on the output operand,
3478 instead use a pseudo and subreg it in a move.
3479 (p9_xxbrd_<mode>): Changed to ...
3480 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3481 (p9_xxbrd_v2df): New expander.
3482 (p9_xxbrw_<mode>): Changed to ...
3483 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3484 (p9_xxbrw_v4sf): New expander.
3486 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3488 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3490 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3493 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3494 multiply in 32-bit mode.
3495 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3496 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3499 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3501 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3502 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3503 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3504 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3506 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3509 * dwarf2out.c (is_trivial_indirect_ref): New function.
3510 (dwarf2out_late_global_decl): Do not generate a location
3511 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3512 and that are not defined in the current unit.
3514 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3516 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3517 instead of a libcall for UNORDERED.
3519 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3522 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3523 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3525 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3528 * config/s390/s390.c (s390_set_current_function): Invoke
3529 s390_indirect_branch_settings also if fndecl didn't change.
3531 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3533 * config/rs6000/rs6000.md (blockage): Set length to zero.
3535 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3537 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3539 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3542 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3543 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3547 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3548 PARALLEL incoming that failed vt_get_decl_and_offset check.
3551 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3552 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3554 (get_variable_section): For decls in named .bss* sections pass true as
3555 second argument to bss_initializer_p.
3557 2018-02-09 Marek Polacek <polacek@redhat.com>
3558 Jakub Jelinek <jakub@redhat.com>
3561 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3562 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3563 Sync some changes from cxx_fold_indirect_ref.
3565 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3567 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3569 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3570 (BLOCK_INLINE_ENTRY_LABEL): New.
3571 (dwarf2out_var_location): Disregard inline entry markers.
3572 (inline_entry_data): New struct.
3573 (inline_entry_data_hasher): New hashtable type.
3574 (inline_entry_data_hasher::hash): New.
3575 (inline_entry_data_hasher::equal): New.
3576 (inline_entry_data_table): New variable.
3577 (add_high_low_attributes): Add DW_AT_entry_pc and
3578 DW_AT_GNU_entry_view attributes if a pending entry is found
3579 in inline_entry_data_table. Add old entry_pc attribute only
3580 if debug nonbinding markers are disabled.
3581 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3582 markers are enabled.
3583 (block_within_block_p, dwarf2out_inline_entry): New.
3584 (dwarf2out_finish): Check that no entries remained in
3585 inline_entry_data_table.
3586 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3587 (final_scan_insn, notice_source_line): Likewise.
3588 (rest_of_clean_state): Skip inline entry markers.
3589 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3591 * gimple.c (gimple_build_debug_inline_entry): New.
3592 * gimple.h (enum gimple_debug_subcode): Add
3593 GIMPLE_DEBUG_INLINE_ENTRY.
3594 (gimple_build_debug_inline_entry): Declare.
3595 (gimple_debug_inline_entry_p): New.
3596 (gimple_debug_nonbind_marker_p): Adjust.
3597 * insn-notes.def (INLINE_ENTRY): New.
3598 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3599 inline entry marker notes.
3600 (print_insn): Likewise.
3601 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3602 (INSN_DEBUG_MARKER_KIND): Likewise.
3603 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3604 * tree-inline.c (expand_call_inline): Build and insert
3605 debug_inline_entry stmt.
3606 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3607 inline entry blocks early, if nonbind markers are enabled.
3608 (dump_scope_block): Dump fragment info.
3609 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3610 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3611 (gimple_build_debug_inline_entry): New.
3612 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3613 Enable/disable inline entry points too.
3614 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3615 (DEBUG_INSN): Describe inline entry markers.
3617 * common.opt (gvariable-location-views): New.
3618 (gvariable-location-views=incompat5): New.
3619 * config.in: Rebuilt.
3620 * configure: Rebuilt.
3621 * configure.ac: Test assembler for view support.
3622 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3623 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3624 * dwarf2out.c (var_loc_view): New typedef.
3625 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3626 (dwarf2out_locviews_in_attribute): New.
3627 (dwarf2out_locviews_in_loclist): New.
3628 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3629 (enum dw_line_info_opcode): Add LI_adv_address.
3630 (struct dw_line_info_table): Add view.
3631 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3632 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3633 (zero_view_p): New variable.
3634 (ZERO_VIEW_P): New macro.
3635 (output_asm_line_debug_info): New.
3636 (struct var_loc_node): Add view.
3637 (add_AT_view_list, AT_loc_list): New.
3638 (add_var_loc_to_decl): Add view param. Test it against last.
3639 (new_loc_list): Add view params. Record them.
3640 (AT_loc_list_ptr): Handle loc and view lists.
3641 (view_list_to_loc_list_val_node): New.
3642 (print_dw_val): Handle dw_val_class_view_list.
3643 (size_of_die): Likewise.
3644 (value_format): Likewise.
3645 (loc_list_has_views): New.
3646 (gen_llsym): Set vl_symbol too.
3647 (maybe_gen_llsym, skip_loc_list_entry): New.
3648 (dwarf2out_maybe_output_loclist_view_pair): New.
3649 (output_loc_list): Output view list or entries too.
3650 (output_view_list_offset): New.
3651 (output_die): Handle dw_val_class_view_list.
3652 (output_dwarf_version): New.
3653 (output_compilation_unit_header): Use it.
3654 (output_skeleton_debug_sections): Likewise.
3655 (output_rnglists, output_line_info): Likewise.
3656 (output_pubnames, output_aranges): Update version comments.
3657 (output_one_line_info_table): Output view numbers in asm comments.
3658 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3659 Call maybe_gen_llsym.
3660 (loc_list_from_tree_1): Adjust.
3661 (add_AT_location_description): Create view list attribute if
3662 needed, check it's absent otherwise.
3663 (convert_cfa_to_fb_loc_list): Adjust.
3664 (maybe_emit_file): Call output_asm_line_debug_info for test.
3665 (dwarf2out_var_location): Reset views as needed. Precompute
3666 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3667 attribute. Set view.
3668 (new_line_info_table): Reset next view.
3669 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3670 (dwarf2out_source_line): Likewise. Output view resets and labels to
3671 the assembler, or select appropriate line info opcodes.
3672 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3673 (optimize_string_length): Catch it. Adjust.
3674 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3675 dw_val_class_view_list, and remove it if no longer needed.
3676 (hash_loc_list): Hash view numbers.
3677 (loc_list_hasher::equal): Compare them.
3678 (optimize_location_lists): Check whether a view list symbol is
3679 needed, and whether the locview attribute is present, and
3680 whether they match. Remove the locview attribute if no longer
3682 (index_location_lists): Call skip_loc_list_entry for test.
3683 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3684 Use output_dwarf_version.
3685 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3686 (struct dw_val_node): Add val_view_list.
3687 * final.c (SEEN_NEXT_VIEW): New.
3688 (set_next_view_needed): New.
3689 (clear_next_view_needed): New.
3690 (maybe_output_next_view): New.
3691 (final_start_function): Rename to...
3692 (final_start_function_1): ... this. Take pointer to FIRST,
3693 add SEEN parameter. Emit param bindings in the initial view.
3694 (final_start_function): Reintroduce SEEN-less interface.
3695 (final): Rename to...
3696 (final_1): ... this. Take SEEN parameter. Output final pending
3697 next view at the end.
3698 (final): Reintroduce seen-less interface.
3699 (final_scan_insn): Output pending next view before switching
3700 sections or ending a block. Mark the next view as needed when
3701 outputting variable locations. Notify debug backend of section
3702 changes, and of location view changes.
3703 (rest_of_handle_final): Adjust.
3704 * toplev.c (process_options): Autodetect value for debug variable
3705 location views option. Warn on incompat5 without -gdwarf-5.
3706 * doc/invoke.texi (gvariable-location-views): New.
3707 (gvariable-location-views=incompat5): New.
3708 (gno-variable-location-views): New.
3710 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3712 PR tree-optimization/84136
3713 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3714 that the result of find_edge is non-NULL.
3716 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3719 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3720 storing integer register in SImode. Fix cost of 256 and 512
3721 byte aligned SSE register store.
3723 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3725 * config/i386/i386.c (ix86_multiplication_cost): Fix
3726 multiplication cost for TARGET_AVX512DQ.
3728 2018-02-08 Marek Polacek <polacek@redhat.com>
3730 PR tree-optimization/84238
3731 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3734 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3736 PR tree-optimization/84265
3737 * tree-vect-stmts.c (vectorizable_store): Don't treat
3738 VMAT_CONTIGUOUS accesses as grouped.
3739 (vectorizable_load): Likewise.
3741 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3743 PR tree-optimization/81635
3744 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3745 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3746 (test_round_for_mask): New functions.
3747 (wide_int_cc_tests): Call test_round_for_mask.
3748 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3749 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3750 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3751 range returned by get_range_info.
3753 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3756 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3757 * symtab.c: Include builtins.h
3758 (symtab_node::output_to_lto_symbol_table_p): Move here
3759 from lto-streamer-out.c:output_symbol_p.
3760 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3761 (output_symbol_p): Move all logic to symtab.c
3762 (produce_symtab): Update.
3764 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3766 * config/s390/s390-opts.h (enum indirect_branch): Define.
3767 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3768 (s390_indirect_branch_via_thunk)
3769 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3770 (enum s390_indirect_branch_type): Define.
3771 * config/s390/s390.c (struct s390_frame_layout, struct
3772 machine_function): Remove.
3773 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3774 (indirect_branch_table_label_no, indirect_branch_table_name):
3776 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3777 (enum s390_indirect_branch_option): Define.
3778 (s390_return_addr_from_memory): New function.
3779 (s390_handle_string_attribute): New function.
3780 (s390_attribute_table): Add new attribute handler.
3781 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3782 (s390_indirect_branch_via_thunk): New function.
3783 (s390_indirect_branch_via_inline_thunk): New function.
3784 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3785 sibling call optimization for non z10 compiles.
3786 (s390_emit_call): Force indirect branch target to be a single
3787 register. Add r1 clobber for non-z10 compiles.
3788 (s390_emit_epilogue): Emit return jump via return_use expander.
3789 (s390_reorg): Handle JUMP_INSNs as execute targets.
3790 (s390_option_override_internal): Perform validity checks for the
3791 new command line options.
3792 (s390_indirect_branch_attrvalue): New function.
3793 (s390_indirect_branch_settings): New function.
3794 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3795 (s390_output_indirect_thunk_function): New function.
3796 (s390_code_end): Implement target hook.
3797 (s390_case_values_threshold): Implement target hook.
3798 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3800 * config/s390/s390.h (struct s390_frame_layout)
3801 (struct machine_function): Move here from s390.c.
3802 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3803 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3804 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3805 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3806 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3807 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3808 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3809 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3810 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3811 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3812 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3813 (mnemonic attribute): Add values which aren't recognized
3815 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3816 pattern for branch conversion. Fix mnemonic attribute.
3817 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3818 indirect branch via thunk if requested.
3819 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3820 ("*indirect_jump"): Disable for branch conversion using out of
3822 ("indirect_jump_via_thunk<mode>_z10")
3823 ("indirect_jump_via_thunk<mode>")
3824 ("indirect_jump_via_inlinethunk<mode>_z10")
3825 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3826 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3827 ("casesi_jump_via_inlinethunk<mode>_z10")
3828 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3829 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3830 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3831 ("*indirect2_jump"): Disable for branch conversion.
3832 ("casesi_jump"): Turn into expander and expand patterns for branch
3834 ("return_use"): New expander.
3835 ("*return"): Emit return via thunk and rename it to ...
3836 ("*return<mode>"): ... this one.
3837 * config/s390/s390.opt: Add new options and and enum for the
3840 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3842 * lra-constraints.c (match_reload): Unconditionally use
3843 gen_lowpart_SUBREG, rather than selecting between that
3844 and equivalent gen_rtx_SUBREG code.
3846 2018-02-08 Richard Biener <rguenther@suse.de>
3848 PR tree-optimization/84233
3849 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3850 changed flag instead of boguously re-using phi_inserted.
3852 2018-02-08 Martin Jambor <mjambor@suse.cz>
3854 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3855 static local variables.
3857 2018-02-08 Richard Biener <rguenther@suse.de>
3859 PR tree-optimization/84278
3860 * tree-vect-stmts.c (vectorizable_store): When looking for
3861 smaller vector types to perform grouped strided loads/stores
3862 make sure the mode is supported by the target.
3863 (vectorizable_load): Likewise.
3865 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3867 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3868 Increase LDP/STP opportunities by adding adjacent callee-saves.
3870 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3872 PR rtl-optimization/84068
3873 PR rtl-optimization/83459
3874 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3876 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3878 PR tree-optimization/84224
3879 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3880 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3883 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3886 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3887 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3888 count, remove one USE.
3890 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3892 * doc/install.texi (Configuration): Document the
3893 --with-long-double-format={ibm,ieee} PowerPC configuration
3897 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3898 Convert from define_expand to be define_insn_and_split. Rework
3899 float/double/_Float128 conversions to QI/HI/SImode to work with
3900 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3901 conversions to QI/HImode types did a store and then a load to
3902 truncate the value. For conversions to VSX registers, don't split
3903 the insn, instead emit the code directly. Use the code iterator
3904 any_fix to combine signed and unsigned conversions.
3905 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3906 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3907 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3908 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3909 (fix_<mode>di2_hw): Likewise.
3910 (fixuns_<mode>di2_hw): Likewise.
3911 (fix_<mode>si2_hw): Likewise.
3912 (fixuns_<mode>si2_hw): Likewise.
3913 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3914 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3915 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3916 fix<uns>_trunc<SFDF:mode>si2_p8.
3917 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3919 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3920 (fix<uns>_<mode>_mem): Likewise.
3921 (fctiw<u>z_<mode>_mem): Likewise.
3922 (fix<uns>_<mode>_mem): Likewise.
3923 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3924 the register allocator from doing a direct move to the GPRs to do
3925 a store, and instead use the ISA 3.0 store byte/half-word from
3926 vector register instruction. For IEEE 128-bit floating point,
3927 also optimize stores of 32-bit ints.
3928 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3930 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3932 * genextract.c (push_pathstr_operand): New function to support
3934 (walk_rtx): Call push_pathstr_operand.
3935 (print_path): Support [a-zA-Z].
3937 2018-02-07 Richard Biener <rguenther@suse.de>
3939 PR tree-optimization/84037
3940 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3941 (cse_and_gimplify_to_preheader): Declare.
3942 (vect_get_place_in_interleaving_chain): Likewise.
3943 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3945 (_loop_vec_info::~_loop_vec_info): Delete it.
3946 (cse_and_gimplify_to_preheader): New function.
3947 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3948 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3949 (vectorizable_load): Likewise. For grouped stores always base
3950 the IV on the first element.
3951 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3952 condition before gimplifying.
3954 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3956 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3957 *DIV_EXPR and *MOD_EXPR.
3959 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3962 * config/i386/i386.c (ix86_option_override_internal): Mask out
3963 the CF_SET bit when checking -fcf-protection.
3965 2018-02-07 Tom de Vries <tom@codesourcery.com>
3968 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3971 2018-02-07 Richard Biener <rguenther@suse.de>
3973 PR tree-optimization/84204
3974 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3977 PR tree-optimization/84205
3978 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3979 special-case isl_ast_op_zdiv_r.
3981 PR tree-optimization/84223
3982 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3983 Only add conditions from within the region.
3984 (gather_bbs::after_dom_children): Adjust.
3986 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3989 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3990 * config/avr/avr.md: Only post-reload split REG-REG moves if
3991 either register is GENERAL_REG_P.
3993 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3995 PR tree-optimization/84235
3996 * tree-ssa-scopedtables.c
3997 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3998 if the subtraction is performed in floating point type where NaNs are
3999 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4000 build 1. Formatting fix.
4002 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4005 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4006 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4007 and skip it regardless of bb boundaries. Use CALL_P macro,
4008 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4009 unnecessarily, formatting fix.
4011 2018-02-06 Michael Collison <michael.collison@arm.com>
4013 * config/arm/thumb2.md:
4014 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4015 (*thumb_mov_notscc): Ditto.
4017 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4020 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4021 unsigned_fix, not "s".
4023 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4025 * configure.ac (gcc_fn_eh_frame_ro): New function.
4026 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4027 correct .eh_frame permissions.
4028 * configure: Regenerate.
4030 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4032 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4035 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4037 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4038 Display warning message for -mno-speculate-indirect-jumps.
4040 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4042 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4044 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4046 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4048 PR tree-optimization/84225
4049 * tree-eh.c (find_trapping_overflow): Only call
4050 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4052 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4055 * config/i386/i386.c: Reimplement the check of possible options
4056 -mibt/-mshstk conbination. Change error messages.
4057 * doc/invoke.texi: Fix a typo: remove extra '='.
4059 2018-02-06 Marek Polacek <polacek@redhat.com>
4061 PR tree-optimization/84228
4062 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4064 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4067 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4068 emitted arch directives.
4069 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4070 __ARM_FEATURE_COPROC before changing architectures.
4072 2018-02-06 Richard Biener <rguenther@suse.de>
4074 * config/i386/i386.c (print_reg): Fix typo.
4075 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4077 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4079 * configure: Regenerate.
4081 2018-02-05 Martin Sebor <msebor@redhat.com>
4083 PR tree-optimization/83369
4084 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4087 2018-02-05 Martin Liska <mliska@suse.cz>
4089 * doc/invoke.texi: Cherry-pick upstream r323995.
4091 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4093 * ira.c (ira_init_register_move_cost): Adjust comment.
4095 2018-02-05 Martin Liska <mliska@suse.cz>
4097 PR gcov-profile/84137
4098 * doc/gcov.texi: Fix typo in documentation.
4100 2018-02-05 Martin Liska <mliska@suse.cz>
4102 PR gcov-profile/83879
4103 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4104 using dlopen functionality.
4106 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4108 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4109 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4110 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4111 _mm_maskz_range_round_ss): New intrinsics.
4112 (__builtin_ia32_rangesd128_round)
4113 (__builtin_ia32_rangess128_round): Remove.
4114 (__builtin_ia32_rangesd128_mask_round,
4115 __builtin_ia32_rangess128_mask_round): New builtins.
4116 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4117 __builtin_ia32_rangess128_round): Remove.
4118 (__builtin_ia32_rangesd128_mask_round,
4119 __builtin_ia32_rangess128_mask_round): New builtins.
4120 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4121 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4122 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4123 "<round_saeonly_constraint>")): Changed to ...
4124 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4125 "<round_saeonly_scalar_constraint>")): ... this.
4126 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4127 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4128 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4129 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4130 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4132 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4134 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4136 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4137 Remove all values except native, 8540 and 8548.
4139 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4141 * config/i386/i386.c (ix86_output_function_return): Pass
4142 INVALID_REGNUM, instead of -1, as invalid register number to
4143 indirect_thunk_name and output_indirect_thunk.
4145 2018-02-02 Julia Koval <julia.koval@intel.com>
4147 * config.gcc: Add -march=icelake.
4148 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4149 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4150 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4151 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4152 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4153 (processor_target_table): Add icelake.
4154 (ix86_option_override_internal): Handle new PTAs.
4155 (get_builtin_code_for_version): Handle icelake.
4156 (M_INTEL_COREI7_ICELAKE): New.
4157 (fold_builtin_cpu): Handle icelake.
4158 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4159 * doc/invoke.texi: Add -march=icelake.
4161 2018-02-02 Julia Koval <julia.koval@intel.com>
4163 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4164 to wide_int_bitmask.
4165 * wide-int-bitmask.h: New.
4167 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4170 * config/i386/i386.md: Replace Pmode with word_mode in
4171 builtin_setjmp_setup and builtin_longjmp to support x32.
4173 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4177 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4179 (rs6000_supported_cpu_names): New static variable.
4180 (linux_cpu_translation_table): Likewise.
4181 (elf_platform) <cpu>: Define new static variable and use it.
4182 Translate kernel AT_PLATFORM name to canonical name if needed.
4183 Error if platform name is unknown.
4185 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4188 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4190 2018-02-01 Jeff Law <law@redhat.com>
4193 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4194 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4195 the scratch if RELEASE_VIA_POP is false.
4196 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4197 If we have to save a temporary register, decrement SIZE appropriately.
4198 Pass new arguments to release_scratch_register_on_entry.
4199 (ix86_adjust_stack_and_probe): Likewise.
4200 (ix86_emit_probe_stack_range): Pass new arguments to
4201 release_scratch_register_on_entry.
4203 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4205 PR rtl-optimization/84157
4206 * combine.c (change_zero_ext): Use REG_P predicate in
4207 front of HARD_REGISTER_P predicate.
4209 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4211 * config/avr/avr.c (avr_option_override): Move disabling of
4212 -fdelete-null-pointer-checks to...
4213 * common/config/avr/avr-common.c (avr_option_optimization_table):
4216 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4218 PR tree-optimization/81635
4219 * tree-data-ref.c (split_constant_offset_1): For types that
4220 wrap on overflow, try to use range info to prove that wrapping
4223 2018-02-01 Renlin Li <renlin.li@arm.com>
4226 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4228 (aarch64_register_move_cost): Likewise.
4229 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4231 (REG_CLASS_NAMES): Likewise.
4232 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4233 TAILCALL_ADDR_REGS. Remove IP registers.
4234 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4236 2018-02-01 Richard Biener <rguenther@suse.de>
4238 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4239 for specifying RPO order and allow NULL for that.
4240 * domwalk.c (dom_walker::dom_walker): Likewise.
4241 (dom_walker::walk): Handle NULL RPO order.
4242 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4244 (rewrite_update_dom_walker): Likewise.
4245 (mark_def_dom_walker): Likewise.
4247 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4249 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4250 (aarch64_maybe_expand_sve_subreg_move): Declare.
4251 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4252 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4254 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4255 that are semantically a reverse operation.
4256 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4257 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4258 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4260 (aarch64_can_change_mode_class): For big-endian, forbid changes
4261 between two SVE modes if they have different element sizes.
4263 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4265 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4266 the TImode handling for big-endian targets.
4268 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4270 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4271 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4273 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4274 Remove BSWAP handing for big-endian targets and use the form of
4275 LD1RQ appropariate for the mode.
4277 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4279 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4280 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4283 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4286 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4287 check for operands that need to go through aarch64_sve_reload_be.
4289 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4291 PR tree-optimization/81661
4292 PR tree-optimization/84117
4293 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4294 * tree-eh.c: Include gimplify.h.
4295 (find_trapping_overflow, replace_trapping_overflow,
4296 rewrite_to_non_trapping_overflow): New functions.
4297 * tree-vect-loop.c: Include tree-eh.h.
4298 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4299 * tree-data-ref.c: Include tree-eh.h.
4300 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4302 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4304 PR rtl-optimization/84123
4305 * combine.c (change_zero_ext): Check if hard register satisfies
4306 can_change_dest_mode before calling gen_lowpart_SUBREG.
4308 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4311 * ira.c (ira_init_register_move_cost): Remove assert.
4313 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4315 PR rtl-optimization/84071
4316 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4317 * doc/tm.texi: Regenerate.
4319 2018-01-31 Richard Biener <rguenther@suse.de>
4321 PR tree-optimization/84132
4322 * tree-data-ref.c (analyze_miv_subscript): Properly
4323 check whether evolution_function_is_affine_multivariate_p
4324 before calling gcd_of_steps_may_divide_p.
4326 2018-01-31 Julia Koval <julia.koval@intel.com>
4329 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4330 * config/i386/i386.md (rdpid_rex64) New.
4331 (rdpid): Make 32bit only.
