1 2018-05-09 Jim Wilson <jimw@sifive.com>
4 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5 * config/riscv/t-withmultilib: New.
6 * config/riscv/withmultilib.h: New.
7 * doc/install.texi: Document RISC-V --with-multilib-list support.
9 2018-05-09 Richard Biener <rguenther@suse.de>
11 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13 (vect_bb_vectorization_profitable_p): Adjust. Compute
14 actual scalar cost using the cost vector and the add_stmt_cost
17 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
19 PR rtl-optimization/85645
20 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
21 in the REG_CFA_REGISTER note for LR, don't leave it empty.
23 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
25 PR rtl-optimization/85645
26 * shrink-wrap.c (spread_components): Return a boolean saying if
28 (try_shrink_wrapping_separate): Iterate spread_components until
29 nothing changes anymore.
31 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
33 PR rtl-optimization/85645
34 * regrename.c (build_def_use): Also kill the chains that include the
35 destination of a REG_CFA_REGISTER note.
37 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
39 PR rtl-optimization/85645
40 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
41 insn that has a REG_CFA_REGISTER note.
43 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
45 * cfgexpand.c (expand_clobber): New function.
46 (expand_gimple_stmt_1): Use it.
47 * tree-vect-stmts.c (vect_clobber_variable): New function,
49 (vectorizable_simd_clone_call): ...here.
50 (vectorizable_store): Emit a clobber either side of an
51 IFN_STORE_LANES sequence.
52 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
54 2018-05-09 Tom de Vries <tom@codesourcery.com>
57 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
58 (define_insn "trap_if_false"): Add exit after trap.
60 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
62 PR rtl-optimization/85638
63 * bb-reorder.c: Include common/common-target.h.
64 (create_forwarder_block): New function extracted from...
65 (fix_up_crossing_landing_pad): ...here. Rename into...
66 (dw2_fix_up_crossing_landing_pad): ...this.
67 (sjlj_fix_up_crossing_landing_pad): New function.
68 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
69 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
70 from both partitions and exit the loop after one iteration.
72 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
75 * doc/extend.texi (PowerPC Built-in Functions): Rename this
77 (Basic PowerPC Built-in Functions): The new name of the
78 subsection previously known as "PowerPC Built-in Functions".
79 (Basic PowerPC Built-in Functions Available on all Configurations):
81 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
83 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
85 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
87 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
90 2018-05-08 Jim Wilson <jimw@sifive.com>
92 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
93 (LD_EMUL_SUFFIX): New.
96 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
98 * doc/extend.texi (PowerPC Built-in Functions): Rename this
100 (Basic PowerPC Built-in Functions): The new name of the
101 subsection previously known as "PowerPC Built-in Functions".
102 (Basic PowerPC Built-in Functions Available on all Configurations):
104 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
106 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
108 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
110 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
113 2018-05-08 Jakub Jelinek <jakub@redhat.com>
116 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
117 after cmpelim optimization.
119 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
121 * config.gcc: Support "goldmont".
122 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
123 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
125 * config/i386/i386.c (m_GOLDMONT): Define.
126 (processor_target_table): Add "goldmont".
127 (PTA_GOLDMONT): Define.
128 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
129 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
130 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
131 (fold_builtin_cpu): Add "goldmont".
132 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
133 (ix86_option_override_internal): Add "goldmont".
134 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
135 (processor_type): Add PROCESSOR_GOLDMONT.
136 * config/i386/i386.md: Add CPU "glm".
137 * config/i386/glm.md: New file.
138 * config/i386/x86-tune.def: Add m_GOLDMONT.
139 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
141 2018-05-08 Jakub Jelinek <jakub@redhat.com>
144 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
146 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
147 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
148 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
151 * config/i386/i386.c (ix86_fold_builtin): Handle
152 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
155 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
156 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
158 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
161 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
162 (check_arch): Likewise.
163 (check_fpu): Return the result rather than printing it.
164 (end arch): Fix operator precedence.
166 (END): Print the result from check_fpu.
168 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
169 Alan Hayward <alan.hayward@arm.com>
170 David Sherwood <david.sherwood@arm.com>
172 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
173 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
174 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
175 (*fcmuo<mode>_and): New patterns.
177 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
179 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
180 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
181 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
182 (cmp_op, sve_imm_con): New code attributes.
183 (SVE_COND_INT_CMP, imm_con): Delete.
184 (cmp_op): Remove above unspecs from int attribute.
185 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
187 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
188 comparison-specific unspecs.
189 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
190 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
191 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
192 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
193 (*vec_fcm<cmp_op><mode>): Rename to...
194 (*fcm<cmp_op><mode>): ...this and adjust likewise.
195 (*vec_fcmuo<mode>): Rename to...
196 (*fcmuo<mode>): ...this and adjust likewise.
197 (*pred_fcm<cmp_op><mode>): New pattern.
198 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
199 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
201 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
203 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
204 (aarch64_emit_sve_predicated_cond): New function.
205 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
206 (aarch64_emit_unspec_cond_or): Replace with...
207 (aarch64_emit_sve_or_conds): ...this new function. Use
208 aarch64_emit_sve_ptrue_op for the individual comparisons and
209 aarch64_emit_binop to OR them together.
210 (aarch64_emit_inverted_unspec_cond): Replace with...
211 (aarch64_emit_sve_inverted_cond): ...this new function. Use
212 aarch64_emit_sve_ptrue_op for the comparison and
213 aarch64_emit_unop to invert the result.
214 (aarch64_expand_sve_vec_cmp_float): Update after the above
215 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
217 2018-05-07 Nathan Sidwell <nathan@acm.org>
219 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
220 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
221 (Backwards Compatibility): Likewise.
223 2018-05-07 Luis Machado <luis.machado@linaro.org>
227 2018-05-07 Luis Machado <luis.machado@linaro.org>
229 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
230 <prefetch_dynamic_strides>: New const bool field.
231 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
232 prefetch_dynamic_strides.
233 (exynosm1_prefetch_tune): Likewise.
234 (thunderxt88_prefetch_tune): Likewise.
235 (thunderx_prefetch_tune): Likewise.
236 (thunderx2t99_prefetch_tune): Likewise.
237 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
238 (aarch64_override_options_internal): Update to set
239 PARAM_PREFETCH_DYNAMIC_STRIDES.
240 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
241 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
242 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
243 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
244 prefetch-dynamic-strides setting.
246 2018-05-07 Luis Machado <luis.machado@linaro.org>
248 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
249 <minimum_stride>: New const int field.
250 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
251 minimum_stride field.
252 (exynosm1_prefetch_tune): Likewise.
253 (thunderxt88_prefetch_tune): Likewise.
254 (thunderx_prefetch_tune): Likewise.
255 (thunderx2t99_prefetch_tune): Likewise.
256 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
257 (aarch64_override_options_internal): Update to set
258 PARAM_PREFETCH_MINIMUM_STRIDE.
259 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
260 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
261 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
262 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
263 stride is constant and is below the minimum stride threshold.
265 2018-05-07 Luis Machado <luis.machado@linaro.org>
267 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
270 2018-05-07 Luis Machado <luis.machado@linaro.org>
272 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
273 <prefetch_dynamic_strides>: New const bool field.
274 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
275 prefetch_dynamic_strides.
276 (exynosm1_prefetch_tune): Likewise.
277 (thunderxt88_prefetch_tune): Likewise.
278 (thunderx_prefetch_tune): Likewise.
279 (thunderx2t99_prefetch_tune): Likewise.
280 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
281 (aarch64_override_options_internal): Update to set
282 PARAM_PREFETCH_DYNAMIC_STRIDES.
283 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
284 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
285 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
286 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
287 prefetch-dynamic-strides setting.
289 2018-05-07 Luis Machado <luis.machado@linaro.org>
291 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
292 <minimum_stride>: New const int field.
293 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
294 minimum_stride field.
295 (exynosm1_prefetch_tune): Likewise.
296 (thunderxt88_prefetch_tune): Likewise.
297 (thunderx_prefetch_tune): Likewise.
298 (thunderx2t99_prefetch_tune): Likewise.
299 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
300 (aarch64_override_options_internal): Update to set
301 PARAM_PREFETCH_MINIMUM_STRIDE.
302 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
303 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
304 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
305 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
306 stride is constant and is below the minimum stride threshold.
308 2018-05-06 Jakub Jelinek <jakub@redhat.com>
311 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
312 the type is addressable. Don't force op into register if it has
315 2018-05-05 Roland McGrath <mcgrathr@google.com>
318 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
319 any section for which we don't know a specific type it should have,
320 regardless of name. Previously this was done only for the exact
321 names ".init_array", ".fini_array", and ".preinit_array".
322 (default_elf_asm_named_section): Add comment about
323 relationship with default_section_type_flags and SECTION_NOTYPE.
324 (get_section): Don't consider it a type conflict if one side has
325 SECTION_NOTYPE and the other doesn't, as long as neither has the
326 SECTION_BSS et al used in the default_section_type_flags logic.
328 2018-05-05 Tom de Vries <tom@codesourcery.com>
331 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
332 (workaround_barsyncs): New function.
333 (nvptx_reorg): Use workaround_barsyncs.
334 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
335 (define_expand "nvptx_membar_cta"): New define_expand.
336 (define_insn "*nvptx_membar_cta"): New insn.
338 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
340 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
341 To improve optimization opportunities.
342 * builtin-types.def: The new needed builtin types for the above.
344 2018-05-04 Richard Biener <rguenther@suse.de>
346 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
347 * gimple-ssa-store-merging.c
348 (imm_store_chain_info::output_merged_store): Remove redundant create,
349 release split_store vector contents on failure.
350 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
351 scalar stmt vector on cache hit.
353 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
355 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
357 * config.gcc (powerpc-xilinx-eabi*): Remove.
358 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
360 (fusion_addis_mem_combo_load): Ditto.
361 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
363 (rs6000_cpu_cpp_builtins): Ditto.
364 * config/rs6000/rs6000-linux.c
365 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
366 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
367 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
369 (rs6000_setup_reg_addr_masks): Ditto.
370 (rs6000_init_hard_regno_mode_ok): Ditto.
371 (rs6000_option_override_internal): Ditto.
372 (legitimate_lo_sum_address_p): Ditto.
373 (rs6000_legitimize_address): Ditto.
374 (rs6000_legitimize_reload_address): Ditto.
375 (rs6000_legitimate_address_p): Ditto.
376 (abi_v4_pass_in_fpr): Ditto.
377 (setup_incoming_varargs): Ditto.
378 (rs6000_gimplify_va_arg): Ditto.
379 (rs6000_split_multireg_move): Ditto.
380 (rs6000_savres_strategy): Ditto.
381 (rs6000_emit_prologue_components): Ditto.
382 (rs6000_emit_epilogue_components): Ditto.
383 (rs6000_emit_prologue): Ditto.
384 (rs6000_emit_epilogue): Ditto.
385 (rs6000_elf_file_end): Ditto.
386 (rs6000_function_value): Ditto.
387 (rs6000_libcall_value): Ditto.
388 * config/rs6000/rs6000.h: Ditto.
389 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
390 (TARGET_MINMAX): ... this. New.
391 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
392 * config/rs6000/rs6000.md: Remove Xilinx FP support.
393 (*movsi_internal1_single): Delete.
394 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
395 mfpu=, mxilinx-fpu): Delete.
396 * config/rs6000/singlefp.h: Delete.
397 * config/rs6000/sysv4.h: Remove Xilinx FP support.
398 * config/rs6000/t-rs6000: Ditto.
399 * config/rs6000/t-xilinx: Delete.
400 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
401 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
402 (VStype_simple): Delete.
403 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
404 * config/rs6000/xfpu.h: Delete.
405 * config/rs6000/xfpu.md: Delete.
406 * config/rs6000/xilinx.h: Delete.
407 * config/rs6000/xilinx.opt: Delete.
408 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
409 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
411 2018-05-04 Tom de Vries <tom@codesourcery.com>
414 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
417 2018-05-04 Richard Biener <rguenther@suse.de>
420 * tree-complex.c (update_complex_assignment): We are always in SSA form.
421 (expand_complex_div_wide): Likewise.
422 (expand_complex_operations_1): Likewise.
423 (expand_complex_libcall): Preserve EH info of the original stmt.
424 (tree_lower_complex): Handle removed blocks.
425 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
426 on complex multiplication and division libcall builtins.
428 2018-05-04 Richard Biener <rguenther@suse.de>
431 * fold-const.c (negate_expr_p): Restrict negation of operand
432 zero of a division to when we know that can happen without
434 (fold_negate_expr_1): Likewise.
436 2018-05-04 Jakub Jelinek <jakub@redhat.com>
439 * real.h (real_nextafter): Declare.
440 * real.c (real_nextafter): New function.
441 * fold-const-call.c (fold_const_nextafter): New function.
442 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
444 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
445 even when arg1_mode is different from arg0_mode.
447 2018-05-03 Nathan Sidwell <nathan@acm.org>
449 * doc/extend.texi (Deprecated Features): Remove
451 (Backwards Compatibility): Likewise.
452 * doc/invoke.texi (C++ Language Options): Likewise.
453 (C++ Dialect Options): Likewise.
455 2018-05-03 Jakub Jelinek <jakub@redhat.com>
458 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
459 _mm512_mask_mullox_epi64): New intrinsics.
461 2018-05-03 Tom de Vries <tom@codesourcery.com>
464 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
465 dump files): Add offload-tree.
467 2018-05-03 Richard Biener <rguenther@suse.de>
469 PR tree-optimization/85615
470 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
471 to loops not nested in BBs loop father to avoid creating multi-entry
474 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
476 PR tree-optimization/70291
477 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
478 arguments. Change return type to tree. Emit libcall as a new
479 statement rather than replacing existing one when inplace_p is true.
480 (expand_complex_multiplication_components): New function.
481 (expand_complex_multiplication): Expand floating-point complex
482 multiplication using the above.
483 (expand_complex_division): Rename inner_type parameter to type.
484 Update expand_complex_libcall call-site.
485 (expand_complex_operations_1): Update expand_complex_multiplication
486 and expand_complex_division call-sites.
488 2018-05-02 Jakub Jelinek <jakub@redhat.com>
491 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
492 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
493 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
494 the highest significant bit of the shift count mask is clear. In
495 check whether and[sq]i3 is needed verify that all significant bits
496 of the shift count other than the highest are set.
498 2018-05-02 Tom de Vries <tom@codesourcery.com>
501 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
502 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
503 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
504 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
505 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
506 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
507 __builtin_goacc_parlevel_size.
509 2018-05-02 Richard Biener <rguenther@suse.de>
511 PR tree-optimization/85597
512 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
513 do not use split vect_get_vec_defs call but call vect_get_slp_defs
516 2018-05-02 Tom de Vries <tom@codesourcery.com>
519 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
520 dump files): Add ltrans-tree.
522 2018-05-02 Tom de Vries <tom@codesourcery.com>
525 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
526 dump files): Add wpa-ipa.
528 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
530 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
531 powerpc*-*-linux*paired* target.
532 * config/rs6000/750cl.h: Delete.
533 * config/rs6000/paired.h: Delete.
534 * config/rs6000/paired.md: Delete.
535 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
537 * config/rs6000/rs6000-builtin.def: Remove paired float support.
538 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
539 comment. Remove paired float support.
540 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
541 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
543 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
544 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
546 * config/rs6000/rs6000.c: Remove paired float support.
547 (paired_expand_vector_init, paired_expand_vector_move,
548 paired_emit_vector_compare, paired_emit_vector_cond_expr,
549 (paired_expand_lv_builtin, paired_expand_stv_builtin,
550 paired_expand_builtin, paired_expand_predicate_builtin,
551 paired_init_builtins): Delete.
552 * config/rs6000/rs6000.h: Remove paired float support.
553 * config/rs6000/rs6000.md: Remove paired float support.
554 (move_from_CR_ov_bit): Delete.
555 * config/rs6000/rs6000.opt (mpaired): Delete.
556 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
557 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
559 2018-05-02 Richard Biener <rguenther@suse.de>
562 * gimplify.c (gimplify_save_expr): When in SSA form allow
563 SAVE_EXPRs to compute to SSA vars.
565 2018-05-02 Jakub Jelinek <jakub@redhat.com>
568 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
569 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
570 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
571 clobber operands[2], instead use a new pseudo. Formatting fixes.
573 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
575 PR tree-optimization/85586
576 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
577 exit early for statements in the same group if the accesses are
580 2018-05-02 Tom de Vries <tom@codesourcery.com>
583 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
586 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
588 PR tree-optimization/85143
589 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
591 2018-05-01 Tom de Vries <tom@codesourcery.com>
594 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
595 not found" error message.
597 2018-05-01 Tom de Vries <tom@codesourcery.com>
600 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
601 * vec.c (test_ordered_remove_if): New function.
602 (vec_c_tests): Call test_ordered_remove_if.
603 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
604 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
605 * tree-vect-patterns.c (vect_pattern_recog_1): Use
606 VEC_ORDERED_REMOVE_IF.
608 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
610 PR tree-optimization/82665
611 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
612 pointer subtraction where arguments come from a memchr call.
614 2018-05-01 Jakub Jelinek <jakub@redhat.com>
616 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
617 --push-state --as-needed and --pop-state instead of --as-needed and
618 --no-as-needed if ld supports it.
619 * configure: Regenerated.
622 * doc/install.texi2html: Replace _002d with - and _002a with * in
623 generated html files using sed.
625 2018-04-30 David Malcolm <dmalcolm@redhat.com>
628 * gcc-rich-location.c (blank_line_before_p): New function.
629 (use_new_line): New function.
630 (gcc_rich_location::add_fixit_insert_formatted): New function.
631 * gcc-rich-location.h
632 (gcc_rich_location::add_fixit_insert_formatted): New function.
634 2018-04-30 David Malcolm <dmalcolm@redhat.com>
636 * selftest.c (assert_streq): Rename "expected" and "actual" to
637 "val1" and "val2". Extend NULL-handling to cover both inputs
638 symmetrically, while still requiring both to be non-NULL for a pass.
639 * selftest.h (assert_streq): Rename "expected" and "actual" to
641 (ASSERT_EQ): Likewise.
642 (ASSERT_EQ_AT): Likewise.
643 (ASSERT_KNOWN_EQ): Likewise.
644 (ASSERT_KNOWN_EQ_AT): Likewise.
645 (ASSERT_NE): Likewise.
646 (ASSERT_MAYBE_NE): Likewise.
647 (ASSERT_MAYBE_NE_AT): Likewise.
648 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
649 the assertion to pass.
650 (ASSERT_STREQ_AT): Likewise.
652 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
654 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
655 interaction with -pie.
657 2018-04-30 David Malcolm <dmalcolm@redhat.com>
659 * selftest.h: Fix alphabetization of per-source-file selftest
662 2018-04-30 Jason Merrill <jason@redhat.com>
664 PR c++/61982 - dead stores to destroyed objects.
665 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
668 2018-04-30 Jason Merrill <jason@redhat.com>
670 * tree.c (build_clobber): New.
671 * tree.h: Declare it.
672 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
674 2018-04-30 David Malcolm <dmalcolm@redhat.com>
676 * diagnostic-show-locus.c (layout::layout): Update for
677 location_get_source_line returning a char_span.
678 (struct char_span): Move to input.h.
679 (struct correction): Update for fields in char_span becoming
681 (struct source_line): Update for location_get_source_line
682 returning a char_span.
683 (layout::print_line): Likewise.
684 * edit-context.c (edited_file::print_content): Likewise.
685 (edited_file::print_diff_hunk): Likewise.
686 (edited_file::print_run_of_changed_lines): Likewise.
687 (edited_file::get_num_lines): Likewise.
688 (edited_line::edited_line): Likewise.
689 * final.c (asm_show_source): Likewise.
690 * input.c (location_get_source_line): Convert return type
691 from const char * to char_span, losing the final "line_len"
693 (dump_location_info): Update for the above.
694 (get_substring_ranges_for_loc): Likewise. Use a char_span
695 when handling the literal within the line.
696 (test_reading_source_line): Update for location_get_source_line
697 returning a char_span.
698 * input.h (class char_span): Move here from
699 diagnostic-show-locus.c, converting from a struct to a class.
700 Make data members private.
701 (char_span::operator bool): New.
702 (char_span::length): New.
703 (char_span::get_buffer): New.
704 (char_span::operator[]): New.
705 (char_span::subspan): Make const.
706 (char_span::xstrdup): New.
707 (location_get_source_line): Convert return type from const char *
708 to char_span, losing the final "line_size" param.
710 2018-04-30 Jan Hubicka <jh@suse.cz>
712 * lto-wrapper.c (ltrans_priorities): New static var.
714 (run_gcc): Read priorities and if doing parallel build order
715 the Makefile by them.
717 2018-04-30 David Malcolm <dmalcolm@redhat.com>
719 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
721 2018-04-30 Richard Biener <rguenther@suse.de>
723 * tree-cfg.c (verify_address): Remove base argument, add
724 flag whether to check TREE_ADDRESSABLE and do that.
725 (verify_expr): Remove.
726 (verify_types_in_gimple_reference): Add pieces from verify_expr.
727 (verify_gimple_assign_single): Likewise.
728 (verify_gimple_switch): Likewise.
729 (verify_expr_location_1): Dereference tp once. Add (disabled)
730 piece from verify_expr.
731 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
733 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
735 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
737 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
739 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
740 (small_data_pattern): Likewise.
741 (arc_rewrite_small_data): Likewise.
742 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
743 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
744 (get_symbol_alignment): New function.
745 (legitimate_small_data_address_p): Likewise.
746 (legitimate_scaled_address): Update, call
747 legitimate_small_data_address_p.
748 (output_sdata): New static variable.
749 (arc_print_operand): Update how we handle small data operands.
750 (arc_print_operand_address): Likewise.
751 (arc_legitimate_address_p): Update, use
752 legitimate_small_data_address_p.
753 (arc_rewrite_small_data_p): Remove.
754 (arc_rewrite_small_data_1): Likewise.
755 (arc_rewrite_small_data): Likewise.
756 (small_data_pattern): Likewise.
757 (compact_sda_memory_operand): Update to use
758 legitimate_small_data_address_p and get_symbol_alignment.
759 (prepare_move_operands): Don't rewite sdata pattern.
760 (prepare_extend_operands): Remove.
761 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
763 (zero_extendqisi2): Likewise.
764 (zero_extendhisi2): Likewise.
765 (extendqihi2): Likewise.
766 (extendqisi2): Likewise.
767 (extendhisi2): Likewise.
771 * config/arc/constraints.md (Usd): Change it to memory constraint.
773 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
775 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
776 as source of std instructions.
777 * config/arc/arc.md (movsi_insn): Update pattern predicate to
778 allow 6-bit constants as source for store instructions.
779 (movdi_insn): Update instruction pattern to allow 6-bit constants
780 as source for store instructions.
782 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
784 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
786 2018-04-30 Nathan Sidwell <nathan@acm.org>
787 Sandra Loosemore <sandra@codesourcery.com>
789 * dumpfile.c (dump_open): Allow '-' for stdout.
790 * doc/invoke.texi (Developer Options): Document dump filename
791 determination early. Document stdin/stdout selection.
793 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
795 Microblaze Target: PIC data text relative
797 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
798 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
800 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
801 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
802 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
803 New addressing mode for data-text relative position indepenedent code.
804 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
805 'ADDRESS_SYMBOLIC_TXT_REL'.
806 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
807 (microblaze_legitimate_pic_operand): Exclude function calls from
808 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
809 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
811 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
812 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
813 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
814 for 'address + offset'.
815 (microblaze_expand_prologue): Add new function prologue call for
817 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
818 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
819 table in case of TARGET_PIC_DATA_TEXT_REL.
820 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
821 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
822 Add new macros 'UNSPEC_TEXT',
823 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
824 + exclude function calls from 'UNSPEC_PLT' in case of data text
826 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
827 new target hook for generating address diff vector tables in case of
829 * doc/tm.texi : Regenerate.
830 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
831 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
832 of addr diff vector generation.
833 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
834 target hook definition.
835 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
836 Add default function for generate_pic_addr_diff_vec -> flag_pic.
837 * doc/invoke.texi (Add new pic option): Add new microblaze pic
838 option for data text relative.
840 2018-04-30 Richard Biener <rguenther@suse.de>
842 * tree-chrec.h (evolution_function_is_constant_p): Remove
844 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
846 2018-04-30 Richard Biener <rguenther@suse.de>
849 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
851 2018-04-30 Richard Biener <rguenther@suse.de>
853 PR tree-optimization/28364
854 PR tree-optimization/85275
855 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
856 copying first exit test.
858 2018-04-28 Mark Wielaard <mark@klomp.org>
860 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
862 (dwarf_AT): Handle DW_AT_addr_base.
863 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
865 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
868 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
869 (*ashl<dwi>3_doubleword_mask_1): Ditto.
870 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
871 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
873 2018-04-28 Richard Biener <rguenther@suse.de>
875 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
876 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
877 to reflect use. Only add interesting stmts.
879 2018-04-27 Martin Jambor <mjambor@suse.cz>
882 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
883 the jump function allows for passing through aggregate values.
885 2018-04-27 David Malcolm <dmalcolm@redhat.com>
887 * input.h (in_system_header_at): Convert from macro to inline
889 (from_macro_expansion_at): Likewise.
890 (from_macro_definition_at): Likewise.
892 2018-04-27 Jeff Law <law@redhat.com>
894 * config.gcc: Mark tile* targets as deprecated/obsolete.
896 2018-04-27 Richard Biener <rguenther@suse.de>
898 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
901 2018-04-27 Richard Biener <rguenther@suse.de>
903 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
905 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
907 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
908 with Yd constraint. Set "preferred_for_speed" attribute from
909 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
911 (*movdi_internal): Ditto.
912 (movti_interunit splitters): Remove
913 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
914 (movdi_interunit splitters): Ditto.
915 * config/i386/constraints.md (Ye): Remove.
916 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
918 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
921 * config/aarch64/constraints.md (Usg): Limit to 31.
924 2018-04-27 Jakub Jelinek <jakub@redhat.com>
926 PR tree-optimization/85529
927 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
928 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
929 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
930 zero extension or masking of the MSB bit.
931 (optimize_range_tests): Add FIRST_BB argument, pass it through
932 to optimize_range_tests_var_bound.
933 (maybe_optimize_range_tests, reassociate_bb): Adjust
934 optimize_range_tests callers.
936 2018-04-26 Richard Biener <rguenther@suse.de>
937 Jakub Jelinek <jakub@redhat.com>
939 * cgraph.h (symbol_table): Just declare debug method here.
940 * symtab.c (symbol_table::debug): Define.
942 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
944 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
946 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
948 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
949 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
950 (*movdi_internal): Substitute Yi and Yj constraint with x
951 and Ym and Yn constraint with y constraint. Update "isa"
952 attribute and set "preferred_for_speed" attribute from
953 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
954 (*movsi_internal): Ditto.
955 (*movdf_internal): Ditto.
956 (*movsf_internal): Ditto.
957 (*zero_extendsidi2): Ditto.
958 * config/i386/sse.md (vec_set<mode>_0): Ditto.
959 (sse2_loadld): Ditto.
960 (*vec_extract<ssevecmodelower>_0): Ditto.
961 (*vec_extractv4si_0_zext_sse4): Ditto.
962 (vec_concatv2di): Ditto.
963 (*vec_dup<mode>): Ditto.
964 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
965 * config/i386/constraints.md (Yi): Remove.
970 2018-04-26 Nathan Sidwell <nathan@acm.org>
972 * dumpfile.c (dump_open): New.
973 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
974 (dump_finish): Detect stdio/stderr by value not name.
976 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
978 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
980 2018-04-26 Tom de Vries <tom@codesourcery.com>
983 * config/nvptx/nvptx.c (verify_neutering_jumps)
984 (verify_neutering_labels): New function
985 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
987 2018-04-26 Tom de Vries <tom@codesourcery.com>
990 * config/nvptx/nvptx.c (needs_neutering_p): New function.
991 (nvptx_single): Use needs_neutering_p to skip over insns that do not
994 2018-04-26 Richard Biener <rguenther@suse.de>
995 Tom de Vries <tom@codesourcery.com>
998 * lto-streamer-out.c (output_function): Fixup loops if required to match
999 discovery done in the reader.
1001 2018-04-26 Richard Biener <rguenther@suse.de>
1003 PR tree-optimization/85116
1004 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1005 have a loop exit from the single latch predecessor. Remove
1006 case of header with just condition.
1007 (ch_base::copy_headers): Exclude infinite loops from any
1009 (pass_ch::execute): Record exits.
1011 2018-04-26 Richard Biener <rguenther@suse.de>
1013 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1014 prologue cost vector and pass it to vect_get_load_cost.
1015 (vect_get_peeling_costs_all_drs): Likewise.
1016 (vect_peeling_hash_get_lowest_cost): Likewise.
1017 (vect_enhance_data_refs_alignment): Likewise.
1019 2018-04-26 Richard Biener <rguenther@suse.de>
1022 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1023 checking of integer<->pointer conversions.
1024 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1025 sign-/zero-extending pointer types.
1026 (expand_omp_for_static_chunk): Likewise.
1028 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1029 Jean Lee <xiaoyur347@gmail.com>
1031 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1032 (TARGET_ASAN_SHADOW_OFFSET): Define.
1033 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1034 true for -fsanitize=address.
1036 2018-04-25 Mark Wielaard <mark@klomp.org>
1038 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1041 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1043 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1044 than "alu", remove explicit "memory" and "imm_disp" attributes.
1045 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1048 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1049 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1050 gen_lowpart_no_emit.
1052 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1055 * config/i386/i386.c (ix86_expand_builtin): Change memory
1056 operand to XI, extend p0 to Pmode.
1057 * config/i386/i386.md: Change unspec volatile and operand
1058 1 mode to XI, change operand 0 mode to P.
1060 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1062 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1063 GET_MODE_MASK before any checking.
1064 (nds32_can_use_bset_p): Likewise.
1065 (nds32_can_use_btgl_p): Likewise.
1067 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1069 * config/nds32/nds32-doubleword.md: New define_split pattern for
1070 illegal register number.
1072 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1074 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1076 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1078 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1080 2018-04-25 Richard Biener <rguenther@suse.de>
1082 * lto-streamer.h (LTO_major_version): Bump to 8.
1084 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1086 * BASE-VER: Set to 9.0.0.
1088 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1090 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1091 in __abskf2 and __powikf2.
1093 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1096 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1097 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1098 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1099 Use the above on operand 2. Reindent.
1100 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1102 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1105 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1107 * config/i386/i386.opt (mcet): Removed.
1108 * doc/install.texi: Remove -mcet documentation.
1109 * doc/invoke.texi: Likewise.
1111 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1114 * doc/install.texi: Remove -mcet from bootstrap-cet.
1116 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1119 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1120 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1124 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1125 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1126 containing a CONST_VECTOR.
1128 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1130 * doc/install.texi: Update newlib dependency for nvptx.
1132 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1135 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1136 instead of INTVAL when shifting x left.
1138 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1140 PR tree-optimization/85478
1141 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1142 vect_grouped_store_supported for single element vectors.
1144 2018-04-24 Richard Biener <rguenther@suse.de>
1147 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1148 load cost increase to the case of non-constant step.
1150 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1153 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1154 destination if any_malformed_asm.
1156 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1159 * expr.c (store_field): In the bitfield case, if the value comes from
1160 a function call and is returned in registers by means of a PARALLEL,
1161 do not change the mode of the temporary unless BLKmode and VOIDmode.
1163 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1165 PR rtl-optimization/85423
1166 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1167 dependencies to debug insns when the previous insn is non-debug.
1169 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1171 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1172 enums into a single definition.
1173 (fls): Fix predicates and printing.
1176 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1178 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1179 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1180 and short u6 immediate.
1181 (check_if_valid_sleep_operand): Remove.
1182 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1184 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1186 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1187 flag_always_save_lp condition.
1188 * config/nds32/nds32.opt (malways-save-lp): New option.
1190 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1192 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1193 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1194 * config/nds32/nds32.h
1195 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1196 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1198 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1200 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1201 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1203 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1204 Chung-Ju Wu <jasonwucj@gmail.com>
1206 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1208 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1209 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1211 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1213 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1215 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1217 * config/nds32/nds32-protos.h (nds32_data_alignment,
1218 nds32_local_alignment): Declare.
1219 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1220 nds32_local_alignment): New functions.
1221 (TARGET_CONSTANT_ALIGNMENT): Define.
1222 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1224 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1226 * config/nds32/nds32.c
1227 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1228 (TARGET_MODES_TIEABLE_P): Likewise.
1230 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1232 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1235 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1236 Chung-Ju Wu <jasonwucj@gmail.com>
1238 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1239 for unaligned access.
1240 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1242 * config/nds32/nds32-intrinsic.md: Likewise.
1243 * config/nds32/nds32_intrinsic.h: Likewise.
1244 * config/nds32/nds32.h (nds32_builtins): Likewise.
1245 * config/nds32/nds32.opt (munaligned-access): New option.
1246 * config/nds32/nds32.c (nds32_asm_file_start): Display
1247 flag_unaligned_access status.
1249 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1251 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1252 -mno-relax is present.
1253 * config/riscv/linux.h (LINK_SPEC): Ditto.
1255 2018-04-20 Martin Sebor <msebor@redhat.com>
1258 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1260 (gimple_fold_builtin_stxcpy_chk): Same.
1261 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1263 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1266 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1267 __powikf2 when long double is IEEE 128-bit.
1269 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1271 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1272 step to make sure stack always aligned.
1274 2018-04-20 Carl Love <cel@us.ibm.com>
1277 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1278 size check for arg0.
1280 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1281 Tom de Vries <tom@codesourcery.com>
1284 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1285 Emit insns for calls too.
1286 (nvptx_find_par): Always look for worker-level predecessor insn.
1287 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1289 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1290 (nvptx_process_pars): Propagate frames for calls.
1292 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1295 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1297 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1298 (ix86_handle_option): Don't handle OPT_mibt.
1299 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1301 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1303 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1304 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1305 (ix86_target_macros): Define __CET__ with flag_cf_protection
1306 for -fcf-protection.
1307 * config/i386/i386.c (isa2_opts): Remove -mibt.
1308 * config/i386/i386.h (TARGET_IBT): Removed.
1309 (TARGET_IBT_P): Likewise.
1310 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1311 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1312 * config/i386/i386.opt (mcet): Update help message.
1315 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1316 -mcet as an alias for -mshstk.
1318 2018-04-20 Richard Biener <rguenther@suse.de>
1321 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1322 complexity by forcing a single use of the multiply operand.
1324 2018-04-20 Martin Jambor <mjambor@suse.cz>
1327 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1328 recursion dependency to only apply to non-clones.
1330 2018-04-20 Martin Jambor <mjambor@suse.cz>
1333 * ipa-cp.c (create_specialized_node): Check that clones of
1334 self-recursive edges exist during IPA-CP.
1336 2018-04-19 Toon Moene <toon@moene.org>
1338 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1341 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1343 PR tree-optimization/85467
1344 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1345 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1346 VECTOR_CST element to type.
1348 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1351 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1352 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1353 (builtin_longjmp): Likewise.
1354 (save_stack_nonlocal): New pattern.
1355 (restore_stack_nonlocal): Likewise.
1357 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1360 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1361 Replace ASM_OUTPUT_LABEL with fprintf.
1363 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1366 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1367 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1368 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1369 define __IBT__ and __SHSTK__ for -fcf-protection.
1370 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1372 (ix86_trampoline_init): Likewise.
1373 (x86_output_mi_thunk): Likewise.
1374 (ix86_notrack_prefixed_insn_p): Likewise.
1375 (ix86_option_override_internal): Don't disallow -fcf-protection.
1376 * config/i386/i386.md (rdssp<mode>): Also enable for
1378 (incssp<mode>): Likewise.
1379 (nop_endbr): Likewise.
1380 * config/i386/i386.opt (mcet): Change help message to built-in
1384 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1385 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1386 enable CET built-in functions.
1388 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1390 * common/config/i386/i386-common.c
1391 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1392 OPTION_MASK_ISA_MOVDIRI_UNSET,
1393 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1394 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1395 * config.gcc (movdirintrin.h): New header.
1396 * config/i386/cpuid.h (bit_MOVDIRI,
1397 bit_MOVDIR64B): New bits.
1398 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1400 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1401 (VOID, PVOID, PCVOID)): New function types.
1402 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1403 __builtin_ia32_directstoreu_u64,
1404 __builtin_ia32_movdir64b): New builtins.
1405 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1406 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1408 (ix86_valid_target_attribute_inner_p): Ditto.
1409 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1410 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1411 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1412 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1413 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1414 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1415 (movdiri<mode>, movdir64b_<mode>): New.
1416 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1417 * config/i386/immintrin.h: Include movdirintrin.h.
1418 * config/i386/movdirintrin.h: New file.
1419 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1421 2018-04-19 Richard Biener <rguenther@suse.de>
1424 * cfg.c (clear_bb_flags): When loop state says we have
1425 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1427 2018-04-19 Richard Biener <rguenther@suse.de>
1429 PR tree-optimization/84737
1430 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1431 copying restrict info.
1432 (vect_setup_realignment): Use it.
1433 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1434 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1435 the first DR to all generated stores.
1436 (vectorizable_load): Likewise for loads.
1438 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1440 PR tree-optimization/85446
1441 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1442 the integral and pointer types to have the same precision.
1444 * doc/install.texi: Document --disable-cet being the default and
1447 2018-04-18 Martin Liska <mliska@suse.cz>
1449 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1452 2018-04-18 Martin Liska <mliska@suse.cz>
1455 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1458 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1459 arguments if they are comparable.
1461 2018-04-18 Martin Liska <mliska@suse.cz>
1464 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1467 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1470 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1473 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1474 ENDBR after calling __morestack.
1476 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1479 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1480 by using gcc_base_ver to generate a gcc_driver_version, and use
1481 it when generating GCC_DRIVER_NAME.
1482 * configure: Regenerate.
1484 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1487 * config.gcc: Obsolete powerpc*-*-*spe*.
1489 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1492 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1493 (stabstr_D): Change type of unum from unsigned int to
1494 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1497 2018-04-17 Jim Wilson <jimw@sifive.com>
1500 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1501 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1502 Set arg_pointer_offset after using pretend_args_size.
1504 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1506 PR rtl-optimization/85431
1507 * dse.c (record_store): Ignore zero width stores.
1510 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1511 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1512 __builtin_stack_restore rather than after it.
1513 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1514 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1515 argument instead of virtual_dynamic_stack_rtx.
1517 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1519 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1521 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1522 Add note to error message to explain internal mapping of overloaded
1523 built-in function name to non-overloaded built-in function name.
1524 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1527 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1530 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1531 where the inputs overlap with the output.
1533 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1535 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1536 (=v, v) alternative and explicit "memory" attribute.
1537 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1538 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1540 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1541 "sselog1" type instead of "sselog".
1542 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1543 "sselog". Remove explicit "memory" attribute.
1544 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1545 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1547 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1548 "sselog1" type instead of "sselog". Remove explicit "memory"
1550 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1551 use "sselog1" type instead of "sselog". Remove explicit "memory"
1553 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1554 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1556 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1557 "sselog1" type instead of "sselog". Remove explicit "memory"
1559 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1560 use "sselog1" type instead of "sselog". Remove explicit "memory"
1564 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1567 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1570 2018-04-17 Martin Jambor <mjambor@suse.cz>
1573 * ipa-cp.c (create_specialized_node): Call
1574 expand_all_artificial_thunks if necessary.
1576 2018-04-17 Martin Liska <mliska@suse.cz>
1579 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1580 in message, remote space in between '_G' and '('.
1582 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1585 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1586 avx512f_vmcmp<mode>3<round_saeonly_name>,
1587 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1588 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1589 avx512f_rndscale<mode><round_saeonly_name>,
1590 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1591 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1592 Use %<iptr>2 instead of %2 for -masm=intel.
1593 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1594 avx512f_vcvttss2usi<round_saeonly_name>,
1595 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1597 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1598 avx512f_vcvttsd2usi<round_saeonly_name>,
1599 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1600 Use %q1 instead of %1 for -masm=intel.
1601 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1602 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1603 of %3 for -masm=intel.
1604 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1606 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1608 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1610 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1611 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1613 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1615 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1616 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1617 %g1 and one with %0 and %1.
1618 (avx512er_vmrcp28<mode><round_saeonly_name>,
1619 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1621 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1622 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1623 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1624 of %0 and %{%4%} for -masm=intel.
1625 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1626 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1627 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1628 order of %0 and %{%5%}%{z%} for -masm=intel.
1630 2018-04-17 Jan Hubicka <jh@suse.cz>
1633 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1635 2018-04-17 Martin Liska <mliska@suse.cz>
1638 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1639 for target_clone error message. Make default implementation
1640 clone to be a local declaration.
1641 (separate_attrs): Add new argument and check for an empty
1643 (expand_target_clones): Handle it.
1644 (ipa_target_clone): Make redirection just for target_clones
1647 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1648 Tom de Vries <tom@codesourcery.com>
1651 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1652 tiled basic blocks without omp continue statements.
1654 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1657 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1658 vec_extract expression as having side effects to make sure it gets
1661 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1664 * config/i386/i386.c (get_builtin_code_for_version): Check
1667 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1670 * gcc/config.gcc: Support "skylake".
1671 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1673 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1674 (processor_target_table): Add "skylake".
1675 (ix86_option_override_internal): Add "skylake".
1676 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1677 PROCESSOR_CANNONLAKE.
1678 (get_builtin_code_for_version): Fix priority for
1679 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1680 PROCESSOR_SKYLAKE-AVX512.
1681 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1682 (processor_type): Add PROCESSOR_SKYLAKE.
1684 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1685 Jason Merrill <jason@redhat.com>
1688 * convert.c (convert_to_integer_1): Use direct recursion for
1689 enumeral types and types with a precision less than the number
1690 of bits in their mode.
1692 2018-04-16 Julia Koval <julia.koval@intel.com>
1695 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1696 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1698 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1701 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1702 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1703 and -mno-direct-move.
1705 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1708 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1709 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1710 Ensure negative shifts result in {0}.
1712 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1714 PR rtl-optimization/79916
1715 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1716 regs (if any) to define how to gnerate SD moves when LRA is in
1719 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1721 PR rtl-optimization/85393
1722 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1723 * except.c (expand_dw2_landing_pad_for_region): Make static.
1724 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1725 a label and unconditional jump to old_bb, rather than
1726 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1729 PR rtl-optimization/85376
1730 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1731 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1732 instead of a specific value.
1734 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1735 Bin Cheng <bin.cheng@arm.com>
1737 PR tree-optimization/82965
1738 PR tree-optimization/83991
1739 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1740 by_profile_only parameter.
1741 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1742 information if the loop was predicted to iterate too many times.
1743 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1745 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1748 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1751 2018-04-13 Martin Liska <mliska@suse.cz>
1752 Jakub Jelinek <jakub@redhat.com>
1755 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1756 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1757 * builtins.c (expand_builtin_memory_copy_args): Use
1758 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1759 handle dest_addr == pc_rtx.
1761 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1764 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1765 asked to not generate direct moves.
1766 (fix_trunc<mode>si2_stfiwx): Similar.
1767 (fix_trunc<mode>si2_internal): Similar.
1769 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1772 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1773 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1774 lookup if dest in some wider mode is known to be const0_rtx and
1775 if so, record permanent equivalence for it to be ZERO_EXTEND of
1776 the narrower mode destination.
1778 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1780 * lto-streamer-out.c (output_function): Revert 259346.
1781 * omp-expand.c (expand_oacc_for): Likewise.
1783 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1785 PR rtl-optimization/85354
1786 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1787 * sel-sched.c (sel_global_init): ... here.
1789 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1792 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1793 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1794 mode for PE-COFF targets.
1795 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1796 (i386_pe_asm_lto_end): Likewise.
1797 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1798 (TARGET_ASM_LTO_END): Likewise.
1799 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1800 (i386_pe_asm_lto_start): New function.
1801 (i386_pe_asm_lto_end): Likewise.
1803 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1804 Richard Biener <rguenther@suse.de>
1807 * lto-streamer-out.c (output_function): Fix CFG loop state before
1809 * omp-expand.c (expand_oacc_for): Handle calls to internal
1810 functions like regular functions.
1812 2018-04-12 Richard Biener <rguenther@suse.de>
1815 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1816 for the early LTO debug to properly generate references to it
1817 during DIE emission. Do not re-use that for the skeleton for
1819 (dwarf2out_early_finish): Likewise.
1821 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1824 * config/i386/sse.md
1825 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1826 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1827 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1828 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1829 and output is a reg, avoid creating invalid lowpart subreg, but
1830 instead split into a 512-bit move. Don't split if not AVX512VL,
1831 input is xmm16+ reg and output is a mem.
1832 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1833 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1834 xmm16+ reg and output is a mem.
1836 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1838 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1839 also for flag_dwarf2_cfi_asm.
1841 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1843 PR rtl-optimization/85342
1844 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1845 a bool scalar var inside of the loop instead. Don't try to update
1846 recog_data.operand after failed apply_change_group.
1848 2018-04-12 Tom de Vries <tom@codesourcery.com>
1851 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1852 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1853 array with flexible array member as array without given dimension.
1854 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1855 argument for undefined param to true.
1857 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1860 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1861 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1862 from PowerPC section.
1863 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1864 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1865 help text that is too long.
1866 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1867 help text that is too long.
1868 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1869 help text that is too long.
1871 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1873 * config/alpha/alpha.md (stack_probe_internal): Rename
1874 from "probe_stack". Update all callers.
1876 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1878 PR rtl-optimization/84566
1879 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1880 sched_macro_fuse_insns.
1882 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1885 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1886 (compute_block_dependences): ... from here.
1888 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1890 PR tree-optimization/85331
1891 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1892 from int to HOST_WIDE_INT.
1894 2018-04-11 Martin Jambor <mjambor@suse.cz>
1897 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1898 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1899 not the same as the source val.
1900 (cgraph_edge_brings_value_p): New parameter.
1901 (gather_edges_for_value): Pass destination value to
1902 cgraph_edge_brings_value_p.
1903 (perhaps_add_new_callers): Likewise.
1904 (get_info_about_necessary_edges): Likewise and exclude values brought
1905 only by self-recursive edges.
1906 (create_specialized_node): Redirect only clones of self-calling edges.
1907 (+self_recursive_pass_through_p): New function.
1908 (find_more_scalar_values_for_callers_subset): Use it.
1909 (find_aggregate_values_for_callers_subset): Likewise.
1910 (known_aggs_to_agg_replacement_list): Removed.
1911 (decide_whether_version_node): Re-calculate known constants for all
1912 remaining context clones.
1914 2018-04-11 Richard Biener <rguenther@suse.de>
1917 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1918 from early DWARF output.
1919 (dwarf2out_early_finish): Output line info unconditionally into
1920 early DWARF and add reference to it.
1922 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1925 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1926 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1927 other than V2DFmode using iptr mode attribute.
1928 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1930 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1932 PR rtl-optimization/84659
1933 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1935 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1938 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1940 (output_loc_list): Pass address of a dummy size variable even in the
1941 locview handling loop.
1942 (index_location_lists): Add comment on why skip_loc_list_entry can't
1945 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1948 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1951 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1954 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1955 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1956 and -mstring-compare-inline-limit.
1958 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1961 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1962 for stack clash protection in a register whenever we need it to be in
1965 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1967 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1968 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1970 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1973 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1975 (mlong-double-): Ditto.
1976 * config/rs6000/sysv4.opt (msdata=): Ditto.
1977 (mtls-size=): Ditto.
1979 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1981 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1982 erroneous entries for
1983 "vector int vec_ldl (int, long int *)", and
1984 "vector unsigned int vec_ldl (int, unsigned long int *)".
1985 Add comments and entries for
1986 "vector bool char vec_ldl (int, bool char *)",
1987 "vector bool short vec_ldl (int, bool short *)",
1988 "vector bool int vec_ldl (int, bool int *)",
1989 "vector bool long long vec_ldl (int, bool long long *)",
1990 "vector pixel vec_ldl (int, pixel *)",
1991 "vector long long vec_ldl (int, long long *)",
1992 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1993 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1994 type tree bool_long_long_type_node and correct definition of
1995 bool_V2DI_type_node to make reference to this new type tree.
1996 (rs6000_mangle_type): Replace erroneous reference to
1997 bool_long_type_node with bool_long_long_type_node.
1998 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1999 comments to emphasize sign distinctions for char and int types and
2000 replace RS6000_BTI_bool_long constant with
2001 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2002 use of RS6000_BTI_pixel.
2003 (bool_long_type_node): Remove this macro definition.
2004 (bool_long_long_type_node): New macro definition
2006 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2008 PR rtl-optimization/85300
2009 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2010 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2011 simplify_unary_operation fails.
2013 2018-04-10 Martin Liska <mliska@suse.cz>
2015 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2016 cgraph_edge and ipa_ref.
2018 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2022 * config/i386/sse.md
2023 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2024 computation of the VEC_MERGE selector from mask.
2025 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2026 Fix decoding of the VEC_MERGE selector into mask.
2028 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2030 PR tree-optimization/85286
2031 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2033 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2035 * final.c (final_1): Set insn_last_address as well as
2036 insn_current_address.
2038 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2041 * explow.c (emit_stack_probe): Call validize_mem on memory location
2042 before passing it to gen_probe_stack. Create address operand and
2043 legitimize it for the probe_stack_address case.
2045 2018-04-09 Jan Hubicka <jh@suse.cz>
2048 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2049 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2050 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2052 (free_lang_data): Rebuild type inheritance graph.
2054 2018-04-09 Martin Sebor <msebor@redhat.com>
2056 * invoke.texi (-finline-small-functions): Mention other optimization
2058 (-findirect-inlining, -fpartial-inlining): Same.
2059 (-finline-functions-called-once): Same.
2060 (-freorder-blocks-and-partition): Same.
2062 2018-04-09 Jan Hubicka <jh@suse.cz>
2065 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2066 jumps; choose last target that matches the criteria (i.e.
2067 no partition changes for non-crossing jumps).
2068 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2069 support for redirecting crossing jumps to non-crossing.
2071 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2073 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2074 also for naked functions.
2076 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2078 * config/arc/arc.md (add_shift): New pattern.
2079 (add_shift2): Likewise.
2080 (sub_shift): Likewise.
2081 (sub_shift_cmp0_noout): Likewise.
2082 (compare_si_ashiftsi): Likewise.
2083 (xbfu_cmp0_noout): New combine pattern.
2084 (xbfu_cmp0"): Likewise.
2085 (movsi_set_cc_insn): Place the predicable variant first.
2086 (commutative_binary_cmp0_noout): Remove clobber.
2087 (commutative_binary_cmp0): New pattern.
2088 (noncommutative_binary_cmp0): Likewise.
2089 (noncommutative_binary_cmp0_noout): Likewise.
2090 (noncommutative_binary_comparison_result_used): Removed.
2091 (rsub_cmp0): New pattern.
2092 (rsub_cmp0_noout): Likewise.
2093 (extzvsi): Changed, keep only meaningful variants.
2094 (SQH, SEZ): New iterators.
2095 (SQH_postfix): New mode attribute.
2096 (SEZ_prefix): New code attribute.
2097 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2098 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2099 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2101 (noncommutative_operator): Check the availability of barrel
2104 2018-04-09 Richard Biener <rguenther@suse.de>
2106 PR tree-optimization/85284
2107 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2108 Only use the niter constraining form of simple_iv when the exit
2111 2018-04-09 Tom de Vries <tom@codesourcery.com>
2114 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2115 (define_expand "*memory_barrier"): New define_expand.
2116 (define_insn "memory_barrier"): New insn.
2118 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2120 PR rtl-optimization/80463
2121 PR rtl-optimization/83972
2122 PR rtl-optimization/83480
2124 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2125 correct producer for the insn.
2126 (tidy_control_flow): Fixup seqnos in case of debug insns.
2128 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2130 PR rtl-optimization/83913
2132 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2133 different sched-times when merging exprs.
2135 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2137 PR rtl-optimization/83962
2139 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2140 tidy_fallthru_edge and tidy_control_flow.
2142 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2144 PR rtl-optimization/83530
2146 * sel-sched.c (force_next_insn): New global variable.
2147 (remove_insn_for_debug): When force_next_insn is true, also leave only
2148 next insn in the ready list.
2149 (sel_sched_region): When the region wasn't scheduled, make another pass
2150 over it with force_next_insn set to 1.
2152 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2154 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2156 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2157 for interrupt control.
2158 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2159 functions for interrupt control.
2160 * config/nds32/nds32-intrinsic.md: Likewise.
2161 * config/nds32/nds32_intrinsic.h: Likewise.
2162 * config/nds32/nds32.h (nds32_builtins): Likewise.
2164 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2166 * config/nds32/nds32.c (nds32_init_machine_status,
2167 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2168 strict_aligned_p field.
2169 (nds32_expand_to_rtl_hook): New function.
2170 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2171 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2173 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2174 Chung-Ju Wu <jasonwucj@gmail.com>
2176 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2177 * config/nds32/nds32-n7.md: New file.
2178 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2179 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2181 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2182 * config/nds32/nds32.md (pipeline_model): Add n7.
2183 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2184 * config/nds32/pipelines.md: Include n7 settings.
2186 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2187 Chung-Ju Wu <jasonwucj@gmail.com>
2189 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2190 * config/nds32/nds32-e8.md: New file.
2191 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2192 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2194 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2195 * config/nds32/nds32.md (pipeline_model): Add e8.
2196 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2197 * config/nds32/pipelines.md: Include e8 settings.
2199 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2200 Chung-Ju Wu <jasonwucj@gmail.com>
2202 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2203 * config/nds32/nds32-n8.md: New file.
2204 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2205 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2207 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2208 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2209 * config/nds32/nds32.md (pipeline_model): Add n8.
2210 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2211 * config/nds32/pipelines.md: Include n8 settings.
2213 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2214 Chung-Ju Wu <jasonwucj@gmail.com>
2216 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2217 * config/nds32/nds32-n9-2r1w.md: New file.
2218 * config/nds32/nds32-n9-3r2w.md: New file.
2219 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2220 nds32_register_ports): New or modify for cpu n9.
2221 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2223 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2224 * config/nds32/nds32-utils.c: New file.
2225 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2226 TARGET_MUL_SLOW): Define.
2227 * config/nds32/nds32.md (pipeline_model): New attribute.
2228 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2229 New options that support cpu n9.
2230 * config/nds32/pipelines.md: Include n9 settings.
2231 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2233 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2235 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2236 information if necessary.
2237 (output_cond_branch_compare_zero): Likewise.
2238 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2239 (nds32_target_alignment): Refine for alignment.
2240 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2241 (FUNCTION_BOUNDARY): Modify.
2242 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2244 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2246 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2248 * config/nds32/constants.md (unspec_volatile_element): Add values for
2249 TLB operation and data prefetch.
2250 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2251 functions for TLB operation and data prefetch.
2252 * config/nds32/nds32-intrinsic.md: Likewise.
2253 * config/nds32/nds32_intrinsic.h: Likewise.
2254 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2255 (nds32_print_operand): Likewise.
2256 * config/nds32/nds32.h (nds32_builtins): Likewise.
2258 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2259 Andrew Pinski <pinsika@gcc.gnu.org>
2262 * match.pd: Use constant_boolean_node of correct type instead of
2263 boolean_true_node or boolean_false_node for simplifying
2264 pointer comparisons to zero.
2266 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2268 PR tree-optimization/80021
2269 * tree.c (verify_type_variant): Make error call in verify_variant_match
2270 translatable and remove final full stop.
2272 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2274 * config/nds32/constants.md (unspec_volatile_element): Add
2275 UNSPEC_VOLATILE_EH_RETURN.
2276 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2277 nds32_output_stack_pop): Support dwarf exception handling process.
2278 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2279 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2280 exception handling process.
2281 (nds32_compute_stack_frame): Likewise.
2282 (nds32_return_addr_rtx): Likewise.
2283 (nds32_initial_elimination_offset): Likewise.
2284 (nds32_expand_prologue): Likewise.
2285 (nds32_expand_epilogue): Likewise.
2286 (nds32_dynamic_chain_address): New function.
2287 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2289 (DYNAMIC_CHAIN_ADDRESS): Define.
2290 (EH_RETURN_DATA_REGNO): Define.
2291 (EH_RETURN_STACKADJ_RTX): Define.
2292 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2293 patterns for dwarf exception handling.
2295 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2297 * config/nds32/nds32.h: Clean up obsolete macros.
2299 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2301 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2302 Add enum values for particular instructions.
2303 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2304 particular intrinsic functions.
2305 * config/nds32/nds32-intrinsic.md: Likewise.
2306 * config/nds32/nds32_intrinsic.h: Likewise.
2307 * config/nds32/nds32.h (nds32_builtins): Likewise.
2308 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2309 (btst, ave): New patterns for particular instructions.
2311 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2313 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2314 Add enum values for atomic load/store and memory sync.
2315 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2317 * config/nds32/nds32-intrinsic.md: Likewise.
2318 * config/nds32/nds32_intrinsic.h: Likewise.
2319 * config/nds32/nds32.h (nds32_builtins): Likewise.
2321 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2323 PR tree-optimization/85257
2324 * fold-const.c (native_encode_vector): If not all elts could fit
2325 and off is -1, return 0 rather than offset.
2326 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2327 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2328 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2329 adjust buffer in native_interpret_expr call.
2331 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2333 * config/nds32/constants.md (unspec_volatile_element): Add cache
2334 control enum values.
2335 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2336 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2337 * config/nds32/nds32.c (nds32_cctl_names): New.
2338 (nds32_print_operand): Handle cache control register names.
2339 * config/nds32/nds32.h (nds32_builtins): New enum values.
2340 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2342 * config/nds32/nds32.md (type): Add mmu.
2343 * config/nds32/pipelines.md (simple_insn): Add mmu.
2345 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2347 * config/nds32/nds32.md (type): Remove call.
2348 * config/nds32/pipelines.md (simple_insn): Likewise.
2350 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2352 * config/nds32/constants.md (unspec_volatile_element): Add
2353 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2354 UNSPEC_VOLATILE_FMFCFG.
2355 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2356 description for fmfcfg and fmfcsr.
2357 (bdesc_1arg): Add fmtcsr.
2358 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2359 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2360 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2361 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2362 unspec_fmfcfg): New patterns.
2363 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2364 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2365 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2366 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2367 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2368 __nds32__fmfcfg): Define.
2370 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2372 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2373 intrinsic register names.
2374 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2375 intrinsic register enum values and macros.
2377 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2379 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2380 for load/store addressing form.
2381 (nds32_print_operand_address): Likewise.
2383 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2386 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2387 based on LABEL_REF. Remove useless assertion.
2388 (pic_address_needs_scratch): Fix formatting.
2389 (sparc_legitimize_pic_address): Minor tweaks.
2390 (sparc_delegitimize_address): Adjust assertion accordingly.
2391 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2392 into symbolic_operand.
2393 (movsi_high_pic_label_ref): Likewise.
2394 (movsi_lo_sum_pic_label_ref): Likewise.
2395 (movdi_pic_label_ref): Likewise.
2396 (movdi_high_pic_label_ref): Likewise.
2397 (movdi_lo_sum_pic_label_ref): Likewise.
2399 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2401 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2402 custom LIB_SPEC setup.
2404 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2405 Kito Cheng <kito.cheng@gmail.com>
2407 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2408 * config/riscv/freebsd.h: New.
2410 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2412 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2413 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2416 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2417 Kito Cheng <kito.cheng@gmail.com>
2419 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2420 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2421 * config/nds32/nds32-protos.h (nds32_output_call,
2422 nds32_output_return): Declare.
2423 * config/nds32/nds32.md: Refine all the call and return patterns.
2425 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2428 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2429 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2431 PR rtl-optimization/84872
2432 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2433 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2436 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2438 * expr.c (copy_blkmode_to_reg): Revert 254862.
2439 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2441 2018-04-06 Richard Biener <rguenther@suse.de>
2444 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2445 after seeing a component reference with an adjacent field. Treat
2446 refs to arrays at struct end of external decls similar to
2447 refs to unconstrained commons.
2449 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2452 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2453 look through SAVE_EXPRs with non-side-effects argument. Adjust
2455 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2458 2018-04-06 Richard Biener <rguenther@suse.de>
2461 * alias.c (find_base_term): New wrapper around find_base_term
2462 unwinding CSELIB_VAL_PTR changes.
2463 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2466 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2468 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2470 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2471 constant definitions.
2472 ("nop"): lr 0,0 -> nopr r0
2473 ("nop_lr0", "nop_lr1"): New insn definitions.
2475 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2477 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2478 NDS32_V3PUSH_AVAILABLE_P macro.
2480 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2481 Chung-Ju Wu <jasonwucj@gmail.com>
2483 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2484 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2485 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2486 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2487 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2488 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2489 * config/nds32/constraints.md: New constraints and checking for hard
2490 float configuration.
2491 * config/nds32/iterators.md: New mode iterator and attribute for hard
2492 float configuration.
2493 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2495 * config/nds32/nds32-fpu.md: New file.
2496 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2497 deal with hard float code generation.
2498 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2500 (abi_type, float_reg_number): New enum type.
2501 * config/nds32/nds32-predicates.c: New predicates for hard float.
2502 * config/nds32/nds32-protos.h: Declare functions for hard float.
2503 * config/nds32/nds32.c: Implementation for hard float configuration.
2504 * config/nds32/nds32.h: Definitions for hard float configuration.
2505 * config/nds32/nds32.md: Include hard float machine description and
2506 modify patterns for hard float configuration.
2507 * config/nds32/nds32.opt: New options for hard float configuration.
2508 * config/nds32/predicates.md: New predicates for hard float
2511 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2513 * common/config/nds32/nds32-common.c
2514 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2516 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2519 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2520 CONSTRUCTOR_ELT (ctor, ...)->value.
2522 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2525 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2527 2018-04-05 Tom de Vries <tom@codesourcery.com>
2530 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2533 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2534 Kito Cheng <kito.cheng@gmail.com>
2536 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2537 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2538 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2539 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2541 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2542 Kito Cheng <kito.cheng@gmail.com>
2544 * config/nds32/constraints.md (Ufe): New memory constraint.
2545 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2546 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2547 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2549 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2550 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2552 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2554 * config/nds32/nds32.md: Use optimize_size in the condition for
2555 alu-shift instructions.
2557 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2559 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2561 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2563 * config/nds32/nds32.md (negsi2): Refine pattern.
2565 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2566 Chung-Ju Wu <jasonwucj@gmail.com>
2568 * config/nds32/iterators.md (shift_rotate): New code iterator.
2569 (shift): New code attribute.
2570 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2571 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2572 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2573 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2574 bit-wise operations.
2575 (andsi3, *andsi3): Ditto.
2576 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2577 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2578 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2579 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2580 nds32_ior_operand, nds32_xor_operand): New predicates.
2582 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2584 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2585 (addsi3, subsi3): ... this.
2587 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2589 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2591 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2593 * config/nds32/nds32.md: Adjust indention.
2595 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2597 * config/nds32/nds32.md (feature): New attribute.
2599 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2601 * config/nds32/nds32.md (subtype): New attribute.
2603 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2606 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2607 expansion to perform a bitwise AND of the argument followed by a
2608 boolean negation of the result.
2610 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2612 PR rtl-optimization/84878
2613 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2614 the basic block. Assert the use reference is not artificial and that
2615 it has an associated insn.
2617 2018-04-04 Michael Matz <matz@suse.de>
2619 * builtins.c (compute_objsize): Pass correct operand
2620 to array_at_struct_end_p.
2622 2018-04-04 Richard Biener <rguenther@suse.de>
2625 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2626 from contexts for DINFO_LEVEL_TERSE and below.
2628 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2630 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2631 resiter_operand condition.
2632 * config/nds32/nds32.md (*move<mode>): Ditto.
2634 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2635 Monk Chiang <sh.chiang04@gmail.com>
2637 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2639 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2641 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2643 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2644 Kito Cheng <kito.cheng@gmail.com>
2646 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2647 nds32_cond_code_str, output_cond_branch,
2648 output_cond_branch_compare_zero, nds32_expand_cbranch,
2649 nds32_expand_cstore, nds32_expand_movcc,
2650 nds32_output_cbranchsi4_equality_zero,
2651 nds32_output_cbranchsi4_equality_reg,
2652 nds32_output_cbranchsi4_equality_reg_or_const_int,
2653 nds32_output_cbranchsi4_greater_less_zero: New functions.
2654 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2655 nds32_expand_cstore, nds32_expand_movcc,
2656 nds32_output_cbranchsi4_equality_zero,
2657 nds32_output_cbranchsi4_equality_reg,
2658 nds32_output_cbranchsi4_equality_reg_or_const_int,
2659 nds32_output_cbranchsi4_greater_less_zero): Declare.
2660 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2661 nds32_rimm11s_operand): New predicates.
2662 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2663 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2666 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2668 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2669 * config/nds32/nds32.md: Ditto.
2670 * config/nds32/pipelines.md: Ditto.
2672 2018-04-04 Richard Biener <rguenther@suse.de>
2674 PR tree-optimization/85168
2675 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2676 propagating abnormals.
2678 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2680 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2682 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2683 Kito Cheng <kito.cheng@gmail.com>
2685 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2686 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2687 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2688 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2689 * config/nds32/nds32.md (sibcall_internal): New.
2690 (sibcall_register): Remove.
2691 (sibcall_immediate): Remove.
2692 (sibcall_value_internal): New.
2693 (sibcall_value_register): Remove.
2694 (sibcall_value_immediate): Remove.
2695 * config/nds32/predicates.md (nds32_general_register_operand): New.
2696 (nds32_call_address_operand): New.
2698 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2700 PR rtl-optimization/85167
2701 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2702 bb_defs if *split_p, instead preinitialize it to NULL.
2704 PR tree-optimization/85156
2705 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2706 evaluating the argument multiple times.
2708 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2710 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2712 (_mm_cvtpd_ps): Likewise.
2713 (_mm_cvttpd_epi32): Likewise.
2714 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2715 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2716 vector, pixel, and bool following altivec.h include.
2718 2018-04-03 Martin Sebor <msebor@redhat.com>
2720 * doc/extend.texi (Common Function Attributes): Clarify.
2721 (const attribute): Likewise.
2722 (pure attribute): Likewise.
2724 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2727 * config/i386/i386.c (ix86_expand_vector_set): Use
2728 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2730 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2732 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2733 instructions when changing rounding bits to preserve precision bits
2734 in the x87 control word.
2736 2018-04-03 Martin Liska <mliska@suse.cz>
2738 PR tree-optimization/82491
2739 * rtl.h (strip_offset_and_add): Replace += suboffset with
2740 poly_uint64 () + suboffset.
2742 2018-03-29 Martin Liska <mliska@suse.cz>
2743 Martin Jambor <mjambor@suse.cz>
2746 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2747 param_type is not an integral or pointer type.
2749 2018-04-03 Richard Biener <rguenther@suse.de>
2751 * sese.h (recompute_all_dominators): Remove.
2753 2018-04-02 Martin Sebor <msebor@redhat.com>
2755 * doc/invoke.texi (-Wrestrict): Fix typos.
2757 2018-04-02 Jim Wilson <jimw@sifive.com>
2759 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2760 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2761 (<optab>di3, <optab>si3_extend): Likewise.
2762 (<optab>si3_mask, <optab>si3_mask_1): New.
2763 (<optab>di3_mask, <optab>di3_mask_1): New.
2764 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2765 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2766 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2768 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2770 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2773 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2775 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2776 (nds32_canonicalize_comparison): New function.
2778 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2779 Kito Cheng <kito.cheng@gmail.com>
2780 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2782 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2783 * config/nds32/constants.md (unspec_volatile_element): Add
2784 UNSPEC_VOLATILE_RELAX_GROUP.
2785 * config/nds32/nds32-relax-opt.c: New file.
2786 * config/nds32/nds32-predicates.c
2787 (nds32_symbol_load_store_p): New function.
2788 * config/nds32/nds32-protos.h
2789 (nds32_symbol_load_store_p): Declare function.
2790 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2791 * config/nds32/nds32.c
2792 (nds32_register_pass): New function to register pass.
2793 (nds32_register_passes): New function to register passes.
2794 * config/nds32/nds32.md (relax_group): New pattern.
2795 * config/nds32/nds32.opt (mrelax-hint): New option.
2796 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2798 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2800 * config/nds32/t-nds32: Modify files dependency.
2802 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2804 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2805 (PROFILE_HOOK): Define its implementation.
2807 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2809 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2810 type and 32-bit size.
2812 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2815 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2816 (V_128_256): New mode iterator.
2817 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2818 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2819 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2821 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2822 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2824 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2827 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2828 NaN inputs correctly.
2830 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2833 * config/rs6000/vsx.md (??r): New mode attribute.
2834 (*vsx_mov<mode>_64bit): Use it.
2835 (*vsx_mov<mode>_32bit): Likewise.
2837 2018-03-30 Martin Sebor <msebor@redhat.com>
2839 PR tree-optimization/84818
2840 * builtins.c (check_access): Use warning_n.
2842 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2845 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2847 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2850 2018-03-30 Julia Koval <julia.koval@intel.com>
2853 * x86-tune.def (movx, partial_reg_dependency): Enable for
2856 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2859 * lra-constraints.c (process_alt_operands): Move setting
2860 this_alternative_matches below.
2862 2018-03-29 Martin Liska <mliska@suse.cz>
2865 * doc/invoke.texi: Document how LTO works with debug info.
2866 Describe auto-load support of binutils. Mention 'x86-64'
2867 as valid option value of -march option.
2869 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2871 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2874 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2875 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2876 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2879 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2882 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2883 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2884 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2885 for RS6000_BTM_POWERPC64.
2886 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2887 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2888 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2893 2018-03-28 Carl Love <cel@us.ibm.com>
2896 2017-09-27 Carl Love <cel@us.ibm.com>
2898 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2899 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2900 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2903 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2905 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2906 instead of __vector bool.
2907 (_mm_max_pu8): Likewise.
2908 (_mm_min_pi16): Likewise.
2910 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2913 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2914 (DIVWEUO): Likewise.
2916 (DIVDEUO): Likewise.
2917 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2918 DIVWEUO and DIVDEUO.
2919 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2920 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2921 (div_extend): Likewise.
2922 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2924 (__builtin_divweuo): Likewise.
2925 (__builtin_divdeo): Likewise.
2926 (__builtin_divdeuo): Likewise.
2928 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2931 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2932 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2934 PR tree-optimization/82004
2935 * gimple-match-head.c (optimize_pow_to_exp): New function.
2936 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2937 Don't fold to exp if optimize_pow_to_exp is false.
2939 2018-03-28 Martin Liska <mliska@suse.cz>
2942 * calls.c (initialize_argument_information): Fix trailing space.
2943 * common.opt: Fix typo and provide better explanation for
2944 -fsanitize-coverage option.
2945 * config/i386/i386.opt: Fix typo.
2947 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2948 Martin Liska <mliska@suse.cz>
2951 * gimplify.c (asan_poison_variable): Don't do the check for
2952 gimplify_omp_ctxp here.
2953 (gimplify_decl_expr): Do it here.
2954 (gimplify_target_expr): Likewise.
2956 2018-03-28 Martin Liska <mliska@suse.cz>
2959 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2960 chkp_type_bounds_count if MPX is not enabled.
2962 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2964 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2966 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2969 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2970 function to create the function decl for complex long double
2971 multiply and divide for -mabi=ieeelongdouble.
2972 (init_float128_ieee): Call it.
2974 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2977 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2978 -fcf-protection=branch -mibt.