4333 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4336 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4337 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4339 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4342 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4344 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4346 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4348 PR rtl-optimization/84071
4349 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4350 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4352 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4354 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4355 (arc_attribute_table): Add 'aux' attribute.
4356 (arc_in_small_data_p): Consider aux like variables.
4357 (arc_is_aux_reg_p): New function.
4358 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4359 (arc_get_aux_arg): New function.
4360 (prepare_move_operands): Handle aux-register access.
4361 (arc_handle_aux_attribute): New function.
4362 * doc/extend.texi (ARC Variable attributes): Add subsection.
4364 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4366 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4367 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4368 (arc_attribute_table): Add 'uncached' attribute.
4369 (arc_print_operand): Print '.di' flag for uncached memory
4371 (arc_in_small_data_p): Do not consider for small data the uncached
4373 (arc_is_uncached_mem_p): New function.
4374 * config/arc/predicates.md (compact_store_memory_operand): Check
4375 for uncached memory accesses.
4376 (nonvol_nonimm_operand): Likewise.
4377 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4379 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4382 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4383 falign-loops=): Add Optimization flag.
4385 2018-01-30 Jeff Law <law@redhat.com>
4388 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4389 INT_REGISTERS_SAVED. Check it prior to calling
4390 get_scratch_register_on_entry.
4391 (ix86_adjust_stack_and_probe): Similarly.
4392 (ix86_emit_probe_stack_range): Similarly.
4393 (ix86_expand_prologue): Corresponding changes.
4395 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4398 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4399 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4401 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4404 * lra-constraints.c (curr_insn_transform): Process AND in the
4407 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4409 PR rtl-optimization/83986
4410 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4411 dependence against last_pending_memory_flush in addition to
4414 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4416 PR tree-optimization/81611
4417 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4420 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4423 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4426 2018-01-30 Richard Biener <rguenther@suse.de>
4427 Jakub Jelinek <jakub@redhat.com>
4429 PR tree-optimization/84111
4430 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4431 inner loops added during recursion, as they don't have up-to-date
4434 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4437 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4438 (can_inline_edge_by_limits_p): ... here.
4439 (can_early_inline_edge_p, check_callers,
4440 update_caller_keys, update_callee_keys, recursive_inlining,
4441 add_new_edges_to_heap, speculation_useful_p,
4442 inline_small_functions,
4443 inline_small_functions, flatten_function,
4444 inline_to_all_callers_1): Update.
4446 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4448 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4451 2018-01-30 Richard Biener <rguenther@suse.de>
4453 PR tree-optimization/83008
4454 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4455 invariant and constant vector uses in stmts when they need
4458 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4461 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4462 * configure: Regenerate.
4464 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4466 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4468 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4469 Use gen_rtx_REG rather than gen_lowpart.
4471 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4473 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4474 rather than 0 when creating partial subregs.
4476 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4478 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4481 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4484 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4485 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4486 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4487 flags. This restores the settings used before the 2017-07-24.
4488 Turning off pre increment/decrement/modify allows IVOPTS to
4489 optimize DF/SF loops where the index is an int.
4491 2018-01-29 Richard Biener <rguenther@suse.de>
4492 Kelvin Nilsen <kelvin@gcc.gnu.org>
4495 * tree-vect-stmts.c (vectorizable_call): Don't call
4496 targetm.vectorize_builtin_md_vectorized_function if callee is
4499 2018-01-22 Carl Love <cel@us.ibm.com>
4501 * doc/extend.tex: Fix typo in second arg in
4502 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4504 2018-01-29 Richard Biener <rguenther@suse.de>
4506 PR tree-optimization/84086
4507 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4508 (flush_ssaname_freelist): When SSA names were released reset
4509 the SCEV hash table.
4511 2018-01-29 Richard Biener <rguenther@suse.de>
4513 PR tree-optimization/84057
4514 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4515 removed paths when removing edges.
4517 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4519 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4520 -mfunction-return=@var{choice}.
4522 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4525 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4527 (layout::print_source_line): Likewise.
4528 (test_get_line_width_without_trailing_whitespace): Add test cases.
4530 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4533 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4536 2018-01-26 Jim Wilson <jimw@sifive.com>
4538 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4540 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4543 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4545 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4546 and CMP + SUB-immediate -> SUBS.
4548 2018-01-26 Martin Sebor <msebor@redhat.com>
4550 PR tree-optimization/83896
4551 * tree-ssa-strlen.c (get_string_len): Rename...
4552 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4553 Avoid assuming length is constant.
4554 (handle_char_store): Use HOST_WIDE_INT for string length.
4556 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4559 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4560 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4562 2018-01-26 Richard Biener <rguenther@suse.de>
4564 PR rtl-optimization/84003
4565 * dse.c (record_store): Only record redundant stores when
4566 the earlier store aliases at least all accesses the later one does.
4568 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4570 PR rtl-optimization/83985
4571 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4572 REG_CFA_RESTORE insns.
4573 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4574 REG_CFA_RESTORE insns here.
4577 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4578 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4580 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4582 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4583 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4584 (arc_init): Likewise.
4585 (arc_override_options): Likewise.
4586 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4588 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4590 * config/arc/arc.h (TARGET_DBNZ): Define.
4591 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4592 properly set the tune attribute.
4593 (dbnz): Use TARGET_DBNZ guard.
4594 * config/arc/arc.opt (mtune): Add core3 option.
4596 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4598 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4599 recognize new pic like addresses.
4600 (arc_delegitimize_address): Clean up.
4602 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4604 * config/arc/arc-arches.def: Option mrf16 valid for all
4606 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4607 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4608 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4609 * config/arc/arc-tables.opt: Regenerate.
4610 * config/arc/arc.c (arc_conditional_register_usage): Handle
4611 reduced register file case.
4612 (arc_file_start): Set must have build attributes.
4613 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4615 * config/arc/arc.opt (mrf16): Add new option.
4616 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4617 * config/arc/genmultilib.awk: Handle new mrf16 option.
4618 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4619 * config/arc/t-multilib: Regenerate.
4620 * doc/invoke.texi (ARC Options): Document mrf16 option.
4622 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4624 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4625 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4626 (arc_attribute_table): Add 'secure_call' attribute.
4627 (arc_print_operand): Print secure call operand.
4628 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4630 (arc_is_secure_call_p): New function. * config/arc/arc.md
4631 (call_i): Add support for sjli instruction.
4632 (call_value_i): Likewise.
4633 * config/arc/constraints.md (Csc): New constraint.
4635 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4636 John Eric Martin <John.Martin@emmicro-us.com>
4638 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4639 * config/arc/arc.c (_arc_jli_section): New struct.
4640 (arc_jli_section): New type.
4641 (rc_jli_sections): New static variable.
4642 (arc_handle_jli_attribute): New function.
4643 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4644 (arc_file_end): New function.
4645 (TARGET_ASM_FILE_END): Define.
4646 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4647 (arc_add_jli_section): New function.
4648 (jli_call_scan): Likewise.
4649 (arc_reorg): Call jli_call_scan.
4650 (arc_output_addsi): Remove 'S' from printing asm operand.
4651 (arc_is_jli_call_p): New function.
4652 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4654 (movhi_insn): Likewise.
4655 (movsi_insn): Likewise.
4656 (movsi_set_cc_insn): Likewise.
4657 (loadqi_update): Likewise.
4658 (load_zeroextendqisi_update): Likewise.
4659 (load_signextendqisi_update): Likewise.
4660 (loadhi_update): Likewise.
4661 (load_zeroextendhisi_update): Likewise.
4662 (load_signextendhisi_update): Likewise.
4663 (loadsi_update): Likewise.
4664 (loadsf_update): Likewise.
4665 (movsicc_insn): Likewise.
4666 (bset_insn): Likewise.
4667 (bxor_insn): Likewise.
4668 (bclr_insn): Likewise.
4669 (bmsk_insn): Likewise.
4670 (bicsi3_insn): Likewise.
4671 (cmpsi_cc_c_insn): Likewise.
4672 (movsi_ne): Likewise.
4673 (movsi_cond_exec): Likewise.
4674 (clrsbsi2): Likewise.
4685 (call_i): Remove 'S' asm letter, add jli instruction.
4686 (call_value_i): Likewise.
4687 * config/arc/arc.op (mjli-always): New option.
4688 * config/arc/constraints.md (Cji): New constraint.
4689 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4691 (subsf3_fpx): Likewise.
4692 (mulsf3_fpx): Likewise.
4693 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4695 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4697 * doc/invoke.texi (ARC): Document mjli-always option.
4699 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4701 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4702 avoid addition with 0 and use incw and decw where possible.
4704 2018-01-26 Richard Biener <rguenther@suse.de>
4706 PR tree-optimization/81082
4707 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4708 association if it requires casting to unsigned.
4709 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4710 from fold_plusminus_mult_expr to catch important cases late when
4711 range info is available.
4713 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4715 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4716 * configure.ac (hidden_linkonce): New test.
4717 * configure: Regenerate.
4718 * config.in: Regenerate.
4720 2018-01-26 Julia Koval <julia.koval@intel.com>
4722 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4723 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4724 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4725 _mm_mask_bitshuffle_epi64_mask): Fix type.
4726 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4727 USI_FTYPE_V4DI_V4DI_USI): Remove.
4728 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4729 __builtin_ia32_vpshufbitqmb256_mask,
4730 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4731 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4732 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4734 2018-01-26 Alan Modra <amodra@gmail.com>
4737 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4738 UNSPEC_VBPERMQ. Sort other unspecs.
4740 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4742 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4744 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4747 * predict.c (drop_profile): Do not push/pop cfun; update also
4749 (handle_missing_profiles): Fix logic looking for zero profiles.
4751 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4754 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4755 on functions with #pragma omp declare simd or functions with simd
4757 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4758 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4759 Remove trailing \n from warning_at calls.
4761 2018-01-25 Tom de Vries <tom@codesourcery.com>
4764 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4765 for neutered workers.
4767 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4770 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4771 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4773 2018-01-24 Jeff Law <law@redhat.com>
4776 * i386.c (get_probe_interval): Move to earlier point.
4777 (ix86_compute_frame_layout): If -fstack-clash-protection and
4778 the frame is larger than the probe interval, then use pushes
4779 to save registers rather than reg->mem moves.
4780 (ix86_expand_prologue): Remove conditional for int_registers_saved
4783 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4786 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4787 min/max for never referenced object.
4789 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4792 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4794 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4795 attributes from DECL_ATTRIBUTES (decl) without affecting
4796 DECL_ATTRIBUTES (current_function_decl).
4797 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4798 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4800 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4802 PR tree-optimization/83979
4803 * fold-const.c (fold_comparison): Use constant_boolean_node
4804 instead of boolean_{true,false}_node.
4806 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4808 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4811 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4813 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4814 Simplify the clause that sets the length attribute.
4815 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4816 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4817 clause that sets the length attribute.
4818 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4820 2018-01-24 Tom de Vries <tom@codesourcery.com>
4823 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4824 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4825 Add strict parameter.
4826 (prevent_branch_around_nothing): Insert dummy insn between branch to
4827 label and label with no ptx insn inbetween.
4828 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4830 2018-01-24 Tom de Vries <tom@codesourcery.com>
4833 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4834 for neutered threads in warp.
4835 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4837 2018-01-24 Richard Biener <rguenther@suse.de>
4839 PR tree-optimization/83176
4840 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4843 2018-01-24 Richard Biener <rguenther@suse.de>
4845 PR tree-optimization/82819
4846 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4847 code generating pluses that are no-ops in the target precision.
4849 2018-01-24 Richard Biener <rguenther@suse.de>
4852 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4854 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4856 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4857 to merge probabilities.
4858 * predict.c (probably_never_executed): Also mark as cold functions
4859 with global 0 profile and guessed local profile.
4860 * profile-count.c (profile_probability::combine_with_count): New
4862 * profile-count.h (profile_probability::operator*,
4863 profile_probability::operator*=, profile_probability::operator/,
4864 profile_probability::operator/=): Reduce precision to adjusted
4865 and set value to guessed on contradictory divisions.
4866 (profile_probability::combine_with_freq): Remove.
4867 (profile_probability::combine_wiht_count): Declare.
4868 (profile_count::force_nonzero):: Set to adjusted.
4869 (profile_count::probability_in):: Set quality to adjusted.
4870 * tree-ssa-tail-merge.c (replace_block_by): Use
4873 2018-01-23 Andrew Waterman <andrew@sifive.com>
4874 Jim Wilson <jimw@sifive.com>
4876 * config/riscv/riscv.c (riscv_stack_boundary): New.
4877 (riscv_option_override): Set riscv_stack_boundary. Handle
4878 riscv_preferred_stack_boundary_arg.
4879 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4880 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4881 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4882 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4883 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4884 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4886 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4889 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4890 of struct ix86_frame.
4891 (ix86_expand_epilogue): Likewise. Add a local variable for
4892 the reg_save_offset field in struct ix86_frame.
4894 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4896 PR tree-optimization/82604
4897 * tree-loop-distribution.c (enum partition_kind): New enum item
4898 PKIND_PARTIAL_MEMSET.
4899 (partition_builtin_p): Support above new enum item.
4900 (generate_code_for_partition): Ditto.
4901 (compute_access_range): Differentiate cases that equality can be
4902 proven at all loops, the innermost loops or no loops.
4903 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4904 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4905 (finalize_partitions, distribute_loop): Don't fuse partition of
4906 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4907 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4910 2018-01-23 Martin Liska <mliska@suse.cz>
4912 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4913 order to ignore the predictor.
4914 (PRED_POLYMORPHIC_CALL): Likewise.
4915 (PRED_RECURSIVE_CALL): Likewise.
4917 2018-01-23 Martin Liska <mliska@suse.cz>
4919 * tree-profile.c (tree_profiling): Print function header to
4920 aware reader which function we are working on.
4921 * value-prof.c (gimple_find_values_to_profile): Do not print
4922 not interesting value histograms.
4924 2018-01-23 Martin Liska <mliska@suse.cz>
4926 * profile-count.h (enum profile_quality): Add
4927 profile_uninitialized as the first value. Do not number values
4928 as they are zero based.
4929 (profile_count::verify): Update sanity check.
4930 (profile_probability::verify): Likewise.
4932 2018-01-23 Nathan Sidwell <nathan@acm.org>
4934 * doc/invoke.texi (ffor-scope): Deprecate.
4936 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4938 PR tree-optimization/83510
4939 * domwalk.c (set_all_edges_as_executable): New function.
4940 (dom_walker::dom_walker): Convert bool param
4941 "skip_unreachable_blocks" to enum reachability. Move setup of
4942 edge flags to set_all_edges_as_executable and only do it when
4943 reachability is REACHABLE_BLOCKS.
4944 * domwalk.h (enum dom_walker::reachability): New enum.
4945 (dom_walker::dom_walker): Convert bool param
4946 "skip_unreachable_blocks" to enum reachability.
4947 (set_all_edges_as_executable): New decl.
4948 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4949 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4951 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4952 but converting true to REACHABLE_BLOCKS.
4953 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4955 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4956 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4957 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4959 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4960 if check_all_array_refs will be called.
4962 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4964 * tree.c (selftest::test_location_wrappers): Add more test
4967 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4969 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4970 (selftest::test_bit_in_range): Likewise.
4972 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4975 * doc/sourcebuild.texi (vect_float): Say that the selector
4976 only describes the situation when -funsafe-math-optimizations is on.
4977 (vect_float_strict): Document.
4979 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4981 PR tree-optimization/83965
4982 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4983 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4984 instead of checking only for a reduction.
4985 (vect_recog_widen_sum_pattern): Likewise.
4987 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4989 * predict.c (probably_never_executed): Only use precise profile info.
4990 (compute_function_frequency): Skip after inlining hack since we now
4991 have quality checking.
4993 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4995 * profile-count.h (profile_probability::very_unlikely,
4996 profile_probability::unlikely, profile_probability::even): Set
4997 precision to guessed.
4999 2018-01-23 Richard Biener <rguenther@suse.de>
5001 PR tree-optimization/83963
5002 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5003 Properly terminate dominator walk when crossing the exit edge not
5004 when visiting its source block.
5006 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5009 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5010 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5012 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5014 PR tree-optimization/83957
5015 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5016 semicolon after for body surrounded by braces.
5018 PR tree-optimization/83081
5019 * profile-count.h (profile_probability::split): New method.
5020 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5021 Use profile_probability::split.
5022 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5023 when splitting a single conditional jump into 2.
5025 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5027 PR tree-optimization/69452
5028 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5031 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5033 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5034 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5035 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5037 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5039 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5040 * config/rl78/rl78.md (movdi): New define_expand.
5041 * config/rl78/rl78.c (rl78_split_movdi): New function.
5043 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5046 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5048 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5049 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5050 128-bit to produce an UNSPEC move to get the double word with the
5051 signbit and then a shift directly to do signbit.
5052 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5053 implementation with a new version that just does either a direct
5054 move or a regular move. Move memory interface to separate insns.
5055 Move insns so they are next to the expander.
5056 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5057 with signbit move. Split big and little endian case.
5058 (signbit<mode>2_dm_mem_le): Likewise.
5059 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5060 (signbit<mode>2_dm2): Likewise.
5062 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5064 * config/rl78/rl78.md (anddi3): New define_expand.
5066 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5068 * config/rl78/rl78.md (umindi3): New define_expand.
5070 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5072 * config/rl78/rl78.md (smindi3): New define_expand.
5074 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5076 * config/rl78/rl78.md (smaxdi3): New define_expand.
5078 2018-01-22 Carl Love <cel@us.ibm.com>
5080 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5081 LVX_V1TI): Add macro expansion.
5082 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5083 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5084 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5085 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5086 Change check to determine if the instruction is a byte reversing
5087 entry. Fix typo in comment.
5088 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5089 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5090 Add def_builtin calls for new builtins.
5091 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5092 Add define_insn expansion.
5094 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5096 * config/rl78/rl78.md (umaxdi3): New define_expand.
5098 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5100 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5101 for non-QImode registers.
5103 2018-01-22 Richard Biener <rguenther@suse.de>
5105 PR tree-optimization/83963
5106 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5107 including the loop exit block.
5108 (scop_detection::merge_sese): Likewise.
5109 (scop_detection::add_scop): Do it here instead.
5111 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5113 * doc/sourcebuild.texi (arm_softfloat): Document.
5115 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5118 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5119 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5120 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5122 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5123 David Edelsohn <dje.gcc@gmail.com>
5126 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5127 Change "crset eq" to "crset 2".
5128 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5129 (*call_indirect_aix<mode>_nospec): Likewise.
5130 (*call_value_indirect_aix<mode>_nospec): Likewise.
5131 (*call_indirect_elfv2<mode>_nospec): Likewise.
5132 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5133 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5134 change assembly output from . to $.
5135 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5136 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5137 (*tablejump<mode>_internal1_nospec): Likewise.
5139 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5142 * config/sh/sh_optimize_sett_clrt.cc:
5143 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5145 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5147 PR tree-optimization/83940
5148 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5149 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5150 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5151 use it to pass back the definition type.
5152 (vect_check_store_rhs): Likewise rhs_dt_out.
5153 (vect_build_gather_load_calls): Add a mask_dt argument and use
5154 it instead of a call to vect_is_simple_use.
5155 (vectorizable_store): Update calls to vect_check_load_store_mask
5156 and vect_check_store_rhs. Use the dt returned by the latter instead
5157 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5158 instead of calls to vect_is_simple_use. Pass the scalar rather
5159 than the vector operand to vect_is_simple_use when handling
5160 second and subsequent copies of an rhs value.
5161 (vectorizable_load): Update calls to vect_check_load_store_mask
5162 and vect_build_gather_load_calls. Use the cached mask_dt and
5163 gs_info.offset_dt instead of calls to vect_is_simple_use.
5165 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5168 * tree-emutls.c: Include gimplify.h.
5169 (lower_emutls_2): New function.
5170 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5171 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5172 it before further processing.
5175 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5176 UINTVAL (trueop1) instead of INTVAL (op1).
5178 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5182 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5183 INCOMING_FRAME_SP_OFFSET if not defined.
5184 (scan_trace): Add ENTRY argument. If true and
5185 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5186 emit a note to adjust the CFA offset.
5187 (create_cfi_notes): Adjust scan_trace callers.
5188 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5189 INCOMING_FRAME_SP_OFFSET in the CIE.
5190 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5191 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5193 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5194 * doc/tm.texi: Regenerated.
5196 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5198 PR rtl-optimization/83147
5199 * lra-constraints.c (remove_inheritance_pseudos): Use
5200 lra_substitute_pseudo_within_insn.
5202 2018-01-19 Tom de Vries <tom@codesourcery.com>
5203 Cesar Philippidis <cesar@codesourcery.com>
5206 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5208 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5211 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5212 spaces for function labels.
5214 2018-01-19 Martin Liska <mliska@suse.cz>
5216 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5217 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5218 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5219 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5220 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5221 (PRED_CONST_RETURN): Change from 69 to 65.
5222 (PRED_NULL_RETURN): Change from 91 to 71.
5223 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5224 (PRED_LOOP_GUARD): Change from 66 to 73.
5226 2018-01-19 Martin Liska <mliska@suse.cz>
5228 * predict.c (predict_insn_def): Add new assert.
5229 (struct branch_predictor): Change type to signed integer.
5230 (test_prediction_value_range): Amend test to cover
5232 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5233 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5234 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5235 (PRED_LOOP_IV_COMPARE): Likewise.
5236 * predict.h (PROB_UNINITIALIZED): Define new constant.
5238 2018-01-19 Martin Liska <mliska@suse.cz>
5240 * predict.c (dump_prediction): Add new format for
5241 analyze_brprob.py script which is enabled with -details
5243 * profile-count.h (precise_p): New function.
5245 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5247 PR tree-optimization/83922
5248 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5249 there are no statements that need masking.
5250 (vect_active_double_reduction_p): New function.
5251 (vect_analyze_loop_operations): Use it when handling phis that
5252 are not in the loop header.
5254 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5256 PR tree-optimization/83914
5257 * tree-vect-loop.c (vectorizable_induction): Don't convert
5258 init_expr or apply the peeling adjustment for inductions
5259 that are nested within the vectorized loop.
5261 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5263 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5266 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5270 * function.h (gimplify_parameters): Add gimple_seq * argument.
5271 * function.c: Include gimple.h and options.h.
5272 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5273 for the added local temporaries if needed.
5274 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5275 if there are any parameter cleanups, wrap whole body into a
5276 try/finally with the cleanups.
5278 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5281 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5282 Use GET_MODE_CLASS for scalar floating point.
5284 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5288 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5289 Fix call of call_cgraph_insertion_hooks.
5291 2018-01-18 Martin Sebor <msebor@redhat.com>
5293 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5295 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5298 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5301 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5304 * common.opt: (-ffile-prefix-map): New option.
5305 * opts.c (common_handle_option): Defer it.
5306 * opts-global.c (handle_common_deferred_options): Handle it.
5307 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5308 * file-prefix-map.h: New file.
5309 (remap_debug_filename, add_debug_prefix_map): ...here.
5310 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5311 * final.c (debug_prefix_map, add_debug_prefix_map
5312 remap_debug_filename): Move to...
5313 * file-prefix-map.c: New file.
5314 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5315 generalize, get rid of alloca(), use strrchr() instead of strchr().
5316 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5317 Implement in terms of add_prefix_map().
5318 (remap_macro_filename, remap_debug_filename): Implement in term of
5320 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5321 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5322 * dbxout.c: Include file-prefix-map.h.
5323 * varasm.c: Likewise.
5324 * vmsdbgout.c: Likewise.
5325 * xcoffout.c: Likewise.
5326 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5327 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5328 * doc/invoke.texi (-ffile-prefix-map): Document.
5329 (-fdebug-prefix-map): Update description.
5331 2018-01-18 Martin Liska <mliska@suse.cz>
5333 * config/i386/i386.c (indirect_thunk_name): Document that also
5335 (output_indirect_thunk): Document why both instructions
5336 (pause and lfence) are generated.
5338 2018-01-18 Richard Biener <rguenther@suse.de>
5340 PR tree-optimization/83887
5341 * graphite-scop-detection.c
5342 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5343 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5344 (scop_detection::merge_sese): Re-implement with a flood-fill
5345 algorithm that properly finds a SESE region if it exists.
5347 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5350 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5351 pointer_diff optimizations use view_convert instead of convert.
5353 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5355 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5356 Generate different code for -mno-speculate-indirect-jumps.
5357 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5358 (*call_indirect_aix<mode>): Disable for
5359 -mno-speculate-indirect-jumps.
5360 (*call_indirect_aix<mode>_nospec): New define_insn.
5361 (*call_value_indirect_aix<mode>): Disable for
5362 -mno-speculate-indirect-jumps.
5363 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5364 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5365 -mno-speculate-indirect-jumps.
5366 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5368 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5370 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5371 long double type, set the flags for noting the default long double
5372 type, even if we don't pass or return a long double type.
5374 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5377 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5380 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5382 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5383 support for merge[hl].
5384 (fold_mergehl_helper): New helper function.
5385 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5386 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5387 (altivec_vmrglw_direct): Add xxmrglw insn.
5389 2018-01-17 Andrew Waterman <andrew@sifive.com>
5391 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5392 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5394 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5397 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5398 call the lto_location_cache before reading the
5399 DECL_SOURCE_LOCATION of the types.
5401 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5402 Richard Sandiford <richard.sandiford@linaro.org>
5404 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5405 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5406 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5407 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5408 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5410 * config/aarch64/constraints.md (aarch64_movti_operand):
5412 * config/aarch64/predicates.md (Uti): Add new constraint.
5414 2018-01-17 Carl Love <cel@us.ibm.com>
5416 * config/rs6000/vsx.md (define_expand xl_len_r,
5417 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5418 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5420 (define_expand, define_insn): Move the shift left from the
5421 define_insn to the define_expand for lxvl and stxvl instructions.
5422 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5423 and XL_LEN_R definitions to PURE.
5425 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5427 * config/i386/i386.c (indirect_thunk_name): Declare regno
5428 as unsigned int. Compare regno with INVALID_REGNUM.
5429 (output_indirect_thunk): Ditto.
5430 (output_indirect_thunk_function): Ditto.
5431 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5432 in the call to output_indirect_thunk_function.
5434 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5437 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5438 rather than the size of inner_type to determine the stack slot size
5439 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5441 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5444 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5447 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5449 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5450 endian Linux systems to optionally enable multilibs for selecting
5451 the long double type if the user configured an explicit type.
5452 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5453 have no long double multilibs if not defined.
5454 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5455 warn if the user used -mabi={ieee,ibm}longdouble and we built
5456 multilibs for long double.
5457 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5458 appropriate multilib option.
5459 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5461 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5462 for building long double multilibs.
5463 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5465 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5467 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5470 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5472 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5475 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5478 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5481 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5483 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5484 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5486 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5488 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5489 different rtl trees depending on TARGET_64BIT.
5490 (rs6000_gen_lvx): Likewise.
5492 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5494 * config/visium/visium.md (nop): Tweak comment.
5495 (hazard_nop): Likewise.
5497 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5499 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5500 -mspeculate-indirect-jumps.
5501 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5502 for -mno-speculate-indirect-jumps.
5503 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5504 (*call_value_indirect_elfv2<mode>): Disable for
5505 -mno-speculate-indirect-jumps.
5506 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5507 (indirect_jump): Emit different RTL for
5508 -mno-speculate-indirect-jumps.
5509 (*indirect_jump<mode>): Disable for
5510 -mno-speculate-indirect-jumps.
5511 (*indirect_jump<mode>_nospec): New define_insn.
5512 (tablejump): Emit different RTL for
5513 -mno-speculate-indirect-jumps.
5514 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5515 (tablejumpsi_nospec): New define_expand.
5516 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5517 (tablejumpdi_nospec): New define_expand.
5518 (*tablejump<mode>_internal1): Disable for
5519 -mno-speculate-indirect-jumps.
5520 (*tablejump<mode>_internal1_nospec): New define_insn.
5521 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5524 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5526 * caller-save.c (insert_save): Drop unnecessary parameter. All
5529 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5530 Richard Biener <rguenth@suse.de>
5533 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5534 return early, inline manually is_gimple_sizepos. Make sure if we
5535 call gimplify_expr we don't end up with a gimple constant.
5536 * tree.c (variably_modified_type_p): Don't return true for
5537 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5538 * gimplify.h (is_gimple_sizepos): Remove.
5540 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5542 PR tree-optimization/83857
5543 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5544 vectorizable_live_operation for pure SLP statements.
5545 (vectorizable_live_operation): Handle PHIs.
5547 2018-01-16 Richard Biener <rguenther@suse.de>
5549 PR tree-optimization/83867
5550 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5551 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5553 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5556 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5557 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5558 If off is not INTEGER_CST, issue a may not be aligned warning
5559 rather than isn't aligned. Use isn%'t rather than isn't.
5560 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5562 <case MULT_EXPR>: Improve the case when bottom and one of the
5563 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5564 operand, in that case check if the other operand is multiple of
5565 bottom divided by the INTEGER_CST operand.
5567 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5570 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5571 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5572 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5573 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5574 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5575 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5576 (pa_function_arg_size): New function.
5578 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5580 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5581 in a separate statement.
5583 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5585 PR tree-optimization/83847
5586 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5587 group gathers and scatters.
5589 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5591 PR rtl-optimization/86620
5592 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5594 PR rtl-optimization/83213
5595 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5596 to last if both are JUMP_INSNs.
5598 PR tree-optimization/83843
5599 * gimple-ssa-store-merging.c
5600 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5601 store_immediate_info for bswap/nop orig_stores.
5603 2018-01-15 Andrew Waterman <andrew@sifive.com>
5605 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5607 <UDIV>: Increase cost if !TARGET_DIV.
5609 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5611 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5612 (define_attr "cr_logical_3op"): New.
5613 (cceq_ior_compare): Adjust.
5614 (cceq_ior_compare_complement): Adjust.
5615 (*cceq_rev_compare): Adjust.
5616 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5617 (is_cracked_insn): Adjust.
5618 (insn_must_be_first_in_group): Adjust.
5619 * config/rs6000/40x.md: Adjust.
5620 * config/rs6000/440.md: Adjust.
5621 * config/rs6000/476.md: Adjust.
5622 * config/rs6000/601.md: Adjust.
5623 * config/rs6000/603.md: Adjust.
5624 * config/rs6000/6xx.md: Adjust.
5625 * config/rs6000/7450.md: Adjust.
5626 * config/rs6000/7xx.md: Adjust.
5627 * config/rs6000/8540.md: Adjust.
5628 * config/rs6000/cell.md: Adjust.
5629 * config/rs6000/e300c2c3.md: Adjust.
5630 * config/rs6000/e500mc.md: Adjust.
5631 * config/rs6000/e500mc64.md: Adjust.
5632 * config/rs6000/e5500.md: Adjust.
5633 * config/rs6000/e6500.md: Adjust.
5634 * config/rs6000/mpc.md: Adjust.
5635 * config/rs6000/power4.md: Adjust.
5636 * config/rs6000/power5.md: Adjust.
5637 * config/rs6000/power6.md: Adjust.
5638 * config/rs6000/power7.md: Adjust.
5639 * config/rs6000/power8.md: Adjust.
5640 * config/rs6000/power9.md: Adjust.
5641 * config/rs6000/rs64.md: Adjust.
5642 * config/rs6000/titan.md: Adjust.
5644 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5646 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5647 ix86_indirect_branch_register logic.
5649 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5651 * config/i386/constraints.md (Bs): Update
5652 ix86_indirect_branch_register check. Don't check
5653 ix86_indirect_branch_register with GOT_memory_operand.
5655 * config/i386/predicates.md (GOT_memory_operand): Don't check
5656 ix86_indirect_branch_register here.
5657 (GOT32_symbol_operand): Likewise.
5659 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5661 * config/i386/predicates.md (constant_call_address_operand):
5662 Rewrite ix86_indirect_branch_register logic.
5663 (sibcall_insn_operand): Likewise.
5665 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5667 * config/i386/constraints.md (Bs): Replace
5668 ix86_indirect_branch_thunk_register with
5669 ix86_indirect_branch_register.
5671 * config/i386/i386.md (indirect_jump): Likewise.
5672 (tablejump): Likewise.
5673 (*sibcall_memory): Likewise.
5674 (*sibcall_value_memory): Likewise.
5675 Peepholes of indirect call and jump via memory: Likewise.
5676 * config/i386/i386.opt: Likewise.
5677 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5678 (GOT_memory_operand): Likewise.
5679 (call_insn_operand): Likewise.
5680 (sibcall_insn_operand): Likewise.
5681 (GOT32_symbol_operand): Likewise.
5683 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5686 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5687 type rather than type addr's type points to.
5688 (expand_omp_atomic_mutex): Likewise.
5689 (expand_omp_atomic): Likewise.
5691 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5694 * config/i386/i386.c (output_indirect_thunk_function): Use
5695 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5696 for __x86_return_thunk.
5698 2018-01-15 Richard Biener <rguenther@suse.de>
5701 * expmed.c (extract_bit_field_1): Fix typo.
5703 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5706 * config/arm/iterators.md (VF): New mode iterator.
5707 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5708 Remove integer-related logic from pattern.
5709 (neon_vabd<mode>_3): Likewise.
5711 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5714 * common.opt (fstrict-overflow): No longer an alias.
5715 (fwrapv-pointer): New option.
5716 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5717 also for pointer types based on flag_wrapv_pointer.
5718 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5719 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5720 opts->x_flag_wrapv got set.
5721 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5722 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5723 POINTER_TYPE_OVERFLOW_UNDEFINED.
5724 * match.pd: Likewise in address comparison pattern.
5725 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5727 2018-01-15 Richard Biener <rguenther@suse.de>
5730 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5731 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5732 Reset type names to their identifier if their TYPE_DECL doesn't
5733 have linkage (and thus is used for ODR and devirt).
5734 (save_debug_info_for_decl): Remove.
5735 (save_debug_info_for_type): Likewise.
5736 (add_tree_to_fld_list): Adjust.
5737 * tree-pretty-print.c (dump_generic_node): Make dumping of
5738 type names more robust.
5740 2018-01-15 Richard Biener <rguenther@suse.de>
5742 * BASE-VER: Bump to 8.0.1.
5744 2018-01-14 Martin Sebor <msebor@redhat.com>
5747 * builtins.c (check_access): Avoid warning when the no-warning bit
5750 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5752 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5753 * ira-color (allocno_hard_regs_compare): Likewise.
5755 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5758 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5759 Use .pushsection/.popsection.
5761 2018-01-14 Martin Sebor <msebor@redhat.com>
5764 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5766 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5768 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5769 entry from extra_headers.
5770 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5771 extra_headers, make the list bitwise identical to the i?86-*-* one.
5773 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5775 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5776 -mcmodel=large with -mindirect-branch=thunk,
5777 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5778 -mfunction-return=thunk-extern.
5779 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5780 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5781 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5783 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5785 * config/i386/i386.c (print_reg): Print the name of the full
5786 integer register without '%'.
5787 (ix86_print_operand): Handle 'V'.
5788 * doc/extend.texi: Document 'V' modifier.
5790 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5792 * config/i386/constraints.md (Bs): Disallow memory operand for
5793 -mindirect-branch-register.
5795 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5796 (GOT_memory_operand): Likewise.
5797 (call_insn_operand): Likewise.
5798 (sibcall_insn_operand): Likewise.
5799 (GOT32_symbol_operand): Likewise.
5800 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5801 for -mindirect-branch-register.
5802 (tablejump): Likewise.
5803 (*sibcall_memory): Likewise.
5804 (*sibcall_value_memory): Likewise.
5805 Disallow peepholes of indirect call and jump via memory for
5806 -mindirect-branch-register.
5807 (*call_pop): Replace m with Bw.
5808 (*call_value_pop): Likewise.
5809 (*sibcall_pop_memory): Replace m with Bs.
5810 * config/i386/i386.opt (mindirect-branch-register): New option.
5811 * doc/invoke.texi: Document -mindirect-branch-register option.
5813 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5815 * config/i386/i386-protos.h (ix86_output_function_return): New.
5816 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5817 set function_return_type.
5818 (indirect_thunk_name): Add ret_p to indicate thunk for function
5820 (output_indirect_thunk_function): Pass false to
5821 indirect_thunk_name.
5822 (ix86_output_indirect_branch_via_reg): Likewise.
5823 (ix86_output_indirect_branch_via_push): Likewise.
5824 (output_indirect_thunk_function): Create alias for function
5825 return thunk if regno < 0.
5826 (ix86_output_function_return): New function.
5827 (ix86_handle_fndecl_attribute): Handle function_return.
5828 (ix86_attribute_table): Add function_return.
5829 * config/i386/i386.h (machine_function): Add
5830 function_return_type.
5831 * config/i386/i386.md (simple_return_internal): Use
5832 ix86_output_function_return.
5833 (simple_return_internal_long): Likewise.
5834 * config/i386/i386.opt (mfunction-return=): New option.
5835 (indirect_branch): Mention -mfunction-return=.
5836 * doc/extend.texi: Document function_return function attribute.
5837 * doc/invoke.texi: Document -mfunction-return= option.
5839 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5841 * config/i386/i386-opts.h (indirect_branch): New.
5842 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5843 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5844 with local indirect jump when converting indirect call and jump.
5845 (ix86_set_indirect_branch_type): New.
5846 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5847 (indirectlabelno): New.
5848 (indirect_thunk_needed): Likewise.
5849 (indirect_thunk_bnd_needed): Likewise.
5850 (indirect_thunks_used): Likewise.
5851 (indirect_thunks_bnd_used): Likewise.
5852 (INDIRECT_LABEL): Likewise.
5853 (indirect_thunk_name): Likewise.
5854 (output_indirect_thunk): Likewise.
5855 (output_indirect_thunk_function): Likewise.
5856 (ix86_output_indirect_branch_via_reg): Likewise.
5857 (ix86_output_indirect_branch_via_push): Likewise.
5858 (ix86_output_indirect_branch): Likewise.
5859 (ix86_output_indirect_jmp): Likewise.
5860 (ix86_code_end): Call output_indirect_thunk_function if needed.
5861 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5863 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5864 (ix86_attribute_table): Add indirect_branch.
5865 * config/i386/i386.h (machine_function): Add indirect_branch_type
5866 and has_local_indirect_jump.
5867 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5869 (tablejump): Likewise.
5870 (*indirect_jump): Use ix86_output_indirect_jmp.
5871 (*tablejump_1): Likewise.
5872 (simple_return_indirect_internal): Likewise.
5873 * config/i386/i386.opt (mindirect-branch=): New option.
5874 (indirect_branch): New.
5877 (thunk-inline): Likewise.
5878 (thunk-extern): Likewise.
5879 * doc/extend.texi: Document indirect_branch function attribute.
5880 * doc/invoke.texi: Document -mindirect-branch= option.
5882 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5885 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5887 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5889 * ipa-inline.c (want_inline_small_function_p): Return false if
5890 inlining has already failed with CIF_FINAL_ERROR.
5891 (update_caller_keys): Call want_inline_small_function_p before
5893 (update_callee_keys): Likewise.
5895 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5897 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5899 (rs6000_quadword_masked_address_p): Likewise.
5900 (quad_aligned_load_p): Likewise.
5901 (quad_aligned_store_p): Likewise.
5902 (const_load_sequence_p): Add comment to describe the outer-most loop.
5903 (mimic_memory_attributes_and_flags): New function.
5904 (rs6000_gen_stvx): Likewise.
5905 (replace_swapped_aligned_store): Likewise.
5906 (rs6000_gen_lvx): Likewise.
5907 (replace_swapped_aligned_load): Likewise.
5908 (replace_swapped_load_constant): Capitalize argument name in
5909 comment describing this function.
5910 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5911 and stores that access quad-word aligned addresses and replace
5912 with stvx or lvx instructions when appropriate.
5913 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5914 New function prototype.
5915 (rs6000_quadword_masked_address_p): Likewise.
5916 (rs6000_gen_lvx): Likewise.
5917 (rs6000_gen_stvx): Likewise.
5918 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5919 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5920 when memory address is aligned.
5921 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5922 this split to select lvx instruction when memory address is aligned.
5923 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5924 instruction when memory address is aligned.
5925 (*vsx_le_perm_load_v16qi): Likewise.
5926 (four unnamed splitters): Modify to select the stvx instruction
5927 when memory is aligned.
5929 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5931 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5932 which appears in the queue multiple times.
5934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5935 Alan Hayward <alan.hayward@arm.com>
5936 David Sherwood <david.sherwood@arm.com>
5938 * tree-vectorizer.h (vec_lower_bound): New structure.
5939 (_loop_vec_info): Add check_nonzero and lower_bounds.
5940 (LOOP_VINFO_CHECK_NONZERO): New macro.
5941 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5942 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5943 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5944 fields. Make seg_len the distance travelled, not including the
5946 (dr_direction_indicator): Declare.
5947 (dr_zero_step_indicator): Likewise.
5948 (dr_known_forward_stride_p): Likewise.
5949 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5951 (runtime_alias_check_p): Allow runtime alias checks with
5953 (operator ==): Compare access_size and align.
5954 (prune_runtime_alias_test_list): Rework for new distinction between
5955 the access_size and seg_len.
5956 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5958 (get_segment_min_max): New function.
5959 (create_intersect_range_checks): Use it.
5960 (dr_step_indicator): New function.
5961 (dr_direction_indicator): Likewise.
5962 (dr_zero_step_indicator): Likewise.
5963 (dr_known_forward_stride_p): Likewise.
5964 * tree-loop-distribution.c (data_ref_segment_size): Return
5965 DR_STEP * (niters - 1).
5966 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5968 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5969 (vect_preserves_scalar_order_p): New function, split out from...
5970 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5971 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5972 (vect_vfa_access_size): New function.
5973 (vect_vfa_align): Likewise.
5974 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5975 (dump_lower_bound): New function.
5976 (vect_check_lower_bound): Likewise.
5977 (vect_small_gap_p): Likewise.
5978 (vectorizable_with_step_bound_p): Likewise.
5979 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5980 depencies if the vectorization factor is 1. Convert the checks
5981 for nonzero steps into checks on the bounds of DR_STEP. Try using
5982 a bunds check for variable steps if the minimum required step is
5983 relatively small. Update calls to the dr_with_seg_len
5984 constructor and to vect_compile_time_alias.
5985 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5987 (vect_loop_versioning): Call it.
5988 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5990 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5993 Alan Hayward <alan.hayward@arm.com>
5994 David Sherwood <david.sherwood@arm.com>
5996 * doc/sourcebuild.texi (vect_scatter_store): Document.
5997 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5999 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6001 * genopinit.c (main): Add supports_vec_scatter_store and
6002 supports_vec_scatter_store_cached to target_optabs.
6003 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6004 IFN_MASK_SCATTER_STORE.
6005 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6007 * internal-fn.h (internal_store_fn_p): Declare.
6008 (internal_fn_stored_value_index): Likewise.
6009 * internal-fn.c (scatter_store_direct): New macro.
6010 (expand_scatter_store_optab_fn): New function.
6011 (direct_scatter_store_optab_supported_p): New macro.
6012 (internal_store_fn_p): New function.
6013 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6014 IFN_MASK_SCATTER_STORE.
6015 (internal_fn_mask_index): Likewise.
6016 (internal_fn_stored_value_index): New function.
6017 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6019 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6020 * optabs-query.c (supports_vec_scatter_store_p): New function.
6021 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6022 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6023 true for scatter stores.
6024 (vect_gather_scatter_fn_p): Handle scatter stores too.
6025 (vect_check_gather_scatter): Consider using scatter stores if
6026 supports_vec_scatter_store_p.
6027 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6029 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6030 internal_fn_stored_value_index.
6031 (check_load_store_masking): Handle scatter stores too.
6032 (vect_get_store_rhs): Make public.
6033 (vectorizable_call): Use internal_store_fn_p.
6034 (vectorizable_store): Handle scatter store internal functions.
6035 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6036 when deciding whether the end of the group has been reached.
6037 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6038 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6039 (mask_scatter_store<mode>): New insns.
6041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6042 Alan Hayward <alan.hayward@arm.com>
6043 David Sherwood <david.sherwood@arm.com>
6045 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6046 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6047 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6049 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6050 Use vect_truncate_gather_scatter_offset if we can't treat the
6051 operation as a normal gather load or scatter store.
6052 (get_group_load_store_type): Take the gather_scatter_info
6053 as argument. Try using a gather load or scatter store for
6054 single-element groups.
6055 (get_load_store_type): Update calls to get_group_load_store_type
6056 and vect_use_strided_gather_scatters_p.
6058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6059 Alan Hayward <alan.hayward@arm.com>
6060 David Sherwood <david.sherwood@arm.com>
6062 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6063 optional tree argument.
6064 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6066 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6067 but continue to use the current value as a fallback.
6068 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6069 to compare the updates.
6070 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6071 (get_load_store_type): Use it when handling a strided access.
6072 (vect_get_strided_load_store_ops): New function.
6073 (vect_get_data_ptr_increment): Likewise.
6074 (vectorizable_load): Handle strided gather loads. Always pass
6075 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6078 Alan Hayward <alan.hayward@arm.com>
6079 David Sherwood <david.sherwood@arm.com>
6081 * doc/md.texi (gather_load@var{m}): Document.
6082 (mask_gather_load@var{m}): Likewise.
6083 * genopinit.c (main): Add supports_vec_gather_load and
6084 supports_vec_gather_load_cached to target_optabs.
6085 * optabs-tree.c (init_tree_optimization_optabs): Use
6086 ggc_cleared_alloc to allocate target_optabs.
6087 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6088 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6090 * internal-fn.h (internal_load_fn_p): Declare.
6091 (internal_gather_scatter_fn_p): Likewise.
6092 (internal_fn_mask_index): Likewise.
6093 (internal_gather_scatter_fn_supported_p): Likewise.
6094 * internal-fn.c (gather_load_direct): New macro.
6095 (expand_gather_load_optab_fn): New function.
6096 (direct_gather_load_optab_supported_p): New macro.
6097 (direct_internal_fn_optab): New function.
6098 (internal_load_fn_p): Likewise.
6099 (internal_gather_scatter_fn_p): Likewise.
6100 (internal_fn_mask_index): Likewise.
6101 (internal_gather_scatter_fn_supported_p): Likewise.
6102 * optabs-query.c (supports_at_least_one_mode_p): New function.
6103 (supports_vec_gather_load_p): Likewise.
6104 * optabs-query.h (supports_vec_gather_load_p): Declare.
6105 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6106 and memory_type field.
6107 (NUM_PATTERNS): Bump to 15.
6108 * tree-vect-data-refs.c: Include internal-fn.h.
6109 (vect_gather_scatter_fn_p): New function.
6110 (vect_describe_gather_scatter_call): Likewise.
6111 (vect_check_gather_scatter): Try using internal functions for
6112 gather loads. Recognize existing calls to a gather load function.
6113 (vect_analyze_data_refs): Consider using gather loads if
6114 supports_vec_gather_load_p.