2979 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2981 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2984 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2986 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2989 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2990 extern array declarations.
2992 2018-03-27 Richard Biener <rguenther@suse.de>
2995 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2996 explicit single_use checks.
2998 2018-03-27 Richard Biener <rguenther@suse.de>
3000 PR tree-optimization/85082
3001 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3004 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3006 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3007 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3008 Turn on fasynchronous-unwind-tables and funwind-tables.
3010 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3013 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3014 (*bmi_blsr_<mode>_ccz): Ditto.
3016 2018-03-26 Tom de Vries <tom@codesourcery.com>
3018 PR tree-optimization/85063
3019 * omp-general.c (offloading_function_p): New function. Factor out
3021 * omp-offload.c (pass_omp_target_link::gate): ... here.
3022 * omp-general.h (offloading_function_p): Declare.
3023 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3024 with attribute omp declare target for offloading functions.
3026 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3028 PR tree-optimization/84005
3029 * tree-data-ref.h (get_base_for_alignment): Declare.
3030 * tree-data-ref.c (get_base_for_alignment_1): New function.
3031 (get_base_for_alignment): Likewise.
3032 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3033 get_base_for_alignment to find a suitable base object, instead
3034 of always using drb->base_address.
3036 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3039 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3040 known size by default.
3042 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3045 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3046 and non BLKmode operands.
3048 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3051 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3052 Clean up attributes.
3054 2018-03-23 Richard Biener <rguenther@suse.de>
3057 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3058 we are going to emit early debug for LTO.
3060 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3063 * function.c (match_asm_constraints_1): Don't optimize if input
3064 doesn't satisfy general_operand predicate for output's mode.
3067 * alias.c (write_dependence_p): Don't require for x_canonicalized
3068 non-VOIDmode if x has VOIDmode.
3071 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3072 just don't try to optimize it rather than assert it never happens.
3074 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3076 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3077 macro expansions for definition of ST_INTERNAL_<mode> and
3078 LD_INTERNAL_<mode> builtins.
3079 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3081 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3083 (altivec_expand_st_builtin): Likewise.
3084 (altivec_expand_builtin): Remove calls to deleted functions.
3085 (rs6000_address_for_altivec): Delete this function.
3086 * config/rs6000/vector.md: Remove expands for
3087 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3089 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3092 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3093 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3094 re-computing once computed.
3095 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3096 (arm_init_machine_status): Initialize
3097 machine->static_chain_stack_bytes.
3099 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3102 * doc/extend.texi: Add four new prototypes for vec_ld.
3103 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3104 definitions for more logical presentation.
3105 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3106 entries for V1TI variants of __builtin_altivec_ld builtin.
3107 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3108 handling of V1TI variant of LVX icode pattern.
3109 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3110 (rs6000_gimple_fold_builtin): Likewise.
3111 (altivec_init_builtins): Add code to define
3112 __builtin_altivec_lvx_v1ti function.
3114 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3117 * function.c (match_asm_constraints_1): Don't do the optimization
3118 if input isn't a REG, SUBREG, MEM or constant.
3120 2018-03-22 Tom de Vries <tom@codesourcery.com>
3122 PR tree-optimization/84956
3123 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3124 bb_has_abnormal_pred.
3126 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3129 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3130 DECL_INITIAL (decl) to decl at the end.
3131 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3134 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3136 * doc/extend.texi (__builtin_tgmath): Document when complex
3137 integer types are treated as _Complex _Float64.
3139 2018-03-21 Tom de Vries <tom@codesourcery.com>
3141 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3143 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3145 PR tree-optimization/84960
3146 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3147 if it is ENTRY block, move them into single succ of ENTRY in that case.
3149 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3151 PR tree-optimization/84811
3152 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3153 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3154 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3155 (known_subrange_p): Update accordingly. Cast each value involved
3156 in the size comparison, rather than casting the result of the
3159 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3161 PR tree-optimization/84982
3162 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3163 by flipping the least significant bit rather than all bits from
3164 bitpos to bitpos + bitsize - 1.
3166 2018-03-21 Nathan Sidwell <nathan@acm.org>
3168 * doc/extend.texi (Deprecated Features): Remove mention of
3169 long-deleted deprecations.
3171 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3174 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3175 * configure: Regenerate.
3177 2018-03-21 Tom de Vries <tom@codesourcery.com>
3179 PR tree-optimization/83126
3180 * tree-parloops.c (num_phis): New function.
3181 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3183 2018-03-21 Nathan Sidwell <nathan@acm.org>
3185 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3186 mention anon-struct/union members and trailing attributes.
3188 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3190 PR tree-optimization/84969
3191 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3192 builtin memset partitions if they set different rhs values.
3194 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3196 PR rtl-optimization/84989
3197 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3198 VEC_DUPLICATE with scalar result mode.
3200 2018-03-21 Martin Liska <mliska@suse.cz>
3203 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3204 not intended return statement.
3206 2018-03-21 Martin Liska <mliska@suse.cz>
3209 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3210 (chkp_find_bound_slots_1): Limit number of iterations.
3212 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3215 * Minor grammar fixes for x86 options.
3217 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3220 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3221 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3223 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3226 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3227 (altivec_lvx_<mode>_1op): Likewise.
3228 (altivec_stvx_<mode>_2op): Likewise.
3229 (altivec_stvx_<mode>_1op): Likewise.
3230 (altivec_lvx_<VM2:mode>): New define_expand.
3231 (altivec_stvx_<VM2:mode>): Likewise.
3232 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3233 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3234 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3235 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3236 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3237 (rs6000_gen_lvx): Likewise.
3238 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3239 (altivec_expand_stv_builtin): Likewise.
3240 (altivec_expand_builtin): Likewise.
3241 * config/rs6000/vector.md: Likewise.
3243 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3246 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3249 2018-03-20 Richard Biener <rguenther@suse.de>
3252 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3253 sign-conversions as zero, fall back to standard scalar_stmt
3256 2018-03-20 Martin Liska <mliska@suse.cz>
3259 * predict.c (rebuild_frequencies): Handle case when we have
3260 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3262 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3265 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3266 flag_section_anchors.
3267 * varasm.c (use_blocks_for_decl_p): Remove hack for
3268 dw2_force_const_mem.
3271 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3273 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3274 be created, use lowpart_subreg of operands[0] rather than operands[0]
3276 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3277 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3278 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3279 and n constraint instead of aarch64_shift_imm_di and Usd.
3280 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3281 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3283 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3286 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3287 to favor GPR over NEON registers.
3288 (<shift>di3_neon): Likewise.
3290 2018-03-20 Tom de Vries <tom@codesourcery.com>
3293 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3294 (nvptx_process_pars): Emit bar.sync asap and alap.
3296 2018-03-20 Tom de Vries <tom@codesourcery.com>
3299 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3300 seen_label if seen_label is already set.
3302 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3305 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3306 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3307 Use 1U instead of 1. Formatting fixes.
3310 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3311 instead of TREE_TYPE (s1) for the return value.
3313 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3315 PR tree-optimization/84946
3316 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3317 bitsize + bitsize in poly_uint64 rather than poly_int64.
3320 * dwarf2asm.c: Include fold-const.c.
3321 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3322 of decl rather than decl itself.
3324 PR rtl-optimization/84643
3325 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3327 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3330 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3331 calling assemble_variable.
3333 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3336 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3337 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3339 2018-03-19 Jim Wilson <jimw@sifive.com>
3342 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3343 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3344 (riscv_first_stack_step): Likewise.
3345 (riscv_option_override): Use STACK_BOUNDARY instead of
3347 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3349 (BIGGEST_ALIGNMENT): Set to 128.
3350 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3351 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3354 2018-03-19 Richard Biener <rguenther@suse.de>
3356 PR tree-optimization/84933
3357 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3358 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3360 2018-03-19 Richard Biener <rguenther@suse.de>
3362 PR tree-optimization/84859
3363 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3364 (cond_if_else_store_replacement): Perform sinking operation on
3365 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3366 Generalize what a BB with a single eligible store is.
3368 2018-03-19 Richard Biener <rguenther@suse.de>
3370 PR tree-optimization/84929
3371 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3372 chrec_is_positive against non-chrec arg.
3374 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3377 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3379 2018-03-18 Martin Liska <mliska@suse.cz>
3381 PR rtl-optimization/84635
3382 * regrename.c (build_def_use): Use matches_mode only when
3385 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3387 PR tree-optimization/84913
3388 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3389 vectorize chains of COND_EXPRs.
3391 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3393 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3395 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3397 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3399 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3401 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3403 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3404 Kito Cheng <kito.cheng@gmail.com>
3406 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3407 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3408 (nds32_adjust_reg_alloc_order): New function.
3409 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3411 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3413 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3414 nds32_print_operand, nds32_print_operand_address): Use
3415 HOST_WIDE_INT_PRINT_DEC instead.
3417 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3419 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3421 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3424 * config/i386/i386.c (initial_ix86_tune_features,
3425 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3427 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3428 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3429 rather than 1u << ix86_tune. Formatting fix.
3430 (ix86_option_override_internal): Change ix86_arch_mask from
3431 unsigned int to unsigned HOST_WIDE_INT, initialize to
3432 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3433 (ix86_function_specific_restore): Likewise.
3435 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3438 * postreload.c (reload_combine_recognize_pattern): Perform
3439 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3440 truncate_int_for_mode the result for the destination's mode.
3443 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3444 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3446 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3449 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3452 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3453 regno_allocno_class_array and sorted_pseudos.
3454 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3455 insns where regno is used.
3457 2018-03-16 Martin Liska <mliska@suse.cz>
3460 * multiple_target.c (create_dispatcher_calls): Redirect
3461 reference in the symbol table.
3463 2018-03-16 Martin Liska <mliska@suse.cz>
3466 * multiple_target.c (create_dispatcher_calls): Redirect also
3469 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3473 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3474 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3475 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3477 2018-03-16 Julia Koval <julia.koval@intel.com>
3479 * doc/invoke.texi (Skylake Server): Add CLWB.
3480 Cannonlake): Remove CLWB.
3482 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3484 PR tree-optimization/84841
3485 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3487 (FLOAT_ONE_CONST_TYPE): Define.
3488 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3489 (sort_by_operand_rank): Put entries with higher constant_type last
3490 rather than first to match comments.
3492 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3494 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3497 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3500 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3501 check and use address of target always.
3503 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3506 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3507 (indirect_thunk_bnd_needed): Likewise.
3508 (indirect_thunks_used): Likewise.
3509 (indirect_thunks_bnd_used): Likewise.
3510 (indirect_return_needed): New.
3511 (indirect_return_bnd_needed): Likewise.
3512 (output_indirect_thunk_function): Add a bool argument for
3514 (output_indirect_thunk_function): Don't generate alias for
3515 function return thunk.
3516 (ix86_code_end): Call output_indirect_thunk_function to generate
3517 function return thunks.
3518 (ix86_output_function_return): Set indirect_return_bnd_needed
3519 and indirect_return_needed instead of indirect_thunk_bnd_needed
3520 and indirect_thunk_needed.
3522 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3524 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3525 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3526 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3528 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3529 Paul Hua <paul.hua.gm@gmail.com>
3532 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3534 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3536 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3537 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3540 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3543 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3544 instead of GET_MODE_SIZE when comparing Units.
3546 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3549 * varasm.c (hash_section): Return an unchangeble hash value
3550 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3551 Return !aarch64_can_use_per_function_literal_pools_p ().
3553 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3556 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3557 rather than address of cmode as last argument to prepare_cmp_insn.
3559 2018-03-15 Julia Koval <julia.koval@intel.com>
3561 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3562 F_AVX512VNNI, F_AVX512BITALG): New.
3564 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3567 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3568 insn for floating-point loads and stores.
3570 2018-03-14 Carl Love <cel@us.ibm.com>
3572 * config/rs6000/rs6000-c.c: Add macro definitions for
3573 ALTIVEC_BUILTIN_VEC_PERMXOR.
3574 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3575 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3576 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3577 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3579 * config/doc/extend.texi: Add prototypes for vec_permxor.
3581 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3584 * diagnostic-show-locus.c (class layout_point): Convert m_line
3585 from int to linenum_type.
3586 (line_span::comparator): Use linenum "compare" function when
3587 comparing line numbers.
3588 (test_line_span): New function.
3589 (layout_range::contains_point): Convert param "row" from int to
3591 (layout_range::intersects_line_p): Likewise.
3592 (layout::will_show_line_p): Likewise.
3593 (layout::print_source_line): Likewise.
3594 (layout::should_print_annotation_line_p): Likewise.
3595 (layout::print_annotation_line): Likewise.
3596 (layout::print_leading_fixits): Likewise.
3597 (layout::annotation_line_showed_range_p): Likewise.
3598 (struct line_corrections): Likewise for field m_row.
3599 (line_corrections::line_corrections): Likewise for param "row".
3600 (layout::print_trailing_fixits): Likewise.
3601 (layout::get_state_at_point): Likewise.
3602 (layout::get_x_bound_for_row): Likewise.
3603 (layout::print_line): Likewise.
3604 (diagnostic_show_locus): Likewise for locals "last_line" and
3606 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3607 * input.c (selftest::test_linenum_comparisons): New function.
3608 (selftest::input_c_tests): Call it.
3609 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3610 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3611 * selftest.h (ASSERT_GT): New macro.
3612 (ASSERT_GT_AT): New macro.
3613 (ASSERT_LT): New macro.
3614 (ASSERT_LT_AT): New macro.
3616 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3618 PR rtl-optimization/84780
3619 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3621 2018-03-14 Martin Liska <mliska@suse.cz>
3623 * tree.c (record_node_allocation_statistics): Use
3624 get_stats_node_kind.
3625 (get_stats_node_kind): New function extracted from
3626 record_node_allocation_statistics.
3627 (free_node): Use get_stats_node_kind.
3629 2018-03-14 Richard Biener <rguenther@suse.de>
3631 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3632 that the value-set of ANTIC_IN doesn't grow.
3635 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3637 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3638 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3640 2018-03-14 Julia Koval <julia.koval@intel.com>
3642 * config.gcc (icelake-client, icelake-server): New.
3644 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3645 (initial_ix86_arch_features): Ditto.
3646 (PTA_SKYLAKE): Add SGX.
3647 (PTA_ICELAKE): Remove.
3648 (PTA_ICELAKE_CLIENT): New.
3649 (PTA_ICELAKE_SERVER): New.
3650 (ix86_option_override_internal): Split up icelake on icelake client and
3652 (get_builtin_code_for_version): Ditto.
3653 (fold_builtin_cpu): Ditto.
3654 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3655 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3656 * config/i386/i386.h (processor_type): Ditto.
3657 * doc/invoke.texi: Ditto.
3659 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3662 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3663 INTEGER_CST offset, add it together with bitpos / 8 and
3664 sign extend based on POINTER_SIZE.
3668 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3671 * config/i386/constraints.md (Yc): New register constraint.
3672 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3673 Use Yc constraint for alternative 2 of operand 0. Remove
3674 preferred_for_speed attribute.
3676 2018-03-14 Richard Biener <rguenther@suse.de>
3678 PR tree-optimization/84830
3679 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3680 with the old one to avoid oscillations.
3682 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3685 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3687 (assign_by_spills): Return a flag of reload assignment failure.
3688 Do not process the reload assignment failures. Do not spill other
3689 reload pseudos if they has the same reg class. Update n if
3691 (lra_assign): Add a return arg. Set up from the result of
3692 assign_by_spills call.
3693 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3694 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3695 usage_insns if it is not NULL.
3696 (spill_hard_reg_in_range): New function.
3697 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3698 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3699 function prototypes.
3700 (lra_assign): Change prototype.
3701 * lra.c (lra): Add code to deal with fails by splitting hard reg
3704 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3706 * config/riscv/riscv.opt (mrelax): New option.
3707 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3708 "norelax" when riscv_mrelax is disabled.
3709 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3711 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3714 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3715 reassociation for int modes.
3717 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3719 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3720 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3722 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3723 * config/aarch64/aarch64-sve.md
3724 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3725 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3726 (*extend<mode><Vwide>2): Rename to...
3727 (aarch64_sve_extend<mode><Vwide>2): ...this.
3728 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3729 renaming the old pattern to...
3730 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3732 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3733 define_expand, renaming the old pattern to...
3734 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3735 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3736 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3737 account when deciding which SVE instruction the optab should use.
3738 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3740 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3742 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3743 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3744 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3745 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3746 (tlsdesc_small_<mode>): Turn a define_expand and use
3747 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3748 (tlsdesc_small_advsimd_<mode>): ...this.
3749 (tlsdesc_small_sve_<mode>): New pattern.
3751 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3753 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3754 (UNSPEC_UMUL_HIGHPART): New constants.
3755 (MUL_HIGHPART): New int iteraor.
3756 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3757 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3759 (*<su>mul<mode>3_highpart): New define_insn.
3761 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3764 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3767 2018-03-13 Martin Liska <mliska@suse.cz>
3770 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3772 (sem_item_optimizer::~sem_item_optimizer): Release it.
3773 (sem_item_optimizer::merge_classes): Register variable aliases.
3774 (sem_item_optimizer::fixup_pt_set): New function.
3775 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3776 * ipa-icf.h: Declare new variables and functions.
3778 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3781 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3782 integer_pow2p@2 and test integer_pow2p in condition.
3783 (A < 0 ? C : 0): Similarly for @1.
3786 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3787 characters starting at p contain '\0' character, don't look beyond
3791 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3792 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3795 * reg-stack.c (change_stack): Change update_end var from int to
3796 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3797 also call set_block_for_insn on the newly added insns and rescan.
3800 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3801 on the last operand.
3804 * tree.c (stabilize_reference_1): Return save_expr (e) for
3805 STATEMENT_LIST even if it doesn't have side-effects.
3807 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3809 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3811 2018-03-12 Renlin Li <renlin.li@arm.com>
3813 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3814 aarch64_output_scalar_simd_mov_immediate.
3816 2018-03-12 Martin Sebor <msebor@redhat.com>
3818 PR tree-optimization/83456
3819 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3820 for perfectly overlapping calls to memcpy.
3821 (gimple_fold_builtin_memory_chk): Same.
3822 (gimple_fold_builtin_strcpy): Handle no-warning.
3823 (gimple_fold_builtin_stxcpy_chk): Same.
3824 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3826 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3828 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3829 parameter. Use it for SFmode.
3830 (rs6000_function_arg_advance_1): Adjust.
3831 (rs6000_function_arg): Adjust.
3832 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3834 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3836 PR rtl-optimization/84169
3837 PR rtl-optimization/84780
3838 * combine.c (can_combine_p): Check for a 2-insn combination whether
3839 the destination register is used between the two insns, too.
3841 2018-03-12 Richard Biener <rguenther@suse.de>
3843 PR tree-optimization/84803
3844 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3845 for refs DR analysis didn't process.
3847 2018-03-12 Richard Biener <rguenther@suse.de>
3849 PR tree-optimization/84777
3850 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3851 force-vectorize loops ignore whether we are optimizing for size.
3853 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3855 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3856 (TARGET_MD_ASM_ADJUST): Define.
3858 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3859 Kito Cheng <kito.cheng@gmail.com>
3860 Chung-Ju Wu <jasonwucj@gmail.com>
3862 * config/nds32/nds32.c (nds32_compute_stack_frame,
3863 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3864 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3865 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3866 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3867 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3868 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3869 * config/nds32/nds32.md (prologue, epilogue): Use macro
3870 NDS32_V3PUSH_AVAILABLE_P to do checking.
3872 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3875 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3876 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3877 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3878 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3879 addition of most attributes on !orig_type_die or the attribute not
3880 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3882 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3883 Chung-Ju Wu <jasonwucj@gmail.com>
3885 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3887 * config/nds32/nds32.opt (mvh): New option.
3889 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3890 Chung-Ju Wu <jasonwucj@gmail.com>
3892 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3894 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3895 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3898 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3899 Chung-Ju Wu <jasonwucj@gmail.com>
3901 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3903 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3904 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3906 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3907 Kito Cheng <kito.cheng@gmail.com>
3908 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3911 UNSPEC_FFMISM and UNSPEC_FLMISM.
3912 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3913 for ffb, ffmism and flmism.
3914 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3915 (unspec_ffmism): Ditto.
3916 (unspec_flmism): Ditto.
3917 (nds32_expand_builtin_impl): Check if string extension is available.
3918 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3919 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3921 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3924 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3927 * lra-assigns.c (assign_by_spills): Return a flag of reload
3928 assignment failure. Do not process the reload assignment
3929 failures. Do not spill other reload pseudos if they has the same
3931 (lra_assign): Add a return arg. Set up from the result of
3932 assign_by_spills call.
3933 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3934 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3935 usage_insns if it is not NULL.
3936 (spill_hard_reg_in_range): New function.
3937 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3938 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3939 function prototypes.
3940 (lra_assign): Change prototype.
3941 * lra.c (lra): Add code to deal with fails by splitting hard reg
3944 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3947 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3949 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3952 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3953 (dw_val_node): Add val_symbolic_view.
3954 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3955 (symview_upper_bound): New.
3956 (new_line_info_table): Initialize symviews_since_reset.
3957 (dwarf2out_source_line): Count symviews_since_reset and set
3958 symview_upper_bound.
3959 (dw_val_equal_p): Handle symview.
3960 (add_AT_symview): New.
3961 (print_dw_val): Handle symview.
3962 (attr_checksum, attr_checksum_ordered): Likewise.
3963 (same_dw_val_p, size_of_die): Likewise.
3964 (value_format, output_die): Likewise.
3965 (add_high_low_attributes): Use add_AT_symview for entry_view.
3966 (dwarf2out_finish): Reset symview_upper_bound, clear
3969 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3972 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3973 Add strict argument and use it.
3974 (rs6000_split_multireg_move): Update for new strict argument.
3975 (mem_operand_gpr): Disallow all non-offsettable addresses.
3976 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3978 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3981 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3982 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3983 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3986 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3987 decl, use remap_type if we want to use the type.
3989 2018-03-09 Martin Sebor <msebor@redhat.com>
3991 PR tree-optimization/84526
3992 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3994 (builtin_access::generic_overlap): Be prepared to handle non-array
3997 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3999 PR rtl-optimization/84682
4000 * lra-constraints.c (process_address_1): Check is_address flag
4001 for address constraints.
4002 (process_alt_operands): Likewise.
4003 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4004 preprocess_constraints.
4005 * recog.h (preprocess_constraints): Add oploc parameter.
4007 * recog.c (preprocess_constraints): Test address_operand for
4008 CT_ADDRESS constraints.
4010 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4013 * lra-assigns.c (assign_by_spills): Return a flag of reload
4014 assignment failure. Do not process the reload assignment
4015 failures. Do not spill other reload pseudos if they has the same
4017 (lra_assign): Add a return arg. Set up from the result of
4018 assign_by_spills call.
4019 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4020 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4021 usage_insns if it is not NULL.
4022 (spill_hard_reg_in_range): New function.
4023 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4024 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4025 function prototypes.
4026 (lra_assign): Change prototype.
4027 * lra.c (lra): Add code to deal with fails by splitting hard reg
4030 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4033 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4034 Accept complain bool parameter. Only emit errors if it is true.
4035 (arm_parse_cpu_option_name): Likewise.
4036 (arm_target_thumb_only): Adjust callers of the above.
4037 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4038 prototype to take a default true bool parameter.
4039 (arm_parse_arch_option_name): Likewise.
4041 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4042 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4046 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4047 * configure: Regenerate.
4048 * configure.ac ("linker --version-script option"): New.
4049 ("linker soname option"): New.
4051 2018-03-09 Richard Biener <rguenther@suse.de>
4053 PR tree-optimization/84775
4054 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4055 immediate uses of predicate stmts and mark them modified.
4058 PR tree-optimization/84178
4059 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4061 (version_loop_for_if_conversion): Delay update_ssa call.
4062 (tree_if_conversion): Delay update_ssa until after predicate
4065 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4068 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4069 when the function accesses prior frames.
4071 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4074 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4075 gen_llsym, otherwise call maybe_gen_llsym.
4078 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4079 has ',' character inside of it.
4081 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4084 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4085 as clobbering CC_REGNUM.
4087 2018-03-08 Richard Biener <rguenther@suse.de>
4090 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4091 (follow_copies_to_constant): Do not follow SSA names registered
4094 2018-03-08 Richard Biener <rguenther@suse.de>
4096 PR tree-optimization/84178
4097 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4099 (version_loop_for_if_conversion): Delay update_ssa call.
4100 (tree_if_conversion): Delay update_ssa until after predicate
4103 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4105 PR tree-optimization/84178
4106 * tree-if-conv.c (release_bb_predicate): Remove the
4107 the assertion that the stmts have NULL use_ops.
4108 Discard the statements, asserting that they haven't
4109 yet been added to a BB.
4111 2018-03-08 Richard Biener <rguenther@suse.de>
4113 PR tree-optimization/84746
4114 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4115 (phi_translate): Pass in destination ANTIC_OUT set.
4116 (phi_translate_1): Likewise. For a simplified result lookup
4117 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4118 (phi_translate_set): Adjust.
4119 (do_pre_regular_insertion): Likewise.
4120 (do_pre_partial_partial_insertion): Likewise.
4122 2018-03-08 Martin Liska <mliska@suse.cz>
4124 PR gcov-profile/84735
4125 * doc/gcov.texi: Document usage of profile files.
4126 * gcov-io.h: Document changes in the format.
4128 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4132 * dwarf2out.c (struct dw_line_info_table): Update comments for
4134 (FORCE_RESET_NEXT_VIEW): New.
4135 (FORCE_RESETTING_VIEW_P): New.
4136 (RESETTING_VIEW_P): Check for -1 too.
4137 (ZERO_VIEW_P): Likewise.
4138 (new_line_info_table): Force-reset next view.
4139 (dwarf2out_begin_function): Likewise.
4140 (dwarf2out_source_line): Simplify zero_view_p initialization.
4141 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4142 view directly. Omit view when omitting .loc at line 0.
4144 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4146 PR tree-optimization/84740
4147 * tree-switch-conversion.c (process_switch): Call build_constructors
4148 only if info.phi_count is non-zero.
4150 PR tree-optimization/84739
4151 * tree-tailcall.c (find_tail_calls): Check call arguments against
4152 DECL_ARGUMENTS (current_function_decl) rather than
4153 DECL_ARGUMENTS (func) when checking for tail recursion.
4155 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4157 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4158 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4159 Volker Reichelt's entry and add entries for people that perform
4160 GCC fuzzy testing and report numerous bugs.
4162 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4165 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4166 readonly data in sdata, if that is disabled.
4167 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4168 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4169 -mreadonly-in-sdata option.
4171 2018-03-07 Martin Sebor <msebor@redhat.com>
4173 PR tree-optimization/84468
4174 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4175 basic block when looking for nul assignment.
4177 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4180 * except.h (output_function_exception_table): Adjust prototype.
4181 * except.c (output_function_exception_table): Remove FNNAME parameter
4182 and add SECTION parameter. Ouput one part of the table at a time.
4183 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4184 the first part of the exception table and emit unwind directives.
4185 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4186 (i386_pe_seh_cold_init): Likewise.
4187 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4188 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4189 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4190 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4191 * config/i386/winnt.c: Include except.h.
4192 (struct seh_frame_state): Add reg_offset, after_prologue and
4193 in_cold_section fields.
4194 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4195 (i386_pe_seh_cold_init): New function.
4196 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4197 to seh->in_cold_section.
4198 (seh_emit_push): Record the offset of the push.
4199 (seh_emit_save): Record the offet of the save.
4200 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4201 Test seh->after_prologue to disregard the epilogue.
4202 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4203 (i386_pe_end_cold_function): New function.
4205 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4208 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4209 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4212 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4214 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4215 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4218 * multiple_target.c: Include tree-inline.h and intl.h.
4219 (expand_target_clones): Diagnose and fail if node->definition and
4220 !tree_versionable_function_p (node->decl).
4222 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4224 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4226 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4227 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4228 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4230 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4233 * combine.c (try_combine): Use reg_or_subregno instead of handling
4234 just paradoxical SUBREGs and REGs.
4236 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4238 * config/arc/arc.c (arc_finalize_pic): Remove function.
4239 (arc_must_save_register): We use single base PIC register, remove
4240 checks to save/restore the PIC register.
4241 (arc_expand_prologue): Likewise.
4242 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4244 (arc_verify_short): Likewise.
4245 (arc_attr_type): Likewise.
4246 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4247 (walk_stores): Likewise.
4248 (arc_address_cost): Make it static.
4249 (arc_verify_short): Likewise.
4250 (branch_dest): Likewise.
4251 (arc_attr_type): Likewise.
4252 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4253 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4254 (arc_final_prescan_insn): Remove inserting the nops due to
4255 hardware hazards. It is done in reorg step.
4256 (insn_length_variant_t): Remove.
4257 (insn_length_parameters_t): Likewise.
4258 (arc_insn_length_parameters): Likewise.
4259 (arc_get_insn_variants): Likewise.
4260 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4262 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4265 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4268 PR tree-optimization/84687
4269 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4271 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4273 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4275 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4276 Rename to ppc_speculation_barrier.
4277 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4278 __builtin_ppc_speculation_barrier.
4280 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4283 * combine.c (combine_simplify_rtx): Don't try to simplify if
4284 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4287 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4289 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4290 to 32 bytes when compiling for POWER9.
4292 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4295 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4296 regparm >= 3 with no arg reg available also for calls with
4297 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4300 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4302 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4304 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4307 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4309 2018-03-05 Richard Biener <rguenther@suse.de>
4311 PR tree-optimization/84486
4312 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4313 When inserting a __builtin_assume_aligned call set the LHS
4314 SSA name alignment info accordingly.
4316 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4318 PR tree-optimization/84114
4319 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4320 Avoid reassociation of FLOAT_MODE addition.
4322 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4324 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4325 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4326 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4327 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4328 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4329 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4330 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4332 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4333 __builtin_ia32_wbinvd): New builtins.
4334 (SPECIAL_ARGS2): New.
4335 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4336 (SPECIAL_ARGS2): New.
4337 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4338 (ix86_valid_target_attribute_inner_p): Ditto.
4339 (ix86_init_mmx_sse_builtins): Add special_args2.
4340 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4341 TARGET_WBNOINVD_P): New.
4342 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4343 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4344 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4345 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4346 * config/i386/pconfigintrin.h: New file.
4347 * config/i386/wbnoinvdintrin.h: Ditto.
4348 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4349 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4351 2018-03-05 Richard Biener <rguenther@suse.de>
4353 PR tree-optimization/84670
4354 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4356 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4357 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4358 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4359 doesn't grow if all successors (recursively) were visited at least
4362 2018-03-05 Richard Biener <rguenther@suse.de>
4364 PR tree-optimization/84650
4365 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4366 if executed in the loop pipeline.
4368 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4370 * doc/configfiles.texi (Configuration Files): Move info about
4371 conditionalizing $target-protos.h to...
4372 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4373 differs from $target-protos.h.
4375 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4376 Chung-Ju Wu <jasonwucj@gmail.com>
4378 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4379 * config/nds32/nds32-multiple.md (setmemsi): Define.
4380 * config/nds32/nds32-memory-manipulation.c
4381 (nds32_gen_dup_4_byte_to_word_value): New.
4382 (emit_setmem_word_loop): New.
4383 (emit_setmem_byte_loop): New.
4384 (nds32_expand_setmem_loop): New.
4385 (nds32_expand_setmem_loop_v3m): New.
4386 (nds32_expand_setmem_unroll): New.
4387 (nds32_expand_setmem): New.
4389 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4390 Chung-Ju Wu <jasonwucj@gmail.com>
4392 * config/nds32/nds32-memory-manipulation.c
4393 (nds32_emit_load_store): New.
4394 (nds32_emit_post_inc_load_store): New.
4395 (nds32_emit_mem_move): New.
4396 (nds32_emit_mem_move_block): New.
4397 (nds32_expand_movmemsi_loop_unknown_size): New.
4398 (nds32_expand_movmemsi_loop_known_size): New.
4399 (nds32_expand_movmemsi_loop): New.
4400 (nds32_expand_movmemsi_unroll): New.
4401 (nds32_expand_movmemqi): Rename ...
4402 (nds32_expand_movmemsi): ... to this.
4403 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4404 (movmemsi): ... to this.
4405 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4406 (nds32_expand_movmemsi): ... to this.
4408 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4409 Monk Chiang <sh.chiang04@gmail.com>
4410 Chung-Ju Wu <jasonwucj@gmail.com>
4412 * config/nds32/nds32-protos.h
4413 (nds32_expand_load_multiple): New arguments.
4414 (nds32_expand_store_multiple): Ditto.
4415 (nds32_valid_multiple_load_store): Rename ...
4416 (nds32_valid_multiple_load_store_p): ... to this.
4417 * config/nds32/nds32-memory-manipulation.c
4418 (nds32_expand_load_multiple): Refine implementation.
4419 (nds32_expand_store_multiple): Ditto.
4420 * config/nds32/nds32-multiple.md
4421 (load_multiple): Update nds32_expand_load_multiple interface.
4422 (store_multiple): Update nds32_expand_store_multiple interface.
4423 * config/nds32/nds32-predicates.c
4424 (nds32_valid_multiple_load_store): Rename ...
4425 (nds32_valid_multiple_load_store_p): ... to this and refine
4427 * config/nds32/predicates.md
4428 (nds32_load_multiple_and_update_address_operation): New predicate.
4429 (nds32_store_multiple_and_update_address_operation): New predicate.
4431 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4432 Chung-Ju Wu <jasonwucj@gmail.com>
4434 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4435 (combo): New attribute.
4436 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4438 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4440 * config/nds32/nds32.opt: Change -mcmodel= default value.
4442 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4443 Monk Chiang <sh.chiang04@gmail.com>
4444 Chung-Ju Wu <jasonwucj@gmail.com>
4446 * config/nds32/constants.md (unspec_element): New enum.
4447 * config/nds32/constraints.md (Umw): New constraint.
4448 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4449 * config/nds32/nds32-intrinsic.md: Likewise.
4450 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4451 (nds32_valid_smw_lwm_base_p): New.
4452 (nds32_output_smw_single_word): New.
4453 (nds32_output_lmw_single_word): New.
4454 (nds32_expand_unaligned_load): New.
4455 (nds32_expand_unaligned_store): New.
4456 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4457 (nds32_output_smw_single_word): Declare.
4458 (nds32_output_lmw_single_word): Declare.
4459 (nds32_expand_unaligned_load): Declare.
4460 (nds32_expand_unaligned_store): Declare.
4461 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4462 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4463 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4464 NDS32_BUILTIN_UASTORE_DW.
4465 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4468 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4469 Kito Cheng <kito.cheng@gmail.com>
4470 Chung-Ju Wu <jasonwucj@gmail.com>
4472 * config/nds32/nds32-intrinsic.c
4473 (nds32_expand_builtin_null_ftype_reg): Delete.
4474 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4475 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4476 (nds32_read_argument): New.
4477 (nds32_legitimize_target): Ditto.
4478 (nds32_legitimize_argument): Ditto.
4479 (nds32_check_constant_argument): Ditto.
4480 (nds32_expand_unop_builtin): Ditto.
4481 (nds32_expand_unopimm_builtin): Ditto.
4482 (nds32_expand_binop_builtin): Ditto.
4483 (nds32_builtin_decl_impl): Ditto.
4484 (builtin_description): Ditto.
4485 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4486 (nds32_init_builtins_impl): Ditto.
4487 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4488 (nds32_builtin_decl): New.
4489 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4490 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4492 2018-03-02 Jeff Law <law@redhat.com>
4494 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4495 (redundant_insn, fill_simple_delay_slots): Likewise.
4496 (fill_slots_from_thread): Likewise.
4497 * resource.c (mark_referenced_resources): Likewise.
4498 (mark_set_resources, find_dead_or_set_registers): Likewise.
4500 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4502 * substring-locations.h (format_warning_va): Formatting fix for
4504 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4506 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4507 * substring-locations.c: Include intl.h.
4508 (format_warning_va): Turned into small wrapper around
4509 format_warning_n_va, renamed to ...
4510 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4511 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4513 (format_warning_at_substring_n): New function.
4514 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4515 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4516 format_warning_at_substring with just a shorter name instead of
4517 const function pointer.
4518 (fmtwarn_n): New function.
4519 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4520 appropriate, get rid of all the fmtstr temporaries, move conditionals
4521 with G_() wrapped string literals directly into fmtwarn arguments,
4522 cast dir.len to (int), formatting fixes.
4524 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4526 * doc/invoke.texi: Remove "Cilk Plus" references.
4528 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4529 Richard Biener <rguenther@suse.de>
4532 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4533 for error or warning attributes if CALL_FROM_THUNK_P is set.
4536 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4539 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4540 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4543 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4544 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4546 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4547 instead of -1U in last predictors element's probability member.
4549 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4552 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4553 arguments if they are comparable.
4555 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4557 PR tree-optimization/84634
4558 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4559 masks and masked_loop_p with a single loop_masks, making sure it's
4560 null for bb vectorization.
4562 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4564 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4565 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4566 loop->force_vectorize to check whether there is no alias.
4568 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4571 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4573 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4575 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4576 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4577 instead of a loop around prev_real_insn.
4578 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4582 * config/i386/i386.c (ix86_print_operand): Use conditional
4583 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4586 2018-03-02 Richard Biener <rguenther@suse.de>
4588 PR tree-optimization/84427
4589 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4590 (bitmap_set_subtract_values): Rewrite to handle multiple
4593 (prune_clobbered_mems): Likewise.
4594 (phi_translate): Take edge instead of pred/phiblock.
4595 (phi_translate_1): Likewise.
4596 (phi_translate_set): Likewise. Insert all translated
4597 exprs for a value into the set, keeping possibly multiple
4598 expressions per value.
4599 (compute_antic_aux): Adjust for phi_translate changes.
4600 When intersecting union the expressions and prune those
4601 not in the final value set, keeping possibly multiple
4602 expressions per value. Do not use value-insertion
4603 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4604 all expressions. Add verification that the value-sets
4605 only shrink during iteration.
4606 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4607 (do_pre_regular_insertion): Likewise.
4608 (do_pre_partial_partial_insertion): Likewise.
4610 2018-03-02 Richard Biener <rguenther@suse.de>
4613 * config/darwin.c (saved_debug_info_level): New static global.
4614 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4615 (darwin_asm_lto_end): Restore debug info generation settings.
4617 2018-03-01 Martin Liska <mliska@suse.cz>
4620 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4623 2018-03-01 Richard Biener <rguenther@suse.de>
4626 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4627 type annotation with LTO when debug was disabled at compile-time.
4629 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4631 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4633 (mips_final_postscan_insn): Likewise.
4635 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4637 PR rtl-optimization/84528
4638 * alias.c (init_alias_target): Add commentary.
4639 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4640 a unique base value if the frame pointer is not eliminated
4641 to the stack pointer.
4643 2018-03-01 Tom de Vries <tom@codesourcery.com>
4645 PR rtl-optimization/83327
4646 * lra-int.h (hard_regs_spilled_into): Declare.
4647 * lra.c (hard_regs_spilled_into): Define.
4648 (init_reg_info): Init hard_regs_spilled_into.
4649 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4650 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4651 (process_bb_lives): Handle hard_regs_spilled_into.
4652 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4653 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4655 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4657 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4658 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4659 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4660 * config/rs6000/aix72.h: New file.
4662 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4664 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4665 instead of warning_at with conditional singular and plural messages
4669 * stor-layout.c (update_alignment_for_field): For
4670 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4671 && !DECL_PACKED (field), do the alignment update, just use
4672 only desired_align instead of MAX (type_align, desired_align)
4674 (place_field): Don't do known_align < desired_align handling
4675 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4676 is non-NULL, instead do it after rli->prev_field handling and
4677 only if not within a bitfield word. For DECL_PACKED (field)
4678 use type_align of BITS_PER_UNIT.
4680 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4682 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4683 superfluous parentheses and trailing spaces.
4685 2018-02-28 Richard Biener <rguenther@suse.de>
4687 PR tree-optimization/84584
4688 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4689 SCoPs with fake exit edge.
4691 2018-02-28 Martin Liska <mliska@suse.cz>
4694 * timevar.c (timer::print): Fix format to properly print 100%
4697 2018-02-28 Richard Biener <rguenther@suse.de>
4700 * genmatch.c (capture_info::walk_match): Do not mark
4701 captured expressions without operands as expr_p given
4702 they act more like predicates and should be subject to
4703 "lost tail" side-effect preserving.
4705 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4707 PR rtl-optimization/81611
4708 * auto-inc-dec.c (attempt_change): Move dead note from
4709 mem_insn if it's the next use of regno
4710 (find_address): Take address use of reg holding
4711 non-incremented value. Add parm to limit search to the named
4713 (merge_in_block): Attempt to use a mem insn that is the next
4714 use of the original regno.
4716 2018-02-27 Martin Sebor <msebor@redhat.com>
4719 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4720 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4722 2018-02-27 Martin Sebor <msebor@redhat.com>
4724 PR translation/84207
4725 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4726 n argument to unsigned HOST_WIDE_INT.
4727 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4728 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4729 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4730 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4732 2018-02-27 Richard Biener <rguenther@suse.de>
4734 PR tree-optimization/84512
4735 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4736 Do not use the estimate returned from record_stmt_cost for
4737 the scalar iteration cost but sum properly using add_stmt_cost.
4739 2018-02-27 Richard Biener <rguenther@suse.de>
4741 PR tree-optimization/84466
4742 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4743 Adjust last change to less strictly validate use operands.
4745 2018-02-27 Martin Liska <mliska@suse.cz>
4747 PR gcov-profile/84548
4748 * gcov.c (process_file): Allow partial overlap and consider it
4749 also as group functions.
4750 (output_lines): Properly calculate range of lines for a group.
4752 2018-02-27 Martin Liska <mliska@suse.cz>
4754 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4755 'ggc' suffixes. Change first column width.
4756 (timer::print): Fix formatting of the column.
4758 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4760 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4761 preserve inline entry blocks for the sake of debug inline
4762 entry point markers alone.
4763 (remove_unused_locals): Suggest in comments a better place to
4764 force the preservation of inline entry blocks that are
4765 otherwise unused, but do not preserve them.
4767 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4769 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4771 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4774 * config/i386/constraints.md (Bs): Replace
4775 ix86_indirect_branch_register with
4776 TARGET_INDIRECT_BRANCH_REGISTER.
4778 * config/i386/i386.md (indirect_jump): Likewise.
4779 (tablejump): Likewise.
4780 (*sibcall_memory): Likewise.
4781 (*sibcall_value_memory): Likewise.
4782 Peepholes of indirect call and jump via memory: Likewise.
4783 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4784 (*sibcall_value_GOT_32): Likewise.
4785 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4786 (GOT_memory_operand): Likewise.
4787 (call_insn_operand): Likewise.
4788 (sibcall_insn_operand): Likewise.
4789 (GOT32_symbol_operand): Likewise.
4790 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4792 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4794 PR rtl-optimization/83496
4795 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4796 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4797 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4798 redundant insn, if any.
4799 (relax_delay_slots): Likewise.
4800 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4802 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4804 PR tree-optimization/83965
4805 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4806 that grouped statements are part of a reduction chain. Return
4807 true if the statement is not marked as a reduction itself but
4809 (vect_recog_dot_prod_pattern): Don't check whether the statement
4810 is part of a group here.
4811 (vect_recog_sad_pattern): Likewise.
4812 (vect_recog_widen_sum_pattern): Likewise.
4814 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4817 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4819 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4822 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4824 (ix86_output_indirect_function_return): New prototype.
4825 (ix86_split_simple_return_pop_internal): Likewise.
4826 * config/i386/i386.c (indirect_return_via_cx): New.
4827 (indirect_return_via_cx_bnd): Likewise.
4828 (indirect_thunk_name): Handle return va CX_REG.
4829 (output_indirect_thunk_function): Create alias for
4830 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4831 (ix86_output_indirect_jmp): Remove the bool argument.
4832 (ix86_output_indirect_function_return): New function.
4833 (ix86_split_simple_return_pop_internal): Likewise.
4834 * config/i386/i386.md (*indirect_jump): Don't pass false
4835 to ix86_output_indirect_jmp.
4836 (*tablejump_1): Likewise.
4837 (simple_return_pop_internal): Change it to define_insn_and_split.
4838 Call ix86_split_simple_return_pop_internal to split it for
4840 (simple_return_indirect_internal): Call
4841 ix86_output_indirect_function_return instead of
4842 ix86_output_indirect_jmp.
4844 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4847 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4848 memset and value initialization afterwards.
4850 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4852 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4854 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4857 * common/config/aarch64/aarch64-common.c
4858 (aarch_option_optimization_table[]): Switch
4859 off fomit-frame-pointer
4861 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4862 Chung-Ju Wu <jasonwucj@gmail.com>
4864 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4866 (store_multiple): Ditto.
4868 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4870 * config.gcc: Add --with-cpu support for nds32 target.
4871 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4872 * config/nds32/nds32.opt: Add -mcpu= option.
4874 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4876 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4877 isel=yes): Warn for these deprecated options.
4879 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4881 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4882 ISA_2_5_MASKS_EMBEDDED.
4884 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4886 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4887 p->max as pointers rather than using iterative_hash_expr.
4889 2018-02-23 Carl Love <cel@us.ibm.com>
4891 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4892 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4894 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4895 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4896 P8V_BUILTIN_VEC_VUNSIGNED2.
4898 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4901 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4902 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4904 * lra-constraints.c (curr_insn_transform): Set up
4905 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4907 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4908 * lra-eliminations.c (spill_pseudos): Ditto.
4909 (process_insn_for_elimination): Ditto.
4910 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4911 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4912 LRA_NON_CLOBBERED_ALT.
4914 2018-02-22 Martin Sebor <msebor@redhat.com>
4916 PR tree-optimization/84480
4917 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4918 to maybe_diag_stxncpy_trunc. Call it.
4919 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4920 from gimple_fold_builtin_strcpy. Print inlining stack.
4921 (handle_builtin_stxncpy): Print inlining stack.
4922 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4924 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4927 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4928 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4929 and -fcheck-pointer-bounds are used together.
4930 (indirect_thunk_prefix): New enum.
4931 (indirect_thunk_need_prefix): New function.
4932 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4933 "_nt" instead of "_bnd" for NOTRACK prefix.
4934 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4935 (output_indirect_thunk_function): Likewise.
4937 (ix86_code_end): Update output_indirect_thunk_function calls.
4938 (ix86_output_indirect_branch_via_reg): Replace
4939 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4940 (ix86_output_indirect_branch_via_push): Likewise.
4941 (ix86_output_function_return): Likewise.
4942 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4943 incompatible with -fcf-protection=branch and
4944 -fcheck-pointer-bounds.
4946 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4949 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4950 Change gcc_assert call to output_operand_lossage.
4952 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4954 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4956 2018-02-22 DJ Delorie <dj@redhat.com>
4957 Sebastian Perta <sebastian.perta@renesas.com>
4958 Oleg Endo <olegendo@gcc.gnu.org>
4960 * config/rx/rx.c (rx_rtx_costs): New function.
4961 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4963 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4965 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4967 2018-02-22 Martin Liska <mliska@suse.cz>
4970 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4971 Add "native" as a possible value.
4973 2018-02-22 Martin Liska <mliska@suse.cz>
4976 * config/i386/i386.c (ix86_option_override_internal):
4977 Add "native" as a possible value for -march and -mtune.
4979 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4982 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4983 to all type variants.
4985 PR tree-optimization/84503
4986 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4987 width as info->bitpos + info->bitsize - start.
4988 (merged_store_group::merge_overlapping): Simplify width computation.
4989 (check_no_overlap): New function.
4990 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4991 start + width and last_order of the group, fail if check_no_overlap
4993 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4994 to group if check_no_overlap fails.
4996 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4998 * config/rs6000/altivec.md: Delete contraint arguments to
4999 define_expand, define_split, and define_peephole2, and in
5000 define_insn_and_split if always unused.
5001 * config/rs6000/darwin.md: Ditto.
5002 * config/rs6000/dfp.md: Ditto.
5003 * config/rs6000/rs6000.md: Ditto.
5004 * config/rs6000/sync.md: Ditto.
5005 * config/rs6000/vector.md: Ditto.
5006 * config/rs6000/vsx.md: Ditto.
5008 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5010 * config/rs6000/altivec.md: Write output control strings as braced
5011 blocks instead of double-quoted strings.
5012 * config/rs6000/darwin.md: Ditto.
5013 * config/rs6000/rs6000.md: Ditto.
5014 * config/rs6000/vector.md: Ditto.
5015 * config/rs6000/vsx.md: Ditto.
5017 2018-02-21 Jason Merrill <jason@redhat.com>
5019 PR c++/84314 - ICE with templates and fastcall attribute.
5020 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5022 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5024 * ipa-cp.c (determine_versionability): Fix comment typos.
5026 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5029 * ipa-cp.c (determine_versionability): Do not version functions caling
5032 2018-02-21 Martin Liska <mliska@suse.cz>
5035 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5036 Add "native" as a possible value.
5037 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5038 the macro when native cpu detection is available.
5040 2018-02-21 Martin Liska <mliska@suse.cz>
5043 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5044 Add "native" as a possible value.
5045 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5046 when native cpu detection is available.
5048 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5049 Martin Sebor <msebor@redhat.com>
5051 PR tree-optimization/84478
5052 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5054 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5055 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5056 support which is conservatively correct, for 2 only stay conservative
5057 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5058 argument to the 2 argument get_range_strlen, adjust 6 arg
5059 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5061 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5062 (gimple_fold_builtin_strlen): Pass true as last argument to
5065 2018-02-20 Martin Sebor <msebor@redhat.com>
5068 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5069 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5070 (builtin_memref::builtin_memref): Factor out parts into
5071 set_base_and_offset and call it.
5073 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5076 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5077 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5078 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5079 search at the associated MODE_INT.
5081 2018-02-20 Jeff Law <law@redhat.com>
5084 PR tree-optimization/81592
5086 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5087 for range data rather than using global data.
5088 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5089 range data rather than using global data.
5090 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5091 pass it to children as needed.
5092 (struct directive::fmtresult): Similarly.
5093 (struct directive::set_width): Similarly.
5094 (struct directive::set_precision): Similarly.
5095 (format_integer, format_directive, parse_directive): Similarly.
5096 (format_none): Accept unnamed vr_values parameter.
5097 (format_percent, format_floating, format_character): Similarly.
5098 (format_string, format_plain): Similarly.
5099 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5100 the EVRP range analyzer for range data rather than using global data.
5101 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5102 gimple-ssa-evrp-analyze.h
5103 (class sprintf_dom_walker): Add after_dom_children member function.
5104 Add evrp_range_analyzer member.
5105 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5106 range analyzer as needed.
5107 (sprintf_dom_walker::after_dom_children): New member function.
5108 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5110 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5111 (evrp_range_analyzer::pop_to_marker): Likewise.
5113 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5115 PR tree-optimization/84419
5116 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5117 with the required type if its current type is compatible but
5120 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5123 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5124 after vectorization.
5126 2018-02-20 Martin Liska <mliska@suse.cz>
5129 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5130 possible values if we don't have a hint.
5132 2018-02-20 Martin Liska <mliska@suse.cz>
5136 * final.c (shorten_branches): Build align_tab array with one
5138 * opts.c (finish_options): Add alignment option limit check.
5139 (MAX_CODE_ALIGN): Likewise.
5140 (MAX_CODE_ALIGN_VALUE): Likewise.
5141 * doc/invoke.texi: Document maximum allowed option value for
5142 all -falign-* options.
5144 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5147 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5148 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5149 * var-tracking.c (emit_note_insn_var_location): Remove all references
5150 to NOTE_INSN_CALL_ARG_LOCATION.
5151 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5152 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5153 Use copy_rtx_if_shared.
5154 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5155 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5156 (dwarf2out_var_location): Remove handling of
5157 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5159 * final.c (final_scan_insn): Remove all references to
5160 NOTE_INSN_CALL_ARG_LOCATION.
5161 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5162 before dumping final insns.
5163 * except.c (emit_note_eh_region_end): Remove all references to
5164 NOTE_INSN_CALL_ARG_LOCATION.
5165 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5166 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5167 * config/arc/arc.c (hwloop_optimize): Likewise.
5168 * config/arm/arm.c (create_fix_barrier): Likewise.
5169 * config/s390/s390.c (s390_chunkify_start): Likewise.
5170 * config/sh/sh.c (find_barrier): Likewise.
5171 * config/i386/i386.c (rest_of_insert_endbranch,
5172 ix86_seh_fixup_eh_fallthru): Likewise.
5173 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5174 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5175 * config/frv/frv.c (frv_function_prologue): Likewise.
5176 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5178 (note_outside_basic_block_p): Remove all references to
5179 NOTE_INSN_CALL_ARG_LOCATION.
5180 * gengtype.c (adjust_field_rtx_def): Likewise.
5181 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5183 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5184 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5187 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5190 PR tree-optimization/84452
5191 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5192 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5195 2018-02-19 Martin Liska <mliska@suse.cz>
5198 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5200 2018-02-19 Martin Liska <mliska@suse.cz>
5201 Richard Sandiford <richard.sandiford@linaro.org>
5203 PR tree-optimization/82491
5204 * gimple-fold.c (get_base_constructor): Make earlier bail out
5207 2018-02-19 Carl Love <cel@us.ibm.com>
5209 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5210 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5212 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5213 P8V_BUILTIN_VEC_NEG.
5215 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5217 * config/rl78/rl78.md (movdf): New define expand.
5219 2018-02-19 Martin Liska <mliska@suse.cz>
5222 * doc/invoke.texi: Fix typo.
5223 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5225 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5227 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5228 handle rs6000_single_float and rs6000_double_float specially for
5231 2018-02-16 Jeff Law <law@redhat.com>
5233 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5236 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5239 * ipa-inline.c (inline_small_functions): Fix a typo.
5241 2018-02-16 Nathan Sidwell <nathan@acm.org>
5243 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5245 2018-02-16 Carl Love <cel@us.ibm.com>
5247 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5248 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5249 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5250 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5251 expansion to P8V_BUILTIN_VEC_FLOAT2.
5253 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5255 PR rtl-optimization/70023
5256 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5257 src_regno into account.
5259 2018-02-16 Carl Love <cel@us.ibm.com>
5261 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5262 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5263 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5264 * config/rs6000/rs6000.c: Remove case statements for
5265 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5266 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5267 and P9V_BUILTIN_VEC_VINSERT4B.
5268 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5269 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5270 * config/rs6000/vsx.md:
5271 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5274 2018-02-16 Carl Love <cel@us.ibm.com>
5276 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5278 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5280 * config/rs6000/rs6000-c.c: Add the definitions for
5281 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5282 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5283 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5284 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5285 definition for insert4b and define insn *insert3b_internal.
5286 * doc/extend.texi: Add documentation for vec_extract4b.
5288 2018-02-16 Nathan Sidwell <nathan@acm.org>
5290 * doc/extend.texi (Backwards Compatibility): Mention friend
5291 injection. Note for-scope is deprecated.
5292 * doc/invoke.texi (-ffriend-injection): Deprecate.
5294 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5296 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5297 that moved to I2, also allow destinations that are a paradoxical
5298 subreg (instead of a normal reg).
5300 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5303 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5306 2018-02-16 Richard Biener <rguenther@suse.de>
5308 PR tree-optimization/84037
5309 PR tree-optimization/84016
5311 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5312 Adjust vec_construct for the fact we need additional higher latency
5313 128bit inserts for AVX256 and AVX512 vector builds.
5314 (ix86_add_stmt_cost): Scale vector construction cost for
5317 2018-02-16 Richard Biener <rguenther@suse.de>
5319 PR tree-optimization/84417
5320 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5321 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5322 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5324 2018-02-16 Martin Liska <mliska@suse.cz>
5327 * internal-fn.def (ASAN_CHECK): Set proper flags.
5328 (ASAN_MARK): Likewise.
5330 2018-02-16 Julia Koval <julia.koval@intel.com>
5332 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5333 from PTA_CANNONLAKE.
5335 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5338 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5339 Use ++iter rather than iter++ for std::list iterators.
5340 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5341 defer deleting them until all nodes in the forest are processed. Do
5342 free even leaf nodes. Change to_process into auto_vec.
5345 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5346 * vec.h (vec_default_construct): Use memset instead of placement new
5347 if BROKEN_VALUE_INITIALIZATION is defined.
5348 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5349 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5352 PR rtl-optimization/83723
5353 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5354 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5355 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5357 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5359 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5361 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5363 PR rtl-optimization/81443
5364 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5365 from inner REGs to paradoxical SUBREGs.
5367 2018-02-16 Richard Biener <rguenther@suse.de>
5369 PR tree-optimization/84399
5370 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5371 For operands we can analyze at their definition make sure we can
5372 analyze them at each use as well.
5374 2018-02-16 Richard Biener <rguenther@suse.de>
5376 PR tree-optimization/84190
5377 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5378 volatile accesses if the decl isn't volatile.
5380 2018-02-15 Jason Merrill <jason@redhat.com>
5382 PR c++/84314 - ICE with templates and fastcall attribute.
5383 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5384 TYPE_CANONICAL on an existing type.
5386 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5388 PR tree-optimization/84383
5389 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5390 dstoff nor call operand_equal_p if dstbase is NULL.
5392 PR tree-optimization/84334
5393 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5394 also a CONSTANT_CLASS_P, punt.
5396 2018-02-14 Jim Wilson <jimw@sifive.com>
5398 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5399 first SMALL_OPERAND check. New local min_second_step. Move assert
5400 to where locals are set. Add TARGET_RVC support.
5401 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5403 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5405 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5407 2018-02-14 Martin Sebor <msebor@redhat.com>
5409 PR tree-optimization/83698
5410 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5411 arrays constrain the offset range to their bounds.
5412 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5413 (builtin_access::overlap): Avoid setting the size of overlap if it's
5415 (maybe_diag_overlap): Also consider arrays when deciding what values
5416 of offsets to include in diagnostics.
5418 2018-02-14 Martin Sebor <msebor@redhat.com>
5421 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5422 that correspond to the kind of a declaration.
5424 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5427 * config/pa/pa.md: Load address of PIC label using the linkage table
5428 if the label is nonlocal.
5430 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5432 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5433 warning message if user requests -maltivec=be.
5434 * doc/invoke.texi: Document deprecation of -maltivec=be.
5436 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5439 * config/rs6000/rs6000-c.c: Update definitions for
5440 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5441 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5443 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5446 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5447 add _get_ssp intrinsics. Remove argument from
5448 __builtin_ia32_rdssp[d|q].
5449 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5450 * config/i386/i386-builtin.def: Remove argument from
5451 __builtin_ia32_rdssp[d|q].
5452 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5453 ix86_expand_special_args_builtin for _rdssp[d|q].
5454 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5455 Clear register before usage.
5456 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5457 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5459 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5461 PR tree-optimization/84357
5462 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5463 operand 1 of an ARRAY_REF too.
5465 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5468 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5469 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5471 (set_of_reg): New struct.
5472 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5473 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5474 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5476 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5477 Split into bitclr, bitset, bitinvert patterns if appropriate.
5478 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5479 use rx_fuse_in_memory_bitop.
5480 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5481 to named insn, correct maximum insn length.
5483 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5486 * machmode.def: Define a complex mode for PARTIAL_INT.
5487 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5489 * doc/rtl.texi: Document CSPImode.
5490 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5492 (msp430_hard_regno_nregs_with_padding): Likewise.
5494 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5497 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5499 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5501 PR rtl-optimization/84169
5502 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5503 we generated a parallel as new i3 and we split that to new i2 and i3
5504 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5505 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5506 those to i2, not i1. Partially rewrite this scan code.
5508 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5511 * stor-layout.c (place_field): For variable length fields, adjust
5512 offset_align afterwards not just based on the field's alignment,
5513 but also on the size.
5516 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5517 of exps and logs in the use_exp2 case.
5519 2018-02-13 Jeff Law <law@redhat.com>
5521 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5524 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5527 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5531 * common.opt (gas-loc-support, gas-locview-support): New.
5532 (ginline-points, ginternal-reset-location-views): New.
5533 * doc/invoke.texi: Document them. Use @itemx where intended.
5534 (gvariable-location-views): Adjust.
5535 * target.def (reset_location_view): New.
5536 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5537 (TARGET_RESET_LOCATION_VIEW): New.
5538 * doc/tm.texi: Rebuilt.
5539 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5540 (dwarf2out_default_as_locview_support): New.
5541 (output_asm_line_debug_info): Use option variables.
5542 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5543 (output_loc_list): Likewise.
5544 (add_high_low_attributes): Check option variables.
5545 Don't output entry view attribute in strict mode.
5546 (gen_inlined_subroutine_die): Check option variables.
5547 (dwarf2out_inline_entry): Likewise.
5548 (init_sections_and_labels): Likewise.
5549 (dwarf2out_early_finish): Likewise.
5550 (maybe_reset_location_view): New, from...
5551 (dwarf2out_var_location): ... here. Call it.
5552 * debug.h (dwarf2out_default_as_loc_support): Declare.
5553 (dwarf2out_default_as_locview_support): Declare.
5554 * hooks.c (hook_int_rtx_insn_0): New.
5555 * hooks.h (hook_int_rtx_insn_0): Declare.
5556 * toplev.c (process_options): Take -gas-loc-support and
5557 -gas-locview-support from dwarf2out. Enable
5558 -gvariable-location-views by default only with locview
5559 assembler support. Enable -ginternal-reset-location-views by
5560 default only if the target defines the corresponding hook.
5561 Enable -ginline-points by default if location views are
5562 enabled; force it disabled if statement frontiers are
5564 * tree-inline.c (expand_call_inline): Check option variables.
5565 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5567 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5569 PR tree-optimization/84321
5570 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5571 handling. Also check whether the anti-range contains any values
5572 that satisfy the mask; switch to a VR_RANGE if not.
5574 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5577 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5579 2018-02-13 Martin Jambor <mjambor@suse.cz>
5582 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5583 of call statements, also set location of a load to a temporary.
5585 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5587 * config/rl78/rl78.c (add_vector_labels): New function.
5588 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5589 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5590 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5591 which checks that no arguments are passed.
5592 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5593 * doc/extend.texi: Documentation for the new attribute.
5595 2018-02-13 Andreas Schwab <schwab@suse.de>
5597 * config/riscv/linux.h (CPP_SPEC): Define.
5599 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5602 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5603 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5604 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5605 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5606 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5607 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5608 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5609 temporarily for AES and PCLMUL builtins.
5611 PR tree-optimization/84339
5612 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5613 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5617 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5618 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5619 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5621 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5623 * omp-simd-clone.h: New file.
5624 * omp-simd-clone.c: Include omp-simd-clone.h.
5625 (expand_simd_clones): No longer static.
5626 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5627 cgraph.h and omp-simd-clone.h.
5628 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5629 (vect_recog_widen_shift_pattern): Formatting fix.
5630 (vect_pattern_recog_1): Don't check optab for calls.
5633 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5634 operands[2] into a REG before using gen_lowpart on it.
5636 2018-02-12 Jeff Law <law@redhat.com>
5639 * config/sh/sh.c (find_barrier): Consider a sibling call
5642 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5643 successfully back substituting a reg.
5645 2018-02-12 Richard Biener <rguenther@suse.de>
5647 PR tree-optimization/84037
5648 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5649 parameter, move visited init to caller.
5650 (vect_slp_analyze_operations): Separate cost from validity
5651 check, initialize visited once for all instances.
5652 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5654 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5656 (vect_model_promotion_demotion_cost): Likewise.
5657 (vectorizable_bswap): Guard cost modeling with !slp_node
5658 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5660 (vectorizable_call): Likewise.
5661 (vectorizable_conversion): Likewise.
5662 (vectorizable_assignment): Likewise.
5663 (vectorizable_shift): Likewise.
5664 (vectorizable_operation): Likewise.
5665 (vectorizable_store): Likewise.
5666 (vectorizable_load): Likewise.
5667 (vectorizable_condition): Likewise.
5668 (vectorizable_comparison): Likewise.
5670 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5673 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5674 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5675 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5677 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5680 * params.def (inline-min-speedup): Increase from 8 to 15.
5681 (max-inline-insns-auto): Decrease from 40 to 30.
5682 * ipa-split.c (consider_split): Add some buffer for function to
5683 be considered inlining candidate.
5684 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5687 2018-02-12 Richard Biener <rguenther@suse.de>
5689 PR tree-optimization/84037
5690 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5691 matched stmts if we cannot swap the non-matched ones.
5693 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5695 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5696 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5697 _mm_maskz_scalef_round_ss): New intrinsics.
5698 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5699 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5700 __builtin_ia32_scalefss_round): Remove.
5701 (__builtin_ia32_scalefsd_mask_round,
5702 __builtin_ia32_scalefss_mask_round): New intrinsics.
5703 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5704 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5705 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5706 "<round_constraint>")): Changed to ...
5707 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5708 "<round_scalar_constraint>")): ... this.
5709 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5710 %0, %1, %2<round_op3>}"): Changed to ...
5711 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5712 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5713 %2<round_scalar_mask_op3>}"): ... this.
5714 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5716 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5718 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5719 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5720 (_mm_maskz_sqrt_round_ss): New intrinsics.
5721 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5722 (__builtin_ia32_sqrtsd_mask_round)
5723 (__builtin_ia32_sqrtss_mask_round): New builtins.
5724 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5725 (__builtin_ia32_sqrtss_round): Remove.
5726 (__builtin_ia32_sqrtsd_mask_round)
5727 (__builtin_ia32_sqrtss_mask_round): New builtins.
5728 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5729 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5730 ((match_operand:VF_128 1 "vector_operand"
5731 "xBm,<round_constraint>")): Changed to ...
5732 ((match_operand:VF_128 1 "vector_operand"
5733 "xBm,<round_scalar_constraint>")): ... this.
5734 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5735 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5736 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5737 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5738 %<iptr>1<round_scalar_mask_op3>}): ... this.
5739 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5740 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5742 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5745 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5746 Cast vec_cmpeq result to correct type.
5747 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5748 Cast vec_cmpgt result to correct type.
5750 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5752 * final.c (final_scan_insn_1): Renamed from...
5753 (final_scan_insn): ... this. New wrapper, to recover
5754 seen from the outermost call in recursive ones.
5755 * config/sparc/sparc.c (output_return): Drop seen from call.
5756 (output_sibcall): Likewise.
5757 * config/visium/visium.c (output_branch): Likewise.
5759 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5761 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5764 2018-02-10 Alan Modra <amodra@gmail.com>
5767 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5768 Specify LR as an input.
5770 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5773 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5774 remove_member_access_dummy_vars): New functions.
5775 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5776 lower_omp_1, execute_lower_omp): Use them.
5778 PR rtl-optimization/84308
5779 * shrink-wrap.c (spread_components): Release todo vector.
5781 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5783 PR rtl-optimization/57193
5784 * ira-color.c (struct allocno_color_data): Add member
5785 conflict_allocno_hard_prefs.
5786 (update_conflict_allocno_hard_prefs): New.
5787 (bucket_allocno_compare_func): Add a preference based on
5788 conflict_allocno_hard_prefs.
5789 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5790 (color_allocnos): Remove a dead code. Initiate
5791 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5793 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5796 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5797 constraint from =wa to wa. Avoid a subreg on the output operand,
5798 instead use a pseudo and subreg it in a move.
5799 (p9_xxbrd_<mode>): Changed to ...
5800 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5801 (p9_xxbrd_v2df): New expander.
5802 (p9_xxbrw_<mode>): Changed to ...
5803 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5804 (p9_xxbrw_v4sf): New expander.
5806 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5808 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5810 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5813 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5814 multiply in 32-bit mode.
5815 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5816 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5819 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5821 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5822 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5823 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5824 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5826 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5829 * dwarf2out.c (is_trivial_indirect_ref): New function.
5830 (dwarf2out_late_global_decl): Do not generate a location
5831 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5832 and that are not defined in the current unit.
5834 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5836 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5837 instead of a libcall for UNORDERED.
5839 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5842 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5843 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5845 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5848 * config/s390/s390.c (s390_set_current_function): Invoke
5849 s390_indirect_branch_settings also if fndecl didn't change.
5851 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5853 * config/rs6000/rs6000.md (blockage): Set length to zero.
5855 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5857 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5859 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5862 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5863 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5867 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5868 PARALLEL incoming that failed vt_get_decl_and_offset check.
5871 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5872 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5874 (get_variable_section): For decls in named .bss* sections pass true as
5875 second argument to bss_initializer_p.
5877 2018-02-09 Marek Polacek <polacek@redhat.com>
5878 Jakub Jelinek <jakub@redhat.com>
5881 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5882 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5883 Sync some changes from cxx_fold_indirect_ref.
5885 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5887 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5889 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5890 (BLOCK_INLINE_ENTRY_LABEL): New.
5891 (dwarf2out_var_location): Disregard inline entry markers.
5892 (inline_entry_data): New struct.
5893 (inline_entry_data_hasher): New hashtable type.
5894 (inline_entry_data_hasher::hash): New.
5895 (inline_entry_data_hasher::equal): New.
5896 (inline_entry_data_table): New variable.
5897 (add_high_low_attributes): Add DW_AT_entry_pc and
5898 DW_AT_GNU_entry_view attributes if a pending entry is found
5899 in inline_entry_data_table. Add old entry_pc attribute only
5900 if debug nonbinding markers are disabled.
5901 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5902 markers are enabled.
5903 (block_within_block_p, dwarf2out_inline_entry): New.
5904 (dwarf2out_finish): Check that no entries remained in
5905 inline_entry_data_table.
5906 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5907 (final_scan_insn, notice_source_line): Likewise.
5908 (rest_of_clean_state): Skip inline entry markers.
5909 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5911 * gimple.c (gimple_build_debug_inline_entry): New.
5912 * gimple.h (enum gimple_debug_subcode): Add
5913 GIMPLE_DEBUG_INLINE_ENTRY.
5914 (gimple_build_debug_inline_entry): Declare.
5915 (gimple_debug_inline_entry_p): New.
5916 (gimple_debug_nonbind_marker_p): Adjust.
5917 * insn-notes.def (INLINE_ENTRY): New.
5918 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5919 inline entry marker notes.
5920 (print_insn): Likewise.
5921 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5922 (INSN_DEBUG_MARKER_KIND): Likewise.
5923 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5924 * tree-inline.c (expand_call_inline): Build and insert
5925 debug_inline_entry stmt.
5926 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5927 inline entry blocks early, if nonbind markers are enabled.
5928 (dump_scope_block): Dump fragment info.
5929 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5930 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5931 (gimple_build_debug_inline_entry): New.
5932 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5933 Enable/disable inline entry points too.
5934 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5935 (DEBUG_INSN): Describe inline entry markers.
5937 * common.opt (gvariable-location-views): New.
5938 (gvariable-location-views=incompat5): New.
5939 * config.in: Rebuilt.
5940 * configure: Rebuilt.
5941 * configure.ac: Test assembler for view support.
5942 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5943 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5944 * dwarf2out.c (var_loc_view): New typedef.
5945 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5946 (dwarf2out_locviews_in_attribute): New.
5947 (dwarf2out_locviews_in_loclist): New.
5948 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5949 (enum dw_line_info_opcode): Add LI_adv_address.
5950 (struct dw_line_info_table): Add view.
5951 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5952 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5953 (zero_view_p): New variable.
5954 (ZERO_VIEW_P): New macro.
5955 (output_asm_line_debug_info): New.
5956 (struct var_loc_node): Add view.
5957 (add_AT_view_list, AT_loc_list): New.
5958 (add_var_loc_to_decl): Add view param. Test it against last.