6115 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6116 (vect_get_gather_scatter_offset_type): Likewise.
6117 (vect_convert_mask_for_vectype): Likewise.
6118 (vect_add_conversion_to_patterm): Likewise.
6119 (vect_try_gather_scatter_pattern): Likewise.
6120 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6121 (vect_vect_recog_func_ptrs): Add it.
6122 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6123 internal_fn_mask_index and internal_gather_scatter_fn_p.
6124 (check_load_store_masking): Take the gather_scatter_info as an
6125 argument and handle gather loads.
6126 (vect_get_gather_scatter_ops): New function.
6127 (vectorizable_call): Check internal_load_fn_p.
6128 (vectorizable_load): Likewise. Handle gather load internal
6130 (vectorizable_store): Update call to check_load_store_masking.
6131 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6132 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6133 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6134 (aarch64_gather_scale_operand_d): New predicates.
6135 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6136 (mask_gather_load<mode>): New insns.
6138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6139 Alan Hayward <alan.hayward@arm.com>
6140 David Sherwood <david.sherwood@arm.com>
6142 * optabs.def (fold_left_plus_optab): New optab.
6143 * doc/md.texi (fold_left_plus_@var{m}): Document.
6144 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6145 * internal-fn.c (fold_left_direct): Define.
6146 (expand_fold_left_optab_fn): Likewise.
6147 (direct_fold_left_optab_supported_p): Likewise.
6148 * fold-const-call.c (fold_const_fold_left): New function.
6149 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6150 * tree-parloops.c (valid_reduction_p): New function.
6151 (gather_scalar_reductions): Use it.
6152 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6153 (vect_finish_replace_stmt): Declare.
6154 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6155 (needs_fold_left_reduction_p): New function, split out from...
6156 (vect_is_simple_reduction): ...here. Accept reductions that
6157 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6158 (vect_force_simple_reduction): Also store the reduction type in
6159 the assignment's STMT_VINFO_REDUC_TYPE.
6160 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6161 (merge_with_identity): New function.
6162 (vect_expand_fold_left): Likewise.
6163 (vectorize_fold_left_reduction): Likewise.
6164 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6165 scalar phi in place for it. Check for target support and reject
6166 cases that would reassociate the operation. Defer the transform
6167 phase to vectorize_fold_left_reduction.
6168 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6169 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6170 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6172 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6174 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6175 call to ifc_temp_var.
6177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6178 Alan Hayward <alan.hayward@arm.com>
6179 David Sherwood <david.sherwood@arm.com>
6181 * target.def (legitimize_address_displacement): Take the original
6182 offset as a poly_int.
6183 * targhooks.h (default_legitimize_address_displacement): Update
6185 * targhooks.c (default_legitimize_address_displacement): Likewise.
6186 * doc/tm.texi: Regenerate.
6187 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6188 as an argument, moving assert of ad->disp == ad->disp_term to...
6189 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6190 Try calling targetm.legitimize_address_displacement before expanding
6191 the address rather than afterwards, and adjust for the new interface.
6192 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6193 Match the new hook interface. Handle SVE addresses.
6194 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6199 * Makefile.in (OBJS): Add early-remat.o.
6200 * target.def (select_early_remat_modes): New hook.
6201 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6202 * doc/tm.texi: Regenerate.
6203 * targhooks.h (default_select_early_remat_modes): Declare.
6204 * targhooks.c (default_select_early_remat_modes): New function.
6205 * timevar.def (TV_EARLY_REMAT): New timevar.
6206 * passes.def (pass_early_remat): New pass.
6207 * tree-pass.h (make_pass_early_remat): Declare.
6208 * early-remat.c: New file.
6209 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6211 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6214 Alan Hayward <alan.hayward@arm.com>
6215 David Sherwood <david.sherwood@arm.com>
6217 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6218 vfm1 with a bound_epilog parameter.
6219 (vect_do_peeling): Update calls accordingly, and move the prologue
6220 call earlier in the function. Treat the base bound_epilog as 0 for
6221 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6222 this base when peeling for gaps.
6223 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6224 with fully-masked loops.
6225 (vect_estimate_min_profitable_iters): Handle the single peeled
6226 iteration in that case.
6228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6229 Alan Hayward <alan.hayward@arm.com>
6230 David Sherwood <david.sherwood@arm.com>
6232 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6233 single-element interleaving even if the size is not a power of 2.
6234 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6235 accesses for single-element interleaving if the group size is
6238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6239 Alan Hayward <alan.hayward@arm.com>
6240 David Sherwood <david.sherwood@arm.com>
6242 * doc/md.texi (fold_extract_last_@var{m}): Document.
6243 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6244 * optabs.def (fold_extract_last_optab): New optab.
6245 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6246 * internal-fn.c (fold_extract_direct): New macro.
6247 (expand_fold_extract_optab_fn): Likewise.
6248 (direct_fold_extract_optab_supported_p): Likewise.
6249 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6250 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6251 EXTRACT_LAST_REDUCTION.
6252 (get_initial_def_for_reduction): Do not create an initial vector
6253 for EXTRACT_LAST_REDUCTION reductions.
6254 (vectorizable_reduction): Leave the scalar phi in place for
6255 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6256 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6257 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6258 transform phase to vectorizable_condition.
6259 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6261 (vect_finish_stmt_generation): ...here.
6262 (vect_finish_replace_stmt): New function.
6263 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6264 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6266 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6269 Alan Hayward <alan.hayward@arm.com>
6270 David Sherwood <david.sherwood@arm.com>
6272 * doc/md.texi (extract_last_@var{m}): Document.
6273 * optabs.def (extract_last_optab): New optab.
6274 * internal-fn.def (EXTRACT_LAST): New internal function.
6275 * internal-fn.c (cond_unary_direct): New macro.
6276 (expand_cond_unary_optab_fn): Likewise.
6277 (direct_cond_unary_optab_supported_p): Likewise.
6278 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6279 loops using EXTRACT_LAST.
6280 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6281 (extract_last_<mode>): ...this optab.
6282 (vec_extract<mode><Vel>): Update accordingly.
6284 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6285 Alan Hayward <alan.hayward@arm.com>
6286 David Sherwood <david.sherwood@arm.com>
6288 * target.def (empty_mask_is_expensive): New hook.
6289 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6290 * doc/tm.texi: Regenerate.
6291 * targhooks.h (default_empty_mask_is_expensive): Declare.
6292 * targhooks.c (default_empty_mask_is_expensive): New function.
6293 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6294 if the target says that empty masks are expensive.
6295 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6297 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6300 Alan Hayward <alan.hayward@arm.com>
6301 David Sherwood <david.sherwood@arm.com>
6303 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6304 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6305 (vect_use_loop_mask_for_alignment_p): New function.
6306 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6307 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6308 niters_skip argument. Make sure that the first niters_skip elements
6309 of the first iteration are inactive.
6310 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6311 Update call to vect_set_loop_masks_directly.
6312 (get_misalign_in_elems): New function, split out from...
6313 (vect_gen_prolog_loop_niters): ...here.
6314 (vect_update_init_of_dr): Take a code argument that specifies whether
6315 the adjustment should be added or subtracted.
6316 (vect_update_init_of_drs): Likewise.
6317 (vect_prepare_for_masked_peels): New function.
6318 (vect_do_peeling): Skip prologue peeling if we're using a mask
6319 instead. Update call to vect_update_inits_of_drs.
6320 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6322 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6323 alignment. Do not include the number of peeled iterations in
6324 the minimum threshold in that case.
6325 (vectorizable_induction): Adjust the start value down by
6326 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6327 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6328 Take the number of skipped iterations into account when calculating
6330 * tree-vect-stmts.c (vect_gen_while_not): New function.
6332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6333 Alan Hayward <alan.hayward@arm.com>
6334 David Sherwood <david.sherwood@arm.com>
6336 * doc/sourcebuild.texi (vect_fully_masked): Document.
6337 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6339 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6341 (vect_analyze_loop_2): ...here. Don't check the vectorization
6342 factor against the number of loop iterations if the loop is
6345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6346 Alan Hayward <alan.hayward@arm.com>
6347 David Sherwood <david.sherwood@arm.com>
6349 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6350 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6351 (dump_groups): Update accordingly.
6352 (iv_use::mem_type): New member variable.
6353 (address_p): New function.
6354 (record_use): Add a mem_type argument and initialize the new
6356 (record_group_use): Add a mem_type argument. Use address_p.
6357 Remove obsolete null checks of base_object. Update call to record_use.
6358 (find_interesting_uses_op): Update call to record_group_use.
6359 (find_interesting_uses_cond): Likewise.
6360 (find_interesting_uses_address): Likewise.
6361 (get_mem_type_for_internal_fn): New function.
6362 (find_address_like_use): Likewise.
6363 (find_interesting_uses_stmt): Try find_address_like_use before
6364 calling find_interesting_uses_op.
6365 (addr_offset_valid_p): Use the iv mem_type field as the type
6366 of the addressed memory.
6367 (add_autoinc_candidates): Likewise.
6368 (get_address_cost): Likewise.
6369 (split_small_address_groups_p): Use address_p.
6370 (split_address_groups): Likewise.
6371 (add_iv_candidate_for_use): Likewise.
6372 (autoinc_possible_for_pair): Likewise.
6373 (rewrite_groups): Likewise.
6374 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6375 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6376 (get_alias_ptr_type_for_ptr_address): New function.
6377 (rewrite_use_address): Rewrite address uses in calls that were
6378 identified by find_address_like_use.
6380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6381 Alan Hayward <alan.hayward@arm.com>
6382 David Sherwood <david.sherwood@arm.com>
6384 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6386 * gimple-expr.h (is_gimple_addressable: Likewise.
6387 * gimple-expr.c (is_gimple_address): Likewise.
6388 * internal-fn.c (expand_call_mem_ref): New function.
6389 (expand_mask_load_optab_fn): Use it.
6390 (expand_mask_store_optab_fn): Likewise.
6392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6393 Alan Hayward <alan.hayward@arm.com>
6394 David Sherwood <david.sherwood@arm.com>
6396 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6397 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6398 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6399 (cond_umax@var{mode}): Document.
6400 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6401 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6402 (cond_umin_optab, cond_umax_optab): New optabs.
6403 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6404 (COND_IOR, COND_XOR): New internal functions.
6405 * internal-fn.h (get_conditional_internal_fn): Declare.
6406 * internal-fn.c (cond_binary_direct): New macro.
6407 (expand_cond_binary_optab_fn): Likewise.
6408 (direct_cond_binary_optab_supported_p): Likewise.
6409 (get_conditional_internal_fn): New function.
6410 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6411 Cope with reduction statements that are vectorized as calls rather
6413 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6414 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6415 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6416 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6417 (UNSPEC_COND_EOR): New unspecs.
6418 (optab): Add mappings for them.
6419 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6420 (sve_int_op, sve_fp_op): New int attributes.
6422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6423 Alan Hayward <alan.hayward@arm.com>
6424 David Sherwood <david.sherwood@arm.com>
6426 * optabs.def (while_ult_optab): New optab.
6427 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6428 * internal-fn.def (WHILE_ULT): New internal function.
6429 * internal-fn.h (direct_internal_fn_supported_p): New override
6430 that takes two types as argument.
6431 * internal-fn.c (while_direct): New macro.
6432 (expand_while_optab_fn): New function.
6433 (convert_optab_supported_p): Likewise.
6434 (direct_while_optab_supported_p): New macro.
6435 * wide-int.h (wi::udiv_ceil): New function.
6436 * tree-vectorizer.h (rgroup_masks): New structure.
6437 (vec_loop_masks): New typedef.
6438 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6440 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6441 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6442 (vect_max_vf): New function.
6443 (slpeel_make_loop_iterate_ntimes): Delete.
6444 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6445 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6446 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6447 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6448 internal-fn.h, stor-layout.h and optabs-query.h.
6449 (vect_set_loop_mask): New function.
6450 (add_preheader_seq): Likewise.
6451 (add_header_seq): Likewise.
6452 (interleave_supported_p): Likewise.
6453 (vect_maybe_permute_loop_masks): Likewise.
6454 (vect_set_loop_masks_directly): Likewise.
6455 (vect_set_loop_condition_masked): Likewise.
6456 (vect_set_loop_condition_unmasked): New function, split out from
6457 slpeel_make_loop_iterate_ntimes.
6458 (slpeel_make_loop_iterate_ntimes): Rename to..
6459 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6460 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6461 (vect_do_peeling): Update call accordingly.
6462 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6464 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6465 mask_compare_type, can_fully_mask_p and fully_masked_p.
6466 (release_vec_loop_masks): New function.
6467 (_loop_vec_info): Use it to free the loop masks.
6468 (can_produce_all_loop_masks_p): New function.
6469 (vect_get_max_nscalars_per_iter): Likewise.
6470 (vect_verify_full_masking): Likewise.
6471 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6472 retries, and free the mask rgroups before retrying. Check loop-wide
6473 reasons for disallowing fully-masked loops. Make the final decision
6474 about whether use a fully-masked loop or not.
6475 (vect_estimate_min_profitable_iters): Do not assume that peeling
6476 for the number of iterations will be needed for fully-masked loops.
6477 (vectorizable_reduction): Disable fully-masked loops.
6478 (vectorizable_live_operation): Likewise.
6479 (vect_halve_mask_nunits): New function.
6480 (vect_double_mask_nunits): Likewise.
6481 (vect_record_loop_mask): Likewise.
6482 (vect_get_loop_mask): Likewise.
6483 (vect_transform_loop): Handle the case in which the final loop
6484 iteration might handle a partial vector. Call vect_set_loop_condition
6485 instead of slpeel_make_loop_iterate_ntimes.
6486 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6487 (check_load_store_masking): New function.
6488 (prepare_load_store_mask): Likewise.
6489 (vectorizable_store): Handle fully-masked loops.
6490 (vectorizable_load): Likewise.
6491 (supportable_widening_operation): Use vect_halve_mask_nunits for
6493 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6494 (vect_gen_while): New function.
6495 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6496 (aarch64_uqdec<mode>): New insn.
6498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6499 Alan Hayward <alan.hayward@arm.com>
6500 David Sherwood <david.sherwood@arm.com>
6502 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6503 (reduc_xor_scal_optab): New optabs.
6504 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6505 (reduc_xor_scal_@var{m}): Document.
6506 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6507 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6509 * fold-const-call.c (fold_const_call): Handle them.
6510 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6511 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6512 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6513 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6514 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6515 (UNSPEC_XORV): New unspecs.
6516 (optab): Add entries for them.
6517 (BITWISEV): New int iterator.
6518 (bit_reduc_op): New int attributes.
6520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6521 Alan Hayward <alan.hayward@arm.com>
6522 David Sherwood <david.sherwood@arm.com>
6524 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6525 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6526 * optabs.def (vec_shl_insert_optab): New optab.
6527 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6528 (duplicate_and_interleave): Likewise.
6529 * tree-vect-loop.c: Include internal-fn.h.
6530 (neutral_op_for_slp_reduction): New function, split out from
6531 get_initial_defs_for_reduction.
6532 (get_initial_def_for_reduction): Handle option 2 for variable-length
6533 vectors by loading the neutral value into a vector and then shifting
6534 the initial value into element 0.
6535 (get_initial_defs_for_reduction): Replace the code argument with
6536 the neutral value calculated by neutral_op_for_slp_reduction.
6537 Use gimple_build_vector for constant-length vectors.
6538 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6539 but the first group_size elements have a neutral value.
6540 Use duplicate_and_interleave otherwise.
6541 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6542 Update call to get_initial_defs_for_reduction. Handle SLP
6543 reductions for variable-length vectors by creating one vector
6544 result for each scalar result, with the elements associated
6545 with other scalar results stubbed out with the neutral value.
6546 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6547 Require IFN_VEC_SHL_INSERT for double reductions on
6548 variable-length vectors, or SLP reductions that have
6549 a neutral value. Require can_duplicate_and_interleave_p
6550 support for variable-length unchained SLP reductions if there
6551 is no neutral value, such as for MIN/MAX reductions. Also require
6552 the number of vector elements to be a multiple of the number of
6553 SLP statements when doing variable-length unchained SLP reductions.
6554 Update call to vect_create_epilog_for_reduction.
6555 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6556 and remove initial values.
6557 (duplicate_and_interleave): Make public.
6558 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6559 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6561 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6562 Alan Hayward <alan.hayward@arm.com>
6563 David Sherwood <david.sherwood@arm.com>
6565 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6566 (can_duplicate_and_interleave_p): New function.
6567 (vect_get_and_check_slp_defs): Take the vector of statements
6568 rather than just the current one. Remove excess parentheses.
6569 Restriction rejectinon of vect_constant_def and vect_external_def
6570 for variable-length vectors to boolean types, or types for which
6571 can_duplicate_and_interleave_p is false.
6572 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6573 (duplicate_and_interleave): New function.
6574 (vect_get_constant_vectors): Use gimple_build_vector for
6575 constant-length vectors and suitable variable-length constant
6576 vectors. Use duplicate_and_interleave for other variable-length
6577 vectors. Don't defer the update when inserting new statements.
6579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6580 Alan Hayward <alan.hayward@arm.com>
6581 David Sherwood <david.sherwood@arm.com>
6583 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6584 min_profitable_iters doesn't go negative.
6586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6587 Alan Hayward <alan.hayward@arm.com>
6588 David Sherwood <david.sherwood@arm.com>
6590 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6591 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6592 * optabs.def (vec_mask_load_lanes_optab): New optab.
6593 (vec_mask_store_lanes_optab): Likewise.
6594 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6595 (MASK_STORE_LANES): Likewise.
6596 * internal-fn.c (mask_load_lanes_direct): New macro.
6597 (mask_store_lanes_direct): Likewise.
6598 (expand_mask_load_optab_fn): Handle masked operations.
6599 (expand_mask_load_lanes_optab_fn): New macro.
6600 (expand_mask_store_optab_fn): Handle masked operations.
6601 (expand_mask_store_lanes_optab_fn): New macro.
6602 (direct_mask_load_lanes_optab_supported_p): Likewise.
6603 (direct_mask_store_lanes_optab_supported_p): Likewise.
6604 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6606 (vect_load_lanes_supported): Likewise.
6607 * tree-vect-data-refs.c (strip_conversion): New function.
6608 (can_group_stmts_p): Likewise.
6609 (vect_analyze_data_ref_accesses): Use it instead of checking
6610 for a pair of assignments.
6611 (vect_store_lanes_supported): Take a masked_p parameter.
6612 (vect_load_lanes_supported): Likewise.
6613 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6614 vect_store_lanes_supported and vect_load_lanes_supported.
6615 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6616 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6617 parameter. Don't allow gaps for masked accesses.
6618 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6619 and vect_load_lanes_supported.
6620 (get_load_store_type): Take a masked_p parameter and update
6621 call to get_group_load_store_type.
6622 (vectorizable_store): Update call to get_load_store_type.
6623 Handle IFN_MASK_STORE_LANES.
6624 (vectorizable_load): Update call to get_load_store_type.
6625 Handle IFN_MASK_LOAD_LANES.
6627 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6628 Alan Hayward <alan.hayward@arm.com>
6629 David Sherwood <david.sherwood@arm.com>
6631 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6633 * config/aarch64/aarch64-protos.h
6634 (aarch64_sve_struct_memory_operand_p): Declare.
6635 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6636 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6637 (VPRED, vpred): Handle SVE structure modes.
6638 * config/aarch64/constraints.md (Utx): New constraint.
6639 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6640 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6641 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6642 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6643 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6644 structure modes. Split into pieces after RA.
6645 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6646 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6648 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6649 SVE structure modes.
6650 (aarch64_classify_address): Likewise.
6651 (sizetochar): Move earlier in file.
6652 (aarch64_print_operand): Handle SVE register lists.
6653 (aarch64_array_mode): New function.
6654 (aarch64_sve_struct_memory_operand_p): Likewise.
6655 (TARGET_ARRAY_MODE): Redefine.
6657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6658 Alan Hayward <alan.hayward@arm.com>
6659 David Sherwood <david.sherwood@arm.com>
6661 * target.def (array_mode): New target hook.
6662 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6663 * doc/tm.texi: Regenerate.
6664 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6665 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6666 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6668 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6672 Alan Hayward <alan.hayward@arm.com>
6673 David Sherwood <david.sherwood@arm.com>
6675 * fold-const.c (fold_binary_loc): Check the argument types
6676 rather than the result type when testing for a vector operation.
6678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6680 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6681 * doc/tm.texi: Regenerate.
6683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6684 Alan Hayward <alan.hayward@arm.com>
6685 David Sherwood <david.sherwood@arm.com>
6687 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6688 (sve): Document new AArch64 extension.
6689 * doc/md.texi (w): Extend the description of the AArch64
6690 constraint to include SVE vectors.
6691 (Upl, Upa): Document new AArch64 predicate constraints.
6692 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6694 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6695 (msve-vector-bits=): New option.
6696 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6697 SVE when these are disabled.
6698 (sve): New extension.
6699 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6700 modes. Adjust their number of units based on aarch64_sve_vg.
6701 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6702 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6703 aarch64_addr_query_type.
6704 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6705 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6706 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6707 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6708 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6709 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6710 (aarch64_simd_imm_zero_p): Delete.
6711 (aarch64_check_zero_based_sve_index_immediate): Declare.
6712 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6713 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6714 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6715 (aarch64_sve_float_mul_immediate_p): Likewise.
6716 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6718 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6719 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6720 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6721 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6722 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6723 (aarch64_regmode_natural_size): Likewise.
6724 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6725 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6727 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6728 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6729 for VG and the SVE predicate registers.
6730 (V_ALIASES): Add a "z"-prefixed alias.
6731 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6732 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6733 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6734 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6735 (REG_CLASS_NAMES): Add entries for them.
6736 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6737 and the predicate registers.
6738 (aarch64_sve_vg): Declare.
6739 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6740 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6741 (REGMODE_NATURAL_SIZE): Define.
6742 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6744 * config/aarch64/aarch64.c: Include cfgrtl.h.
6745 (simd_immediate_info): Add a constructor for series vectors,
6746 and an associated step field.
6747 (aarch64_sve_vg): New variable.
6748 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6749 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6750 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6751 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6752 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6753 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6754 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6755 (aarch64_get_mask_mode): New functions.
6756 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6757 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6758 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6759 predicate modes and predicate registers. Explicitly restrict
6760 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6761 to store a vector mode if it is recognized by
6762 aarch64_classify_vector_mode.
6763 (aarch64_regmode_natural_size): New function.
6764 (aarch64_hard_regno_caller_save_mode): Return the original mode
6766 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6767 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6768 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6769 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6771 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6772 does not overlap dest if the function is frame-related. Handle
6774 (aarch64_split_add_offset): New function.
6775 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6776 them aarch64_add_offset.
6777 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6778 and update call to aarch64_sub_sp.
6779 (aarch64_add_cfa_expression): New function.
6780 (aarch64_expand_prologue): Pass extra temporary registers to the
6781 functions above. Handle the case in which we need to emit new
6782 DW_CFA_expressions for registers that were originally saved
6783 relative to the stack pointer, but now have to be expressed
6784 relative to the frame pointer.
6785 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6787 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6788 IP0 and IP1 values for SVE frames.
6789 (aarch64_expand_vec_series): New function.
6790 (aarch64_expand_sve_widened_duplicate): Likewise.
6791 (aarch64_expand_sve_const_vector): Likewise.
6792 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6793 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6794 into the register, rather than emitting a SET directly.
6795 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6796 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6797 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6798 (offset_9bit_signed_scaled_p): New functions.
6799 (aarch64_replicate_bitmask_imm): New function.
6800 (aarch64_bitmask_imm): Use it.
6801 (aarch64_cannot_force_const_mem): Reject expressions involving
6802 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6803 (aarch64_classify_index): Handle SVE indices, by requiring
6804 a plain register index with a scale that matches the element size.
6805 (aarch64_classify_address): Handle SVE addresses. Assert that
6806 the mode of the address is VOIDmode or an integer mode.
6807 Update call to aarch64_classify_symbol.
6808 (aarch64_classify_symbolic_expression): Update call to
6809 aarch64_classify_symbol.
6810 (aarch64_const_vec_all_in_range_p): New function.
6811 (aarch64_print_vector_float_operand): Likewise.
6812 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6813 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6814 and the FP immediates 1.0 and 0.5.
6815 (aarch64_print_address_internal): Handle SVE addresses.
6816 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6817 (aarch64_regno_regclass): Handle predicate registers.
6818 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6820 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6821 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6822 (aarch64_convert_sve_vector_bits): New function.
6823 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6824 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6826 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6827 Handle SVE vector and predicate modes. Accept VL-based constants
6828 that need only one temporary register, and VL offsets that require
6829 no temporary registers.
6830 (aarch64_conditional_register_usage): Mark the predicate registers
6831 as fixed if SVE isn't available.
6832 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6833 Return true for SVE vector and predicate modes.
6834 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6835 rather than an unsigned int. Handle SVE modes.
6836 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6838 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6840 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6841 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6842 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6843 (aarch64_sve_float_mul_immediate_p): New functions.
6844 (aarch64_sve_valid_immediate): New function.
6845 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6846 Explicitly reject structure modes. Check for INDEX constants.
6847 Handle PTRUE and PFALSE constants.
6848 (aarch64_check_zero_based_sve_index_immediate): New function.
6849 (aarch64_simd_imm_zero_p): Delete.
6850 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6851 vector modes. Accept constants in the range of CNT[BHWD].
6852 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6853 ask for an Advanced SIMD mode.
6854 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6855 (aarch64_simd_vector_alignment): Handle SVE predicates.
6856 (aarch64_vectorize_preferred_vector_alignment): New function.
6857 (aarch64_simd_vector_alignment_reachable): Use it instead of
6859 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6860 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6862 (MAX_VECT_LEN): Delete.
6863 (expand_vec_perm_d): Add a vec_flags field.
6864 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6865 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6866 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6868 (aarch64_evpc_rev): Rename to...
6869 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6870 (aarch64_evpc_rev_global): New function.
6871 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6872 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6874 (aarch64_evpc_sve_tbl): New function.
6875 (aarch64_expand_vec_perm_const_1): Update after rename of
6876 aarch64_evpc_rev. Handle SVE permutes too, trying
6877 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6878 than aarch64_evpc_tbl.
6879 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6880 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6881 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6882 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6883 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6884 (aarch64_expand_sve_vcond): New functions.
6885 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6886 of aarch64_vector_mode_p.
6887 (aarch64_dwarf_poly_indeterminate_value): New function.
6888 (aarch64_compute_pressure_classes): Likewise.
6889 (aarch64_can_change_mode_class): Likewise.
6890 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6891 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6892 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6893 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6894 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6895 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6896 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6897 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6899 (Dn, Dl, Dr): Accept const as well as const_vector.
6900 (Dz): Likewise. Compare against CONST0_RTX.
6901 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6902 of "vector" where appropriate.
6903 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6904 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6905 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6906 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6907 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6908 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6909 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6910 (v_int_equiv): Extend to SVE modes.
6911 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6913 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6914 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6915 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6916 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6917 (SVE_COND_FP_CMP): New int iterators.
6918 (perm_hilo): Handle the new unpack unspecs.
6919 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6921 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6922 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6923 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6924 (aarch64_equality_operator, aarch64_constant_vector_operand)
6925 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6926 (aarch64_sve_nonimmediate_operand): Likewise.
6927 (aarch64_sve_general_operand): Likewise.
6928 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6929 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6930 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6931 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6932 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6933 (aarch64_sve_float_arith_immediate): Likewise.
6934 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6935 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6936 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6937 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6938 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6939 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6940 (aarch64_sve_float_arith_operand): Likewise.
6941 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6942 (aarch64_sve_float_mul_operand): Likewise.
6943 (aarch64_sve_vec_perm_operand): Likewise.
6944 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6945 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6946 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6947 as well as const_vector.
6948 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6949 in file. Use CONST0_RTX and CONSTM1_RTX.
6950 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6951 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6952 Use aarch64_simd_imm_zero.
6953 * config/aarch64/aarch64-sve.md: New file.
6954 * config/aarch64/aarch64.md: Include it.
6955 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6956 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6957 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6958 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6959 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6960 (sve): New attribute.
6961 (enabled): Disable instructions with the sve attribute unless
6963 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6964 aarch64_expand_mov_immediate.
6965 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6966 CNT[BHSD] immediates.
6967 (movti): Split CONST_POLY_INT moves into two halves.
6968 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6969 Split additions that need a temporary here if the destination
6970 is the stack pointer.
6971 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6972 (*add<mode>3_poly_1): New instruction.
6973 (set_clobber_cc): New expander.
6975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6977 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6978 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6979 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6980 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6981 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6982 Change innermode from fixed_mode_size to machine_mode.
6983 (simplify_subreg): Update call accordingly. Handle a constant-sized
6984 subreg of a variable-length CONST_VECTOR.
6986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6987 Alan Hayward <alan.hayward@arm.com>
6988 David Sherwood <david.sherwood@arm.com>
6990 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6991 (add_offset_to_base): New function, split out from...
6992 (create_mem_ref): ...here. When handling a scale other than 1,
6993 check first whether the address is valid without the offset.
6994 Add it into the base if so, leaving the index and scale as-is.
6996 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6999 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7000 fold_for_warn before checking if arg2 is INTEGER_CST.
7002 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7004 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7005 (store_multiple_operation): Delete.
7006 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7007 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7008 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7009 guarded by TARGET_STRING.
7010 (rs6000_output_load_multiple): Delete.
7011 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7012 OPTION_MASK_STRING / TARGET_STRING handling.
7013 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7014 (const rs6000_opt_masks) <"string">: Change mask to 0.
7015 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7016 (MASK_STRING): Delete.
7017 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7019 (load_multiple): Delete.
7026 (store_multiple): Delete.
7033 (movmemsi_8reg): Delete.
7034 (corresponding unnamed define_insn): Delete.
7035 (movmemsi_6reg): Delete.
7036 (corresponding unnamed define_insn): Delete.
7037 (movmemsi_4reg): Delete.
7038 (corresponding unnamed define_insn): Delete.
7039 (movmemsi_2reg): Delete.
7040 (corresponding unnamed define_insn): Delete.
7041 (movmemsi_1reg): Delete.
7042 (corresponding unnamed define_insn): Delete.
7043 * config/rs6000/rs6000.opt (mno-string): New.
7044 (mstring): Replace by deprecation warning stub.
7045 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7047 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7049 * regrename.c (regrename_do_replace): If replacing the same
7050 reg multiple times, try to reuse last created gen_raw_REG.
7053 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7054 main to workaround a bug in GDB.
7056 2018-01-12 Tom de Vries <tom@codesourcery.com>
7059 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7061 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7063 PR rtl-optimization/80481
7064 * ira-color.c (get_cap_member): New function.
7065 (allocnos_conflict_by_live_ranges_p): Use it.
7066 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7067 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7069 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7072 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7073 (*saddl_se_1): Ditto.
7075 (*ssubl_se_1): Ditto.
7077 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7079 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7080 rather than wi::to_widest for DR_INITs.
7081 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7082 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7083 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7085 (vect_analyze_group_access_1): Note that here.
7087 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7089 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7090 polynomial type sizes.
7092 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7094 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7095 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7096 (gimple_add_tmp_var): Likewise.
7098 2018-01-12 Martin Liska <mliska@suse.cz>
7100 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7101 (gimple_alloc_sizes): Likewise.
7102 (dump_gimple_statistics): Use PRIu64 in printf format.
7103 * gimple.h: Change uint64_t to int.
7105 2018-01-12 Martin Liska <mliska@suse.cz>
7107 * tree-core.h: Use uint64_t instead of int.
7108 * tree.c (tree_node_counts): Likewise.
7109 (tree_node_sizes): Likewise.
7110 (dump_tree_statistics): Use PRIu64 in printf format.
7112 2018-01-12 Martin Liska <mliska@suse.cz>
7114 * Makefile.in: As qsort_chk is implemented in vec.c, add
7115 vec.o to linkage of gencfn-macros.
7116 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7117 passing the info to record_node_allocation_statistics.
7118 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7120 * ggc-common.c (struct ggc_usage): Add operator== and use
7121 it in operator< and compare function.
7122 * mem-stats.h (struct mem_usage): Likewise.
7123 * vec.c (struct vec_usage): Remove operator< and compare
7124 function. Can be simply inherited.
7126 2018-01-12 Martin Jambor <mjambor@suse.cz>
7129 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7130 * tree-ssa-math-opts.c: Include domwalk.h.
7131 (convert_mult_to_fma_1): New function.
7132 (fma_transformation_info): New type.
7133 (fma_deferring_state): Likewise.
7134 (cancel_fma_deferring): New function.
7135 (result_of_phi): Likewise.
7136 (last_fma_candidate_feeds_initial_phi): Likewise.
7137 (convert_mult_to_fma): Added deferring logic, split actual
7138 transformation to convert_mult_to_fma_1.
7139 (math_opts_dom_walker): New type.
7140 (math_opts_dom_walker::after_dom_children): New method, body moved
7141 here from pass_optimize_widening_mul::execute, added deferring logic
7143 (pass_optimize_widening_mul::execute): Moved most of code to
7144 math_opts_dom_walker::after_dom_children.
7145 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7146 * config/i386/i386.c (ix86_option_override_internal): Added
7147 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7149 2018-01-12 Richard Biener <rguenther@suse.de>
7152 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7153 inline instance vars.
7155 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7158 * config/rx/rx.c (rx_is_restricted_memory_address):
7161 2018-01-12 Richard Biener <rguenther@suse.de>
7163 PR tree-optimization/80846
7164 * target.def (split_reduction): New target hook.
7165 * targhooks.c (default_split_reduction): New function.
7166 * targhooks.h (default_split_reduction): Declare.
7167 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7168 target requests first reduce vectors by combining low and high
7170 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7171 (get_vectype_for_scalar_type_and_size): Export.
7172 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7173 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7174 * doc/tm.texi: Regenerate.
7175 * config/i386/i386.c (ix86_split_reduction): Implement
7176 TARGET_VECTORIZE_SPLIT_REDUCTION.
7178 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7181 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7182 in PIC mode except for TARGET_VXWORKS_RTP.
7183 * config/sparc/sparc.c: Include cfgrtl.h.
7184 (TARGET_INIT_PIC_REG): Define.
7185 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7186 (sparc_pic_register_p): New predicate.
7187 (sparc_legitimate_address_p): Use it.
7188 (sparc_legitimize_pic_address): Likewise.
7189 (sparc_delegitimize_address): Likewise.
7190 (sparc_mode_dependent_address_p): Likewise.
7191 (gen_load_pcrel_sym): Remove 4th parameter.
7192 (load_got_register): Adjust call to above. Remove obsolete stuff.
7193 (sparc_expand_prologue): Do not call load_got_register here.
7194 (sparc_flat_expand_prologue): Likewise.
7195 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7196 (sparc_use_pseudo_pic_reg): New function.
7197 (sparc_init_pic_reg): Likewise.
7198 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7199 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7201 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7203 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7204 Add item for branch_cost.
7206 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7208 PR rtl-optimization/83565
7209 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7210 not extend the result to a larger mode for rotate operations.
7211 (num_sign_bit_copies1): Likewise.
7213 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7216 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7218 Use values-Xc.o for -pedantic.
7219 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7221 2018-01-12 Martin Liska <mliska@suse.cz>
7224 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7226 (possible_polymorphic_call_targets): Use it.
7227 (ipa_devirt): Likewise.
7229 2018-01-12 Martin Liska <mliska@suse.cz>
7231 * profile-count.h (enum profile_quality): Use 0 as invalid
7232 enum value of profile_quality.
7234 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7236 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7237 -mext-string options.
7239 2018-01-12 Richard Biener <rguenther@suse.de>
7241 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7242 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7243 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7245 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7247 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7249 * configure.ac (--with-long-double-format): Add support for the
7250 configuration option to change the default long double format on
7252 * config.gcc (powerpc*-linux*-*): Likewise.
7253 * configure: Regenerate.
7254 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7255 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7256 used without modification.
7258 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7260 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7261 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7262 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7263 MISC_BUILTIN_SPEC_BARRIER.
7264 (rs6000_init_builtins): Likewise.
7265 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7267 (speculation_barrier): New define_insn.
7268 * doc/extend.texi: Document __builtin_speculation_barrier.
7270 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7273 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7274 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7275 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7277 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7278 integral modes instead of "ss" and "sd".
7279 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7280 vectors with 32-bit and 64-bit elements.
7281 (vecdupssescalarmodesuffix): New mode attribute.
7282 (vec_dup<mode>): Use it.
7284 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7287 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7288 frame if argument is passed on stack.
7290 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7293 * ree.c (combine_reaching_defs): Optimize also
7294 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7295 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7297 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7300 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7302 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7305 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7306 after they are computed.
7308 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7310 PR tree-optimization/83695
7311 * gimple-loop-linterchange.cc
7312 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7313 reset cached scev information after interchange.
7314 (pass_linterchange::execute): Remove call to scev_reset_htab.
7316 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7318 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7319 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7320 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7321 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7322 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7323 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7324 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7325 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7326 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7327 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7328 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7329 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7330 (V_lane_reg): Likewise.
7331 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7333 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7334 (vfmal_lane_low<mode>_intrinsic,
7335 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7336 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7337 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7338 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7339 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7340 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7342 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7344 * config/arm/arm-cpus.in (fp16fml): New feature.
7345 (ALL_SIMD): Add fp16fml.
7346 (armv8.2-a): Add fp16fml as an option.
7347 (armv8.3-a): Likewise.
7348 (armv8.4-a): Add fp16fml as part of fp16.
7349 * config/arm/arm.h (TARGET_FP16FML): Define.
7350 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7352 * config/arm/arm-modes.def (V2HF): Define.
7353 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7354 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7355 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7356 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7357 vfmsl_low, vfmsl_high): New set of builtins.
7358 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7359 (vfml_op): New code attribute.
7360 (VFMLHALVES): New int iterator.
7361 (VFML, VFMLSEL): New mode attributes.
7362 (V_reg): Define mapping for V2HF.
7363 (V_hi, V_lo): New mode attributes.
7364 (VF_constraint): Likewise.
7365 (vfml_half, vfml_half_selector): New int attributes.
7366 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7368 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7369 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7371 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7372 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7373 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7374 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7376 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7377 Document new effective target and option set.
7379 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7381 * config/arm/arm-cpus.in (armv8_4): New feature.
7382 (ARMv8_4a): New fgroup.
7383 (armv8.4-a): New arch.
7384 * config/arm/arm-tables.opt: Regenerate.
7385 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7386 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7387 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7388 Add matching rules for -march=armv8.4-a and extensions.
7389 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7391 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7394 * config/rx/rx.md (BW): New mode attribute.
7395 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7397 2018-01-11 Richard Biener <rguenther@suse.de>
7399 PR tree-optimization/83435
7400 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7401 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7402 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7404 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7405 Alan Hayward <alan.hayward@arm.com>
7406 David Sherwood <david.sherwood@arm.com>
7408 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7410 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7411 (aarch64_print_address_internal): Use it to check for a zero offset.
7413 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7414 Alan Hayward <alan.hayward@arm.com>
7415 David Sherwood <david.sherwood@arm.com>
7417 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7418 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7419 Return a poly_int64 rather than a HOST_WIDE_INT.
7420 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7421 rather than a HOST_WIDE_INT.
7422 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7423 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7424 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7425 final_offset from HOST_WIDE_INT to poly_int64.
7426 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7427 to_constant when getting the number of units in an Advanced SIMD
7429 (aarch64_builtin_vectorized_function): Check for a constant number
7431 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7433 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7434 attribute instead of GET_MODE_NUNITS.
7435 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7436 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7437 GET_MODE_SIZE for fixed-size registers.
7438 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7439 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7440 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7441 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7442 (aarch64_print_operand, aarch64_print_address_internal)
7443 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7444 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7445 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7446 Handle polynomial GET_MODE_SIZE.
7447 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7448 wider than SImode without modification.
7449 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7450 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7451 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7452 passing and returning SVE modes.
7453 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7454 rather than GEN_INT.
7455 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7456 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7457 (aarch64_allocate_and_probe_stack_space): Likewise.
7458 (aarch64_layout_frame): Cope with polynomial offsets.
7459 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7460 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7462 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7463 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7464 poly_int64 rather than a HOST_WIDE_INT.
7465 (aarch64_get_separate_components, aarch64_process_components)
7466 (aarch64_expand_prologue, aarch64_expand_epilogue)
7467 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7468 (aarch64_anchor_offset): New function, split out from...
7469 (aarch64_legitimize_address): ...here.
7470 (aarch64_builtin_vectorization_cost): Handle polynomial
7471 TYPE_VECTOR_SUBPARTS.
7472 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7474 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7475 number of elements from the PARALLEL rather than the mode.
7476 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7477 rather than GET_MODE_BITSIZE.
7478 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7479 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7480 (aarch64_expand_vec_perm_const_1): Handle polynomial
7481 d->perm.length () and d->perm elements.
7482 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7483 Apply to_constant to d->perm elements.
7484 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7485 polynomial CONST_VECTOR_NUNITS.
7486 (aarch64_move_pointer): Take amount as a poly_int64 rather
7488 (aarch64_progress_pointer): Avoid temporary variable.
7489 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7490 the mode attribute instead of GET_MODE.
7492 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7493 Alan Hayward <alan.hayward@arm.com>
7494 David Sherwood <david.sherwood@arm.com>
7496 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7497 x exists before using it.
7498 (aarch64_add_constant_internal): Rename to...
7499 (aarch64_add_offset_1): ...this. Replace regnum with separate
7500 src and dest rtxes. Handle the case in which they're different,
7501 including when the offset is zero. Replace scratchreg with an rtx.
7502 Use 2 additions if there is no spare register into which we can
7503 move a 16-bit constant.
7504 (aarch64_add_constant): Delete.
7505 (aarch64_add_offset): Replace reg with separate src and dest
7506 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7507 Use aarch64_add_offset_1.
7508 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7509 an rtx rather than an int. Take the delta as a poly_int64
7510 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7511 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7512 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7513 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7514 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7516 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7517 aarch64_add_constant.
7519 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7521 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7522 Use scalar_float_mode.
7524 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7526 * config/aarch64/aarch64-simd.md
7527 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7528 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7529 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7530 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7531 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7532 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7533 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7534 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7535 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7536 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7538 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7541 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7542 targ_options->x_arm_arch_string is non NULL.
7544 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7546 * config/aarch64/aarch64.h
7547 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7549 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7552 * expmed.c (emit_store_flag_force): Swap if const op0
7553 and change VOIDmode to mode of op0.
7555 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7557 PR rtl-optimization/83761
7558 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7559 than bytes to mode_for_size.
7561 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7564 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7565 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7568 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7571 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7572 when in layout mode.
7573 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7574 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7577 2018-01-10 Michael Collison <michael.collison@arm.com>
7579 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7580 * config/aarch64/aarch64-option-extension.def: Add
7581 AARCH64_OPT_EXTENSION of 'fp16fml'.
7582 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7583 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7584 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7585 * config/aarch64/constraints.md (Ui7): New constraint.
7586 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7587 (VFMLA_SEL_W): Ditto.
7590 (VFMLA16_LOW): New int iterator.
7591 (VFMLA16_HIGH): Ditto.
7592 (UNSPEC_FMLAL): New unspec.
7593 (UNSPEC_FMLSL): Ditto.
7594 (UNSPEC_FMLAL2): Ditto.
7595 (UNSPEC_FMLSL2): Ditto.
7596 (f16mac): New code attribute.
7597 * config/aarch64/aarch64-simd-builtins.def
7598 (aarch64_fmlal_lowv2sf): Ditto.
7599 (aarch64_fmlsl_lowv2sf): Ditto.
7600 (aarch64_fmlalq_lowv4sf): Ditto.
7601 (aarch64_fmlslq_lowv4sf): Ditto.
7602 (aarch64_fmlal_highv2sf): Ditto.
7603 (aarch64_fmlsl_highv2sf): Ditto.
7604 (aarch64_fmlalq_highv4sf): Ditto.
7605 (aarch64_fmlslq_highv4sf): Ditto.
7606 (aarch64_fmlal_lane_lowv2sf): Ditto.
7607 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7608 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7609 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7610 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7611 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7612 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7613 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7614 (aarch64_fmlal_lane_highv2sf): Ditto.
7615 (aarch64_fmlsl_lane_highv2sf): Ditto.
7616 (aarch64_fmlal_laneq_highv2sf): Ditto.
7617 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7618 (aarch64_fmlalq_lane_highv4sf): Ditto.
7619 (aarch64_fmlsl_lane_highv4sf): Ditto.
7620 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7621 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7622 * config/aarch64/aarch64-simd.md:
7623 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7624 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7625 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7626 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7627 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7628 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7629 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7630 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7631 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7632 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7633 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7634 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7635 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7636 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7637 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7638 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7639 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7640 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7641 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7642 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7643 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7644 (vfmlsl_low_u32): Ditto.
7645 (vfmlalq_low_u32): Ditto.
7646 (vfmlslq_low_u32): Ditto.
7647 (vfmlal_high_u32): Ditto.
7648 (vfmlsl_high_u32): Ditto.
7649 (vfmlalq_high_u32): Ditto.
7650 (vfmlslq_high_u32): Ditto.
7651 (vfmlal_lane_low_u32): Ditto.
7652 (vfmlsl_lane_low_u32): Ditto.
7653 (vfmlal_laneq_low_u32): Ditto.
7654 (vfmlsl_laneq_low_u32): Ditto.
7655 (vfmlalq_lane_low_u32): Ditto.
7656 (vfmlslq_lane_low_u32): Ditto.
7657 (vfmlalq_laneq_low_u32): Ditto.
7658 (vfmlslq_laneq_low_u32): Ditto.
7659 (vfmlal_lane_high_u32): Ditto.
7660 (vfmlsl_lane_high_u32): Ditto.
7661 (vfmlal_laneq_high_u32): Ditto.
7662 (vfmlsl_laneq_high_u32): Ditto.
7663 (vfmlalq_lane_high_u32): Ditto.
7664 (vfmlslq_lane_high_u32): Ditto.
7665 (vfmlalq_laneq_high_u32): Ditto.
7666 (vfmlslq_laneq_high_u32): Ditto.
7667 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7668 (AARCH64_FL_FOR_ARCH8_4): New.
7669 (AARCH64_ISA_F16FML): New ISA flag.
7670 (TARGET_F16FML): New feature flag for fp16fml.
7671 (doc/invoke.texi): Document new fp16fml option.
7673 2018-01-10 Michael Collison <michael.collison@arm.com>
7675 * config/aarch64/aarch64-builtins.c:
7676 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7677 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7678 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7679 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7680 (AARCH64_ISA_SHA3): New ISA flag.
7681 (TARGET_SHA3): New feature flag for sha3.
7682 * config/aarch64/iterators.md (sha512_op): New int attribute.
7683 (CRYPTO_SHA512): New int iterator.
7684 (UNSPEC_SHA512H): New unspec.