5959 (new_loc_list): Add view params. Record them.
5960 (AT_loc_list_ptr): Handle loc and view lists.
5961 (view_list_to_loc_list_val_node): New.
5962 (print_dw_val): Handle dw_val_class_view_list.
5963 (size_of_die): Likewise.
5964 (value_format): Likewise.
5965 (loc_list_has_views): New.
5966 (gen_llsym): Set vl_symbol too.
5967 (maybe_gen_llsym, skip_loc_list_entry): New.
5968 (dwarf2out_maybe_output_loclist_view_pair): New.
5969 (output_loc_list): Output view list or entries too.
5970 (output_view_list_offset): New.
5971 (output_die): Handle dw_val_class_view_list.
5972 (output_dwarf_version): New.
5973 (output_compilation_unit_header): Use it.
5974 (output_skeleton_debug_sections): Likewise.
5975 (output_rnglists, output_line_info): Likewise.
5976 (output_pubnames, output_aranges): Update version comments.
5977 (output_one_line_info_table): Output view numbers in asm comments.
5978 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5979 Call maybe_gen_llsym.
5980 (loc_list_from_tree_1): Adjust.
5981 (add_AT_location_description): Create view list attribute if
5982 needed, check it's absent otherwise.
5983 (convert_cfa_to_fb_loc_list): Adjust.
5984 (maybe_emit_file): Call output_asm_line_debug_info for test.
5985 (dwarf2out_var_location): Reset views as needed. Precompute
5986 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5987 attribute. Set view.
5988 (new_line_info_table): Reset next view.
5989 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5990 (dwarf2out_source_line): Likewise. Output view resets and labels to
5991 the assembler, or select appropriate line info opcodes.
5992 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5993 (optimize_string_length): Catch it. Adjust.
5994 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5995 dw_val_class_view_list, and remove it if no longer needed.
5996 (hash_loc_list): Hash view numbers.
5997 (loc_list_hasher::equal): Compare them.
5998 (optimize_location_lists): Check whether a view list symbol is
5999 needed, and whether the locview attribute is present, and
6000 whether they match. Remove the locview attribute if no longer
6002 (index_location_lists): Call skip_loc_list_entry for test.
6003 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6004 Use output_dwarf_version.
6005 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6006 (struct dw_val_node): Add val_view_list.
6007 * final.c (SEEN_NEXT_VIEW): New.
6008 (set_next_view_needed): New.
6009 (clear_next_view_needed): New.
6010 (maybe_output_next_view): New.
6011 (final_start_function): Rename to...
6012 (final_start_function_1): ... this. Take pointer to FIRST,
6013 add SEEN parameter. Emit param bindings in the initial view.
6014 (final_start_function): Reintroduce SEEN-less interface.
6015 (final): Rename to...
6016 (final_1): ... this. Take SEEN parameter. Output final pending
6017 next view at the end.
6018 (final): Reintroduce seen-less interface.
6019 (final_scan_insn): Output pending next view before switching
6020 sections or ending a block. Mark the next view as needed when
6021 outputting variable locations. Notify debug backend of section
6022 changes, and of location view changes.
6023 (rest_of_handle_final): Adjust.
6024 * toplev.c (process_options): Autodetect value for debug variable
6025 location views option. Warn on incompat5 without -gdwarf-5.
6026 * doc/invoke.texi (gvariable-location-views): New.
6027 (gvariable-location-views=incompat5): New.
6028 (gno-variable-location-views): New.
6030 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6032 PR tree-optimization/84136
6033 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6034 that the result of find_edge is non-NULL.
6036 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6039 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6040 storing integer register in SImode. Fix cost of 256 and 512
6041 byte aligned SSE register store.
6043 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6045 * config/i386/i386.c (ix86_multiplication_cost): Fix
6046 multiplication cost for TARGET_AVX512DQ.
6048 2018-02-08 Marek Polacek <polacek@redhat.com>
6050 PR tree-optimization/84238
6051 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6054 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6056 PR tree-optimization/84265
6057 * tree-vect-stmts.c (vectorizable_store): Don't treat
6058 VMAT_CONTIGUOUS accesses as grouped.
6059 (vectorizable_load): Likewise.
6061 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6063 PR tree-optimization/81635
6064 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6065 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6066 (test_round_for_mask): New functions.
6067 (wide_int_cc_tests): Call test_round_for_mask.
6068 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6069 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6070 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6071 range returned by get_range_info.
6073 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6076 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6077 * symtab.c: Include builtins.h
6078 (symtab_node::output_to_lto_symbol_table_p): Move here
6079 from lto-streamer-out.c:output_symbol_p.
6080 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6081 (output_symbol_p): Move all logic to symtab.c
6082 (produce_symtab): Update.
6084 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6086 * config/s390/s390-opts.h (enum indirect_branch): Define.
6087 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6088 (s390_indirect_branch_via_thunk)
6089 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6090 (enum s390_indirect_branch_type): Define.
6091 * config/s390/s390.c (struct s390_frame_layout, struct
6092 machine_function): Remove.
6093 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6094 (indirect_branch_table_label_no, indirect_branch_table_name):
6096 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6097 (enum s390_indirect_branch_option): Define.
6098 (s390_return_addr_from_memory): New function.
6099 (s390_handle_string_attribute): New function.
6100 (s390_attribute_table): Add new attribute handler.
6101 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6102 (s390_indirect_branch_via_thunk): New function.
6103 (s390_indirect_branch_via_inline_thunk): New function.
6104 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6105 sibling call optimization for non z10 compiles.
6106 (s390_emit_call): Force indirect branch target to be a single
6107 register. Add r1 clobber for non-z10 compiles.
6108 (s390_emit_epilogue): Emit return jump via return_use expander.
6109 (s390_reorg): Handle JUMP_INSNs as execute targets.
6110 (s390_option_override_internal): Perform validity checks for the
6111 new command line options.
6112 (s390_indirect_branch_attrvalue): New function.
6113 (s390_indirect_branch_settings): New function.
6114 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6115 (s390_output_indirect_thunk_function): New function.
6116 (s390_code_end): Implement target hook.
6117 (s390_case_values_threshold): Implement target hook.
6118 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6120 * config/s390/s390.h (struct s390_frame_layout)
6121 (struct machine_function): Move here from s390.c.
6122 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6123 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6124 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6125 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6126 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6127 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6128 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6129 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6130 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6131 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6132 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6133 (mnemonic attribute): Add values which aren't recognized
6135 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6136 pattern for branch conversion. Fix mnemonic attribute.
6137 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6138 indirect branch via thunk if requested.
6139 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6140 ("*indirect_jump"): Disable for branch conversion using out of
6142 ("indirect_jump_via_thunk<mode>_z10")
6143 ("indirect_jump_via_thunk<mode>")
6144 ("indirect_jump_via_inlinethunk<mode>_z10")
6145 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6146 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6147 ("casesi_jump_via_inlinethunk<mode>_z10")
6148 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6149 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6150 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6151 ("*indirect2_jump"): Disable for branch conversion.
6152 ("casesi_jump"): Turn into expander and expand patterns for branch
6154 ("return_use"): New expander.
6155 ("*return"): Emit return via thunk and rename it to ...
6156 ("*return<mode>"): ... this one.
6157 * config/s390/s390.opt: Add new options and and enum for the
6160 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6162 * lra-constraints.c (match_reload): Unconditionally use
6163 gen_lowpart_SUBREG, rather than selecting between that
6164 and equivalent gen_rtx_SUBREG code.
6166 2018-02-08 Richard Biener <rguenther@suse.de>
6168 PR tree-optimization/84233
6169 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6170 changed flag instead of boguously re-using phi_inserted.
6172 2018-02-08 Martin Jambor <mjambor@suse.cz>
6174 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6175 static local variables.
6177 2018-02-08 Richard Biener <rguenther@suse.de>
6179 PR tree-optimization/84278
6180 * tree-vect-stmts.c (vectorizable_store): When looking for
6181 smaller vector types to perform grouped strided loads/stores
6182 make sure the mode is supported by the target.
6183 (vectorizable_load): Likewise.
6185 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6187 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6188 Increase LDP/STP opportunities by adding adjacent callee-saves.
6190 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6192 PR rtl-optimization/84068
6193 PR rtl-optimization/83459
6194 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6196 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6198 PR tree-optimization/84224
6199 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6200 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6203 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6206 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6207 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6208 count, remove one USE.
6210 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6212 * doc/install.texi (Configuration): Document the
6213 --with-long-double-format={ibm,ieee} PowerPC configuration
6217 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6218 Convert from define_expand to be define_insn_and_split. Rework
6219 float/double/_Float128 conversions to QI/HI/SImode to work with
6220 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6221 conversions to QI/HImode types did a store and then a load to
6222 truncate the value. For conversions to VSX registers, don't split
6223 the insn, instead emit the code directly. Use the code iterator
6224 any_fix to combine signed and unsigned conversions.
6225 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6226 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6227 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6228 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6229 (fix_<mode>di2_hw): Likewise.
6230 (fixuns_<mode>di2_hw): Likewise.
6231 (fix_<mode>si2_hw): Likewise.
6232 (fixuns_<mode>si2_hw): Likewise.
6233 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6234 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6235 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6236 fix<uns>_trunc<SFDF:mode>si2_p8.
6237 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6239 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6240 (fix<uns>_<mode>_mem): Likewise.
6241 (fctiw<u>z_<mode>_mem): Likewise.
6242 (fix<uns>_<mode>_mem): Likewise.
6243 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6244 the register allocator from doing a direct move to the GPRs to do
6245 a store, and instead use the ISA 3.0 store byte/half-word from
6246 vector register instruction. For IEEE 128-bit floating point,
6247 also optimize stores of 32-bit ints.
6248 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6250 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6252 * genextract.c (push_pathstr_operand): New function to support
6254 (walk_rtx): Call push_pathstr_operand.
6255 (print_path): Support [a-zA-Z].
6257 2018-02-07 Richard Biener <rguenther@suse.de>
6259 PR tree-optimization/84037
6260 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6261 (cse_and_gimplify_to_preheader): Declare.
6262 (vect_get_place_in_interleaving_chain): Likewise.
6263 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6265 (_loop_vec_info::~_loop_vec_info): Delete it.
6266 (cse_and_gimplify_to_preheader): New function.
6267 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6268 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6269 (vectorizable_load): Likewise. For grouped stores always base
6270 the IV on the first element.
6271 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6272 condition before gimplifying.
6274 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6276 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6277 *DIV_EXPR and *MOD_EXPR.
6279 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6282 * config/i386/i386.c (ix86_option_override_internal): Mask out
6283 the CF_SET bit when checking -fcf-protection.
6285 2018-02-07 Tom de Vries <tom@codesourcery.com>
6288 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6291 2018-02-07 Richard Biener <rguenther@suse.de>
6293 PR tree-optimization/84204
6294 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6297 PR tree-optimization/84205
6298 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6299 special-case isl_ast_op_zdiv_r.
6301 PR tree-optimization/84223
6302 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6303 Only add conditions from within the region.
6304 (gather_bbs::after_dom_children): Adjust.
6306 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6309 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6310 * config/avr/avr.md: Only post-reload split REG-REG moves if
6311 either register is GENERAL_REG_P.
6313 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6315 PR tree-optimization/84235
6316 * tree-ssa-scopedtables.c
6317 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6318 if the subtraction is performed in floating point type where NaNs are
6319 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6320 build 1. Formatting fix.
6322 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6325 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6326 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6327 and skip it regardless of bb boundaries. Use CALL_P macro,
6328 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6329 unnecessarily, formatting fix.
6331 2018-02-06 Michael Collison <michael.collison@arm.com>
6333 * config/arm/thumb2.md:
6334 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6335 (*thumb_mov_notscc): Ditto.
6337 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6340 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6341 unsigned_fix, not "s".
6343 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6345 * configure.ac (gcc_fn_eh_frame_ro): New function.
6346 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6347 correct .eh_frame permissions.
6348 * configure: Regenerate.
6350 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6352 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6355 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6357 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6358 Display warning message for -mno-speculate-indirect-jumps.
6360 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6362 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6364 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6366 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6368 PR tree-optimization/84225
6369 * tree-eh.c (find_trapping_overflow): Only call
6370 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6372 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6375 * config/i386/i386.c: Reimplement the check of possible options
6376 -mibt/-mshstk conbination. Change error messages.
6377 * doc/invoke.texi: Fix a typo: remove extra '='.
6379 2018-02-06 Marek Polacek <polacek@redhat.com>
6381 PR tree-optimization/84228
6382 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6384 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6387 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6388 emitted arch directives.
6389 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6390 __ARM_FEATURE_COPROC before changing architectures.
6392 2018-02-06 Richard Biener <rguenther@suse.de>
6394 * config/i386/i386.c (print_reg): Fix typo.
6395 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6397 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6399 * configure: Regenerate.
6401 2018-02-05 Martin Sebor <msebor@redhat.com>
6403 PR tree-optimization/83369
6404 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6407 2018-02-05 Martin Liska <mliska@suse.cz>
6409 * doc/invoke.texi: Cherry-pick upstream r323995.
6411 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6413 * ira.c (ira_init_register_move_cost): Adjust comment.
6415 2018-02-05 Martin Liska <mliska@suse.cz>
6417 PR gcov-profile/84137
6418 * doc/gcov.texi: Fix typo in documentation.
6420 2018-02-05 Martin Liska <mliska@suse.cz>
6422 PR gcov-profile/83879
6423 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6424 using dlopen functionality.
6426 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6428 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6429 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6430 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6431 _mm_maskz_range_round_ss): New intrinsics.
6432 (__builtin_ia32_rangesd128_round)
6433 (__builtin_ia32_rangess128_round): Remove.
6434 (__builtin_ia32_rangesd128_mask_round,
6435 __builtin_ia32_rangess128_mask_round): New builtins.
6436 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6437 __builtin_ia32_rangess128_round): Remove.
6438 (__builtin_ia32_rangesd128_mask_round,
6439 __builtin_ia32_rangess128_mask_round): New builtins.
6440 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6441 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6442 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6443 "<round_saeonly_constraint>")): Changed to ...
6444 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6445 "<round_saeonly_scalar_constraint>")): ... this.
6446 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6447 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6448 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6449 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6450 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6452 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6454 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6456 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6457 Remove all values except native, 8540 and 8548.
6459 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6461 * config/i386/i386.c (ix86_output_function_return): Pass
6462 INVALID_REGNUM, instead of -1, as invalid register number to
6463 indirect_thunk_name and output_indirect_thunk.
6465 2018-02-02 Julia Koval <julia.koval@intel.com>
6467 * config.gcc: Add -march=icelake.
6468 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6469 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6470 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6471 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6472 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6473 (processor_target_table): Add icelake.
6474 (ix86_option_override_internal): Handle new PTAs.
6475 (get_builtin_code_for_version): Handle icelake.
6476 (M_INTEL_COREI7_ICELAKE): New.
6477 (fold_builtin_cpu): Handle icelake.
6478 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6479 * doc/invoke.texi: Add -march=icelake.
6481 2018-02-02 Julia Koval <julia.koval@intel.com>
6483 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6484 to wide_int_bitmask.
6485 * wide-int-bitmask.h: New.
6487 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6490 * config/i386/i386.md: Replace Pmode with word_mode in
6491 builtin_setjmp_setup and builtin_longjmp to support x32.
6493 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6497 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6499 (rs6000_supported_cpu_names): New static variable.
6500 (linux_cpu_translation_table): Likewise.
6501 (elf_platform) <cpu>: Define new static variable and use it.
6502 Translate kernel AT_PLATFORM name to canonical name if needed.
6503 Error if platform name is unknown.
6505 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6508 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6510 2018-02-01 Jeff Law <law@redhat.com>
6513 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6514 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6515 the scratch if RELEASE_VIA_POP is false.
6516 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6517 If we have to save a temporary register, decrement SIZE appropriately.
6518 Pass new arguments to release_scratch_register_on_entry.
6519 (ix86_adjust_stack_and_probe): Likewise.
6520 (ix86_emit_probe_stack_range): Pass new arguments to
6521 release_scratch_register_on_entry.
6523 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6525 PR rtl-optimization/84157
6526 * combine.c (change_zero_ext): Use REG_P predicate in
6527 front of HARD_REGISTER_P predicate.
6529 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6531 * config/avr/avr.c (avr_option_override): Move disabling of
6532 -fdelete-null-pointer-checks to...
6533 * common/config/avr/avr-common.c (avr_option_optimization_table):
6536 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6538 PR tree-optimization/81635
6539 * tree-data-ref.c (split_constant_offset_1): For types that
6540 wrap on overflow, try to use range info to prove that wrapping
6543 2018-02-01 Renlin Li <renlin.li@arm.com>
6546 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6548 (aarch64_register_move_cost): Likewise.
6549 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6551 (REG_CLASS_NAMES): Likewise.
6552 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6553 TAILCALL_ADDR_REGS. Remove IP registers.
6554 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6556 2018-02-01 Richard Biener <rguenther@suse.de>
6558 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6559 for specifying RPO order and allow NULL for that.
6560 * domwalk.c (dom_walker::dom_walker): Likewise.
6561 (dom_walker::walk): Handle NULL RPO order.
6562 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6564 (rewrite_update_dom_walker): Likewise.
6565 (mark_def_dom_walker): Likewise.
6567 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6569 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6570 (aarch64_maybe_expand_sve_subreg_move): Declare.
6571 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6572 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6574 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6575 that are semantically a reverse operation.
6576 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6577 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6578 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6580 (aarch64_can_change_mode_class): For big-endian, forbid changes
6581 between two SVE modes if they have different element sizes.
6583 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6585 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6586 the TImode handling for big-endian targets.
6588 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6590 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6591 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6593 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6594 Remove BSWAP handing for big-endian targets and use the form of
6595 LD1RQ appropariate for the mode.
6597 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6599 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6600 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6603 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6606 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6607 check for operands that need to go through aarch64_sve_reload_be.
6609 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6611 PR tree-optimization/81661
6612 PR tree-optimization/84117
6613 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6614 * tree-eh.c: Include gimplify.h.
6615 (find_trapping_overflow, replace_trapping_overflow,
6616 rewrite_to_non_trapping_overflow): New functions.
6617 * tree-vect-loop.c: Include tree-eh.h.
6618 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6619 * tree-data-ref.c: Include tree-eh.h.
6620 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6622 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6624 PR rtl-optimization/84123
6625 * combine.c (change_zero_ext): Check if hard register satisfies
6626 can_change_dest_mode before calling gen_lowpart_SUBREG.
6628 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6631 * ira.c (ira_init_register_move_cost): Remove assert.
6633 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6635 PR rtl-optimization/84071
6636 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6637 * doc/tm.texi: Regenerate.
6639 2018-01-31 Richard Biener <rguenther@suse.de>
6641 PR tree-optimization/84132
6642 * tree-data-ref.c (analyze_miv_subscript): Properly
6643 check whether evolution_function_is_affine_multivariate_p
6644 before calling gcd_of_steps_may_divide_p.
6646 2018-01-31 Julia Koval <julia.koval@intel.com>
6649 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6650 * config/i386/i386.md (rdpid_rex64) New.
6651 (rdpid): Make 32bit only.
6653 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6656 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6657 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6659 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6662 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6664 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6666 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6668 PR rtl-optimization/84071
6669 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6670 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6672 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6674 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6675 (arc_attribute_table): Add 'aux' attribute.
6676 (arc_in_small_data_p): Consider aux like variables.
6677 (arc_is_aux_reg_p): New function.
6678 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6679 (arc_get_aux_arg): New function.
6680 (prepare_move_operands): Handle aux-register access.
6681 (arc_handle_aux_attribute): New function.
6682 * doc/extend.texi (ARC Variable attributes): Add subsection.
6684 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6686 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6687 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6688 (arc_attribute_table): Add 'uncached' attribute.
6689 (arc_print_operand): Print '.di' flag for uncached memory
6691 (arc_in_small_data_p): Do not consider for small data the uncached
6693 (arc_is_uncached_mem_p): New function.
6694 * config/arc/predicates.md (compact_store_memory_operand): Check
6695 for uncached memory accesses.
6696 (nonvol_nonimm_operand): Likewise.
6697 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6699 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6702 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6703 falign-loops=): Add Optimization flag.
6705 2018-01-30 Jeff Law <law@redhat.com>
6708 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6709 INT_REGISTERS_SAVED. Check it prior to calling
6710 get_scratch_register_on_entry.
6711 (ix86_adjust_stack_and_probe): Similarly.
6712 (ix86_emit_probe_stack_range): Similarly.
6713 (ix86_expand_prologue): Corresponding changes.
6715 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6718 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6719 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6721 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6724 * lra-constraints.c (curr_insn_transform): Process AND in the
6727 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6729 PR rtl-optimization/83986
6730 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6731 dependence against last_pending_memory_flush in addition to
6734 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6736 PR tree-optimization/81611
6737 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6740 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6743 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6746 2018-01-30 Richard Biener <rguenther@suse.de>
6747 Jakub Jelinek <jakub@redhat.com>
6749 PR tree-optimization/84111
6750 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6751 inner loops added during recursion, as they don't have up-to-date
6754 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6757 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6758 (can_inline_edge_by_limits_p): ... here.
6759 (can_early_inline_edge_p, check_callers,
6760 update_caller_keys, update_callee_keys, recursive_inlining,
6761 add_new_edges_to_heap, speculation_useful_p,
6762 inline_small_functions,
6763 inline_small_functions, flatten_function,
6764 inline_to_all_callers_1): Update.
6766 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6768 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6771 2018-01-30 Richard Biener <rguenther@suse.de>
6773 PR tree-optimization/83008
6774 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6775 invariant and constant vector uses in stmts when they need
6778 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6781 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6782 * configure: Regenerate.
6784 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6786 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6788 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6789 Use gen_rtx_REG rather than gen_lowpart.
6791 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6793 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6794 rather than 0 when creating partial subregs.
6796 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6798 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6801 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6804 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6805 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6806 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6807 flags. This restores the settings used before the 2017-07-24.
6808 Turning off pre increment/decrement/modify allows IVOPTS to
6809 optimize DF/SF loops where the index is an int.
6811 2018-01-29 Richard Biener <rguenther@suse.de>
6812 Kelvin Nilsen <kelvin@gcc.gnu.org>
6815 * tree-vect-stmts.c (vectorizable_call): Don't call
6816 targetm.vectorize_builtin_md_vectorized_function if callee is
6819 2018-01-22 Carl Love <cel@us.ibm.com>
6821 * doc/extend.tex: Fix typo in second arg in
6822 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6824 2018-01-29 Richard Biener <rguenther@suse.de>
6826 PR tree-optimization/84086
6827 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6828 (flush_ssaname_freelist): When SSA names were released reset
6829 the SCEV hash table.
6831 2018-01-29 Richard Biener <rguenther@suse.de>
6833 PR tree-optimization/84057
6834 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6835 removed paths when removing edges.
6837 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6839 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6840 -mfunction-return=@var{choice}.
6842 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6845 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6847 (layout::print_source_line): Likewise.
6848 (test_get_line_width_without_trailing_whitespace): Add test cases.
6850 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6853 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6856 2018-01-26 Jim Wilson <jimw@sifive.com>
6858 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6860 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6863 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6865 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6866 and CMP + SUB-immediate -> SUBS.
6868 2018-01-26 Martin Sebor <msebor@redhat.com>
6870 PR tree-optimization/83896
6871 * tree-ssa-strlen.c (get_string_len): Rename...
6872 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6873 Avoid assuming length is constant.
6874 (handle_char_store): Use HOST_WIDE_INT for string length.
6876 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6879 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6880 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6882 2018-01-26 Richard Biener <rguenther@suse.de>
6884 PR rtl-optimization/84003
6885 * dse.c (record_store): Only record redundant stores when
6886 the earlier store aliases at least all accesses the later one does.
6888 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6890 PR rtl-optimization/83985
6891 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6892 REG_CFA_RESTORE insns.
6893 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6894 REG_CFA_RESTORE insns here.
6897 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6898 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6900 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6902 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6903 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6904 (arc_init): Likewise.
6905 (arc_override_options): Likewise.
6906 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6908 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6910 * config/arc/arc.h (TARGET_DBNZ): Define.
6911 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6912 properly set the tune attribute.
6913 (dbnz): Use TARGET_DBNZ guard.
6914 * config/arc/arc.opt (mtune): Add core3 option.
6916 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6918 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6919 recognize new pic like addresses.
6920 (arc_delegitimize_address): Clean up.
6922 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6924 * config/arc/arc-arches.def: Option mrf16 valid for all
6926 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6927 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6928 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6929 * config/arc/arc-tables.opt: Regenerate.
6930 * config/arc/arc.c (arc_conditional_register_usage): Handle
6931 reduced register file case.
6932 (arc_file_start): Set must have build attributes.
6933 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6935 * config/arc/arc.opt (mrf16): Add new option.
6936 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6937 * config/arc/genmultilib.awk: Handle new mrf16 option.
6938 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6939 * config/arc/t-multilib: Regenerate.
6940 * doc/invoke.texi (ARC Options): Document mrf16 option.
6942 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6944 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6945 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6946 (arc_attribute_table): Add 'secure_call' attribute.
6947 (arc_print_operand): Print secure call operand.
6948 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6950 (arc_is_secure_call_p): New function. * config/arc/arc.md
6951 (call_i): Add support for sjli instruction.
6952 (call_value_i): Likewise.
6953 * config/arc/constraints.md (Csc): New constraint.
6955 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6956 John Eric Martin <John.Martin@emmicro-us.com>
6958 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6959 * config/arc/arc.c (_arc_jli_section): New struct.
6960 (arc_jli_section): New type.
6961 (rc_jli_sections): New static variable.
6962 (arc_handle_jli_attribute): New function.
6963 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6964 (arc_file_end): New function.
6965 (TARGET_ASM_FILE_END): Define.
6966 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6967 (arc_add_jli_section): New function.
6968 (jli_call_scan): Likewise.
6969 (arc_reorg): Call jli_call_scan.
6970 (arc_output_addsi): Remove 'S' from printing asm operand.
6971 (arc_is_jli_call_p): New function.
6972 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6974 (movhi_insn): Likewise.
6975 (movsi_insn): Likewise.
6976 (movsi_set_cc_insn): Likewise.
6977 (loadqi_update): Likewise.
6978 (load_zeroextendqisi_update): Likewise.
6979 (load_signextendqisi_update): Likewise.
6980 (loadhi_update): Likewise.
6981 (load_zeroextendhisi_update): Likewise.
6982 (load_signextendhisi_update): Likewise.
6983 (loadsi_update): Likewise.
6984 (loadsf_update): Likewise.
6985 (movsicc_insn): Likewise.
6986 (bset_insn): Likewise.
6987 (bxor_insn): Likewise.
6988 (bclr_insn): Likewise.
6989 (bmsk_insn): Likewise.
6990 (bicsi3_insn): Likewise.
6991 (cmpsi_cc_c_insn): Likewise.
6992 (movsi_ne): Likewise.
6993 (movsi_cond_exec): Likewise.
6994 (clrsbsi2): Likewise.
7005 (call_i): Remove 'S' asm letter, add jli instruction.
7006 (call_value_i): Likewise.
7007 * config/arc/arc.op (mjli-always): New option.
7008 * config/arc/constraints.md (Cji): New constraint.
7009 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7011 (subsf3_fpx): Likewise.
7012 (mulsf3_fpx): Likewise.
7013 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7015 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7017 * doc/invoke.texi (ARC): Document mjli-always option.
7019 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7021 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7022 avoid addition with 0 and use incw and decw where possible.
7024 2018-01-26 Richard Biener <rguenther@suse.de>
7026 PR tree-optimization/81082
7027 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7028 association if it requires casting to unsigned.
7029 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7030 from fold_plusminus_mult_expr to catch important cases late when
7031 range info is available.
7033 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7035 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7036 * configure.ac (hidden_linkonce): New test.
7037 * configure: Regenerate.
7038 * config.in: Regenerate.
7040 2018-01-26 Julia Koval <julia.koval@intel.com>
7042 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7043 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7044 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7045 _mm_mask_bitshuffle_epi64_mask): Fix type.
7046 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7047 USI_FTYPE_V4DI_V4DI_USI): Remove.
7048 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7049 __builtin_ia32_vpshufbitqmb256_mask,
7050 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7051 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7052 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7054 2018-01-26 Alan Modra <amodra@gmail.com>
7057 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7058 UNSPEC_VBPERMQ. Sort other unspecs.
7060 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7062 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7064 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7067 * predict.c (drop_profile): Do not push/pop cfun; update also
7069 (handle_missing_profiles): Fix logic looking for zero profiles.
7071 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7074 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7075 on functions with #pragma omp declare simd or functions with simd
7077 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7078 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7079 Remove trailing \n from warning_at calls.
7081 2018-01-25 Tom de Vries <tom@codesourcery.com>
7084 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7085 for neutered workers.
7087 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7090 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7091 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7093 2018-01-24 Jeff Law <law@redhat.com>
7096 * i386.c (get_probe_interval): Move to earlier point.
7097 (ix86_compute_frame_layout): If -fstack-clash-protection and
7098 the frame is larger than the probe interval, then use pushes
7099 to save registers rather than reg->mem moves.
7100 (ix86_expand_prologue): Remove conditional for int_registers_saved
7103 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7106 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7107 min/max for never referenced object.
7109 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7112 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7114 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7115 attributes from DECL_ATTRIBUTES (decl) without affecting
7116 DECL_ATTRIBUTES (current_function_decl).
7117 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7118 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7120 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7122 PR tree-optimization/83979
7123 * fold-const.c (fold_comparison): Use constant_boolean_node
7124 instead of boolean_{true,false}_node.
7126 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7128 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7131 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7133 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7134 Simplify the clause that sets the length attribute.
7135 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7136 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7137 clause that sets the length attribute.
7138 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7140 2018-01-24 Tom de Vries <tom@codesourcery.com>
7143 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7144 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7145 Add strict parameter.
7146 (prevent_branch_around_nothing): Insert dummy insn between branch to
7147 label and label with no ptx insn inbetween.
7148 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7150 2018-01-24 Tom de Vries <tom@codesourcery.com>
7153 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7154 for neutered threads in warp.
7155 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7157 2018-01-24 Richard Biener <rguenther@suse.de>
7159 PR tree-optimization/83176
7160 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7163 2018-01-24 Richard Biener <rguenther@suse.de>
7165 PR tree-optimization/82819
7166 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7167 code generating pluses that are no-ops in the target precision.
7169 2018-01-24 Richard Biener <rguenther@suse.de>
7172 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7174 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7176 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7177 to merge probabilities.
7178 * predict.c (probably_never_executed): Also mark as cold functions
7179 with global 0 profile and guessed local profile.
7180 * profile-count.c (profile_probability::combine_with_count): New
7182 * profile-count.h (profile_probability::operator*,
7183 profile_probability::operator*=, profile_probability::operator/,
7184 profile_probability::operator/=): Reduce precision to adjusted
7185 and set value to guessed on contradictory divisions.
7186 (profile_probability::combine_with_freq): Remove.
7187 (profile_probability::combine_wiht_count): Declare.
7188 (profile_count::force_nonzero):: Set to adjusted.
7189 (profile_count::probability_in):: Set quality to adjusted.
7190 * tree-ssa-tail-merge.c (replace_block_by): Use
7193 2018-01-23 Andrew Waterman <andrew@sifive.com>
7194 Jim Wilson <jimw@sifive.com>
7196 * config/riscv/riscv.c (riscv_stack_boundary): New.
7197 (riscv_option_override): Set riscv_stack_boundary. Handle
7198 riscv_preferred_stack_boundary_arg.
7199 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7200 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7201 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7202 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7203 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7204 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7206 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7209 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7210 of struct ix86_frame.
7211 (ix86_expand_epilogue): Likewise. Add a local variable for
7212 the reg_save_offset field in struct ix86_frame.
7214 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7216 PR tree-optimization/82604
7217 * tree-loop-distribution.c (enum partition_kind): New enum item
7218 PKIND_PARTIAL_MEMSET.
7219 (partition_builtin_p): Support above new enum item.
7220 (generate_code_for_partition): Ditto.
7221 (compute_access_range): Differentiate cases that equality can be
7222 proven at all loops, the innermost loops or no loops.
7223 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7224 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7225 (finalize_partitions, distribute_loop): Don't fuse partition of
7226 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7227 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7230 2018-01-23 Martin Liska <mliska@suse.cz>
7232 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7233 order to ignore the predictor.
7234 (PRED_POLYMORPHIC_CALL): Likewise.
7235 (PRED_RECURSIVE_CALL): Likewise.
7237 2018-01-23 Martin Liska <mliska@suse.cz>
7239 * tree-profile.c (tree_profiling): Print function header to
7240 aware reader which function we are working on.
7241 * value-prof.c (gimple_find_values_to_profile): Do not print
7242 not interesting value histograms.
7244 2018-01-23 Martin Liska <mliska@suse.cz>
7246 * profile-count.h (enum profile_quality): Add
7247 profile_uninitialized as the first value. Do not number values
7248 as they are zero based.
7249 (profile_count::verify): Update sanity check.
7250 (profile_probability::verify): Likewise.
7252 2018-01-23 Nathan Sidwell <nathan@acm.org>
7254 * doc/invoke.texi (ffor-scope): Deprecate.
7256 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7258 PR tree-optimization/83510
7259 * domwalk.c (set_all_edges_as_executable): New function.
7260 (dom_walker::dom_walker): Convert bool param
7261 "skip_unreachable_blocks" to enum reachability. Move setup of
7262 edge flags to set_all_edges_as_executable and only do it when
7263 reachability is REACHABLE_BLOCKS.
7264 * domwalk.h (enum dom_walker::reachability): New enum.
7265 (dom_walker::dom_walker): Convert bool param
7266 "skip_unreachable_blocks" to enum reachability.
7267 (set_all_edges_as_executable): New decl.
7268 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7269 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7271 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7272 but converting true to REACHABLE_BLOCKS.
7273 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7275 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7276 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7277 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7279 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7280 if check_all_array_refs will be called.
7282 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7284 * tree.c (selftest::test_location_wrappers): Add more test
7287 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7289 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7290 (selftest::test_bit_in_range): Likewise.
7292 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7295 * doc/sourcebuild.texi (vect_float): Say that the selector
7296 only describes the situation when -funsafe-math-optimizations is on.
7297 (vect_float_strict): Document.
7299 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7301 PR tree-optimization/83965
7302 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7303 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7304 instead of checking only for a reduction.
7305 (vect_recog_widen_sum_pattern): Likewise.
7307 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7309 * predict.c (probably_never_executed): Only use precise profile info.
7310 (compute_function_frequency): Skip after inlining hack since we now
7311 have quality checking.
7313 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7315 * profile-count.h (profile_probability::very_unlikely,
7316 profile_probability::unlikely, profile_probability::even): Set
7317 precision to guessed.
7319 2018-01-23 Richard Biener <rguenther@suse.de>
7321 PR tree-optimization/83963
7322 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7323 Properly terminate dominator walk when crossing the exit edge not
7324 when visiting its source block.
7326 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7329 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7330 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7332 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7334 PR tree-optimization/83957
7335 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7336 semicolon after for body surrounded by braces.
7338 PR tree-optimization/83081
7339 * profile-count.h (profile_probability::split): New method.
7340 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7341 Use profile_probability::split.
7342 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7343 when splitting a single conditional jump into 2.
7345 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7347 PR tree-optimization/69452
7348 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7351 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7353 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7354 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7355 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7357 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7359 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7360 * config/rl78/rl78.md (movdi): New define_expand.
7361 * config/rl78/rl78.c (rl78_split_movdi): New function.
7363 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7366 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7368 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7369 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7370 128-bit to produce an UNSPEC move to get the double word with the
7371 signbit and then a shift directly to do signbit.
7372 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7373 implementation with a new version that just does either a direct
7374 move or a regular move. Move memory interface to separate insns.