7685 (UNSPEC_SHA512H2): Ditto.
7686 (UNSPEC_SHA512SU0): Ditto.
7687 (UNSPEC_SHA512SU1): Ditto.
7688 * config/aarch64/aarch64-simd-builtins.def
7689 (aarch64_crypto_sha512hqv2di): New builtin.
7690 (aarch64_crypto_sha512h2qv2di): Ditto.
7691 (aarch64_crypto_sha512su0qv2di): Ditto.
7692 (aarch64_crypto_sha512su1qv2di): Ditto.
7693 (aarch64_eor3qv8hi): Ditto.
7694 (aarch64_rax1qv2di): Ditto.
7695 (aarch64_xarqv2di): Ditto.
7696 (aarch64_bcaxqv8hi): Ditto.
7697 * config/aarch64/aarch64-simd.md:
7698 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7699 (aarch64_crypto_sha512su0qv2di): Ditto.
7700 (aarch64_crypto_sha512su1qv2di): Ditto.
7701 (aarch64_eor3qv8hi): Ditto.
7702 (aarch64_rax1qv2di): Ditto.
7703 (aarch64_xarqv2di): Ditto.
7704 (aarch64_bcaxqv8hi): Ditto.
7705 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7706 (vsha512h2q_u64): Ditto.
7707 (vsha512su0q_u64): Ditto.
7708 (vsha512su1q_u64): Ditto.
7709 (veor3q_u16): Ditto.
7710 (vrax1q_u64): Ditto.
7712 (vbcaxq_u16): Ditto.
7713 * config/arm/types.md (crypto_sha512): New type attribute.
7714 (crypto_sha3): Ditto.
7715 (doc/invoke.texi): Document new sha3 option.
7717 2018-01-10 Michael Collison <michael.collison@arm.com>
7719 * config/aarch64/aarch64-builtins.c:
7720 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7721 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7722 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7723 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7724 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7725 (AARCH64_ISA_SM4): New ISA flag.
7726 (TARGET_SM4): New feature flag for sm4.
7727 * config/aarch64/aarch64-simd-builtins.def
7728 (aarch64_sm3ss1qv4si): Ditto.
7729 (aarch64_sm3tt1aq4si): Ditto.
7730 (aarch64_sm3tt1bq4si): Ditto.
7731 (aarch64_sm3tt2aq4si): Ditto.
7732 (aarch64_sm3tt2bq4si): Ditto.
7733 (aarch64_sm3partw1qv4si): Ditto.
7734 (aarch64_sm3partw2qv4si): Ditto.
7735 (aarch64_sm4eqv4si): Ditto.
7736 (aarch64_sm4ekeyqv4si): Ditto.
7737 * config/aarch64/aarch64-simd.md:
7738 (aarch64_sm3ss1qv4si): Ditto.
7739 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7740 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7741 (aarch64_sm4eqv4si): Ditto.
7742 (aarch64_sm4ekeyqv4si): Ditto.
7743 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7744 (sm3part_op): Ditto.
7745 (CRYPTO_SM3TT): Ditto.
7746 (CRYPTO_SM3PART): Ditto.
7747 (UNSPEC_SM3SS1): New unspec.
7748 (UNSPEC_SM3TT1A): Ditto.
7749 (UNSPEC_SM3TT1B): Ditto.
7750 (UNSPEC_SM3TT2A): Ditto.
7751 (UNSPEC_SM3TT2B): Ditto.
7752 (UNSPEC_SM3PARTW1): Ditto.
7753 (UNSPEC_SM3PARTW2): Ditto.
7754 (UNSPEC_SM4E): Ditto.
7755 (UNSPEC_SM4EKEY): Ditto.
7756 * config/aarch64/constraints.md (Ui2): New constraint.
7757 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7758 * config/arm/types.md (crypto_sm3): New type attribute.
7759 (crypto_sm4): Ditto.
7760 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7761 (vsm3tt1aq_u32): Ditto.
7762 (vsm3tt1bq_u32): Ditto.
7763 (vsm3tt2aq_u32): Ditto.
7764 (vsm3tt2bq_u32): Ditto.
7765 (vsm3partw1q_u32): Ditto.
7766 (vsm3partw2q_u32): Ditto.
7767 (vsm4eq_u32): Ditto.
7768 (vsm4ekeyq_u32): Ditto.
7769 (doc/invoke.texi): Document new sm4 option.
7771 2018-01-10 Michael Collison <michael.collison@arm.com>
7773 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7774 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7775 (AARCH64_FL_FOR_ARCH8_4): New.
7776 (AARCH64_FL_V8_4): New flag.
7777 (doc/invoke.texi): Document new armv8.4-a option.
7779 2018-01-10 Michael Collison <michael.collison@arm.com>
7781 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7782 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7783 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7784 * config/aarch64/aarch64-option-extension.def: Add
7785 AARCH64_OPT_EXTENSION of 'sha2'.
7786 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7787 (crypto): Disable sha2 and aes if crypto disabled.
7788 (crypto): Enable aes and sha2 if enabled.
7789 (simd): Disable sha2 and aes if simd disabled.
7790 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7792 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7793 (TARGET_SHA2): New feature flag for sha2.
7794 (TARGET_AES): New feature flag for aes.
7795 * config/aarch64/aarch64-simd.md:
7796 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7797 conditional on TARGET_AES.
7798 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7799 (aarch64_crypto_sha1hsi): Make pattern conditional
7801 (aarch64_crypto_sha1hv4si): Ditto.
7802 (aarch64_be_crypto_sha1hv4si): Ditto.
7803 (aarch64_crypto_sha1su1v4si): Ditto.
7804 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7805 (aarch64_crypto_sha1su0v4si): Ditto.
7806 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7807 (aarch64_crypto_sha256su0v4si): Ditto.
7808 (aarch64_crypto_sha256su1v4si): Ditto.
7809 (doc/invoke.texi): Document new aes and sha2 options.
7811 2018-01-10 Martin Sebor <msebor@redhat.com>
7813 PR tree-optimization/83781
7814 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7817 2018-01-11 Martin Sebor <msebor@gmail.com>
7818 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7820 PR tree-optimization/83501
7821 PR tree-optimization/81703
7823 * tree-ssa-strlen.c (get_string_cst): Rename...
7824 (get_string_len): ...to this. Handle global constants.
7825 (handle_char_store): Adjust.
7827 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7828 Jim Wilson <jimw@sifive.com>
7830 * config/riscv/riscv-protos.h (riscv_output_return): New.
7831 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7832 (riscv_attribute_table, riscv_output_return),
7833 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7834 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7835 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7837 (riscv_expand_prologue): Add early return for naked function.
7838 (riscv_expand_epilogue): Likewise.
7839 (riscv_function_ok_for_sibcall): Return false for naked function.
7840 (riscv_set_current_function): New.
7841 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7842 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7843 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7844 * doc/extend.texi (RISC-V Function Attributes): New.
7846 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7848 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7849 check for 128-bit long double before checking TCmode.
7850 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7851 128-bit long doubles before checking TFmode or TCmode.
7852 (FLOAT128_IBM_P): Likewise.
7854 2018-01-10 Martin Sebor <msebor@redhat.com>
7856 PR tree-optimization/83671
7857 * builtins.c (c_strlen): Unconditionally return zero for the empty
7859 Use -Warray-bounds for warnings.
7860 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7861 for non-constant array indices with COMPONENT_REF, arrays of
7862 arrays, and pointers to arrays.
7863 (gimple_fold_builtin_strlen): Determine and set length range for
7864 non-constant character arrays.
7866 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7869 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7872 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7874 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7876 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7879 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7880 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7881 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7882 indexed_or_indirect_operand predicate.
7883 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7884 (*vsx_le_perm_load_v8hi): Likewise.
7885 (*vsx_le_perm_load_v16qi): Likewise.
7886 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7887 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7888 (*vsx_le_perm_store_v8hi): Likewise.
7889 (*vsx_le_perm_store_v16qi): Likewise.
7890 (eight unnamed splitters): Likewise.
7892 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7894 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7895 * config/rs6000/emmintrin.h: Likewise.
7896 * config/rs6000/mmintrin.h: Likewise.
7897 * config/rs6000/xmmintrin.h: Likewise.
7899 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7902 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7904 * tree.c (tree_nop_conversion): Return true for location wrapper
7906 (maybe_wrap_with_location): New function.
7907 (selftest::check_strip_nops): New function.
7908 (selftest::test_location_wrappers): New function.
7909 (selftest::tree_c_tests): Call it.
7910 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7911 (maybe_wrap_with_location): New decl.
7912 (EXPR_LOCATION_WRAPPER_P): New macro.
7913 (location_wrapper_p): New inline function.
7914 (tree_strip_any_location_wrapper): New inline function.
7916 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7919 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7920 stack_realign_offset for the largest alignment of stack slot
7922 (ix86_find_max_used_stack_alignment): New function.
7923 (ix86_finalize_stack_frame_flags): Use it. Set
7924 max_used_stack_alignment if we don't realign stack.
7925 * config/i386/i386.h (machine_function): Add
7926 max_used_stack_alignment.
7928 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7930 * config/arm/arm.opt (-mbranch-cost): New option.
7931 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7934 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7937 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7938 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7940 2018-01-10 Richard Biener <rguenther@suse.de>
7943 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7944 early out so it also covers the case where we have a non-NULL
7947 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7949 PR tree-optimization/83753
7950 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7951 for non-strided grouped accesses if the number of elements is 1.
7953 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7956 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7957 * i386.h (TARGET_USE_GATHER): Define.
7958 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7960 2018-01-10 Martin Liska <mliska@suse.cz>
7963 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7964 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7966 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7967 CLEANUP_NO_PARTITIONING is not set.
7969 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7971 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7972 for vectors, as a partial revert of r254296.
7973 * rtl.h (const_vec_p): Delete.
7974 (const_vec_duplicate_p): Don't test for vector CONSTs.
7975 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7976 * expmed.c (make_tree): Likewise.
7979 * common.md (E, F): Use CONSTANT_P instead of checking for
7981 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7982 checking for CONST_VECTOR.
7984 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7987 * predict.c (force_edge_cold): Handle in more sane way edges
7990 2018-01-09 Carl Love <cel@us.ibm.com>
7992 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7994 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7995 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7996 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7997 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7998 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7999 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8000 * config/rs6000/rs6000-protos.h: Add extern defition for
8001 rs6000_generate_float2_double_code.
8002 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8004 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8005 (float2_v2df): Add define_expand.
8007 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8010 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8011 op_mode in the force_to_mode call.
8013 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8015 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8016 instead of checking each element individually.
8017 (aarch64_evpc_uzp): Likewise.
8018 (aarch64_evpc_zip): Likewise.
8019 (aarch64_evpc_ext): Likewise.
8020 (aarch64_evpc_rev): Likewise.
8021 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8022 instead of checking each element individually. Return true without
8024 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8025 whether all selected elements come from the same input, instead of
8026 checking each element individually. Remove calls to gen_rtx_REG,
8027 start_sequence and end_sequence and instead assert that no rtl is
8030 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8032 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8033 order of HIGH and CONST checks.
8035 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8037 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8038 if the destination isn't an SSA_NAME.
8040 2018-01-09 Richard Biener <rguenther@suse.de>
8042 PR tree-optimization/83668
8043 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8045 (canonicalize_loop_form): ... here, renamed from ...
8046 (canonicalize_loop_closed_ssa_form): ... this and amended to
8047 swap successor edges for loop exit blocks to make us use
8048 the RPO order we need for initial schedule generation.
8050 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8052 PR tree-optimization/64811
8053 * match.pd: When optimizing comparisons with Inf, avoid
8054 introducing or losing exceptions from comparisons with NaN.
8056 2018-01-09 Martin Liska <mliska@suse.cz>
8059 * asan.c (shadow_mem_size): Add gcc_assert.
8061 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8063 Don't save registers in main().
8066 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8067 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8068 * config/avr/avr.c (avr_set_current_function): Don't error if
8069 naked, OS_task or OS_main are specified at the same time.
8070 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8072 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8074 * common/config/avr/avr-common.c (avr_option_optimization_table):
8075 Switch on -mmain-is-OS_task for optimizing compilations.
8077 2018-01-09 Richard Biener <rguenther@suse.de>
8079 PR tree-optimization/83572
8080 * graphite.c: Include cfganal.h.
8081 (graphite_transform_loops): Connect infinite loops to exit
8082 and remove fake edges at the end.
8084 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8086 * ipa-inline.c (edge_badness): Revert accidental checkin.
8088 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8091 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8092 symbols; not inline clones.
8094 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8097 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8098 hard registers. Formatting fixes.
8100 PR preprocessor/83722
8101 * gcc.c (try_generate_repro): Pass
8102 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8103 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8106 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8107 Kito Cheng <kito.cheng@gmail.com>
8109 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8110 (riscv_leaf_function_p): Delete.
8111 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8113 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8115 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8117 (do_ifelse): New function.
8118 (do_isel): New function.
8119 (do_sub3): New function.
8120 (do_add3): New function.
8121 (do_load_mask_compare): New function.
8122 (do_overlap_load_compare): New function.
8123 (expand_compare_loop): New function.
8124 (expand_block_compare): Call expand_compare_loop() when appropriate.
8125 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8127 (-mblock-compare-inline-loop-limit): New option.
8129 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8132 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8133 Reverse order of second and third operands in first alternative.
8134 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8135 of first and second elements in UNSPEC_VPERMR vector.
8136 (altivec_expand_vec_perm_le): Likewise.
8138 2018-01-08 Jeff Law <law@redhat.com>
8140 PR rtl-optimizatin/81308
8141 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8142 (process_switch): If group_case_labels makes a change, then set
8144 (pass_convert_switch::execute): If a switch is converted, then
8145 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8147 PR rtl-optimization/81308
8148 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8151 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8153 PR target/83663 - Revert r255946
8154 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8155 generation for cases where splatting a value is not useful.
8156 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8157 across a vec_duplicate and a paradoxical subreg forming a vector
8158 mode to a vec_concat.
8160 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8162 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8163 -march=armv8.3-a variants.
8164 * config/arm/t-multilib: Likewise.
8165 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8167 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8169 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8171 (cceq_ior_compare_complement): Give it a name so I can use it, and
8172 change boolean_or_operator predicate to boolean_operator so it can
8173 be used to generate a crand.
8174 (eqne): New code iterator.
8175 (bd/bd_neg): New code_attrs.
8176 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8177 a single define_insn.
8178 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8179 decrement (bdnzt/bdnzf/bdzt/bdzf).
8180 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8181 with the new names of the branch decrement patterns, and added the
8182 names of the branch decrement conditional patterns.
8184 2018-01-08 Richard Biener <rguenther@suse.de>
8186 PR tree-optimization/83563
8187 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8190 2018-01-08 Richard Biener <rguenther@suse.de>
8193 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8195 2018-01-08 Richard Biener <rguenther@suse.de>
8197 PR tree-optimization/83685
8198 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8199 references to abnormals.
8201 2018-01-08 Richard Biener <rguenther@suse.de>
8204 * dwarf2out.c (output_indirect_strings): Handle empty
8205 skeleton_debug_str_hash.
8206 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8208 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8210 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8211 (emit_store_direct): Likewise.
8212 (arc_trampoline_adjust_address): Likewise.
8213 (arc_asm_trampoline_template): New function.
8214 (arc_initialize_trampoline): Use asm_trampoline_template.
8215 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8216 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8217 * config/arc/arc.md (flush_icache): Delete pattern.
8219 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8221 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8222 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8225 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8228 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8229 by not USED_FOR_TARGET.
8230 (make_pass_resolve_sw_modes): Likewise.
8232 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8234 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8237 2018-01-08 Richard Biener <rguenther@suse.de>
8240 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8242 2018-01-08 Richard Biener <rguenther@suse.de>
8245 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8247 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8250 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8251 basic blocks with a small number of successors.
8252 (convert_control_dep_chain_into_preds): Improve handling of
8254 (dump_predicates): Split apart into...
8255 (dump_pred_chain): ...here...
8256 (dump_pred_info): ...and here.
8257 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8258 (can_chain_union_be_invalidated_p): Improve check for invalidation
8260 (uninit_uses_cannot_happen): Avoid unnecessary if
8261 convert_control_dep_chain_into_preds yielded nothing.
8263 2018-01-06 Martin Sebor <msebor@redhat.com>
8265 PR tree-optimization/83640
8266 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8267 subtracting negative offset from size.
8268 (builtin_access::overlap): Adjust offset bounds of the access to fall
8269 within the size of the object if possible.
8271 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8273 PR rtl-optimization/83699
8274 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8275 extract_bit_field_as_subreg to cases in which the extracted
8276 value is also a vector.
8278 * lra-constraints.c (process_alt_operands): Test for the equivalence
8279 substitutions when detecting a possible reload cycle.
8281 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8284 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8285 by default if flag_selective_schedling{,2}. Formatting fixes.
8287 PR rtl-optimization/83682
8288 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8289 if it has non-VECTOR_MODE element mode.
8290 (vec_duplicate_p): Likewise.
8293 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8294 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8296 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8299 * config/i386/i386-builtin.def
8300 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8301 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8302 Require also OPTION_MASK_ISA_AVX512F in addition to
8303 OPTION_MASK_ISA_GFNI.
8304 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8305 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8306 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8307 to OPTION_MASK_ISA_GFNI.
8308 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8309 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8310 OPTION_MASK_ISA_AVX512BW.
8311 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8312 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8313 addition to OPTION_MASK_ISA_GFNI.
8314 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8315 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8316 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8317 to OPTION_MASK_ISA_GFNI.
8318 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8319 a requirement for all ISAs rather than any of them with a few
8321 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8323 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8324 bitmasks to be enabled with 3 exceptions, instead of requiring any
8325 enabled ISA with lots of exceptions.
8326 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8327 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8328 Change avx512bw in isa attribute to avx512f.
8329 * config/i386/sgxintrin.h: Add license boilerplate.
8330 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8331 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8332 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8333 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8335 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8336 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8337 temporarily sse2 rather than sse if not enabled already.
8340 * config/i386/sse.md (VI248_VLBW): Rename to ...
8341 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8342 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8343 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8344 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8345 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8346 mode iterator instead of VI248_VLBW.
8348 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8350 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8351 (record_modified): Skip clobbers; add debug output.
8352 (param_change_prob): Use sreal frequencies.
8354 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8356 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8357 punt for user-aligned variables.
8359 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8361 * tree-chrec.c (chrec_contains_symbols): Return true for
8364 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8367 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8368 of (x|y) == x for BICS pattern.
8370 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8372 PR tree-optimization/83605
8373 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8374 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8377 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8379 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8380 * config/epiphany/rtems.h: New file.
8382 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8383 Uros Bizjak <ubizjak@gmail.com>
8386 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8387 QIreg_operand instead of register_operand predicate.
8388 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8389 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8390 comments instead of -fmitigate[-_]rop.
8392 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8395 * cgraphunit.c (symbol_table::compile): Switch to text_section
8396 before calling assembly_start debug hook.
8397 * run-rtl-passes.c (run_rtl_passes): Likewise.
8400 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8402 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8403 range_int_cst_p rather than !symbolic_range_p before calling
8404 extract_range_from_multiplicative_op_1.
8406 2018-01-04 Jeff Law <law@redhat.com>
8408 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8409 redundant test in assertion.
8411 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8413 * doc/rtl.texi: Document machine_mode wrapper classes.
8415 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8417 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8420 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8422 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8423 the VEC_PERM_EXPR fold to fail.
8425 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8428 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8429 to switched_sections.
8431 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8434 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8437 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8440 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8441 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8443 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8446 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8447 is BLKmode and bitpos not zero or mode change is needed.
8449 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8452 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8455 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8458 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8459 instead of MULT rtx. Update all corresponding splitters.
8461 (*ssub<modesuffix>): Ditto.
8463 (*cmp_sadd_di): Update split patterns.
8464 (*cmp_sadd_si): Ditto.
8465 (*cmp_sadd_sidi): Ditto.
8466 (*cmp_ssub_di): Ditto.
8467 (*cmp_ssub_si): Ditto.
8468 (*cmp_ssub_sidi): Ditto.
8469 * config/alpha/predicates.md (const23_operand): New predicate.
8470 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8471 Look for ASHIFT, not MULT inner operand.
8472 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8474 2018-01-04 Martin Liska <mliska@suse.cz>
8476 PR gcov-profile/83669
8477 * gcov.c (output_intermediate_file): Add version to intermediate
8479 * doc/gcov.texi: Document new field 'version' in intermediate
8480 file format. Fix location of '-k' option of gcov command.
8482 2018-01-04 Martin Liska <mliska@suse.cz>
8485 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8487 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8489 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8491 2018-01-03 Martin Sebor <msebor@redhat.com>
8493 PR tree-optimization/83655
8494 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8495 checking calls with invalid arguments.
8497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8499 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8500 (vectorizable_mask_load_store): Delete.
8501 (vectorizable_call): Return false for masked loads and stores.
8502 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8503 instead of gimple_assign_rhs1.
8504 (vectorizable_load): Handle IFN_MASK_LOAD.
8505 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8509 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8511 (vectorizable_mask_load_store): ...here.
8512 (vectorizable_load): ...and here.
8514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8516 * tree-vect-stmts.c (vect_build_all_ones_mask)
8517 (vect_build_zero_merge_argument): New functions, split out from...
8518 (vectorizable_load): ...here.
8520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8522 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8524 (vectorizable_mask_load_store): ...here.
8525 (vectorizable_store): ...and here.
8527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8529 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8531 (vectorizable_mask_load_store): ...here.
8533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8535 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8536 (vect_model_store_cost): Take a vec_load_store_type instead of a
8538 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8539 (vect_model_store_cost): Take a vec_load_store_type instead of a
8541 (vectorizable_mask_load_store): Update accordingly.
8542 (vectorizable_store): Likewise.
8543 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8547 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8548 IFN_MASK_LOAD calls here rather than...
8549 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8552 Alan Hayward <alan.hayward@arm.com>
8553 David Sherwood <david.sherwood@arm.com>
8555 * expmed.c (extract_bit_field_1): For vector extracts,
8556 fall back to extract_bit_field_as_subreg if vec_extract
8559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8560 Alan Hayward <alan.hayward@arm.com>
8561 David Sherwood <david.sherwood@arm.com>
8563 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8564 they are variable or constant sized.
8565 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8566 slots for constant-sized data.
8568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8569 Alan Hayward <alan.hayward@arm.com>
8570 David Sherwood <david.sherwood@arm.com>
8572 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8573 handling COND_EXPRs with boolean comparisons, try to find a better
8574 basis for the mask type than the boolean itself.
8576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8578 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8579 is calculated and how it can be overridden.
8580 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8581 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8583 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8587 Alan Hayward <alan.hayward@arm.com>
8588 David Sherwood <david.sherwood@arm.com>
8590 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8591 Remove the mode argument.
8592 (aarch64_simd_valid_immediate): Remove the mode and inverse
8594 * config/aarch64/iterators.md (bitsize): New iterator.
8595 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8596 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8597 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8598 aarch64_simd_valid_immediate.
8599 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8600 (aarch64_reg_or_bic_imm): Likewise.
8601 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8602 with an insn_type enum and msl with a modifier_type enum.
8603 Replace element_width with a scalar_mode. Change the shift
8604 to unsigned int. Add constructors for scalar_float_mode and
8605 scalar_int_mode elements.
8606 (aarch64_vect_float_const_representable_p): Delete.
8607 (aarch64_can_const_movi_rtx_p)
8608 (aarch64_simd_scalar_immediate_valid_for_move)
8609 (aarch64_simd_make_constant): Update call to
8610 aarch64_simd_valid_immediate.