7375 Move insns so they are next to the expander.
7376 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7377 with signbit move. Split big and little endian case.
7378 (signbit<mode>2_dm_mem_le): Likewise.
7379 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7380 (signbit<mode>2_dm2): Likewise.
7382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7384 * config/rl78/rl78.md (anddi3): New define_expand.
7386 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7388 * config/rl78/rl78.md (umindi3): New define_expand.
7390 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7392 * config/rl78/rl78.md (smindi3): New define_expand.
7394 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7396 * config/rl78/rl78.md (smaxdi3): New define_expand.
7398 2018-01-22 Carl Love <cel@us.ibm.com>
7400 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7401 LVX_V1TI): Add macro expansion.
7402 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7403 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7404 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7405 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7406 Change check to determine if the instruction is a byte reversing
7407 entry. Fix typo in comment.
7408 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7409 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7410 Add def_builtin calls for new builtins.
7411 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7412 Add define_insn expansion.
7414 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7416 * config/rl78/rl78.md (umaxdi3): New define_expand.
7418 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7420 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7421 for non-QImode registers.
7423 2018-01-22 Richard Biener <rguenther@suse.de>
7425 PR tree-optimization/83963
7426 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7427 including the loop exit block.
7428 (scop_detection::merge_sese): Likewise.
7429 (scop_detection::add_scop): Do it here instead.
7431 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7433 * doc/sourcebuild.texi (arm_softfloat): Document.
7435 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7438 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7439 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7440 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7442 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7443 David Edelsohn <dje.gcc@gmail.com>
7446 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7447 Change "crset eq" to "crset 2".
7448 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7449 (*call_indirect_aix<mode>_nospec): Likewise.
7450 (*call_value_indirect_aix<mode>_nospec): Likewise.
7451 (*call_indirect_elfv2<mode>_nospec): Likewise.
7452 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7453 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7454 change assembly output from . to $.
7455 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7456 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7457 (*tablejump<mode>_internal1_nospec): Likewise.
7459 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7462 * config/sh/sh_optimize_sett_clrt.cc:
7463 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7465 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7467 PR tree-optimization/83940
7468 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7469 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7470 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7471 use it to pass back the definition type.
7472 (vect_check_store_rhs): Likewise rhs_dt_out.
7473 (vect_build_gather_load_calls): Add a mask_dt argument and use
7474 it instead of a call to vect_is_simple_use.
7475 (vectorizable_store): Update calls to vect_check_load_store_mask
7476 and vect_check_store_rhs. Use the dt returned by the latter instead
7477 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7478 instead of calls to vect_is_simple_use. Pass the scalar rather
7479 than the vector operand to vect_is_simple_use when handling
7480 second and subsequent copies of an rhs value.
7481 (vectorizable_load): Update calls to vect_check_load_store_mask
7482 and vect_build_gather_load_calls. Use the cached mask_dt and
7483 gs_info.offset_dt instead of calls to vect_is_simple_use.
7485 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7488 * tree-emutls.c: Include gimplify.h.
7489 (lower_emutls_2): New function.
7490 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7491 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7492 it before further processing.
7495 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7496 UINTVAL (trueop1) instead of INTVAL (op1).
7498 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7502 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7503 INCOMING_FRAME_SP_OFFSET if not defined.
7504 (scan_trace): Add ENTRY argument. If true and
7505 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7506 emit a note to adjust the CFA offset.
7507 (create_cfi_notes): Adjust scan_trace callers.
7508 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7509 INCOMING_FRAME_SP_OFFSET in the CIE.
7510 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7511 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7513 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7514 * doc/tm.texi: Regenerated.
7516 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7518 PR rtl-optimization/83147
7519 * lra-constraints.c (remove_inheritance_pseudos): Use
7520 lra_substitute_pseudo_within_insn.
7522 2018-01-19 Tom de Vries <tom@codesourcery.com>
7523 Cesar Philippidis <cesar@codesourcery.com>
7526 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7528 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7531 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7532 spaces for function labels.
7534 2018-01-19 Martin Liska <mliska@suse.cz>
7536 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7537 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7538 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7539 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7540 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7541 (PRED_CONST_RETURN): Change from 69 to 65.
7542 (PRED_NULL_RETURN): Change from 91 to 71.
7543 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7544 (PRED_LOOP_GUARD): Change from 66 to 73.
7546 2018-01-19 Martin Liska <mliska@suse.cz>
7548 * predict.c (predict_insn_def): Add new assert.
7549 (struct branch_predictor): Change type to signed integer.
7550 (test_prediction_value_range): Amend test to cover
7552 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7553 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7554 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7555 (PRED_LOOP_IV_COMPARE): Likewise.
7556 * predict.h (PROB_UNINITIALIZED): Define new constant.
7558 2018-01-19 Martin Liska <mliska@suse.cz>
7560 * predict.c (dump_prediction): Add new format for
7561 analyze_brprob.py script which is enabled with -details
7563 * profile-count.h (precise_p): New function.
7565 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7567 PR tree-optimization/83922
7568 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7569 there are no statements that need masking.
7570 (vect_active_double_reduction_p): New function.
7571 (vect_analyze_loop_operations): Use it when handling phis that
7572 are not in the loop header.
7574 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7576 PR tree-optimization/83914
7577 * tree-vect-loop.c (vectorizable_induction): Don't convert
7578 init_expr or apply the peeling adjustment for inductions
7579 that are nested within the vectorized loop.
7581 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7583 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7586 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7590 * function.h (gimplify_parameters): Add gimple_seq * argument.
7591 * function.c: Include gimple.h and options.h.
7592 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7593 for the added local temporaries if needed.
7594 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7595 if there are any parameter cleanups, wrap whole body into a
7596 try/finally with the cleanups.
7598 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7601 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7602 Use GET_MODE_CLASS for scalar floating point.
7604 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7608 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7609 Fix call of call_cgraph_insertion_hooks.
7611 2018-01-18 Martin Sebor <msebor@redhat.com>
7613 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7615 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7618 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7621 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7624 * common.opt: (-ffile-prefix-map): New option.
7625 * opts.c (common_handle_option): Defer it.
7626 * opts-global.c (handle_common_deferred_options): Handle it.
7627 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7628 * file-prefix-map.h: New file.
7629 (remap_debug_filename, add_debug_prefix_map): ...here.
7630 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7631 * final.c (debug_prefix_map, add_debug_prefix_map
7632 remap_debug_filename): Move to...
7633 * file-prefix-map.c: New file.
7634 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7635 generalize, get rid of alloca(), use strrchr() instead of strchr().
7636 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7637 Implement in terms of add_prefix_map().
7638 (remap_macro_filename, remap_debug_filename): Implement in term of
7640 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7641 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7642 * dbxout.c: Include file-prefix-map.h.
7643 * varasm.c: Likewise.
7644 * vmsdbgout.c: Likewise.
7645 * xcoffout.c: Likewise.
7646 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7647 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7648 * doc/invoke.texi (-ffile-prefix-map): Document.
7649 (-fdebug-prefix-map): Update description.
7651 2018-01-18 Martin Liska <mliska@suse.cz>
7653 * config/i386/i386.c (indirect_thunk_name): Document that also
7655 (output_indirect_thunk): Document why both instructions
7656 (pause and lfence) are generated.
7658 2018-01-18 Richard Biener <rguenther@suse.de>
7660 PR tree-optimization/83887
7661 * graphite-scop-detection.c
7662 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7663 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7664 (scop_detection::merge_sese): Re-implement with a flood-fill
7665 algorithm that properly finds a SESE region if it exists.
7667 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7670 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7671 pointer_diff optimizations use view_convert instead of convert.
7673 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7675 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7676 Generate different code for -mno-speculate-indirect-jumps.
7677 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7678 (*call_indirect_aix<mode>): Disable for
7679 -mno-speculate-indirect-jumps.
7680 (*call_indirect_aix<mode>_nospec): New define_insn.
7681 (*call_value_indirect_aix<mode>): Disable for
7682 -mno-speculate-indirect-jumps.
7683 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7684 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7685 -mno-speculate-indirect-jumps.
7686 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7688 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7690 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7691 long double type, set the flags for noting the default long double
7692 type, even if we don't pass or return a long double type.
7694 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7697 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7700 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7702 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7703 support for merge[hl].
7704 (fold_mergehl_helper): New helper function.
7705 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7706 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7707 (altivec_vmrglw_direct): Add xxmrglw insn.
7709 2018-01-17 Andrew Waterman <andrew@sifive.com>
7711 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7712 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7714 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7717 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7718 call the lto_location_cache before reading the
7719 DECL_SOURCE_LOCATION of the types.
7721 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7722 Richard Sandiford <richard.sandiford@linaro.org>
7724 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7725 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7726 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7727 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7728 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7730 * config/aarch64/constraints.md (aarch64_movti_operand):
7732 * config/aarch64/predicates.md (Uti): Add new constraint.
7734 2018-01-17 Carl Love <cel@us.ibm.com>
7736 * config/rs6000/vsx.md (define_expand xl_len_r,
7737 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7738 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7740 (define_expand, define_insn): Move the shift left from the
7741 define_insn to the define_expand for lxvl and stxvl instructions.
7742 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7743 and XL_LEN_R definitions to PURE.
7745 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7747 * config/i386/i386.c (indirect_thunk_name): Declare regno
7748 as unsigned int. Compare regno with INVALID_REGNUM.
7749 (output_indirect_thunk): Ditto.
7750 (output_indirect_thunk_function): Ditto.
7751 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7752 in the call to output_indirect_thunk_function.
7754 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7757 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7758 rather than the size of inner_type to determine the stack slot size
7759 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7761 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7764 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7767 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7769 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7770 endian Linux systems to optionally enable multilibs for selecting
7771 the long double type if the user configured an explicit type.
7772 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7773 have no long double multilibs if not defined.
7774 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7775 warn if the user used -mabi={ieee,ibm}longdouble and we built
7776 multilibs for long double.
7777 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7778 appropriate multilib option.
7779 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7781 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7782 for building long double multilibs.
7783 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7785 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7787 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7790 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7792 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7795 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7798 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7801 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7803 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7804 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7806 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7808 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7809 different rtl trees depending on TARGET_64BIT.
7810 (rs6000_gen_lvx): Likewise.
7812 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7814 * config/visium/visium.md (nop): Tweak comment.
7815 (hazard_nop): Likewise.
7817 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7819 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7820 -mspeculate-indirect-jumps.
7821 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7822 for -mno-speculate-indirect-jumps.
7823 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7824 (*call_value_indirect_elfv2<mode>): Disable for
7825 -mno-speculate-indirect-jumps.
7826 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7827 (indirect_jump): Emit different RTL for
7828 -mno-speculate-indirect-jumps.
7829 (*indirect_jump<mode>): Disable for
7830 -mno-speculate-indirect-jumps.
7831 (*indirect_jump<mode>_nospec): New define_insn.
7832 (tablejump): Emit different RTL for
7833 -mno-speculate-indirect-jumps.
7834 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7835 (tablejumpsi_nospec): New define_expand.
7836 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7837 (tablejumpdi_nospec): New define_expand.
7838 (*tablejump<mode>_internal1): Disable for
7839 -mno-speculate-indirect-jumps.
7840 (*tablejump<mode>_internal1_nospec): New define_insn.
7841 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7844 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7846 * caller-save.c (insert_save): Drop unnecessary parameter. All
7849 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7850 Richard Biener <rguenth@suse.de>
7853 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7854 return early, inline manually is_gimple_sizepos. Make sure if we
7855 call gimplify_expr we don't end up with a gimple constant.
7856 * tree.c (variably_modified_type_p): Don't return true for
7857 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7858 * gimplify.h (is_gimple_sizepos): Remove.
7860 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7862 PR tree-optimization/83857
7863 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7864 vectorizable_live_operation for pure SLP statements.
7865 (vectorizable_live_operation): Handle PHIs.
7867 2018-01-16 Richard Biener <rguenther@suse.de>
7869 PR tree-optimization/83867
7870 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7871 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7873 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7876 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7877 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7878 If off is not INTEGER_CST, issue a may not be aligned warning
7879 rather than isn't aligned. Use isn%'t rather than isn't.
7880 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7882 <case MULT_EXPR>: Improve the case when bottom and one of the
7883 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7884 operand, in that case check if the other operand is multiple of
7885 bottom divided by the INTEGER_CST operand.
7887 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7890 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7891 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7892 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7893 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7894 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7895 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7896 (pa_function_arg_size): New function.
7898 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7900 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7901 in a separate statement.
7903 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7905 PR tree-optimization/83847
7906 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7907 group gathers and scatters.
7909 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7911 PR rtl-optimization/86620
7912 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7914 PR rtl-optimization/83213
7915 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7916 to last if both are JUMP_INSNs.
7918 PR tree-optimization/83843
7919 * gimple-ssa-store-merging.c
7920 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7921 store_immediate_info for bswap/nop orig_stores.
7923 2018-01-15 Andrew Waterman <andrew@sifive.com>
7925 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7927 <UDIV>: Increase cost if !TARGET_DIV.
7929 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7931 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7932 (define_attr "cr_logical_3op"): New.
7933 (cceq_ior_compare): Adjust.
7934 (cceq_ior_compare_complement): Adjust.
7935 (*cceq_rev_compare): Adjust.
7936 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7937 (is_cracked_insn): Adjust.
7938 (insn_must_be_first_in_group): Adjust.
7939 * config/rs6000/40x.md: Adjust.
7940 * config/rs6000/440.md: Adjust.
7941 * config/rs6000/476.md: Adjust.
7942 * config/rs6000/601.md: Adjust.
7943 * config/rs6000/603.md: Adjust.
7944 * config/rs6000/6xx.md: Adjust.
7945 * config/rs6000/7450.md: Adjust.
7946 * config/rs6000/7xx.md: Adjust.
7947 * config/rs6000/8540.md: Adjust.
7948 * config/rs6000/cell.md: Adjust.
7949 * config/rs6000/e300c2c3.md: Adjust.
7950 * config/rs6000/e500mc.md: Adjust.
7951 * config/rs6000/e500mc64.md: Adjust.
7952 * config/rs6000/e5500.md: Adjust.
7953 * config/rs6000/e6500.md: Adjust.
7954 * config/rs6000/mpc.md: Adjust.
7955 * config/rs6000/power4.md: Adjust.
7956 * config/rs6000/power5.md: Adjust.
7957 * config/rs6000/power6.md: Adjust.
7958 * config/rs6000/power7.md: Adjust.
7959 * config/rs6000/power8.md: Adjust.
7960 * config/rs6000/power9.md: Adjust.
7961 * config/rs6000/rs64.md: Adjust.
7962 * config/rs6000/titan.md: Adjust.
7964 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7966 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7967 ix86_indirect_branch_register logic.
7969 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7971 * config/i386/constraints.md (Bs): Update
7972 ix86_indirect_branch_register check. Don't check
7973 ix86_indirect_branch_register with GOT_memory_operand.
7975 * config/i386/predicates.md (GOT_memory_operand): Don't check
7976 ix86_indirect_branch_register here.
7977 (GOT32_symbol_operand): Likewise.
7979 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7981 * config/i386/predicates.md (constant_call_address_operand):
7982 Rewrite ix86_indirect_branch_register logic.
7983 (sibcall_insn_operand): Likewise.
7985 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7987 * config/i386/constraints.md (Bs): Replace
7988 ix86_indirect_branch_thunk_register with
7989 ix86_indirect_branch_register.
7991 * config/i386/i386.md (indirect_jump): Likewise.
7992 (tablejump): Likewise.
7993 (*sibcall_memory): Likewise.
7994 (*sibcall_value_memory): Likewise.
7995 Peepholes of indirect call and jump via memory: Likewise.
7996 * config/i386/i386.opt: Likewise.
7997 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7998 (GOT_memory_operand): Likewise.
7999 (call_insn_operand): Likewise.
8000 (sibcall_insn_operand): Likewise.
8001 (GOT32_symbol_operand): Likewise.
8003 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8006 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8007 type rather than type addr's type points to.
8008 (expand_omp_atomic_mutex): Likewise.
8009 (expand_omp_atomic): Likewise.
8011 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8014 * config/i386/i386.c (output_indirect_thunk_function): Use
8015 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8016 for __x86_return_thunk.
8018 2018-01-15 Richard Biener <rguenther@suse.de>
8021 * expmed.c (extract_bit_field_1): Fix typo.
8023 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8026 * config/arm/iterators.md (VF): New mode iterator.
8027 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8028 Remove integer-related logic from pattern.
8029 (neon_vabd<mode>_3): Likewise.
8031 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8034 * common.opt (fstrict-overflow): No longer an alias.
8035 (fwrapv-pointer): New option.
8036 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8037 also for pointer types based on flag_wrapv_pointer.
8038 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8039 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8040 opts->x_flag_wrapv got set.
8041 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8042 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8043 POINTER_TYPE_OVERFLOW_UNDEFINED.
8044 * match.pd: Likewise in address comparison pattern.
8045 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8047 2018-01-15 Richard Biener <rguenther@suse.de>
8050 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8051 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8052 Reset type names to their identifier if their TYPE_DECL doesn't
8053 have linkage (and thus is used for ODR and devirt).
8054 (save_debug_info_for_decl): Remove.
8055 (save_debug_info_for_type): Likewise.
8056 (add_tree_to_fld_list): Adjust.
8057 * tree-pretty-print.c (dump_generic_node): Make dumping of
8058 type names more robust.
8060 2018-01-15 Richard Biener <rguenther@suse.de>
8062 * BASE-VER: Bump to 8.0.1.
8064 2018-01-14 Martin Sebor <msebor@redhat.com>
8067 * builtins.c (check_access): Avoid warning when the no-warning bit
8070 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8072 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8073 * ira-color (allocno_hard_regs_compare): Likewise.
8075 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8078 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8079 Use .pushsection/.popsection.
8081 2018-01-14 Martin Sebor <msebor@redhat.com>
8084 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8086 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8088 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8089 entry from extra_headers.
8090 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8091 extra_headers, make the list bitwise identical to the i?86-*-* one.
8093 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8095 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8096 -mcmodel=large with -mindirect-branch=thunk,
8097 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8098 -mfunction-return=thunk-extern.
8099 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8100 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8101 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8103 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8105 * config/i386/i386.c (print_reg): Print the name of the full
8106 integer register without '%'.
8107 (ix86_print_operand): Handle 'V'.
8108 * doc/extend.texi: Document 'V' modifier.
8110 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8112 * config/i386/constraints.md (Bs): Disallow memory operand for
8113 -mindirect-branch-register.
8115 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8116 (GOT_memory_operand): Likewise.
8117 (call_insn_operand): Likewise.
8118 (sibcall_insn_operand): Likewise.
8119 (GOT32_symbol_operand): Likewise.
8120 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8121 for -mindirect-branch-register.
8122 (tablejump): Likewise.
8123 (*sibcall_memory): Likewise.
8124 (*sibcall_value_memory): Likewise.
8125 Disallow peepholes of indirect call and jump via memory for
8126 -mindirect-branch-register.
8127 (*call_pop): Replace m with Bw.
8128 (*call_value_pop): Likewise.
8129 (*sibcall_pop_memory): Replace m with Bs.
8130 * config/i386/i386.opt (mindirect-branch-register): New option.
8131 * doc/invoke.texi: Document -mindirect-branch-register option.
8133 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8135 * config/i386/i386-protos.h (ix86_output_function_return): New.
8136 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8137 set function_return_type.
8138 (indirect_thunk_name): Add ret_p to indicate thunk for function
8140 (output_indirect_thunk_function): Pass false to
8141 indirect_thunk_name.
8142 (ix86_output_indirect_branch_via_reg): Likewise.
8143 (ix86_output_indirect_branch_via_push): Likewise.
8144 (output_indirect_thunk_function): Create alias for function
8145 return thunk if regno < 0.
8146 (ix86_output_function_return): New function.
8147 (ix86_handle_fndecl_attribute): Handle function_return.
8148 (ix86_attribute_table): Add function_return.
8149 * config/i386/i386.h (machine_function): Add
8150 function_return_type.
8151 * config/i386/i386.md (simple_return_internal): Use
8152 ix86_output_function_return.
8153 (simple_return_internal_long): Likewise.
8154 * config/i386/i386.opt (mfunction-return=): New option.
8155 (indirect_branch): Mention -mfunction-return=.
8156 * doc/extend.texi: Document function_return function attribute.
8157 * doc/invoke.texi: Document -mfunction-return= option.
8159 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8161 * config/i386/i386-opts.h (indirect_branch): New.
8162 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8163 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8164 with local indirect jump when converting indirect call and jump.
8165 (ix86_set_indirect_branch_type): New.
8166 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8167 (indirectlabelno): New.
8168 (indirect_thunk_needed): Likewise.
8169 (indirect_thunk_bnd_needed): Likewise.
8170 (indirect_thunks_used): Likewise.
8171 (indirect_thunks_bnd_used): Likewise.
8172 (INDIRECT_LABEL): Likewise.
8173 (indirect_thunk_name): Likewise.
8174 (output_indirect_thunk): Likewise.
8175 (output_indirect_thunk_function): Likewise.
8176 (ix86_output_indirect_branch_via_reg): Likewise.
8177 (ix86_output_indirect_branch_via_push): Likewise.
8178 (ix86_output_indirect_branch): Likewise.
8179 (ix86_output_indirect_jmp): Likewise.
8180 (ix86_code_end): Call output_indirect_thunk_function if needed.
8181 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8183 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8184 (ix86_attribute_table): Add indirect_branch.
8185 * config/i386/i386.h (machine_function): Add indirect_branch_type
8186 and has_local_indirect_jump.
8187 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8189 (tablejump): Likewise.
8190 (*indirect_jump): Use ix86_output_indirect_jmp.
8191 (*tablejump_1): Likewise.
8192 (simple_return_indirect_internal): Likewise.
8193 * config/i386/i386.opt (mindirect-branch=): New option.
8194 (indirect_branch): New.
8197 (thunk-inline): Likewise.
8198 (thunk-extern): Likewise.
8199 * doc/extend.texi: Document indirect_branch function attribute.
8200 * doc/invoke.texi: Document -mindirect-branch= option.
8202 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8205 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8207 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8209 * ipa-inline.c (want_inline_small_function_p): Return false if
8210 inlining has already failed with CIF_FINAL_ERROR.
8211 (update_caller_keys): Call want_inline_small_function_p before
8213 (update_callee_keys): Likewise.
8215 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8217 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8219 (rs6000_quadword_masked_address_p): Likewise.
8220 (quad_aligned_load_p): Likewise.
8221 (quad_aligned_store_p): Likewise.
8222 (const_load_sequence_p): Add comment to describe the outer-most loop.
8223 (mimic_memory_attributes_and_flags): New function.
8224 (rs6000_gen_stvx): Likewise.
8225 (replace_swapped_aligned_store): Likewise.
8226 (rs6000_gen_lvx): Likewise.
8227 (replace_swapped_aligned_load): Likewise.
8228 (replace_swapped_load_constant): Capitalize argument name in
8229 comment describing this function.
8230 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8231 and stores that access quad-word aligned addresses and replace
8232 with stvx or lvx instructions when appropriate.
8233 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8234 New function prototype.
8235 (rs6000_quadword_masked_address_p): Likewise.
8236 (rs6000_gen_lvx): Likewise.
8237 (rs6000_gen_stvx): Likewise.
8238 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8239 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8240 when memory address is aligned.
8241 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8242 this split to select lvx instruction when memory address is aligned.
8243 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8244 instruction when memory address is aligned.
8245 (*vsx_le_perm_load_v16qi): Likewise.
8246 (four unnamed splitters): Modify to select the stvx instruction
8247 when memory is aligned.
8249 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8251 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8252 which appears in the queue multiple times.
8254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8255 Alan Hayward <alan.hayward@arm.com>
8256 David Sherwood <david.sherwood@arm.com>
8258 * tree-vectorizer.h (vec_lower_bound): New structure.
8259 (_loop_vec_info): Add check_nonzero and lower_bounds.
8260 (LOOP_VINFO_CHECK_NONZERO): New macro.
8261 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8262 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8263 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8264 fields. Make seg_len the distance travelled, not including the
8266 (dr_direction_indicator): Declare.
8267 (dr_zero_step_indicator): Likewise.
8268 (dr_known_forward_stride_p): Likewise.
8269 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8271 (runtime_alias_check_p): Allow runtime alias checks with
8273 (operator ==): Compare access_size and align.
8274 (prune_runtime_alias_test_list): Rework for new distinction between
8275 the access_size and seg_len.
8276 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8278 (get_segment_min_max): New function.
8279 (create_intersect_range_checks): Use it.
8280 (dr_step_indicator): New function.
8281 (dr_direction_indicator): Likewise.
8282 (dr_zero_step_indicator): Likewise.
8283 (dr_known_forward_stride_p): Likewise.
8284 * tree-loop-distribution.c (data_ref_segment_size): Return
8285 DR_STEP * (niters - 1).
8286 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8288 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8289 (vect_preserves_scalar_order_p): New function, split out from...
8290 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8291 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8292 (vect_vfa_access_size): New function.
8293 (vect_vfa_align): Likewise.
8294 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8295 (dump_lower_bound): New function.
8296 (vect_check_lower_bound): Likewise.
8297 (vect_small_gap_p): Likewise.
8298 (vectorizable_with_step_bound_p): Likewise.
8299 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8300 depencies if the vectorization factor is 1. Convert the checks
8301 for nonzero steps into checks on the bounds of DR_STEP. Try using
8302 a bunds check for variable steps if the minimum required step is
8303 relatively small. Update calls to the dr_with_seg_len
8304 constructor and to vect_compile_time_alias.
8305 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8307 (vect_loop_versioning): Call it.
8308 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8310 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8313 Alan Hayward <alan.hayward@arm.com>
8314 David Sherwood <david.sherwood@arm.com>
8316 * doc/sourcebuild.texi (vect_scatter_store): Document.
8317 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8319 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8321 * genopinit.c (main): Add supports_vec_scatter_store and
8322 supports_vec_scatter_store_cached to target_optabs.
8323 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8324 IFN_MASK_SCATTER_STORE.
8325 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8327 * internal-fn.h (internal_store_fn_p): Declare.
8328 (internal_fn_stored_value_index): Likewise.
8329 * internal-fn.c (scatter_store_direct): New macro.
8330 (expand_scatter_store_optab_fn): New function.
8331 (direct_scatter_store_optab_supported_p): New macro.
8332 (internal_store_fn_p): New function.
8333 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8334 IFN_MASK_SCATTER_STORE.
8335 (internal_fn_mask_index): Likewise.
8336 (internal_fn_stored_value_index): New function.
8337 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8339 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8340 * optabs-query.c (supports_vec_scatter_store_p): New function.
8341 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8342 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8343 true for scatter stores.
8344 (vect_gather_scatter_fn_p): Handle scatter stores too.
8345 (vect_check_gather_scatter): Consider using scatter stores if
8346 supports_vec_scatter_store_p.
8347 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8349 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8350 internal_fn_stored_value_index.
8351 (check_load_store_masking): Handle scatter stores too.
8352 (vect_get_store_rhs): Make public.
8353 (vectorizable_call): Use internal_store_fn_p.
8354 (vectorizable_store): Handle scatter store internal functions.
8355 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8356 when deciding whether the end of the group has been reached.
8357 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8358 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8359 (mask_scatter_store<mode>): New insns.
8361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8362 Alan Hayward <alan.hayward@arm.com>
8363 David Sherwood <david.sherwood@arm.com>
8365 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8366 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8367 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8369 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8370 Use vect_truncate_gather_scatter_offset if we can't treat the
8371 operation as a normal gather load or scatter store.
8372 (get_group_load_store_type): Take the gather_scatter_info
8373 as argument. Try using a gather load or scatter store for
8374 single-element groups.
8375 (get_load_store_type): Update calls to get_group_load_store_type
8376 and vect_use_strided_gather_scatters_p.
8378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8379 Alan Hayward <alan.hayward@arm.com>
8380 David Sherwood <david.sherwood@arm.com>
8382 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8383 optional tree argument.
8384 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8386 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8387 but continue to use the current value as a fallback.
8388 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8389 to compare the updates.
8390 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8391 (get_load_store_type): Use it when handling a strided access.
8392 (vect_get_strided_load_store_ops): New function.
8393 (vect_get_data_ptr_increment): Likewise.
8394 (vectorizable_load): Handle strided gather loads. Always pass
8395 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8398 Alan Hayward <alan.hayward@arm.com>
8399 David Sherwood <david.sherwood@arm.com>
8401 * doc/md.texi (gather_load@var{m}): Document.
8402 (mask_gather_load@var{m}): Likewise.
8403 * genopinit.c (main): Add supports_vec_gather_load and
8404 supports_vec_gather_load_cached to target_optabs.
8405 * optabs-tree.c (init_tree_optimization_optabs): Use
8406 ggc_cleared_alloc to allocate target_optabs.
8407 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8408 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8410 * internal-fn.h (internal_load_fn_p): Declare.
8411 (internal_gather_scatter_fn_p): Likewise.
8412 (internal_fn_mask_index): Likewise.
8413 (internal_gather_scatter_fn_supported_p): Likewise.
8414 * internal-fn.c (gather_load_direct): New macro.
8415 (expand_gather_load_optab_fn): New function.
8416 (direct_gather_load_optab_supported_p): New macro.
8417 (direct_internal_fn_optab): New function.
8418 (internal_load_fn_p): Likewise.
8419 (internal_gather_scatter_fn_p): Likewise.
8420 (internal_fn_mask_index): Likewise.
8421 (internal_gather_scatter_fn_supported_p): Likewise.
8422 * optabs-query.c (supports_at_least_one_mode_p): New function.
8423 (supports_vec_gather_load_p): Likewise.
8424 * optabs-query.h (supports_vec_gather_load_p): Declare.
8425 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8426 and memory_type field.
8427 (NUM_PATTERNS): Bump to 15.
8428 * tree-vect-data-refs.c: Include internal-fn.h.
8429 (vect_gather_scatter_fn_p): New function.
8430 (vect_describe_gather_scatter_call): Likewise.
8431 (vect_check_gather_scatter): Try using internal functions for
8432 gather loads. Recognize existing calls to a gather load function.
8433 (vect_analyze_data_refs): Consider using gather loads if
8434 supports_vec_gather_load_p.
8435 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8436 (vect_get_gather_scatter_offset_type): Likewise.
8437 (vect_convert_mask_for_vectype): Likewise.
8438 (vect_add_conversion_to_patterm): Likewise.
8439 (vect_try_gather_scatter_pattern): Likewise.
8440 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8441 (vect_vect_recog_func_ptrs): Add it.
8442 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8443 internal_fn_mask_index and internal_gather_scatter_fn_p.
8444 (check_load_store_masking): Take the gather_scatter_info as an
8445 argument and handle gather loads.
8446 (vect_get_gather_scatter_ops): New function.
8447 (vectorizable_call): Check internal_load_fn_p.
8448 (vectorizable_load): Likewise. Handle gather load internal
8450 (vectorizable_store): Update call to check_load_store_masking.
8451 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8452 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8453 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8454 (aarch64_gather_scale_operand_d): New predicates.
8455 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8456 (mask_gather_load<mode>): New insns.
8458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8459 Alan Hayward <alan.hayward@arm.com>
8460 David Sherwood <david.sherwood@arm.com>
8462 * optabs.def (fold_left_plus_optab): New optab.
8463 * doc/md.texi (fold_left_plus_@var{m}): Document.
8464 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8465 * internal-fn.c (fold_left_direct): Define.
8466 (expand_fold_left_optab_fn): Likewise.
8467 (direct_fold_left_optab_supported_p): Likewise.
8468 * fold-const-call.c (fold_const_fold_left): New function.
8469 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8470 * tree-parloops.c (valid_reduction_p): New function.
8471 (gather_scalar_reductions): Use it.
8472 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8473 (vect_finish_replace_stmt): Declare.
8474 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8475 (needs_fold_left_reduction_p): New function, split out from...
8476 (vect_is_simple_reduction): ...here. Accept reductions that
8477 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8478 (vect_force_simple_reduction): Also store the reduction type in
8479 the assignment's STMT_VINFO_REDUC_TYPE.
8480 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8481 (merge_with_identity): New function.
8482 (vect_expand_fold_left): Likewise.
8483 (vectorize_fold_left_reduction): Likewise.
8484 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8485 scalar phi in place for it. Check for target support and reject
8486 cases that would reassociate the operation. Defer the transform
8487 phase to vectorize_fold_left_reduction.
8488 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8489 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8490 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8494 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8495 call to ifc_temp_var.
8497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8498 Alan Hayward <alan.hayward@arm.com>
8499 David Sherwood <david.sherwood@arm.com>
8501 * target.def (legitimize_address_displacement): Take the original
8502 offset as a poly_int.
8503 * targhooks.h (default_legitimize_address_displacement): Update
8505 * targhooks.c (default_legitimize_address_displacement): Likewise.
8506 * doc/tm.texi: Regenerate.
8507 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8508 as an argument, moving assert of ad->disp == ad->disp_term to...
8509 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8510 Try calling targetm.legitimize_address_displacement before expanding
8511 the address rather than afterwards, and adjust for the new interface.
8512 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8513 Match the new hook interface. Handle SVE addresses.
8514 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8519 * Makefile.in (OBJS): Add early-remat.o.
8520 * target.def (select_early_remat_modes): New hook.
8521 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8522 * doc/tm.texi: Regenerate.
8523 * targhooks.h (default_select_early_remat_modes): Declare.
8524 * targhooks.c (default_select_early_remat_modes): New function.
8525 * timevar.def (TV_EARLY_REMAT): New timevar.
8526 * passes.def (pass_early_remat): New pass.
8527 * tree-pass.h (make_pass_early_remat): Declare.
8528 * early-remat.c: New file.
8529 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8531 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8534 Alan Hayward <alan.hayward@arm.com>
8535 David Sherwood <david.sherwood@arm.com>
8537 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8538 vfm1 with a bound_epilog parameter.
8539 (vect_do_peeling): Update calls accordingly, and move the prologue
8540 call earlier in the function. Treat the base bound_epilog as 0 for
8541 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8542 this base when peeling for gaps.
8543 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8544 with fully-masked loops.
8545 (vect_estimate_min_profitable_iters): Handle the single peeled
8546 iteration in that case.
8548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8549 Alan Hayward <alan.hayward@arm.com>
8550 David Sherwood <david.sherwood@arm.com>
8552 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8553 single-element interleaving even if the size is not a power of 2.
8554 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8555 accesses for single-element interleaving if the group size is
8558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8559 Alan Hayward <alan.hayward@arm.com>
8560 David Sherwood <david.sherwood@arm.com>
8562 * doc/md.texi (fold_extract_last_@var{m}): Document.
8563 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8564 * optabs.def (fold_extract_last_optab): New optab.
8565 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8566 * internal-fn.c (fold_extract_direct): New macro.
8567 (expand_fold_extract_optab_fn): Likewise.
8568 (direct_fold_extract_optab_supported_p): Likewise.
8569 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8570 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8571 EXTRACT_LAST_REDUCTION.
8572 (get_initial_def_for_reduction): Do not create an initial vector
8573 for EXTRACT_LAST_REDUCTION reductions.
8574 (vectorizable_reduction): Leave the scalar phi in place for
8575 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8576 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8577 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8578 transform phase to vectorizable_condition.
8579 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8581 (vect_finish_stmt_generation): ...here.
8582 (vect_finish_replace_stmt): New function.
8583 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8584 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8586 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8589 Alan Hayward <alan.hayward@arm.com>
8590 David Sherwood <david.sherwood@arm.com>
8592 * doc/md.texi (extract_last_@var{m}): Document.
8593 * optabs.def (extract_last_optab): New optab.
8594 * internal-fn.def (EXTRACT_LAST): New internal function.
8595 * internal-fn.c (cond_unary_direct): New macro.
8596 (expand_cond_unary_optab_fn): Likewise.
8597 (direct_cond_unary_optab_supported_p): Likewise.
8598 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8599 loops using EXTRACT_LAST.
8600 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8601 (extract_last_<mode>): ...this optab.
8602 (vec_extract<mode><Vel>): Update accordingly.
8604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8605 Alan Hayward <alan.hayward@arm.com>
8606 David Sherwood <david.sherwood@arm.com>
8608 * target.def (empty_mask_is_expensive): New hook.
8609 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8610 * doc/tm.texi: Regenerate.
8611 * targhooks.h (default_empty_mask_is_expensive): Declare.
8612 * targhooks.c (default_empty_mask_is_expensive): New function.
8613 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8614 if the target says that empty masks are expensive.
8615 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8617 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8620 Alan Hayward <alan.hayward@arm.com>
8621 David Sherwood <david.sherwood@arm.com>
8623 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8624 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8625 (vect_use_loop_mask_for_alignment_p): New function.
8626 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8627 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8628 niters_skip argument. Make sure that the first niters_skip elements
8629 of the first iteration are inactive.
8630 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8631 Update call to vect_set_loop_masks_directly.
8632 (get_misalign_in_elems): New function, split out from...
8633 (vect_gen_prolog_loop_niters): ...here.
8634 (vect_update_init_of_dr): Take a code argument that specifies whether
8635 the adjustment should be added or subtracted.
8636 (vect_update_init_of_drs): Likewise.
8637 (vect_prepare_for_masked_peels): New function.
8638 (vect_do_peeling): Skip prologue peeling if we're using a mask
8639 instead. Update call to vect_update_inits_of_drs.
8640 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8642 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8643 alignment. Do not include the number of peeled iterations in
8644 the minimum threshold in that case.
8645 (vectorizable_induction): Adjust the start value down by
8646 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8647 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8648 Take the number of skipped iterations into account when calculating
8650 * tree-vect-stmts.c (vect_gen_while_not): New function.
8652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8653 Alan Hayward <alan.hayward@arm.com>
8654 David Sherwood <david.sherwood@arm.com>
8656 * doc/sourcebuild.texi (vect_fully_masked): Document.
8657 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8659 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8661 (vect_analyze_loop_2): ...here. Don't check the vectorization
8662 factor against the number of loop iterations if the loop is
8665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8666 Alan Hayward <alan.hayward@arm.com>
8667 David Sherwood <david.sherwood@arm.com>
8669 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8670 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8671 (dump_groups): Update accordingly.
8672 (iv_use::mem_type): New member variable.
8673 (address_p): New function.
8674 (record_use): Add a mem_type argument and initialize the new
8676 (record_group_use): Add a mem_type argument. Use address_p.
8677 Remove obsolete null checks of base_object. Update call to record_use.
8678 (find_interesting_uses_op): Update call to record_group_use.
8679 (find_interesting_uses_cond): Likewise.
8680 (find_interesting_uses_address): Likewise.
8681 (get_mem_type_for_internal_fn): New function.
8682 (find_address_like_use): Likewise.
8683 (find_interesting_uses_stmt): Try find_address_like_use before
8684 calling find_interesting_uses_op.
8685 (addr_offset_valid_p): Use the iv mem_type field as the type
8686 of the addressed memory.
8687 (add_autoinc_candidates): Likewise.
8688 (get_address_cost): Likewise.
8689 (split_small_address_groups_p): Use address_p.
8690 (split_address_groups): Likewise.
8691 (add_iv_candidate_for_use): Likewise.
8692 (autoinc_possible_for_pair): Likewise.
8693 (rewrite_groups): Likewise.
8694 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8695 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8696 (get_alias_ptr_type_for_ptr_address): New function.
8697 (rewrite_use_address): Rewrite address uses in calls that were
8698 identified by find_address_like_use.
8700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8701 Alan Hayward <alan.hayward@arm.com>
8702 David Sherwood <david.sherwood@arm.com>
8704 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8706 * gimple-expr.h (is_gimple_addressable: Likewise.
8707 * gimple-expr.c (is_gimple_address): Likewise.
8708 * internal-fn.c (expand_call_mem_ref): New function.
8709 (expand_mask_load_optab_fn): Use it.
8710 (expand_mask_store_optab_fn): Likewise.
8712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8713 Alan Hayward <alan.hayward@arm.com>
8714 David Sherwood <david.sherwood@arm.com>
8716 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8717 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8718 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8719 (cond_umax@var{mode}): Document.
8720 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8721 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8722 (cond_umin_optab, cond_umax_optab): New optabs.
8723 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8724 (COND_IOR, COND_XOR): New internal functions.
8725 * internal-fn.h (get_conditional_internal_fn): Declare.
8726 * internal-fn.c (cond_binary_direct): New macro.
8727 (expand_cond_binary_optab_fn): Likewise.
8728 (direct_cond_binary_optab_supported_p): Likewise.
8729 (get_conditional_internal_fn): New function.
8730 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8731 Cope with reduction statements that are vectorized as calls rather
8733 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8734 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8735 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8736 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8737 (UNSPEC_COND_EOR): New unspecs.
8738 (optab): Add mappings for them.
8739 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8740 (sve_int_op, sve_fp_op): New int attributes.
8742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8743 Alan Hayward <alan.hayward@arm.com>
8744 David Sherwood <david.sherwood@arm.com>
8746 * optabs.def (while_ult_optab): New optab.
8747 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8748 * internal-fn.def (WHILE_ULT): New internal function.
8749 * internal-fn.h (direct_internal_fn_supported_p): New override
8750 that takes two types as argument.
8751 * internal-fn.c (while_direct): New macro.
8752 (expand_while_optab_fn): New function.
8753 (convert_optab_supported_p): Likewise.
8754 (direct_while_optab_supported_p): New macro.
8755 * wide-int.h (wi::udiv_ceil): New function.
8756 * tree-vectorizer.h (rgroup_masks): New structure.
8757 (vec_loop_masks): New typedef.
8758 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8760 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8761 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8762 (vect_max_vf): New function.
8763 (slpeel_make_loop_iterate_ntimes): Delete.
8764 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8765 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8766 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8767 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8768 internal-fn.h, stor-layout.h and optabs-query.h.
8769 (vect_set_loop_mask): New function.
8770 (add_preheader_seq): Likewise.
8771 (add_header_seq): Likewise.
8772 (interleave_supported_p): Likewise.
8773 (vect_maybe_permute_loop_masks): Likewise.
8774 (vect_set_loop_masks_directly): Likewise.
8775 (vect_set_loop_condition_masked): Likewise.
8776 (vect_set_loop_condition_unmasked): New function, split out from
8777 slpeel_make_loop_iterate_ntimes.
8778 (slpeel_make_loop_iterate_ntimes): Rename to..
8779 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8780 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8781 (vect_do_peeling): Update call accordingly.
8782 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8784 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8785 mask_compare_type, can_fully_mask_p and fully_masked_p.
8786 (release_vec_loop_masks): New function.
8787 (_loop_vec_info): Use it to free the loop masks.
8788 (can_produce_all_loop_masks_p): New function.
8789 (vect_get_max_nscalars_per_iter): Likewise.
8790 (vect_verify_full_masking): Likewise.
8791 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8792 retries, and free the mask rgroups before retrying. Check loop-wide
8793 reasons for disallowing fully-masked loops. Make the final decision
8794 about whether use a fully-masked loop or not.
8795 (vect_estimate_min_profitable_iters): Do not assume that peeling
8796 for the number of iterations will be needed for fully-masked loops.
8797 (vectorizable_reduction): Disable fully-masked loops.
8798 (vectorizable_live_operation): Likewise.
8799 (vect_halve_mask_nunits): New function.
8800 (vect_double_mask_nunits): Likewise.
8801 (vect_record_loop_mask): Likewise.
8802 (vect_get_loop_mask): Likewise.
8803 (vect_transform_loop): Handle the case in which the final loop
8804 iteration might handle a partial vector. Call vect_set_loop_condition
8805 instead of slpeel_make_loop_iterate_ntimes.
8806 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8807 (check_load_store_masking): New function.
8808 (prepare_load_store_mask): Likewise.
8809 (vectorizable_store): Handle fully-masked loops.
8810 (vectorizable_load): Likewise.
8811 (supportable_widening_operation): Use vect_halve_mask_nunits for
8813 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8814 (vect_gen_while): New function.
8815 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8816 (aarch64_uqdec<mode>): New insn.
8818 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8819 Alan Hayward <alan.hayward@arm.com>
8820 David Sherwood <david.sherwood@arm.com>
8822 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8823 (reduc_xor_scal_optab): New optabs.
8824 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8825 (reduc_xor_scal_@var{m}): Document.
8826 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8827 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8829 * fold-const-call.c (fold_const_call): Handle them.
8830 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8831 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8832 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8833 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8834 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8835 (UNSPEC_XORV): New unspecs.
8836 (optab): Add entries for them.
8837 (BITWISEV): New int iterator.
8838 (bit_reduc_op): New int attributes.
8840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8841 Alan Hayward <alan.hayward@arm.com>
8842 David Sherwood <david.sherwood@arm.com>
8844 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8845 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8846 * optabs.def (vec_shl_insert_optab): New optab.
8847 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8848 (duplicate_and_interleave): Likewise.
8849 * tree-vect-loop.c: Include internal-fn.h.
8850 (neutral_op_for_slp_reduction): New function, split out from
8851 get_initial_defs_for_reduction.
8852 (get_initial_def_for_reduction): Handle option 2 for variable-length
8853 vectors by loading the neutral value into a vector and then shifting
8854 the initial value into element 0.
8855 (get_initial_defs_for_reduction): Replace the code argument with
8856 the neutral value calculated by neutral_op_for_slp_reduction.
8857 Use gimple_build_vector for constant-length vectors.
8858 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8859 but the first group_size elements have a neutral value.
8860 Use duplicate_and_interleave otherwise.
8861 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8862 Update call to get_initial_defs_for_reduction. Handle SLP
8863 reductions for variable-length vectors by creating one vector
8864 result for each scalar result, with the elements associated
8865 with other scalar results stubbed out with the neutral value.
8866 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8867 Require IFN_VEC_SHL_INSERT for double reductions on
8868 variable-length vectors, or SLP reductions that have
8869 a neutral value. Require can_duplicate_and_interleave_p
8870 support for variable-length unchained SLP reductions if there
8871 is no neutral value, such as for MIN/MAX reductions. Also require
8872 the number of vector elements to be a multiple of the number of
8873 SLP statements when doing variable-length unchained SLP reductions.
8874 Update call to vect_create_epilog_for_reduction.
8875 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8876 and remove initial values.
8877 (duplicate_and_interleave): Make public.
8878 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8879 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8882 Alan Hayward <alan.hayward@arm.com>
8883 David Sherwood <david.sherwood@arm.com>
8885 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8886 (can_duplicate_and_interleave_p): New function.
8887 (vect_get_and_check_slp_defs): Take the vector of statements
8888 rather than just the current one. Remove excess parentheses.
8889 Restriction rejectinon of vect_constant_def and vect_external_def
8890 for variable-length vectors to boolean types, or types for which
8891 can_duplicate_and_interleave_p is false.
8892 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8893 (duplicate_and_interleave): New function.
8894 (vect_get_constant_vectors): Use gimple_build_vector for
8895 constant-length vectors and suitable variable-length constant
8896 vectors. Use duplicate_and_interleave for other variable-length
8897 vectors. Don't defer the update when inserting new statements.
8899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8900 Alan Hayward <alan.hayward@arm.com>
8901 David Sherwood <david.sherwood@arm.com>
8903 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8904 min_profitable_iters doesn't go negative.
8906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8907 Alan Hayward <alan.hayward@arm.com>
8908 David Sherwood <david.sherwood@arm.com>
8910 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8911 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8912 * optabs.def (vec_mask_load_lanes_optab): New optab.
8913 (vec_mask_store_lanes_optab): Likewise.
8914 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8915 (MASK_STORE_LANES): Likewise.
8916 * internal-fn.c (mask_load_lanes_direct): New macro.
8917 (mask_store_lanes_direct): Likewise.
8918 (expand_mask_load_optab_fn): Handle masked operations.
8919 (expand_mask_load_lanes_optab_fn): New macro.
8920 (expand_mask_store_optab_fn): Handle masked operations.
8921 (expand_mask_store_lanes_optab_fn): New macro.
8922 (direct_mask_load_lanes_optab_supported_p): Likewise.
8923 (direct_mask_store_lanes_optab_supported_p): Likewise.
8924 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8926 (vect_load_lanes_supported): Likewise.
8927 * tree-vect-data-refs.c (strip_conversion): New function.
8928 (can_group_stmts_p): Likewise.
8929 (vect_analyze_data_ref_accesses): Use it instead of checking
8930 for a pair of assignments.
8931 (vect_store_lanes_supported): Take a masked_p parameter.
8932 (vect_load_lanes_supported): Likewise.
8933 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8934 vect_store_lanes_supported and vect_load_lanes_supported.
8935 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8936 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8937 parameter. Don't allow gaps for masked accesses.
8938 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8939 and vect_load_lanes_supported.
8940 (get_load_store_type): Take a masked_p parameter and update
8941 call to get_group_load_store_type.
8942 (vectorizable_store): Update call to get_load_store_type.
8943 Handle IFN_MASK_STORE_LANES.
8944 (vectorizable_load): Update call to get_load_store_type.
8945 Handle IFN_MASK_LOAD_LANES.
8947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8948 Alan Hayward <alan.hayward@arm.com>
8949 David Sherwood <david.sherwood@arm.com>
8951 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8953 * config/aarch64/aarch64-protos.h
8954 (aarch64_sve_struct_memory_operand_p): Declare.
8955 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8956 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8957 (VPRED, vpred): Handle SVE structure modes.
8958 * config/aarch64/constraints.md (Utx): New constraint.
8959 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8960 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8961 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8962 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8963 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8964 structure modes. Split into pieces after RA.
8965 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8966 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8968 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8969 SVE structure modes.
8970 (aarch64_classify_address): Likewise.
8971 (sizetochar): Move earlier in file.
8972 (aarch64_print_operand): Handle SVE register lists.
8973 (aarch64_array_mode): New function.
8974 (aarch64_sve_struct_memory_operand_p): Likewise.
8975 (TARGET_ARRAY_MODE): Redefine.
8977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8978 Alan Hayward <alan.hayward@arm.com>
8979 David Sherwood <david.sherwood@arm.com>
8981 * target.def (array_mode): New target hook.
8982 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8983 * doc/tm.texi: Regenerate.
8984 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8985 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8986 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8988 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8992 Alan Hayward <alan.hayward@arm.com>
8993 David Sherwood <david.sherwood@arm.com>
8995 * fold-const.c (fold_binary_loc): Check the argument types
8996 rather than the result type when testing for a vector operation.
8998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9000 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9001 * doc/tm.texi: Regenerate.
9003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9004 Alan Hayward <alan.hayward@arm.com>
9005 David Sherwood <david.sherwood@arm.com>
9007 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9008 (sve): Document new AArch64 extension.
9009 * doc/md.texi (w): Extend the description of the AArch64
9010 constraint to include SVE vectors.
9011 (Upl, Upa): Document new AArch64 predicate constraints.
9012 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9014 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9015 (msve-vector-bits=): New option.
9016 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9017 SVE when these are disabled.
9018 (sve): New extension.
9019 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9020 modes. Adjust their number of units based on aarch64_sve_vg.
9021 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9022 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9023 aarch64_addr_query_type.
9024 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9025 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9026 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9027 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9028 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9029 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9030 (aarch64_simd_imm_zero_p): Delete.
9031 (aarch64_check_zero_based_sve_index_immediate): Declare.
9032 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9033 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9034 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9035 (aarch64_sve_float_mul_immediate_p): Likewise.
9036 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9038 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9039 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9040 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9041 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9042 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9043 (aarch64_regmode_natural_size): Likewise.
9044 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9045 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9047 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9048 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9049 for VG and the SVE predicate registers.
9050 (V_ALIASES): Add a "z"-prefixed alias.
9051 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9052 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9053 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9054 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9055 (REG_CLASS_NAMES): Add entries for them.
9056 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9057 and the predicate registers.
9058 (aarch64_sve_vg): Declare.
9059 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9060 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9061 (REGMODE_NATURAL_SIZE): Define.
9062 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9064 * config/aarch64/aarch64.c: Include cfgrtl.h.
9065 (simd_immediate_info): Add a constructor for series vectors,
9066 and an associated step field.
9067 (aarch64_sve_vg): New variable.
9068 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9069 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9070 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9071 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9072 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9073 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9074 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9075 (aarch64_get_mask_mode): New functions.
9076 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9077 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9078 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9079 predicate modes and predicate registers. Explicitly restrict
9080 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9081 to store a vector mode if it is recognized by
9082 aarch64_classify_vector_mode.
9083 (aarch64_regmode_natural_size): New function.
9084 (aarch64_hard_regno_caller_save_mode): Return the original mode
9086 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9087 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9088 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9089 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9091 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9092 does not overlap dest if the function is frame-related. Handle
9094 (aarch64_split_add_offset): New function.
9095 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9096 them aarch64_add_offset.
9097 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9098 and update call to aarch64_sub_sp.
9099 (aarch64_add_cfa_expression): New function.
9100 (aarch64_expand_prologue): Pass extra temporary registers to the
9101 functions above. Handle the case in which we need to emit new
9102 DW_CFA_expressions for registers that were originally saved
9103 relative to the stack pointer, but now have to be expressed
9104 relative to the frame pointer.
9105 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9107 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9108 IP0 and IP1 values for SVE frames.
9109 (aarch64_expand_vec_series): New function.
9110 (aarch64_expand_sve_widened_duplicate): Likewise.
9111 (aarch64_expand_sve_const_vector): Likewise.
9112 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9113 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9114 into the register, rather than emitting a SET directly.
9115 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9116 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9117 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9118 (offset_9bit_signed_scaled_p): New functions.
9119 (aarch64_replicate_bitmask_imm): New function.
9120 (aarch64_bitmask_imm): Use it.
9121 (aarch64_cannot_force_const_mem): Reject expressions involving
9122 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9123 (aarch64_classify_index): Handle SVE indices, by requiring
9124 a plain register index with a scale that matches the element size.
9125 (aarch64_classify_address): Handle SVE addresses. Assert that
9126 the mode of the address is VOIDmode or an integer mode.
9127 Update call to aarch64_classify_symbol.
9128 (aarch64_classify_symbolic_expression): Update call to
9129 aarch64_classify_symbol.
9130 (aarch64_const_vec_all_in_range_p): New function.
9131 (aarch64_print_vector_float_operand): Likewise.
9132 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9133 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9134 and the FP immediates 1.0 and 0.5.
9135 (aarch64_print_address_internal): Handle SVE addresses.
9136 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9137 (aarch64_regno_regclass): Handle predicate registers.
9138 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9140 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9141 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9142 (aarch64_convert_sve_vector_bits): New function.
9143 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9144 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9146 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9147 Handle SVE vector and predicate modes. Accept VL-based constants
9148 that need only one temporary register, and VL offsets that require
9149 no temporary registers.
9150 (aarch64_conditional_register_usage): Mark the predicate registers
9151 as fixed if SVE isn't available.
9152 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9153 Return true for SVE vector and predicate modes.
9154 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9155 rather than an unsigned int. Handle SVE modes.
9156 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9158 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9160 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9161 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9162 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9163 (aarch64_sve_float_mul_immediate_p): New functions.
9164 (aarch64_sve_valid_immediate): New function.
9165 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9166 Explicitly reject structure modes. Check for INDEX constants.
9167 Handle PTRUE and PFALSE constants.
9168 (aarch64_check_zero_based_sve_index_immediate): New function.
9169 (aarch64_simd_imm_zero_p): Delete.
9170 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9171 vector modes. Accept constants in the range of CNT[BHWD].
9172 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9173 ask for an Advanced SIMD mode.
9174 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9175 (aarch64_simd_vector_alignment): Handle SVE predicates.
9176 (aarch64_vectorize_preferred_vector_alignment): New function.
9177 (aarch64_simd_vector_alignment_reachable): Use it instead of
9179 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9180 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9182 (MAX_VECT_LEN): Delete.
9183 (expand_vec_perm_d): Add a vec_flags field.
9184 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9185 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9186 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9188 (aarch64_evpc_rev): Rename to...
9189 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9190 (aarch64_evpc_rev_global): New function.
9191 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9192 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9194 (aarch64_evpc_sve_tbl): New function.
9195 (aarch64_expand_vec_perm_const_1): Update after rename of
9196 aarch64_evpc_rev. Handle SVE permutes too, trying
9197 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9198 than aarch64_evpc_tbl.
9199 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9200 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9201 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9202 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9203 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9204 (aarch64_expand_sve_vcond): New functions.
9205 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9206 of aarch64_vector_mode_p.
9207 (aarch64_dwarf_poly_indeterminate_value): New function.
9208 (aarch64_compute_pressure_classes): Likewise.
9209 (aarch64_can_change_mode_class): Likewise.
9210 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9211 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9212 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9213 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9214 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9215 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9216 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9217 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9219 (Dn, Dl, Dr): Accept const as well as const_vector.
9220 (Dz): Likewise. Compare against CONST0_RTX.
9221 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9222 of "vector" where appropriate.
9223 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9224 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9225 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9226 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9227 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9228 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9229 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9230 (v_int_equiv): Extend to SVE modes.
9231 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9233 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9234 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9235 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9236 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9237 (SVE_COND_FP_CMP): New int iterators.
9238 (perm_hilo): Handle the new unpack unspecs.
9239 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9241 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9242 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9243 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9244 (aarch64_equality_operator, aarch64_constant_vector_operand)
9245 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9246 (aarch64_sve_nonimmediate_operand): Likewise.
9247 (aarch64_sve_general_operand): Likewise.
9248 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9249 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9250 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9251 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9252 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9253 (aarch64_sve_float_arith_immediate): Likewise.
9254 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9255 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9256 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9257 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9258 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9259 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9260 (aarch64_sve_float_arith_operand): Likewise.
9261 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9262 (aarch64_sve_float_mul_operand): Likewise.
9263 (aarch64_sve_vec_perm_operand): Likewise.
9264 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9265 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9266 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9267 as well as const_vector.
9268 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9269 in file. Use CONST0_RTX and CONSTM1_RTX.
9270 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9271 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9272 Use aarch64_simd_imm_zero.
9273 * config/aarch64/aarch64-sve.md: New file.
9274 * config/aarch64/aarch64.md: Include it.
9275 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9276 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9277 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9278 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9279 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9280 (sve): New attribute.
9281 (enabled): Disable instructions with the sve attribute unless
9283 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9284 aarch64_expand_mov_immediate.
9285 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9286 CNT[BHSD] immediates.
9287 (movti): Split CONST_POLY_INT moves into two halves.
9288 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9289 Split additions that need a temporary here if the destination
9290 is the stack pointer.
9291 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9292 (*add<mode>3_poly_1): New instruction.
9293 (set_clobber_cc): New expander.
9295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9297 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9298 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9299 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9300 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9301 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9302 Change innermode from fixed_mode_size to machine_mode.
9303 (simplify_subreg): Update call accordingly. Handle a constant-sized
9304 subreg of a variable-length CONST_VECTOR.
9306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9307 Alan Hayward <alan.hayward@arm.com>
9308 David Sherwood <david.sherwood@arm.com>
9310 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9311 (add_offset_to_base): New function, split out from...
9312 (create_mem_ref): ...here. When handling a scale other than 1,
9313 check first whether the address is valid without the offset.
9314 Add it into the base if so, leaving the index and scale as-is.
9316 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9319 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9320 fold_for_warn before checking if arg2 is INTEGER_CST.
9322 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9324 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9325 (store_multiple_operation): Delete.
9326 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9327 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9328 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9329 guarded by TARGET_STRING.
9330 (rs6000_output_load_multiple): Delete.
9331 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9332 OPTION_MASK_STRING / TARGET_STRING handling.
9333 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9334 (const rs6000_opt_masks) <"string">: Change mask to 0.
9335 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9336 (MASK_STRING): Delete.
9337 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9339 (load_multiple): Delete.
9346 (store_multiple): Delete.
9353 (movmemsi_8reg): Delete.
9354 (corresponding unnamed define_insn): Delete.
9355 (movmemsi_6reg): Delete.
9356 (corresponding unnamed define_insn): Delete.
9357 (movmemsi_4reg): Delete.
9358 (corresponding unnamed define_insn): Delete.
9359 (movmemsi_2reg): Delete.
9360 (corresponding unnamed define_insn): Delete.
9361 (movmemsi_1reg): Delete.
9362 (corresponding unnamed define_insn): Delete.
9363 * config/rs6000/rs6000.opt (mno-string): New.
9364 (mstring): Replace by deprecation warning stub.
9365 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9367 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9369 * regrename.c (regrename_do_replace): If replacing the same
9370 reg multiple times, try to reuse last created gen_raw_REG.
9373 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9374 main to workaround a bug in GDB.
9376 2018-01-12 Tom de Vries <tom@codesourcery.com>
9379 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9381 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9383 PR rtl-optimization/80481
9384 * ira-color.c (get_cap_member): New function.
9385 (allocnos_conflict_by_live_ranges_p): Use it.
9386 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9387 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9389 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9392 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9393 (*saddl_se_1): Ditto.
9395 (*ssubl_se_1): Ditto.
9397 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9399 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9400 rather than wi::to_widest for DR_INITs.
9401 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9402 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9403 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9405 (vect_analyze_group_access_1): Note that here.
9407 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9409 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9410 polynomial type sizes.
9412 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9414 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9415 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9416 (gimple_add_tmp_var): Likewise.
9418 2018-01-12 Martin Liska <mliska@suse.cz>
9420 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9421 (gimple_alloc_sizes): Likewise.
9422 (dump_gimple_statistics): Use PRIu64 in printf format.
9423 * gimple.h: Change uint64_t to int.
9425 2018-01-12 Martin Liska <mliska@suse.cz>
9427 * tree-core.h: Use uint64_t instead of int.
9428 * tree.c (tree_node_counts): Likewise.
9429 (tree_node_sizes): Likewise.
9430 (dump_tree_statistics): Use PRIu64 in printf format.
9432 2018-01-12 Martin Liska <mliska@suse.cz>
9434 * Makefile.in: As qsort_chk is implemented in vec.c, add
9435 vec.o to linkage of gencfn-macros.
9436 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9437 passing the info to record_node_allocation_statistics.
9438 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9440 * ggc-common.c (struct ggc_usage): Add operator== and use
9441 it in operator< and compare function.
9442 * mem-stats.h (struct mem_usage): Likewise.
9443 * vec.c (struct vec_usage): Remove operator< and compare
9444 function. Can be simply inherited.
9446 2018-01-12 Martin Jambor <mjambor@suse.cz>
9449 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9450 * tree-ssa-math-opts.c: Include domwalk.h.
9451 (convert_mult_to_fma_1): New function.
9452 (fma_transformation_info): New type.
9453 (fma_deferring_state): Likewise.
9454 (cancel_fma_deferring): New function.
9455 (result_of_phi): Likewise.
9456 (last_fma_candidate_feeds_initial_phi): Likewise.
9457 (convert_mult_to_fma): Added deferring logic, split actual
9458 transformation to convert_mult_to_fma_1.
9459 (math_opts_dom_walker): New type.
9460 (math_opts_dom_walker::after_dom_children): New method, body moved
9461 here from pass_optimize_widening_mul::execute, added deferring logic
9463 (pass_optimize_widening_mul::execute): Moved most of code to
9464 math_opts_dom_walker::after_dom_children.
9465 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9466 * config/i386/i386.c (ix86_option_override_internal): Added
9467 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9469 2018-01-12 Richard Biener <rguenther@suse.de>
9472 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9473 inline instance vars.
9475 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9478 * config/rx/rx.c (rx_is_restricted_memory_address):
9481 2018-01-12 Richard Biener <rguenther@suse.de>
9483 PR tree-optimization/80846
9484 * target.def (split_reduction): New target hook.
9485 * targhooks.c (default_split_reduction): New function.
9486 * targhooks.h (default_split_reduction): Declare.
9487 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9488 target requests first reduce vectors by combining low and high
9490 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9491 (get_vectype_for_scalar_type_and_size): Export.
9492 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9493 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9494 * doc/tm.texi: Regenerate.
9495 * config/i386/i386.c (ix86_split_reduction): Implement
9496 TARGET_VECTORIZE_SPLIT_REDUCTION.
9498 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9501 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9502 in PIC mode except for TARGET_VXWORKS_RTP.
9503 * config/sparc/sparc.c: Include cfgrtl.h.
9504 (TARGET_INIT_PIC_REG): Define.
9505 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9506 (sparc_pic_register_p): New predicate.
9507 (sparc_legitimate_address_p): Use it.
9508 (sparc_legitimize_pic_address): Likewise.
9509 (sparc_delegitimize_address): Likewise.
9510 (sparc_mode_dependent_address_p): Likewise.
9511 (gen_load_pcrel_sym): Remove 4th parameter.
9512 (load_got_register): Adjust call to above. Remove obsolete stuff.
9513 (sparc_expand_prologue): Do not call load_got_register here.
9514 (sparc_flat_expand_prologue): Likewise.
9515 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9516 (sparc_use_pseudo_pic_reg): New function.
9517 (sparc_init_pic_reg): Likewise.
9518 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9519 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9521 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9523 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9524 Add item for branch_cost.
9526 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9528 PR rtl-optimization/83565
9529 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9530 not extend the result to a larger mode for rotate operations.
9531 (num_sign_bit_copies1): Likewise.
9533 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9536 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9538 Use values-Xc.o for -pedantic.
9539 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9541 2018-01-12 Martin Liska <mliska@suse.cz>
9544 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9546 (possible_polymorphic_call_targets): Use it.
9547 (ipa_devirt): Likewise.
9549 2018-01-12 Martin Liska <mliska@suse.cz>
9551 * profile-count.h (enum profile_quality): Use 0 as invalid
9552 enum value of profile_quality.
9554 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9556 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9557 -mext-string options.
9559 2018-01-12 Richard Biener <rguenther@suse.de>
9561 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9562 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9563 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9565 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9567 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9569 * configure.ac (--with-long-double-format): Add support for the
9570 configuration option to change the default long double format on
9572 * config.gcc (powerpc*-linux*-*): Likewise.
9573 * configure: Regenerate.
9574 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9575 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9576 used without modification.
9578 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9580 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9581 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9582 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9583 MISC_BUILTIN_SPEC_BARRIER.
9584 (rs6000_init_builtins): Likewise.
9585 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9587 (speculation_barrier): New define_insn.
9588 * doc/extend.texi: Document __builtin_speculation_barrier.
9590 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9593 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9594 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9595 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9597 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9598 integral modes instead of "ss" and "sd".
9599 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9600 vectors with 32-bit and 64-bit elements.
9601 (vecdupssescalarmodesuffix): New mode attribute.
9602 (vec_dup<mode>): Use it.
9604 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9607 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9608 frame if argument is passed on stack.
9610 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9613 * ree.c (combine_reaching_defs): Optimize also
9614 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9615 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9617 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9620 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9622 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9625 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9626 after they are computed.
9628 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9630 PR tree-optimization/83695
9631 * gimple-loop-linterchange.cc
9632 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9633 reset cached scev information after interchange.
9634 (pass_linterchange::execute): Remove call to scev_reset_htab.
9636 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9638 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9639 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9640 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9641 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9642 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9643 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9644 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9645 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9646 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9647 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9648 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9649 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9650 (V_lane_reg): Likewise.
9651 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9653 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9654 (vfmal_lane_low<mode>_intrinsic,
9655 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9656 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9657 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9658 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9659 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9660 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9662 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9664 * config/arm/arm-cpus.in (fp16fml): New feature.
9665 (ALL_SIMD): Add fp16fml.
9666 (armv8.2-a): Add fp16fml as an option.
9667 (armv8.3-a): Likewise.
9668 (armv8.4-a): Add fp16fml as part of fp16.
9669 * config/arm/arm.h (TARGET_FP16FML): Define.
9670 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9672 * config/arm/arm-modes.def (V2HF): Define.
9673 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9674 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9675 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9676 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9677 vfmsl_low, vfmsl_high): New set of builtins.
9678 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9679 (vfml_op): New code attribute.
9680 (VFMLHALVES): New int iterator.
9681 (VFML, VFMLSEL): New mode attributes.
9682 (V_reg): Define mapping for V2HF.
9683 (V_hi, V_lo): New mode attributes.
9684 (VF_constraint): Likewise.
9685 (vfml_half, vfml_half_selector): New int attributes.
9686 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9688 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9689 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9691 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9692 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9693 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9694 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9696 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9697 Document new effective target and option set.
9699 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9701 * config/arm/arm-cpus.in (armv8_4): New feature.
9702 (ARMv8_4a): New fgroup.
9703 (armv8.4-a): New arch.
9704 * config/arm/arm-tables.opt: Regenerate.
9705 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9706 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9707 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9708 Add matching rules for -march=armv8.4-a and extensions.
9709 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9711 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9714 * config/rx/rx.md (BW): New mode attribute.
9715 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9717 2018-01-11 Richard Biener <rguenther@suse.de>
9719 PR tree-optimization/83435
9720 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9721 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9722 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9724 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9725 Alan Hayward <alan.hayward@arm.com>
9726 David Sherwood <david.sherwood@arm.com>
9728 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9730 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9731 (aarch64_print_address_internal): Use it to check for a zero offset.
9733 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9734 Alan Hayward <alan.hayward@arm.com>
9735 David Sherwood <david.sherwood@arm.com>
9737 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9738 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9739 Return a poly_int64 rather than a HOST_WIDE_INT.
9740 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9741 rather than a HOST_WIDE_INT.
9742 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9743 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9744 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9745 final_offset from HOST_WIDE_INT to poly_int64.
9746 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9747 to_constant when getting the number of units in an Advanced SIMD
9749 (aarch64_builtin_vectorized_function): Check for a constant number
9751 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9753 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9754 attribute instead of GET_MODE_NUNITS.
9755 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9756 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9757 GET_MODE_SIZE for fixed-size registers.
9758 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9759 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9760 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9761 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9762 (aarch64_print_operand, aarch64_print_address_internal)
9763 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9764 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9765 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9766 Handle polynomial GET_MODE_SIZE.
9767 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9768 wider than SImode without modification.
9769 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9770 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9771 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9772 passing and returning SVE modes.
9773 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9774 rather than GEN_INT.
9775 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9776 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9777 (aarch64_allocate_and_probe_stack_space): Likewise.
9778 (aarch64_layout_frame): Cope with polynomial offsets.
9779 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9780 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9782 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9783 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9784 poly_int64 rather than a HOST_WIDE_INT.
9785 (aarch64_get_separate_components, aarch64_process_components)
9786 (aarch64_expand_prologue, aarch64_expand_epilogue)
9787 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9788 (aarch64_anchor_offset): New function, split out from...
9789 (aarch64_legitimize_address): ...here.
9790 (aarch64_builtin_vectorization_cost): Handle polynomial
9791 TYPE_VECTOR_SUBPARTS.
9792 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9794 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9795 number of elements from the PARALLEL rather than the mode.
9796 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9797 rather than GET_MODE_BITSIZE.
9798 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9799 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9800 (aarch64_expand_vec_perm_const_1): Handle polynomial
9801 d->perm.length () and d->perm elements.
9802 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9803 Apply to_constant to d->perm elements.
9804 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9805 polynomial CONST_VECTOR_NUNITS.
9806 (aarch64_move_pointer): Take amount as a poly_int64 rather
9808 (aarch64_progress_pointer): Avoid temporary variable.
9809 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9810 the mode attribute instead of GET_MODE.
9812 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9813 Alan Hayward <alan.hayward@arm.com>
9814 David Sherwood <david.sherwood@arm.com>
9816 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9817 x exists before using it.
9818 (aarch64_add_constant_internal): Rename to...
9819 (aarch64_add_offset_1): ...this. Replace regnum with separate
9820 src and dest rtxes. Handle the case in which they're different,
9821 including when the offset is zero. Replace scratchreg with an rtx.