8611 (aarch64_advsimd_valid_immediate_hs): New function.
8612 (aarch64_advsimd_valid_immediate): Likewise.
8613 (aarch64_simd_valid_immediate): Remove mode and inverse
8614 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8615 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8616 and aarch64_float_const_representable_p on the result.
8617 (aarch64_output_simd_mov_immediate): Remove mode argument.
8618 Update call to aarch64_simd_valid_immediate and use of
8619 simd_immediate_info.
8620 (aarch64_output_scalar_simd_mov_immediate): Update call
8623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8624 Alan Hayward <alan.hayward@arm.com>
8625 David Sherwood <david.sherwood@arm.com>
8627 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8628 (mode_nunits): Likewise CONST_MODE_NUNITS.
8629 * machmode.def (ADJUST_NUNITS): Document.
8630 * genmodes.c (mode_data::need_nunits_adj): New field.
8631 (blank_mode): Update accordingly.
8632 (adj_nunits): New variable.
8633 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8635 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8636 listed in adj_nunits.
8637 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8638 listed in adj_nunits. Don't emit case statements for such modes.
8639 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8640 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8641 nothing if adj_nunits is nonnull.
8642 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8643 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8644 (emit_mode_fbit): Update use of print_maybe_const_decl.
8645 (emit_move_size): Likewise. Treat the array as non-const
8647 (emit_mode_adjustments): Handle adj_nunits.
8649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8651 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8652 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8653 (VECTOR_MODES): Use it.
8654 (make_vector_modes): Take the prefix as an argument.
8656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8657 Alan Hayward <alan.hayward@arm.com>
8658 David Sherwood <david.sherwood@arm.com>
8660 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8661 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8662 for MODE_VECTOR_BOOL.
8663 * machmode.def (VECTOR_BOOL_MODE): Document.
8664 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8665 (make_vector_bool_mode): New function.
8666 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8668 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8669 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8671 * stor-layout.c (int_mode_for_mode): Likewise.
8672 * tree.c (build_vector_type_for_mode): Likewise.
8673 * varasm.c (output_constant_pool_2): Likewise.
8674 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8675 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8676 for MODE_VECTOR_BOOL.
8677 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8678 of mode class checks.
8679 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8680 instead of a list of mode class checks.
8681 (expand_vector_scalar_condition): Likewise.
8682 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8685 Alan Hayward <alan.hayward@arm.com>
8686 David Sherwood <david.sherwood@arm.com>
8688 * machmode.h (mode_size): Change from unsigned short to
8690 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8691 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8692 or if measurement_type is not polynomial.
8693 (fixed_size_mode::includes_p): Check for constant-sized modes.
8694 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8695 return a poly_uint16 rather than an unsigned short.
8696 (emit_mode_size): Change the type of mode_size from unsigned short
8697 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8698 (emit_mode_adjustments): Cope with polynomial vector sizes.
8699 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8701 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8703 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8704 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8705 * caller-save.c (setup_save_areas): Likewise.
8706 (replace_reg_with_saved_mem): Likewise.
8707 * calls.c (emit_library_call_value_1): Likewise.
8708 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8709 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8710 (gen_lowpart_for_combine): Likewise.
8711 * convert.c (convert_to_integer_1): Likewise.
8712 * cse.c (equiv_constant, cse_insn): Likewise.
8713 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8714 (cselib_subst_to_values): Likewise.
8715 * dce.c (word_dce_process_block): Likewise.
8716 * df-problems.c (df_word_lr_mark_ref): Likewise.
8717 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8718 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8719 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8720 (rtl_for_decl_location): Likewise.
8721 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8722 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8723 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8724 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8725 (expand_expr_real_1): Likewise.
8726 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8727 (pad_below): Likewise.
8728 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8729 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8730 * ira.c (get_subreg_tracking_sizes): Likewise.
8731 * ira-build.c (ira_create_allocno_objects): Likewise.
8732 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8733 (ira_sort_regnos_for_alter_reg): Likewise.
8734 * ira-costs.c (record_operand_costs): Likewise.
8735 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8736 (resolve_simple_move): Likewise.
8737 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8738 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8739 (lra_constraints): Likewise.
8740 (CONST_POOL_OK_P): Reject variable-sized modes.
8741 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8742 (add_pseudo_to_slot, lra_spill): Likewise.
8743 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8744 * optabs-query.c (get_best_extraction_insn): Likewise.
8745 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8746 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8747 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8748 * recog.c (offsettable_address_addr_space_p): Likewise.
8749 * regcprop.c (maybe_mode_change): Likewise.
8750 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8751 * regrename.c (build_def_use): Likewise.
8752 * regstat.c (dump_reg_info): Likewise.
8753 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8754 (find_reloads, find_reloads_subreg_address): Likewise.
8755 * reload1.c (eliminate_regs_1): Likewise.
8756 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8757 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8758 (simplify_binary_operation_1, simplify_subreg): Likewise.
8759 * targhooks.c (default_function_arg_padding): Likewise.
8760 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8761 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8762 (verify_gimple_assign_ternary): Likewise.
8763 * tree-inline.c (estimate_move_cost): Likewise.
8764 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8765 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8766 (get_address_cost_ainc): Likewise.
8767 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8768 (vect_supportable_dr_alignment): Likewise.
8769 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8770 (vectorizable_reduction): Likewise.
8771 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8772 (vectorizable_operation, vectorizable_load): Likewise.
8773 * tree.c (build_same_sized_truth_vector_type): Likewise.
8774 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8775 * var-tracking.c (emit_note_insn_var_location): Likewise.
8776 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8777 (ADDR_VEC_ALIGN): Likewise.
8779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8780 Alan Hayward <alan.hayward@arm.com>
8781 David Sherwood <david.sherwood@arm.com>
8783 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8785 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8786 or if measurement_type is polynomial.
8787 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8788 * combine.c (make_extraction): Likewise.
8789 * dse.c (find_shift_sequence): Likewise.
8790 * dwarf2out.c (mem_loc_descriptor): Likewise.
8791 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8792 (extract_bit_field, extract_low_bits): Likewise.
8793 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8794 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8795 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8796 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8797 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8798 * reload.c (find_reloads): Likewise.
8799 * reload1.c (alter_reg): Likewise.
8800 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8801 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8802 * tree-if-conv.c (predicate_mem_writes): Likewise.
8803 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8804 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8805 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8806 * valtrack.c (dead_debug_insert_temp): Likewise.
8807 * varasm.c (mergeable_constant_section): Likewise.
8808 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8811 Alan Hayward <alan.hayward@arm.com>
8812 David Sherwood <david.sherwood@arm.com>
8814 * expr.c (expand_assignment): Cope with polynomial mode sizes
8815 when assigning to a CONCAT.
8817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8818 Alan Hayward <alan.hayward@arm.com>
8819 David Sherwood <david.sherwood@arm.com>
8821 * machmode.h (mode_precision): Change from unsigned short to
8823 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8825 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8826 or if measurement_type is not polynomial.
8827 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8828 in which the mode is already known to be a scalar_int_mode.
8829 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8830 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8832 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8833 for GET_MODE_PRECISION.
8834 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8835 for GET_MODE_PRECISION.
8836 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8838 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8839 (expand_field_assignment, make_extraction): Likewise.
8840 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8841 (get_last_value): Likewise.
8842 * convert.c (convert_to_integer_1): Likewise.
8843 * cse.c (cse_insn): Likewise.
8844 * expr.c (expand_expr_real_1): Likewise.
8845 * lra-constraints.c (simplify_operand_subreg): Likewise.
8846 * optabs-query.c (can_atomic_load_p): Likewise.
8847 * optabs.c (expand_atomic_load): Likewise.
8848 (expand_atomic_store): Likewise.
8849 * ree.c (combine_reaching_defs): Likewise.
8850 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8851 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8852 * tree.h (type_has_mode_precision_p): Likewise.
8853 * ubsan.c (instrument_si_overflow): Likewise.
8855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8856 Alan Hayward <alan.hayward@arm.com>
8857 David Sherwood <david.sherwood@arm.com>
8859 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8860 polynomial numbers of units.
8861 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8862 (valid_vector_subparts_p): New function.
8863 (build_vector_type): Remove temporary shim and take the number
8864 of units as a poly_uint64 rather than an int.
8865 (build_opaque_vector_type): Take the number of units as a
8866 poly_uint64 rather than an int.
8867 * tree.c (build_vector_from_ctor): Handle polynomial
8868 TYPE_VECTOR_SUBPARTS.
8869 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8870 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8871 (build_vector_from_val): If the number of units is variable,
8872 use build_vec_duplicate_cst for constant operands and
8873 VEC_DUPLICATE_EXPR otherwise.
8874 (make_vector_type): Remove temporary is_constant ().
8875 (build_vector_type, build_opaque_vector_type): Take the number of
8876 units as a poly_uint64 rather than an int.
8877 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8879 * cfgexpand.c (expand_debug_expr): Likewise.
8880 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8881 (store_constructor, expand_expr_real_1): Likewise.
8882 (const_scalar_mask_from_tree): Likewise.
8883 * fold-const-call.c (fold_const_reduction): Likewise.
8884 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8885 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8886 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8887 (fold_relational_const): Likewise.
8888 (native_interpret_vector): Likewise. Change the size from an
8889 int to an unsigned int.
8890 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8891 TYPE_VECTOR_SUBPARTS.
8892 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8893 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8894 duplicating a non-constant operand into a variable-length vector.
8895 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8896 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8897 * ipa-icf.c (sem_variable::equals): Likewise.
8898 * match.pd: Likewise.
8899 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8900 * print-tree.c (print_node): Likewise.
8901 * stor-layout.c (layout_type): Likewise.
8902 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8903 * tree-cfg.c (verify_gimple_comparison): Likewise.
8904 (verify_gimple_assign_binary): Likewise.
8905 (verify_gimple_assign_ternary): Likewise.
8906 (verify_gimple_assign_single): Likewise.
8907 * tree-pretty-print.c (dump_generic_node): Likewise.
8908 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8909 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8910 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8911 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8912 (vect_shift_permute_load_chain): Likewise.
8913 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8914 (expand_vector_condition, optimize_vector_constructor): Likewise.
8915 (lower_vec_perm, get_compute_type): Likewise.
8916 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8917 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8918 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8919 (vect_recog_mask_conversion_pattern): Likewise.
8920 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8921 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8922 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8923 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8924 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8925 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8926 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8927 (supportable_widening_operation): Likewise.
8928 (supportable_narrowing_operation): Likewise.
8929 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8931 * varasm.c (output_constant): Likewise.
8933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8934 Alan Hayward <alan.hayward@arm.com>
8935 David Sherwood <david.sherwood@arm.com>
8937 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8938 so that both the length == 3 and length != 3 cases set up their
8939 own permute vectors. Add comments explaining why we know the
8940 number of elements is constant.
8941 (vect_permute_load_chain): Likewise.
8943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8944 Alan Hayward <alan.hayward@arm.com>
8945 David Sherwood <david.sherwood@arm.com>
8947 * machmode.h (mode_nunits): Change from unsigned char to
8949 (ONLY_FIXED_SIZE_MODES): New macro.
8950 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8951 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8952 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8954 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8955 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8956 or if measurement_type is not polynomial.
8957 * genmodes.c (ZERO_COEFFS): New macro.
8958 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8960 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8961 Use ZERO_COEFFS when emitting initializers.
8962 * data-streamer.h (bp_pack_poly_value): New function.
8963 (bp_unpack_poly_value): Likewise.
8964 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8965 for GET_MODE_NUNITS.
8966 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8967 for GET_MODE_NUNITS.
8968 * tree.c (make_vector_type): Remove temporary shim and make
8969 the real function take the number of units as a poly_uint64
8971 (build_vector_type_for_mode): Handle polynomial nunits.
8972 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8973 * emit-rtl.c (const_vec_series_p_1): Likewise.
8974 (gen_rtx_CONST_VECTOR): Likewise.
8975 * fold-const.c (test_vec_duplicate_folding): Likewise.
8976 * genrecog.c (validate_pattern): Likewise.
8977 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8978 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8979 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8980 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8981 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8982 * rtlanal.c (subreg_get_info): Likewise.
8983 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8984 (vect_grouped_load_supported): Likewise.
8985 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8986 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8987 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8988 (simplify_const_unary_operation, simplify_binary_operation_1)
8989 (simplify_const_binary_operation, simplify_ternary_operation)
8990 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8991 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8992 instead of CONST_VECTOR_NUNITS.
8993 * varasm.c (output_constant_pool_2): Likewise.
8994 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8995 explicit-encoded elements in the XVEC for variable-length vectors.
8997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8999 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9002 Alan Hayward <alan.hayward@arm.com>
9003 David Sherwood <david.sherwood@arm.com>
9005 * coretypes.h (fixed_size_mode): Declare.
9006 (fixed_size_mode_pod): New typedef.
9007 * builtins.h (target_builtins::x_apply_args_mode)
9008 (target_builtins::x_apply_result_mode): Change type to
9009 fixed_size_mode_pod.
9010 * builtins.c (apply_args_size, apply_result_size, result_vector)
9011 (expand_builtin_apply_args_1, expand_builtin_apply)
9012 (expand_builtin_return): Update accordingly.
9014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9016 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9017 * cselib.c (cselib_hash_rtx): Likewise.
9018 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9019 CONST_VECTOR encoding.
9021 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9022 Jeff Law <law@redhat.com>
9025 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9026 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9027 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9028 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9031 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9032 explicitly probe *sp in a noreturn function if there were any callee
9033 register saves or frame pointer is needed.
9035 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9038 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9039 BLKmode for ternary, binary or unary expressions.
9042 * var-tracking.c (delete_vta_debug_insn): New inline function.
9043 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9044 insns from get_insns () to NULL instead of each bb separately.
9045 Use delete_vta_debug_insn. No longer static.
9046 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9047 delete_vta_debug_insns callers.
9048 * rtl.h (delete_vta_debug_insns): Declare.
9049 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9050 instead of variable_tracking_main.
9052 2018-01-03 Martin Sebor <msebor@redhat.com>
9054 PR tree-optimization/83603
9055 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9056 arguments past the endof the argument list in functions declared
9057 without a prototype.
9058 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9059 Avoid checking when arguments are null.
9061 2018-01-03 Martin Sebor <msebor@redhat.com>
9064 * doc/extend.texi (attribute const): Fix a typo.
9065 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9066 issuing -Wsuggest-attribute for void functions.
9068 2018-01-03 Martin Sebor <msebor@redhat.com>
9070 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9071 offset_int::from instead of wide_int::to_shwi.
9072 (maybe_diag_overlap): Remove assertion.
9073 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9074 * gimple-ssa-sprintf.c (format_directive): Same.
9075 (parse_directive): Same.
9076 (sprintf_dom_walker::compute_format_length): Same.
9077 (try_substitute_return_value): Same.
9079 2018-01-03 Jeff Law <law@redhat.com>
9082 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9083 non-constant residual for zero at runtime and avoid probing in
9084 that case. Reorganize code for trailing problem to mirror handling
9087 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9089 PR tree-optimization/83501
9090 * tree-ssa-strlen.c (get_string_cst): New.
9091 (handle_char_store): Call get_string_cst.
9093 2018-01-03 Martin Liska <mliska@suse.cz>
9095 PR tree-optimization/83593
9096 * tree-ssa-strlen.c: Include tree-cfg.h.
9097 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9098 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9099 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9101 (strlen_dom_walker::before_dom_children): Call
9102 gimple_purge_dead_eh_edges. Dump tranformation with details
9104 (strlen_dom_walker::before_dom_children): Update call by adding
9105 new argument cleanup_eh.
9106 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9108 2018-01-03 Martin Liska <mliska@suse.cz>
9111 * cif-code.def (VARIADIC_THUNK): New enum value.
9112 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9115 2018-01-03 Jan Beulich <jbeulich@suse.com>
9117 * sse.md (mov<mode>_internal): Tighten condition for when to use
9118 vmovdqu<ssescalarsize> for TI and OI modes.
9120 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9122 Update copyright years.
9124 2018-01-03 Martin Liska <mliska@suse.cz>
9127 * ipa-visibility.c (function_and_variable_visibility): Skip
9128 functions with noipa attribure.
9130 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9132 * gcc.c (process_command): Update copyright notice dates.
9133 * gcov-dump.c (print_version): Ditto.
9134 * gcov.c (print_version): Ditto.
9135 * gcov-tool.c (print_version): Ditto.
9136 * gengtype.c (create_file): Ditto.
9137 * doc/cpp.texi: Bump @copying's copyright year.
9138 * doc/cppinternals.texi: Ditto.
9139 * doc/gcc.texi: Ditto.
9140 * doc/gccint.texi: Ditto.
9141 * doc/gcov.texi: Ditto.
9142 * doc/install.texi: Ditto.
9143 * doc/invoke.texi: Ditto.
9145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9147 * vector-builder.h (vector_builder::m_full_nelts): Change from
9148 unsigned int to poly_uint64.
9149 (vector_builder::full_nelts): Update prototype accordingly.
9150 (vector_builder::new_vector): Likewise.
9151 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9152 (vector_builder::operator ==): Likewise.
9153 (vector_builder::finalize): Likewise.
9154 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9155 Take the number of elements as a poly_uint64 rather than an
9157 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9158 from unsigned int to poly_uint64.
9159 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9160 (vec_perm_indices::new_vector): Likewise.
9161 (vec_perm_indices::length): Likewise.
9162 (vec_perm_indices::nelts_per_input): Likewise.
9163 (vec_perm_indices::input_nelts): Likewise.
9164 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9165 number of elements per input as a poly_uint64 rather than an
9166 unsigned int. Use the original encoding for variable-length
9167 vectors, rather than clamping each individual element.
9168 For the second and subsequent elements in each pattern,
9169 clamp the step and base before clamping their sum.
9170 (vec_perm_indices::series_p): Handle polynomial element counts.
9171 (vec_perm_indices::all_in_range_p): Likewise.
9172 (vec_perm_indices_to_tree): Likewise.
9173 (vec_perm_indices_to_rtx): Likewise.
9174 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9175 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9176 (tree_vector_builder::new_binary_operation): Handle polynomial
9177 element counts. Return false if we need to know the number
9178 of elements at compile time.
9179 * fold-const.c (fold_vec_perm): Punt if the number of elements
9180 isn't known at compile time.
9182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9184 * vec-perm-indices.h (vec_perm_builder): Change element type
9185 from HOST_WIDE_INT to poly_int64.
9186 (vec_perm_indices::element_type): Update accordingly.
9187 (vec_perm_indices::clamp): Handle polynomial element_types.
9188 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9189 (vec_perm_indices::all_in_range_p): Likewise.
9190 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9192 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9193 polynomial vec_perm_indices element types.
9194 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9195 * fold-const.c (fold_vec_perm): Likewise.
9196 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9197 * tree-vect-generic.c (lower_vec_perm): Likewise.
9198 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9199 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9200 element type to HOST_WIDE_INT.
9202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9203 Alan Hayward <alan.hayward@arm.com>
9204 David Sherwood <david.sherwood@arm.com>
9206 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9207 rather than an int. Use plus_constant.
9208 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9209 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9212 Alan Hayward <alan.hayward@arm.com>
9213 David Sherwood <david.sherwood@arm.com>
9215 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9216 a HOST_WIDE_INT to a poly_int64.
9218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9219 Alan Hayward <alan.hayward@arm.com>
9220 David Sherwood <david.sherwood@arm.com>
9222 * calls.c (load_register_parameters): Cope with polynomial
9223 mode sizes. Require a constant size for BLKmode parameters
9224 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9225 forces a parameter to be padded at the lsb end in order to
9226 fill a complete number of words, require the parameter size
9227 to be ordered wrt UNITS_PER_WORD.
9229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9230 Alan Hayward <alan.hayward@arm.com>
9231 David Sherwood <david.sherwood@arm.com>
9233 * reload1.c (spill_stack_slot_width): Change element type
9234 from unsigned int to poly_uint64_pod.
9235 (alter_reg): Treat mode sizes as polynomial.
9237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9238 Alan Hayward <alan.hayward@arm.com>
9239 David Sherwood <david.sherwood@arm.com>
9241 * reload.c (complex_word_subreg_p): New function.
9242 (reload_inner_reg_of_subreg, push_reload): Use it.
9244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9245 Alan Hayward <alan.hayward@arm.com>
9246 David Sherwood <david.sherwood@arm.com>
9248 * lra-constraints.c (process_alt_operands): Reject matched
9249 operands whose sizes aren't ordered.
9250 (match_reload): Refer to this check here.
9252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9253 Alan Hayward <alan.hayward@arm.com>
9254 David Sherwood <david.sherwood@arm.com>
9256 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9257 that the mode size is in the set {1, 2, 4, 8, 16}.
9259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9260 Alan Hayward <alan.hayward@arm.com>
9261 David Sherwood <david.sherwood@arm.com>
9263 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9264 Use plus_constant instead of gen_rtx_PLUS.
9266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9267 Alan Hayward <alan.hayward@arm.com>
9268 David Sherwood <david.sherwood@arm.com>
9270 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9271 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9272 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9273 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9274 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9275 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9276 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9277 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9278 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9279 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9281 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9282 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9283 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9284 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9285 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9286 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9287 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9288 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9289 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9290 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9292 * expr.c (emit_move_resolve_push): Treat the input and result
9293 of PUSH_ROUNDING as a poly_int64.
9294 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9295 (emit_push_insn): Likewise.
9296 * lra-eliminations.c (mark_not_eliminable): Likewise.
9297 * recog.c (push_operand): Likewise.
9298 * reload1.c (elimination_effects): Likewise.
9299 * rtlanal.c (nonzero_bits1): Likewise.
9300 * calls.c (store_one_arg): Likewise. Require the padding to be
9301 known at compile time.
9303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9304 Alan Hayward <alan.hayward@arm.com>
9305 David Sherwood <david.sherwood@arm.com>
9307 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9308 Use plus_constant instead of gen_rtx_PLUS.
9310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9311 Alan Hayward <alan.hayward@arm.com>
9312 David Sherwood <david.sherwood@arm.com>
9314 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9318 Alan Hayward <alan.hayward@arm.com>
9319 David Sherwood <david.sherwood@arm.com>
9321 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9322 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9323 via stack temporaries. Treat the mode size as polynomial too.
9325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9326 Alan Hayward <alan.hayward@arm.com>
9327 David Sherwood <david.sherwood@arm.com>
9329 * expr.c (expand_expr_real_2): When handling conversions involving
9330 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9331 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9332 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9341 Alan Hayward <alan.hayward@arm.com>
9342 David Sherwood <david.sherwood@arm.com>
9344 * combine.c (can_change_dest_mode): Handle polynomial
9345 REGMODE_NATURAL_SIZE.
9346 * expmed.c (store_bit_field_1): Likewise.
9347 * expr.c (store_constructor): Likewise.
9348 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9349 and polynomial REGMODE_NATURAL_SIZE.
9350 (gen_lowpart_common): Likewise.
9351 * reginfo.c (record_subregs_of_mode): Likewise.
9352 * rtlanal.c (read_modify_subreg_p): Likewise.
9354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9355 Alan Hayward <alan.hayward@arm.com>
9356 David Sherwood <david.sherwood@arm.com>
9358 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9359 numbers of elements.
9361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9362 Alan Hayward <alan.hayward@arm.com>
9363 David Sherwood <david.sherwood@arm.com>
9365 * match.pd: Cope with polynomial numbers of vector elements.
9367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9368 Alan Hayward <alan.hayward@arm.com>
9369 David Sherwood <david.sherwood@arm.com>
9371 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9372 in a POINTER_PLUS_EXPR.
9374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9375 Alan Hayward <alan.hayward@arm.com>
9376 David Sherwood <david.sherwood@arm.com>
9378 * omp-simd-clone.c (simd_clone_subparts): New function.
9379 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9380 (ipa_simd_modify_function_body): Likewise.
9382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9383 Alan Hayward <alan.hayward@arm.com>
9384 David Sherwood <david.sherwood@arm.com>
9386 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9387 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9388 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9389 (expand_vector_condition, vector_element): Likewise.
9390 (subparts_gt): New function.
9391 (get_compute_type): Use subparts_gt.
9392 (count_type_subparts): Delete.
9393 (expand_vector_operations_1): Use subparts_gt instead of
9394 count_type_subparts.
9396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9397 Alan Hayward <alan.hayward@arm.com>
9398 David Sherwood <david.sherwood@arm.com>
9400 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9401 (vect_compile_time_alias): ...this new function. Do the calculation
9402 on poly_ints rather than trees.
9403 (vect_prune_runtime_alias_test_list): Update call accordingly.
9405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9406 Alan Hayward <alan.hayward@arm.com>
9407 David Sherwood <david.sherwood@arm.com>
9409 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9411 (vect_schedule_slp_instance): Likewise.
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-vect-slp.c (vect_get_and_check_slp_defs): Reject
9418 constant and extern definitions for variable-length vectors.
9419 (vect_get_constant_vectors): Note that the number of units
9420 is known to be constant.
9422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9423 Alan Hayward <alan.hayward@arm.com>
9424 David Sherwood <david.sherwood@arm.com>
9426 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9427 of units as polynomial. Choose between WIDE and NARROW based
9430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9431 Alan Hayward <alan.hayward@arm.com>
9432 David Sherwood <david.sherwood@arm.com>
9434 * tree-vect-stmts.c (simd_clone_subparts): New function.
9435 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9438 Alan Hayward <alan.hayward@arm.com>
9439 David Sherwood <david.sherwood@arm.com>
9441 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9442 vectors as polynomial. Use build_index_vector for
9445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9446 Alan Hayward <alan.hayward@arm.com>
9447 David Sherwood <david.sherwood@arm.com>
9449 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9450 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9451 for variable-length vectors.
9452 (vectorizable_mask_load_store): Treat the number of units as
9453 polynomial, asserting that it is constant if the condition has
9454 already been enforced.
9455 (vectorizable_store, vectorizable_load): Likewise.
9457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9458 Alan Hayward <alan.hayward@arm.com>
9459 David Sherwood <david.sherwood@arm.com>
9461 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9462 of units as polynomial. Punt if we can't tell at compile time
9463 which vector contains the final result.
9465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9466 Alan Hayward <alan.hayward@arm.com>
9467 David Sherwood <david.sherwood@arm.com>
9469 * tree-vect-loop.c (vectorizable_induction): Treat the number
9470 of units as polynomial. Punt on SLP inductions. Use an integer
9471 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9472 cast of such a series for variable-length floating-point
9475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9476 Alan Hayward <alan.hayward@arm.com>
9477 David Sherwood <david.sherwood@arm.com>
9479 * tree.h (build_index_vector): Declare.
9480 * tree.c (build_index_vector): New function.
9481 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9482 of units as polynomial, forcibly converting it to a constant if
9483 vectorizable_reduction has already enforced the condition.
9484 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9485 to create a {1,2,3,...} vector.
9486 (vectorizable_reduction): Treat the number of units as polynomial.
9487 Choose vectype_in based on the largest scalar element size rather
9488 than the smallest number of units. Enforce the restrictions
9491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9492 Alan Hayward <alan.hayward@arm.com>
9493 David Sherwood <david.sherwood@arm.com>
9495 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9496 number of units as polynomial.
9498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9499 Alan Hayward <alan.hayward@arm.com>
9500 David Sherwood <david.sherwood@arm.com>
9502 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9503 * target.def (autovectorize_vector_sizes): Return the vector sizes
9504 by pointer, using vector_sizes rather than a bitmask.
9505 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9506 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9507 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9509 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9510 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9511 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9512 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9513 * omp-general.c (omp_max_vf): Likewise.
9514 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9515 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9516 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9517 * tree-vect-slp.c (vect_slp_bb): Likewise.
9518 * doc/tm.texi: Regenerate.
9519 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9521 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9522 the vector size as a poly_uint64 rather than an unsigned int.
9523 (current_vector_size): Change from an unsigned int to a poly_uint64.
9524 (get_vectype_for_scalar_type): Update accordingly.
9525 * tree.h (build_truth_vector_type): Take the size and number of
9526 units as a poly_uint64 rather than an unsigned int.
9527 (build_vector_type): Add a temporary overload that takes
9528 the number of units as a poly_uint64 rather than an unsigned int.
9529 * tree.c (make_vector_type): Likewise.
9530 (build_truth_vector_type): Take the number of units as a poly_uint64
9531 rather than an unsigned int.
9533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9534 Alan Hayward <alan.hayward@arm.com>
9535 David Sherwood <david.sherwood@arm.com>
9537 * target.def (get_mask_mode): Take the number of units and length
9538 as poly_uint64s rather than unsigned ints.
9539 * targhooks.h (default_get_mask_mode): Update accordingly.
9540 * targhooks.c (default_get_mask_mode): Likewise.
9541 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9542 * doc/tm.texi: Regenerate.
9544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9545 Alan Hayward <alan.hayward@arm.com>
9546 David Sherwood <david.sherwood@arm.com>
9548 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9549 * omp-general.c (omp_max_vf): Likewise.
9550 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9551 (expand_omp_simd): Handle polynomial safelen.
9552 * omp-low.c (omplow_simd_context): Add a default constructor.
9553 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9554 (lower_rec_simd_input_clauses): Update accordingly.
9555 (lower_rec_input_clauses): Likewise.
9557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9558 Alan Hayward <alan.hayward@arm.com>
9559 David Sherwood <david.sherwood@arm.com>
9561 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9562 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9563 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9564 (vect_analyze_slp_cost): Likewise.
9565 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9566 (vect_model_load_cost): Likewise.
9568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9569 Alan Hayward <alan.hayward@arm.com>
9570 David Sherwood <david.sherwood@arm.com>
9572 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9573 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9574 from an unsigned int * to a poly_uint64_pod *.
9575 (calculate_unrolling_factor): New function.
9576 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9579 Alan Hayward <alan.hayward@arm.com>
9580 David Sherwood <david.sherwood@arm.com>
9582 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9583 from an unsigned int to a poly_uint64.
9584 (_loop_vec_info::slp_unrolling_factor): Likewise.
9585 (_loop_vec_info::vectorization_factor): Change from an int
9587 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9588 (vect_get_num_vectors): New function.
9589 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9590 (vect_get_num_copies): Use vect_get_num_vectors.
9591 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9592 to an unsigned int *.
9593 (vect_analyze_data_refs): Change min_vf from an int * to a
9595 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9596 than an unsigned HOST_WIDE_INT.
9597 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9598 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9599 to an unsigned int *.
9600 (vect_analyze_data_ref_dependences): Likewise.
9601 (vect_compute_data_ref_alignment): Handle polynomial vf.
9602 (vect_enhance_data_refs_alignment): Likewise.
9603 (vect_prune_runtime_alias_test_list): Likewise.
9604 (vect_shift_permute_load_chain): Likewise.
9605 (vect_supportable_dr_alignment): Likewise.
9606 (dependence_distance_ge_vf): Take the vectorization factor as a
9607 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9608 (vect_analyze_data_refs): Change min_vf from an int * to a
9610 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9611 vfm1 as a poly_uint64 rather than an int. Make the same change
9612 for the returned bound_scalar.
9613 (vect_gen_vector_loop_niters): Handle polynomial vf.
9614 (vect_do_peeling): Likewise. Update call to
9615 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9616 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9618 * tree-vect-loop.c (vect_determine_vectorization_factor)
9619 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9620 (vect_get_known_peeling_cost): Likewise.
9621 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9622 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9623 (vect_transform_loop): Likewise. Use the lowest possible VF when
9624 updating the upper bounds of the loop.
9625 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9627 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9628 polynomial unroll factors.
9629 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9630 (vect_make_slp_decision): Likewise.
9631 (vect_supported_load_permutation_p): Likewise, and polynomial
9633 (vect_analyze_slp_cost): Handle polynomial vf.
9634 (vect_slp_analyze_node_operations): Likewise.
9635 (vect_slp_analyze_bb_1): Likewise.
9636 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9637 than an unsigned HOST_WIDE_INT.
9638 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9639 (vectorizable_load): Handle polynomial vf.
9640 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9642 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9645 Alan Hayward <alan.hayward@arm.com>
9646 David Sherwood <david.sherwood@arm.com>
9648 * match.pd: Handle bit operations involving three constants
9649 and try to fold one pair.
9651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9653 * tree-vect-loop-manip.c: Include gimple-fold.h.
9654 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9655 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9656 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9657 Add a path that uses a step of VF instead of 1, but disable it
9659 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9660 and niters_no_overflow parameters. Update calls to
9661 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9662 Create a new SSA name if the latter choses to use a ste other
9663 than zero, and return it via niters_vector_mult_vf_var.
9664 * tree-vect-loop.c (vect_transform_loop): Update calls to
9665 vect_do_peeling, vect_gen_vector_loop_niters and
9666 slpeel_make_loop_iterate_ntimes.
9667 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9668 (vect_gen_vector_loop_niters): Update declarations after above changes.
9670 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9672 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9673 128-bit round to integer instructions.
9674 (ceil<mode>2): Likewise.
9675 (btrunc<mode>2): Likewise.
9676 (round<mode>2): Likewise.
9678 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9680 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9681 unaligned VSX load/store on P8/P9.
9682 (expand_block_clear): Allow the use of unaligned VSX
9683 load/store on P8/P9.
9685 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9687 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9689 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9690 swap associated with both a load and a store.
9692 2018-01-02 Andrew Waterman <andrew@sifive.com>
9694 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9695 * config/riscv/riscv.md (clear_cache): Use it.
9697 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9699 * web.c: Remove out-of-date comment.
9701 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9703 * expr.c (fixup_args_size_notes): Check that any existing
9704 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9705 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9706 (emit_single_push_insn): ...here.
9708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9710 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9711 (const_vector_encoded_nelts): New function.
9712 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9713 (const_vector_int_elt, const_vector_elt): Declare.
9714 * emit-rtl.c (const_vector_int_elt_1): New function.
9715 (const_vector_elt): Likewise.
9716 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9717 of CONST_VECTOR_ELT.
9719 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9721 * expr.c: Include rtx-vector-builder.h.
9722 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9723 directly on the tree encoding.
9724 (const_vector_from_tree): Likewise.
9725 * optabs.c: Include rtx-vector-builder.h.
9726 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9727 sequence of "u" values.
9728 * vec-perm-indices.c: Include rtx-vector-builder.h.
9729 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9730 directly on the vec_perm_indices encoding.
9732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9734 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9735 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9736 * rtx-vector-builder.h: New file.
9737 * rtx-vector-builder.c: Likewise.
9738 * rtl.h (rtx_def::u2): Add a const_vector field.
9739 (CONST_VECTOR_NPATTERNS): New macro.
9740 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9741 (CONST_VECTOR_DUPLICATE_P): Likewise.
9742 (CONST_VECTOR_STEPPED_P): Likewise.
9743 (CONST_VECTOR_ENCODED_ELT): Likewise.
9744 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9745 (unwrap_const_vec_duplicate): Likewise.
9746 (const_vec_series_p): Check for a non-duplicated vector encoding.
9747 Say that the function only returns true for integer vectors.
9748 * emit-rtl.c: Include rtx-vector-builder.h.
9749 (gen_const_vec_duplicate_1): Delete.
9750 (gen_const_vector): Call gen_const_vec_duplicate instead of
9751 gen_const_vec_duplicate_1.
9752 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9753 (gen_const_vec_duplicate): Use rtx_vector_builder.
9754 (gen_const_vec_series): Likewise.
9755 (gen_rtx_CONST_VECTOR): Likewise.
9756 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9757 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9758 Build a new vector rather than modifying a CONST_VECTOR in-place.
9759 (handle_special_swappables): Update call accordingly.
9760 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9761 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9762 Build a new vector rather than modifying a CONST_VECTOR in-place.
9763 (handle_special_swappables): Update call accordingly.
9765 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9767 * simplify-rtx.c (simplify_const_binary_operation): Use
9768 CONST_VECTOR_ELT instead of XVECEXP.
9770 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9772 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9773 the selector elements to be different from the data elements
9774 if the selector is a VECTOR_CST.
9775 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9776 ssizetype for the selector.
9778 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9780 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9781 before testing each element individually.
9782 * tree-vect-generic.c (lower_vec_perm): Likewise.
9784 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9786 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9787 * selftest-run-tests.c (selftest::run_tests): Call it.
9788 * vector-builder.h (vector_builder::operator ==): New function.
9789 (vector_builder::operator !=): Likewise.
9790 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9791 (vec_perm_indices::all_from_input_p): New function.
9792 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9793 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9794 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9795 instead of reading the VECTOR_CST directly. Detect whether both
9796 vector inputs are the same before constructing the vec_perm_indices,
9797 and update the number of inputs argument accordingly. Use the
9798 utility functions added above. Only construct sel2 if we need to.
9800 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9802 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9803 the broadcast of the low byte.
9804 (expand_mult_highpart): Use an explicit encoding for the permutes.
9805 * optabs-query.c (can_mult_highpart_p): Likewise.
9806 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9807 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9808 (vectorizable_bswap): Likewise.
9809 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9810 explicit encoding for the power-of-2 permutes.
9811 (vect_permute_store_chain): Likewise.
9812 (vect_grouped_load_supported): Likewise.
9813 (vect_permute_load_chain): Likewise.
9815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9817 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9818 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9819 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9820 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9821 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9822 (vect_gen_perm_mask_any): Likewise.
9824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9826 * int-vector-builder.h: New file.
9827 * vec-perm-indices.h: Include int-vector-builder.h.
9828 (vec_perm_indices): Redefine as an int_vector_builder.
9829 (auto_vec_perm_indices): Delete.
9830 (vec_perm_builder): Redefine as a stand-alone class.
9831 (vec_perm_indices::vec_perm_indices): New function.
9832 (vec_perm_indices::clamp): Likewise.
9833 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9834 (vec_perm_indices::new_vector): New function.
9835 (vec_perm_indices::new_expanded_vector): Update for new
9836 vec_perm_indices class.
9837 (vec_perm_indices::rotate_inputs): New function.
9838 (vec_perm_indices::all_in_range_p): Operate directly on the
9839 encoded form, without computing elided elements.
9840 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9841 encoding. Update for new vec_perm_indices class.
9842 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9843 the given vec_perm_builder.
9844 (expand_vec_perm_var): Update vec_perm_builder constructor.
9845 (expand_mult_highpart): Use vec_perm_builder instead of
9846 auto_vec_perm_indices.
9847 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9848 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9849 or double series encoding as appropriate.
9850 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9851 vec_perm_indices instead of auto_vec_perm_indices.
9852 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9853 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9854 (vect_permute_store_chain): Likewise.
9855 (vect_grouped_load_supported): Likewise.
9856 (vect_permute_load_chain): Likewise.
9857 (vect_shift_permute_load_chain): Likewise.
9858 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9859 (vect_transform_slp_perm_load): Likewise.
9860 (vect_schedule_slp_instance): Likewise.
9861 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9862 (vectorizable_mask_load_store): Likewise.
9863 (vectorizable_bswap): Likewise.
9864 (vectorizable_store): Likewise.
9865 (vectorizable_load): Likewise.
9866 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9867 vec_perm_indices instead of auto_vec_perm_indices. Use
9868 tree_to_vec_perm_builder to read the vector from a tree.
9869 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9870 vec_perm_builder instead of a vec_perm_indices.
9871 (have_whole_vector_shift): Use vec_perm_builder and
9872 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9873 truncation to calc_vec_perm_mask_for_shift.
9874 (vect_create_epilog_for_reduction): Likewise.
9875 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9876 from auto_vec_perm_indices to vec_perm_indices.
9877 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9878 instead of changing individual elements.
9879 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9880 the vector in d.perm.
9881 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9882 from auto_vec_perm_indices to vec_perm_indices.
9883 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9884 instead of changing individual elements.
9885 (arm_vectorize_vec_perm_const): Use new_vector to install
9886 the vector in d.perm.
9887 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9888 Update vec_perm_builder constructor.
9889 (rs6000_expand_interleave): Likewise.
9890 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9891 (rs6000_expand_interleave): Likewise.
9893 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9895 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9896 to qimode could truncate the indices.
9897 * optabs.c (expand_vec_perm_var): Likewise.
9899 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9901 * Makefile.in (OBJS): Add vec-perm-indices.o.
9902 * vec-perm-indices.h: New file.
9903 * vec-perm-indices.c: Likewise.
9904 * target.h (vec_perm_indices): Replace with a forward class
9906 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9907 * optabs.h: Include vec-perm-indices.h.
9908 (expand_vec_perm): Delete.
9909 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9910 (expand_vec_perm_const): Declare.
9911 * target.def (vec_perm_const_ok): Replace with...
9912 (vec_perm_const): ...this new hook.
9913 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9914 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9915 * doc/tm.texi: Regenerate.
9916 * optabs.def (vec_perm_const): Delete.
9917 * doc/md.texi (vec_perm_const): Likewise.
9918 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9919 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9920 expand_vec_perm for constant permutation vectors. Assert that
9921 the mode of variable permutation vectors is the integer equivalent
9922 of the mode that is being permuted.
9923 * optabs-query.h (selector_fits_mode_p): Declare.
9924 * optabs-query.c: Include vec-perm-indices.h.
9925 (selector_fits_mode_p): New function.
9926 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9927 is defined, instead of checking whether the vec_perm_const_optab
9928 exists. Use targetm.vectorize.vec_perm_const instead of
9929 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9930 fit in the vector mode before using a variable permute.
9931 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9932 vec_perm_indices instead of an rtx.
9933 (expand_vec_perm): Replace with...
9934 (expand_vec_perm_const): ...this new function. Take the selector
9935 as a vec_perm_indices rather than an rtx. Also take the mode of
9936 the selector. Update call to shift_amt_for_vec_perm_mask.
9937 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9938 Use vec_perm_indices::new_expanded_vector to expand the original
9939 selector into bytes. Check whether the indices fit in the vector
9940 mode before using a variable permute.
9941 (expand_vec_perm_var): Make global.
9942 (expand_mult_highpart): Use expand_vec_perm_const.
9943 * fold-const.c: Includes vec-perm-indices.h.
9944 * tree-ssa-forwprop.c: Likewise.
9945 * tree-vect-data-refs.c: Likewise.
9946 * tree-vect-generic.c: Likewise.
9947 * tree-vect-loop.c: Likewise.
9948 * tree-vect-slp.c: Likewise.
9949 * tree-vect-stmts.c: Likewise.
9950 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9952 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9953 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9954 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9955 (aarch64_vectorize_vec_perm_const): ...this new function.
9956 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9957 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9958 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9959 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9960 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9961 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9962 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9964 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9965 check for NEON modes.
9966 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9967 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9968 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9969 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9971 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9972 the old VEC_PERM_CONST conditions.
9973 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9974 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9975 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9976 (ia64_vectorize_vec_perm_const_ok): Merge into...
9977 (ia64_vectorize_vec_perm_const): ...this new function.
9978 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9979 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9980 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9981 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9982 * config/mips/mips.c (mips_expand_vec_perm_const)
9983 (mips_vectorize_vec_perm_const_ok): Merge into...
9984 (mips_vectorize_vec_perm_const): ...this new function.
9985 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9986 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9987 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9988 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9989 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9990 (rs6000_expand_vec_perm_const): Delete.
9991 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9993 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9994 (altivec_expand_vec_perm_const_le): Take each operand individually.
9995 Operate on constant selectors rather than rtxes.
9996 (altivec_expand_vec_perm_const): Likewise. Update call to
9997 altivec_expand_vec_perm_const_le.
9998 (rs6000_expand_vec_perm_const): Delete.
9999 (rs6000_vectorize_vec_perm_const_ok): Delete.
10000 (rs6000_vectorize_vec_perm_const): New function.
10001 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10002 an element count and rtx array.
10003 (rs6000_expand_extract_even): Update call accordingly.
10004 (rs6000_expand_interleave): Likewise.
10005 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10006 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10007 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10008 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10009 (rs6000_expand_vec_perm_const): Delete.
10010 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10011 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10012 (altivec_expand_vec_perm_const_le): Take each operand individually.
10013 Operate on constant selectors rather than rtxes.
10014 (altivec_expand_vec_perm_const): Likewise. Update call to
10015 altivec_expand_vec_perm_const_le.
10016 (rs6000_expand_vec_perm_const): Delete.
10017 (rs6000_vectorize_vec_perm_const_ok): Delete.
10018 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10019 reference to the SPE evmerge intructions.
10020 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10021 an element count and rtx array.
10022 (rs6000_expand_extract_even): Update call accordingly.
10023 (rs6000_expand_interleave): Likewise.
10024 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10025 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10027 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10029 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10031 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10032 vector mode and that that mode matches the mode of the data
10034 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10035 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10036 directly using expand_vec_perm_1 when forcing selectors into
10038 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10040 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10042 * optabs-query.h (can_vec_perm_p): Delete.
10043 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10044 * optabs-query.c (can_vec_perm_p): Split into...
10045 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10046 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10047 particular selector is valid.
10048 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10049 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10050 (vect_grouped_load_supported): Likewise.
10051 (vect_shift_permute_load_chain): Likewise.
10052 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10053 (vect_transform_slp_perm_load): Likewise.
10054 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10055 (vectorizable_bswap): Likewise.
10056 (vect_gen_perm_mask_checked): Likewise.
10057 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10058 implementations of variable permutation vectors into account
10059 when deciding which selector to use.
10060 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10061 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10062 with a false third argument.
10063 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10064 to test whether the constant selector is valid and can_vec_perm_var_p
10065 to test whether a variable selector is valid.
10067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10069 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10070 * optabs-query.c (can_vec_perm_p): Likewise.
10071 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10072 instead of vec_perm_indices.
10073 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10074 (vect_gen_perm_mask_checked): Likewise,
10075 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10076 (vect_gen_perm_mask_checked): Likewise,
10078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10080 * optabs-query.h (qimode_for_vec_perm): Declare.
10081 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10082 (qimode_for_vec_perm): ...this new function.
10083 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10085 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10087 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10088 does not have a conditional at the top.
10090 2018-01-02 Richard Biener <rguenther@suse.de>
10092 * ipa-inline.c (big_speedup_p): Fix expression.
10094 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10097 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10100 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10104 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10105 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10106 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10107 cond_taken_branch_cost 3->4.
10109 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10111 PR tree-optimization/83581
10112 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10113 TODO_cleanup_cfg if any changes have been made.
10115 PR middle-end/83608
10116 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10117 convert_modes if target mode has the right side, but different mode
10120 PR middle-end/83609
10121 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10122 last argument when extracting from CONCAT. If either from_real or
10123 from_imag is NULL, use expansion through memory. If result is not
10124 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10125 the parts directly to inner mode, if even that fails, use expansion
10128 PR middle-end/83623
10129 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10130 check for bswap in mode rather than HImode and use that in expand_unop
10133 Copyright (C) 2018 Free Software Foundation, Inc.
10135 Copying and distribution of this file, with or without modification,
10136 are permitted in any medium without royalty provided the copyright
10137 notice and this notice are preserved.