9822 Use 2 additions if there is no spare register into which we can
9823 move a 16-bit constant.
9824 (aarch64_add_constant): Delete.
9825 (aarch64_add_offset): Replace reg with separate src and dest
9826 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9827 Use aarch64_add_offset_1.
9828 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9829 an rtx rather than an int. Take the delta as a poly_int64
9830 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9831 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9832 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9833 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9834 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9836 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9837 aarch64_add_constant.
9839 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9841 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9842 Use scalar_float_mode.
9844 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9846 * config/aarch64/aarch64-simd.md
9847 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9848 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9849 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9850 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9851 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9852 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9853 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9854 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9855 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9856 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9858 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9861 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9862 targ_options->x_arm_arch_string is non NULL.
9864 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9866 * config/aarch64/aarch64.h
9867 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9869 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9872 * expmed.c (emit_store_flag_force): Swap if const op0
9873 and change VOIDmode to mode of op0.
9875 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9877 PR rtl-optimization/83761
9878 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9879 than bytes to mode_for_size.
9881 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9884 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9885 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9888 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9891 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9892 when in layout mode.
9893 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9894 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9897 2018-01-10 Michael Collison <michael.collison@arm.com>
9899 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9900 * config/aarch64/aarch64-option-extension.def: Add
9901 AARCH64_OPT_EXTENSION of 'fp16fml'.
9902 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9903 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9904 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9905 * config/aarch64/constraints.md (Ui7): New constraint.
9906 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9907 (VFMLA_SEL_W): Ditto.
9910 (VFMLA16_LOW): New int iterator.
9911 (VFMLA16_HIGH): Ditto.
9912 (UNSPEC_FMLAL): New unspec.
9913 (UNSPEC_FMLSL): Ditto.
9914 (UNSPEC_FMLAL2): Ditto.
9915 (UNSPEC_FMLSL2): Ditto.
9916 (f16mac): New code attribute.
9917 * config/aarch64/aarch64-simd-builtins.def
9918 (aarch64_fmlal_lowv2sf): Ditto.
9919 (aarch64_fmlsl_lowv2sf): Ditto.
9920 (aarch64_fmlalq_lowv4sf): Ditto.
9921 (aarch64_fmlslq_lowv4sf): Ditto.
9922 (aarch64_fmlal_highv2sf): Ditto.
9923 (aarch64_fmlsl_highv2sf): Ditto.
9924 (aarch64_fmlalq_highv4sf): Ditto.
9925 (aarch64_fmlslq_highv4sf): Ditto.
9926 (aarch64_fmlal_lane_lowv2sf): Ditto.
9927 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9928 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9929 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9930 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9931 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9932 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9933 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9934 (aarch64_fmlal_lane_highv2sf): Ditto.
9935 (aarch64_fmlsl_lane_highv2sf): Ditto.
9936 (aarch64_fmlal_laneq_highv2sf): Ditto.
9937 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9938 (aarch64_fmlalq_lane_highv4sf): Ditto.
9939 (aarch64_fmlsl_lane_highv4sf): Ditto.
9940 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9941 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9942 * config/aarch64/aarch64-simd.md:
9943 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9944 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9945 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9946 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9947 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9948 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9949 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9950 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9951 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9952 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9953 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9954 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9955 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9956 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9957 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9958 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9959 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9960 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9961 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9962 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9963 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9964 (vfmlsl_low_u32): Ditto.
9965 (vfmlalq_low_u32): Ditto.
9966 (vfmlslq_low_u32): Ditto.
9967 (vfmlal_high_u32): Ditto.
9968 (vfmlsl_high_u32): Ditto.
9969 (vfmlalq_high_u32): Ditto.
9970 (vfmlslq_high_u32): Ditto.
9971 (vfmlal_lane_low_u32): Ditto.
9972 (vfmlsl_lane_low_u32): Ditto.
9973 (vfmlal_laneq_low_u32): Ditto.
9974 (vfmlsl_laneq_low_u32): Ditto.
9975 (vfmlalq_lane_low_u32): Ditto.
9976 (vfmlslq_lane_low_u32): Ditto.
9977 (vfmlalq_laneq_low_u32): Ditto.
9978 (vfmlslq_laneq_low_u32): Ditto.
9979 (vfmlal_lane_high_u32): Ditto.
9980 (vfmlsl_lane_high_u32): Ditto.
9981 (vfmlal_laneq_high_u32): Ditto.
9982 (vfmlsl_laneq_high_u32): Ditto.
9983 (vfmlalq_lane_high_u32): Ditto.
9984 (vfmlslq_lane_high_u32): Ditto.
9985 (vfmlalq_laneq_high_u32): Ditto.
9986 (vfmlslq_laneq_high_u32): Ditto.
9987 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9988 (AARCH64_FL_FOR_ARCH8_4): New.
9989 (AARCH64_ISA_F16FML): New ISA flag.
9990 (TARGET_F16FML): New feature flag for fp16fml.
9991 (doc/invoke.texi): Document new fp16fml option.
9993 2018-01-10 Michael Collison <michael.collison@arm.com>
9995 * config/aarch64/aarch64-builtins.c:
9996 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9997 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9998 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9999 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10000 (AARCH64_ISA_SHA3): New ISA flag.
10001 (TARGET_SHA3): New feature flag for sha3.
10002 * config/aarch64/iterators.md (sha512_op): New int attribute.
10003 (CRYPTO_SHA512): New int iterator.
10004 (UNSPEC_SHA512H): New unspec.
10005 (UNSPEC_SHA512H2): Ditto.
10006 (UNSPEC_SHA512SU0): Ditto.
10007 (UNSPEC_SHA512SU1): Ditto.
10008 * config/aarch64/aarch64-simd-builtins.def
10009 (aarch64_crypto_sha512hqv2di): New builtin.
10010 (aarch64_crypto_sha512h2qv2di): Ditto.
10011 (aarch64_crypto_sha512su0qv2di): Ditto.
10012 (aarch64_crypto_sha512su1qv2di): Ditto.
10013 (aarch64_eor3qv8hi): Ditto.
10014 (aarch64_rax1qv2di): Ditto.
10015 (aarch64_xarqv2di): Ditto.
10016 (aarch64_bcaxqv8hi): Ditto.
10017 * config/aarch64/aarch64-simd.md:
10018 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10019 (aarch64_crypto_sha512su0qv2di): Ditto.
10020 (aarch64_crypto_sha512su1qv2di): Ditto.
10021 (aarch64_eor3qv8hi): Ditto.
10022 (aarch64_rax1qv2di): Ditto.
10023 (aarch64_xarqv2di): Ditto.
10024 (aarch64_bcaxqv8hi): Ditto.
10025 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10026 (vsha512h2q_u64): Ditto.
10027 (vsha512su0q_u64): Ditto.
10028 (vsha512su1q_u64): Ditto.
10029 (veor3q_u16): Ditto.
10030 (vrax1q_u64): Ditto.
10031 (vxarq_u64): Ditto.
10032 (vbcaxq_u16): Ditto.
10033 * config/arm/types.md (crypto_sha512): New type attribute.
10034 (crypto_sha3): Ditto.
10035 (doc/invoke.texi): Document new sha3 option.
10037 2018-01-10 Michael Collison <michael.collison@arm.com>
10039 * config/aarch64/aarch64-builtins.c:
10040 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10041 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10042 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10043 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10044 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10045 (AARCH64_ISA_SM4): New ISA flag.
10046 (TARGET_SM4): New feature flag for sm4.
10047 * config/aarch64/aarch64-simd-builtins.def
10048 (aarch64_sm3ss1qv4si): Ditto.
10049 (aarch64_sm3tt1aq4si): Ditto.
10050 (aarch64_sm3tt1bq4si): Ditto.
10051 (aarch64_sm3tt2aq4si): Ditto.
10052 (aarch64_sm3tt2bq4si): Ditto.
10053 (aarch64_sm3partw1qv4si): Ditto.
10054 (aarch64_sm3partw2qv4si): Ditto.
10055 (aarch64_sm4eqv4si): Ditto.
10056 (aarch64_sm4ekeyqv4si): Ditto.
10057 * config/aarch64/aarch64-simd.md:
10058 (aarch64_sm3ss1qv4si): Ditto.
10059 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10060 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10061 (aarch64_sm4eqv4si): Ditto.
10062 (aarch64_sm4ekeyqv4si): Ditto.
10063 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10064 (sm3part_op): Ditto.
10065 (CRYPTO_SM3TT): Ditto.
10066 (CRYPTO_SM3PART): Ditto.
10067 (UNSPEC_SM3SS1): New unspec.
10068 (UNSPEC_SM3TT1A): Ditto.
10069 (UNSPEC_SM3TT1B): Ditto.
10070 (UNSPEC_SM3TT2A): Ditto.
10071 (UNSPEC_SM3TT2B): Ditto.
10072 (UNSPEC_SM3PARTW1): Ditto.
10073 (UNSPEC_SM3PARTW2): Ditto.
10074 (UNSPEC_SM4E): Ditto.
10075 (UNSPEC_SM4EKEY): Ditto.
10076 * config/aarch64/constraints.md (Ui2): New constraint.
10077 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10078 * config/arm/types.md (crypto_sm3): New type attribute.
10079 (crypto_sm4): Ditto.
10080 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10081 (vsm3tt1aq_u32): Ditto.
10082 (vsm3tt1bq_u32): Ditto.
10083 (vsm3tt2aq_u32): Ditto.
10084 (vsm3tt2bq_u32): Ditto.
10085 (vsm3partw1q_u32): Ditto.
10086 (vsm3partw2q_u32): Ditto.
10087 (vsm4eq_u32): Ditto.
10088 (vsm4ekeyq_u32): Ditto.
10089 (doc/invoke.texi): Document new sm4 option.
10091 2018-01-10 Michael Collison <michael.collison@arm.com>
10093 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10094 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10095 (AARCH64_FL_FOR_ARCH8_4): New.
10096 (AARCH64_FL_V8_4): New flag.
10097 (doc/invoke.texi): Document new armv8.4-a option.
10099 2018-01-10 Michael Collison <michael.collison@arm.com>
10101 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10102 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10103 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10104 * config/aarch64/aarch64-option-extension.def: Add
10105 AARCH64_OPT_EXTENSION of 'sha2'.
10106 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10107 (crypto): Disable sha2 and aes if crypto disabled.
10108 (crypto): Enable aes and sha2 if enabled.
10109 (simd): Disable sha2 and aes if simd disabled.
10110 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10112 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10113 (TARGET_SHA2): New feature flag for sha2.
10114 (TARGET_AES): New feature flag for aes.
10115 * config/aarch64/aarch64-simd.md:
10116 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10117 conditional on TARGET_AES.
10118 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10119 (aarch64_crypto_sha1hsi): Make pattern conditional
10121 (aarch64_crypto_sha1hv4si): Ditto.
10122 (aarch64_be_crypto_sha1hv4si): Ditto.
10123 (aarch64_crypto_sha1su1v4si): Ditto.
10124 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10125 (aarch64_crypto_sha1su0v4si): Ditto.
10126 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10127 (aarch64_crypto_sha256su0v4si): Ditto.
10128 (aarch64_crypto_sha256su1v4si): Ditto.
10129 (doc/invoke.texi): Document new aes and sha2 options.
10131 2018-01-10 Martin Sebor <msebor@redhat.com>
10133 PR tree-optimization/83781
10134 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10137 2018-01-11 Martin Sebor <msebor@gmail.com>
10138 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10140 PR tree-optimization/83501
10141 PR tree-optimization/81703
10143 * tree-ssa-strlen.c (get_string_cst): Rename...
10144 (get_string_len): ...to this. Handle global constants.
10145 (handle_char_store): Adjust.
10147 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10148 Jim Wilson <jimw@sifive.com>
10150 * config/riscv/riscv-protos.h (riscv_output_return): New.
10151 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10152 (riscv_attribute_table, riscv_output_return),
10153 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10154 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10155 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10157 (riscv_expand_prologue): Add early return for naked function.
10158 (riscv_expand_epilogue): Likewise.
10159 (riscv_function_ok_for_sibcall): Return false for naked function.
10160 (riscv_set_current_function): New.
10161 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10162 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10163 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10164 * doc/extend.texi (RISC-V Function Attributes): New.
10166 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10168 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10169 check for 128-bit long double before checking TCmode.
10170 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10171 128-bit long doubles before checking TFmode or TCmode.
10172 (FLOAT128_IBM_P): Likewise.
10174 2018-01-10 Martin Sebor <msebor@redhat.com>
10176 PR tree-optimization/83671
10177 * builtins.c (c_strlen): Unconditionally return zero for the empty
10179 Use -Warray-bounds for warnings.
10180 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10181 for non-constant array indices with COMPONENT_REF, arrays of
10182 arrays, and pointers to arrays.
10183 (gimple_fold_builtin_strlen): Determine and set length range for
10184 non-constant character arrays.
10186 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10188 PR middle-end/81897
10189 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10192 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10194 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10196 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10199 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10200 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10201 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10202 indexed_or_indirect_operand predicate.
10203 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10204 (*vsx_le_perm_load_v8hi): Likewise.
10205 (*vsx_le_perm_load_v16qi): Likewise.
10206 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10207 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10208 (*vsx_le_perm_store_v8hi): Likewise.
10209 (*vsx_le_perm_store_v16qi): Likewise.
10210 (eight unnamed splitters): Likewise.
10212 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10214 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10215 * config/rs6000/emmintrin.h: Likewise.
10216 * config/rs6000/mmintrin.h: Likewise.
10217 * config/rs6000/xmmintrin.h: Likewise.
10219 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10222 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10224 * tree.c (tree_nop_conversion): Return true for location wrapper
10226 (maybe_wrap_with_location): New function.
10227 (selftest::check_strip_nops): New function.
10228 (selftest::test_location_wrappers): New function.
10229 (selftest::tree_c_tests): Call it.
10230 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10231 (maybe_wrap_with_location): New decl.
10232 (EXPR_LOCATION_WRAPPER_P): New macro.
10233 (location_wrapper_p): New inline function.
10234 (tree_strip_any_location_wrapper): New inline function.
10236 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10239 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10240 stack_realign_offset for the largest alignment of stack slot
10242 (ix86_find_max_used_stack_alignment): New function.
10243 (ix86_finalize_stack_frame_flags): Use it. Set
10244 max_used_stack_alignment if we don't realign stack.
10245 * config/i386/i386.h (machine_function): Add
10246 max_used_stack_alignment.
10248 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10250 * config/arm/arm.opt (-mbranch-cost): New option.
10251 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10254 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10257 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10258 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10260 2018-01-10 Richard Biener <rguenther@suse.de>
10263 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10264 early out so it also covers the case where we have a non-NULL
10267 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10269 PR tree-optimization/83753
10270 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10271 for non-strided grouped accesses if the number of elements is 1.
10273 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10276 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10277 * i386.h (TARGET_USE_GATHER): Define.
10278 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10280 2018-01-10 Martin Liska <mliska@suse.cz>
10283 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10284 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10286 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10287 CLEANUP_NO_PARTITIONING is not set.
10289 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10291 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10292 for vectors, as a partial revert of r254296.
10293 * rtl.h (const_vec_p): Delete.
10294 (const_vec_duplicate_p): Don't test for vector CONSTs.
10295 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10296 * expmed.c (make_tree): Likewise.
10299 * common.md (E, F): Use CONSTANT_P instead of checking for
10301 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10302 checking for CONST_VECTOR.
10304 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10306 PR middle-end/83575
10307 * predict.c (force_edge_cold): Handle in more sane way edges
10308 with no prediction.
10310 2018-01-09 Carl Love <cel@us.ibm.com>
10312 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10314 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10315 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10316 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10317 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10318 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10319 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10320 * config/rs6000/rs6000-protos.h: Add extern defition for
10321 rs6000_generate_float2_double_code.
10322 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10324 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10325 (float2_v2df): Add define_expand.
10327 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10330 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10331 op_mode in the force_to_mode call.
10333 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10335 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10336 instead of checking each element individually.
10337 (aarch64_evpc_uzp): Likewise.
10338 (aarch64_evpc_zip): Likewise.
10339 (aarch64_evpc_ext): Likewise.
10340 (aarch64_evpc_rev): Likewise.
10341 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10342 instead of checking each element individually. Return true without
10344 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10345 whether all selected elements come from the same input, instead of
10346 checking each element individually. Remove calls to gen_rtx_REG,
10347 start_sequence and end_sequence and instead assert that no rtl is
10350 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10352 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10353 order of HIGH and CONST checks.
10355 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10357 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10358 if the destination isn't an SSA_NAME.
10360 2018-01-09 Richard Biener <rguenther@suse.de>
10362 PR tree-optimization/83668
10363 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10365 (canonicalize_loop_form): ... here, renamed from ...
10366 (canonicalize_loop_closed_ssa_form): ... this and amended to
10367 swap successor edges for loop exit blocks to make us use
10368 the RPO order we need for initial schedule generation.
10370 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10372 PR tree-optimization/64811
10373 * match.pd: When optimizing comparisons with Inf, avoid
10374 introducing or losing exceptions from comparisons with NaN.
10376 2018-01-09 Martin Liska <mliska@suse.cz>
10379 * asan.c (shadow_mem_size): Add gcc_assert.
10381 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10383 Don't save registers in main().
10386 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10387 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10388 * config/avr/avr.c (avr_set_current_function): Don't error if
10389 naked, OS_task or OS_main are specified at the same time.
10390 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10392 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10394 * common/config/avr/avr-common.c (avr_option_optimization_table):
10395 Switch on -mmain-is-OS_task for optimizing compilations.
10397 2018-01-09 Richard Biener <rguenther@suse.de>
10399 PR tree-optimization/83572
10400 * graphite.c: Include cfganal.h.
10401 (graphite_transform_loops): Connect infinite loops to exit
10402 and remove fake edges at the end.
10404 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10406 * ipa-inline.c (edge_badness): Revert accidental checkin.
10408 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10411 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10412 symbols; not inline clones.
10414 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10417 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10418 hard registers. Formatting fixes.
10420 PR preprocessor/83722
10421 * gcc.c (try_generate_repro): Pass
10422 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10423 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10426 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10427 Kito Cheng <kito.cheng@gmail.com>
10429 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10430 (riscv_leaf_function_p): Delete.
10431 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10433 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10435 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10437 (do_ifelse): New function.
10438 (do_isel): New function.
10439 (do_sub3): New function.
10440 (do_add3): New function.
10441 (do_load_mask_compare): New function.
10442 (do_overlap_load_compare): New function.
10443 (expand_compare_loop): New function.
10444 (expand_block_compare): Call expand_compare_loop() when appropriate.
10445 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10446 option description.
10447 (-mblock-compare-inline-loop-limit): New option.
10449 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10452 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10453 Reverse order of second and third operands in first alternative.
10454 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10455 of first and second elements in UNSPEC_VPERMR vector.
10456 (altivec_expand_vec_perm_le): Likewise.
10458 2018-01-08 Jeff Law <law@redhat.com>
10460 PR rtl-optimizatin/81308
10461 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10462 (process_switch): If group_case_labels makes a change, then set
10464 (pass_convert_switch::execute): If a switch is converted, then
10465 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10467 PR rtl-optimization/81308
10468 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10471 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10473 PR target/83663 - Revert r255946
10474 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10475 generation for cases where splatting a value is not useful.
10476 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10477 across a vec_duplicate and a paradoxical subreg forming a vector
10478 mode to a vec_concat.
10480 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10482 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10483 -march=armv8.3-a variants.
10484 * config/arm/t-multilib: Likewise.
10485 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10487 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10489 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10491 (cceq_ior_compare_complement): Give it a name so I can use it, and
10492 change boolean_or_operator predicate to boolean_operator so it can
10493 be used to generate a crand.
10494 (eqne): New code iterator.
10495 (bd/bd_neg): New code_attrs.
10496 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10497 a single define_insn.
10498 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10499 decrement (bdnzt/bdnzf/bdzt/bdzf).
10500 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10501 with the new names of the branch decrement patterns, and added the
10502 names of the branch decrement conditional patterns.
10504 2018-01-08 Richard Biener <rguenther@suse.de>
10506 PR tree-optimization/83563
10507 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10510 2018-01-08 Richard Biener <rguenther@suse.de>
10512 PR middle-end/83713
10513 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10515 2018-01-08 Richard Biener <rguenther@suse.de>
10517 PR tree-optimization/83685
10518 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10519 references to abnormals.
10521 2018-01-08 Richard Biener <rguenther@suse.de>
10524 * dwarf2out.c (output_indirect_strings): Handle empty
10525 skeleton_debug_str_hash.
10526 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10528 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10530 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10531 (emit_store_direct): Likewise.
10532 (arc_trampoline_adjust_address): Likewise.
10533 (arc_asm_trampoline_template): New function.
10534 (arc_initialize_trampoline): Use asm_trampoline_template.
10535 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10536 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10537 * config/arc/arc.md (flush_icache): Delete pattern.
10539 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10541 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10542 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10545 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10548 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10549 by not USED_FOR_TARGET.
10550 (make_pass_resolve_sw_modes): Likewise.
10552 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10554 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10557 2018-01-08 Richard Biener <rguenther@suse.de>
10559 PR middle-end/83580
10560 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10562 2018-01-08 Richard Biener <rguenther@suse.de>
10564 PR middle-end/83517
10565 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10567 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10569 PR middle-end/81897
10570 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10571 basic blocks with a small number of successors.
10572 (convert_control_dep_chain_into_preds): Improve handling of
10574 (dump_predicates): Split apart into...
10575 (dump_pred_chain): ...here...
10576 (dump_pred_info): ...and here.
10577 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10578 (can_chain_union_be_invalidated_p): Improve check for invalidation
10580 (uninit_uses_cannot_happen): Avoid unnecessary if
10581 convert_control_dep_chain_into_preds yielded nothing.
10583 2018-01-06 Martin Sebor <msebor@redhat.com>
10585 PR tree-optimization/83640
10586 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10587 subtracting negative offset from size.
10588 (builtin_access::overlap): Adjust offset bounds of the access to fall
10589 within the size of the object if possible.
10591 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10593 PR rtl-optimization/83699
10594 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10595 extract_bit_field_as_subreg to cases in which the extracted
10596 value is also a vector.
10598 * lra-constraints.c (process_alt_operands): Test for the equivalence
10599 substitutions when detecting a possible reload cycle.
10601 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10604 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10605 by default if flag_selective_schedling{,2}. Formatting fixes.
10607 PR rtl-optimization/83682
10608 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10609 if it has non-VECTOR_MODE element mode.
10610 (vec_duplicate_p): Likewise.
10612 PR middle-end/83694
10613 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10614 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10616 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10619 * config/i386/i386-builtin.def
10620 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10621 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10622 Require also OPTION_MASK_ISA_AVX512F in addition to
10623 OPTION_MASK_ISA_GFNI.
10624 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10625 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10626 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10627 to OPTION_MASK_ISA_GFNI.
10628 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10629 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10630 OPTION_MASK_ISA_AVX512BW.
10631 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10632 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10633 addition to OPTION_MASK_ISA_GFNI.
10634 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10635 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10636 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10637 to OPTION_MASK_ISA_GFNI.
10638 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10639 a requirement for all ISAs rather than any of them with a few
10641 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10643 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10644 bitmasks to be enabled with 3 exceptions, instead of requiring any
10645 enabled ISA with lots of exceptions.
10646 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10647 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10648 Change avx512bw in isa attribute to avx512f.
10649 * config/i386/sgxintrin.h: Add license boilerplate.
10650 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10651 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10652 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10653 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10655 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10656 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10657 temporarily sse2 rather than sse if not enabled already.
10660 * config/i386/sse.md (VI248_VLBW): Rename to ...
10661 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10662 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10663 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10664 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10665 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10666 mode iterator instead of VI248_VLBW.
10668 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10670 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10671 (record_modified): Skip clobbers; add debug output.
10672 (param_change_prob): Use sreal frequencies.
10674 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10676 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10677 punt for user-aligned variables.
10679 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10681 * tree-chrec.c (chrec_contains_symbols): Return true for
10684 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10687 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10688 of (x|y) == x for BICS pattern.
10690 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10692 PR tree-optimization/83605
10693 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10694 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10697 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10699 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10700 * config/epiphany/rtems.h: New file.
10702 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10703 Uros Bizjak <ubizjak@gmail.com>
10706 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10707 QIreg_operand instead of register_operand predicate.
10708 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10709 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10710 comments instead of -fmitigate[-_]rop.
10712 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10715 * cgraphunit.c (symbol_table::compile): Switch to text_section
10716 before calling assembly_start debug hook.
10717 * run-rtl-passes.c (run_rtl_passes): Likewise.
10720 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10722 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10723 range_int_cst_p rather than !symbolic_range_p before calling
10724 extract_range_from_multiplicative_op_1.
10726 2018-01-04 Jeff Law <law@redhat.com>
10728 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10729 redundant test in assertion.
10731 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10733 * doc/rtl.texi: Document machine_mode wrapper classes.
10735 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10737 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10738 using tree_to_uhwi.
10740 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10742 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10743 the VEC_PERM_EXPR fold to fail.
10745 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10748 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10749 to switched_sections.
10751 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10754 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10755 test for d.testing.
10757 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10760 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10761 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10763 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10766 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10767 is BLKmode and bitpos not zero or mode change is needed.
10769 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10772 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10775 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10778 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10779 instead of MULT rtx. Update all corresponding splitters.
10780 (*saddl_se): Ditto.
10781 (*ssub<modesuffix>): Ditto.
10782 (*ssubl_se): Ditto.
10783 (*cmp_sadd_di): Update split patterns.
10784 (*cmp_sadd_si): Ditto.
10785 (*cmp_sadd_sidi): Ditto.
10786 (*cmp_ssub_di): Ditto.
10787 (*cmp_ssub_si): Ditto.
10788 (*cmp_ssub_sidi): Ditto.
10789 * config/alpha/predicates.md (const23_operand): New predicate.
10790 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10791 Look for ASHIFT, not MULT inner operand.
10792 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10794 2018-01-04 Martin Liska <mliska@suse.cz>
10796 PR gcov-profile/83669
10797 * gcov.c (output_intermediate_file): Add version to intermediate
10799 * doc/gcov.texi: Document new field 'version' in intermediate
10800 file format. Fix location of '-k' option of gcov command.
10802 2018-01-04 Martin Liska <mliska@suse.cz>
10805 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10807 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10809 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10811 2018-01-03 Martin Sebor <msebor@redhat.com>
10813 PR tree-optimization/83655
10814 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10815 checking calls with invalid arguments.
10817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10819 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10820 (vectorizable_mask_load_store): Delete.
10821 (vectorizable_call): Return false for masked loads and stores.
10822 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10823 instead of gimple_assign_rhs1.
10824 (vectorizable_load): Handle IFN_MASK_LOAD.
10825 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10829 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10831 (vectorizable_mask_load_store): ...here.
10832 (vectorizable_load): ...and here.
10834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10836 * tree-vect-stmts.c (vect_build_all_ones_mask)
10837 (vect_build_zero_merge_argument): New functions, split out from...
10838 (vectorizable_load): ...here.
10840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10842 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10844 (vectorizable_mask_load_store): ...here.
10845 (vectorizable_store): ...and here.
10847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10849 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10851 (vectorizable_mask_load_store): ...here.
10853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10855 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10856 (vect_model_store_cost): Take a vec_load_store_type instead of a
10858 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10859 (vect_model_store_cost): Take a vec_load_store_type instead of a
10861 (vectorizable_mask_load_store): Update accordingly.
10862 (vectorizable_store): Likewise.
10863 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10867 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10868 IFN_MASK_LOAD calls here rather than...
10869 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10872 Alan Hayward <alan.hayward@arm.com>
10873 David Sherwood <david.sherwood@arm.com>
10875 * expmed.c (extract_bit_field_1): For vector extracts,
10876 fall back to extract_bit_field_as_subreg if vec_extract
10879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10880 Alan Hayward <alan.hayward@arm.com>
10881 David Sherwood <david.sherwood@arm.com>
10883 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10884 they are variable or constant sized.
10885 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10886 slots for constant-sized data.
10888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10889 Alan Hayward <alan.hayward@arm.com>
10890 David Sherwood <david.sherwood@arm.com>
10892 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10893 handling COND_EXPRs with boolean comparisons, try to find a better
10894 basis for the mask type than the boolean itself.
10896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10898 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10899 is calculated and how it can be overridden.
10900 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10901 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10903 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10907 Alan Hayward <alan.hayward@arm.com>
10908 David Sherwood <david.sherwood@arm.com>
10910 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10911 Remove the mode argument.
10912 (aarch64_simd_valid_immediate): Remove the mode and inverse
10914 * config/aarch64/iterators.md (bitsize): New iterator.
10915 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10916 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10917 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10918 aarch64_simd_valid_immediate.
10919 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10920 (aarch64_reg_or_bic_imm): Likewise.
10921 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10922 with an insn_type enum and msl with a modifier_type enum.
10923 Replace element_width with a scalar_mode. Change the shift
10924 to unsigned int. Add constructors for scalar_float_mode and
10925 scalar_int_mode elements.
10926 (aarch64_vect_float_const_representable_p): Delete.
10927 (aarch64_can_const_movi_rtx_p)
10928 (aarch64_simd_scalar_immediate_valid_for_move)
10929 (aarch64_simd_make_constant): Update call to
10930 aarch64_simd_valid_immediate.
10931 (aarch64_advsimd_valid_immediate_hs): New function.
10932 (aarch64_advsimd_valid_immediate): Likewise.
10933 (aarch64_simd_valid_immediate): Remove mode and inverse
10934 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10935 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10936 and aarch64_float_const_representable_p on the result.
10937 (aarch64_output_simd_mov_immediate): Remove mode argument.
10938 Update call to aarch64_simd_valid_immediate and use of
10939 simd_immediate_info.
10940 (aarch64_output_scalar_simd_mov_immediate): Update call
10943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10944 Alan Hayward <alan.hayward@arm.com>
10945 David Sherwood <david.sherwood@arm.com>
10947 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10948 (mode_nunits): Likewise CONST_MODE_NUNITS.
10949 * machmode.def (ADJUST_NUNITS): Document.
10950 * genmodes.c (mode_data::need_nunits_adj): New field.
10951 (blank_mode): Update accordingly.
10952 (adj_nunits): New variable.
10953 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10955 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10956 listed in adj_nunits.
10957 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10958 listed in adj_nunits. Don't emit case statements for such modes.
10959 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10960 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10961 nothing if adj_nunits is nonnull.
10962 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10963 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10964 (emit_mode_fbit): Update use of print_maybe_const_decl.
10965 (emit_move_size): Likewise. Treat the array as non-const
10967 (emit_mode_adjustments): Handle adj_nunits.
10969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10971 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10972 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10973 (VECTOR_MODES): Use it.
10974 (make_vector_modes): Take the prefix as an argument.
10976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10977 Alan Hayward <alan.hayward@arm.com>
10978 David Sherwood <david.sherwood@arm.com>
10980 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10981 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10982 for MODE_VECTOR_BOOL.
10983 * machmode.def (VECTOR_BOOL_MODE): Document.
10984 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10985 (make_vector_bool_mode): New function.
10986 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10988 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10989 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10991 * stor-layout.c (int_mode_for_mode): Likewise.
10992 * tree.c (build_vector_type_for_mode): Likewise.
10993 * varasm.c (output_constant_pool_2): Likewise.
10994 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10995 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10996 for MODE_VECTOR_BOOL.
10997 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10998 of mode class checks.
10999 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11000 instead of a list of mode class checks.
11001 (expand_vector_scalar_condition): Likewise.
11002 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11005 Alan Hayward <alan.hayward@arm.com>
11006 David Sherwood <david.sherwood@arm.com>
11008 * machmode.h (mode_size): Change from unsigned short to
11010 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11011 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11012 or if measurement_type is not polynomial.
11013 (fixed_size_mode::includes_p): Check for constant-sized modes.
11014 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11015 return a poly_uint16 rather than an unsigned short.
11016 (emit_mode_size): Change the type of mode_size from unsigned short
11017 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11018 (emit_mode_adjustments): Cope with polynomial vector sizes.
11019 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11021 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11023 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11024 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11025 * caller-save.c (setup_save_areas): Likewise.
11026 (replace_reg_with_saved_mem): Likewise.
11027 * calls.c (emit_library_call_value_1): Likewise.
11028 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11029 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11030 (gen_lowpart_for_combine): Likewise.
11031 * convert.c (convert_to_integer_1): Likewise.
11032 * cse.c (equiv_constant, cse_insn): Likewise.
11033 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11034 (cselib_subst_to_values): Likewise.
11035 * dce.c (word_dce_process_block): Likewise.
11036 * df-problems.c (df_word_lr_mark_ref): Likewise.
11037 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11038 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11039 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11040 (rtl_for_decl_location): Likewise.
11041 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11042 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11043 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11044 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11045 (expand_expr_real_1): Likewise.
11046 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11047 (pad_below): Likewise.
11048 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11049 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11050 * ira.c (get_subreg_tracking_sizes): Likewise.
11051 * ira-build.c (ira_create_allocno_objects): Likewise.
11052 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11053 (ira_sort_regnos_for_alter_reg): Likewise.
11054 * ira-costs.c (record_operand_costs): Likewise.
11055 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11056 (resolve_simple_move): Likewise.
11057 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11058 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11059 (lra_constraints): Likewise.
11060 (CONST_POOL_OK_P): Reject variable-sized modes.
11061 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11062 (add_pseudo_to_slot, lra_spill): Likewise.
11063 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11064 * optabs-query.c (get_best_extraction_insn): Likewise.
11065 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11066 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11067 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11068 * recog.c (offsettable_address_addr_space_p): Likewise.
11069 * regcprop.c (maybe_mode_change): Likewise.
11070 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11071 * regrename.c (build_def_use): Likewise.
11072 * regstat.c (dump_reg_info): Likewise.
11073 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11074 (find_reloads, find_reloads_subreg_address): Likewise.
11075 * reload1.c (eliminate_regs_1): Likewise.
11076 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11077 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11078 (simplify_binary_operation_1, simplify_subreg): Likewise.
11079 * targhooks.c (default_function_arg_padding): Likewise.
11080 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11081 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11082 (verify_gimple_assign_ternary): Likewise.
11083 * tree-inline.c (estimate_move_cost): Likewise.
11084 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11085 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11086 (get_address_cost_ainc): Likewise.
11087 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11088 (vect_supportable_dr_alignment): Likewise.
11089 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11090 (vectorizable_reduction): Likewise.
11091 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11092 (vectorizable_operation, vectorizable_load): Likewise.
11093 * tree.c (build_same_sized_truth_vector_type): Likewise.
11094 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11095 * var-tracking.c (emit_note_insn_var_location): Likewise.
11096 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11097 (ADDR_VEC_ALIGN): Likewise.
11099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11100 Alan Hayward <alan.hayward@arm.com>
11101 David Sherwood <david.sherwood@arm.com>
11103 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11105 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11106 or if measurement_type is polynomial.
11107 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11108 * combine.c (make_extraction): Likewise.
11109 * dse.c (find_shift_sequence): Likewise.
11110 * dwarf2out.c (mem_loc_descriptor): Likewise.
11111 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11112 (extract_bit_field, extract_low_bits): Likewise.
11113 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11114 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11115 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11116 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11117 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11118 * reload.c (find_reloads): Likewise.
11119 * reload1.c (alter_reg): Likewise.
11120 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11121 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11122 * tree-if-conv.c (predicate_mem_writes): Likewise.
11123 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11124 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11125 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11126 * valtrack.c (dead_debug_insert_temp): Likewise.
11127 * varasm.c (mergeable_constant_section): Likewise.
11128 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11131 Alan Hayward <alan.hayward@arm.com>
11132 David Sherwood <david.sherwood@arm.com>
11134 * expr.c (expand_assignment): Cope with polynomial mode sizes
11135 when assigning to a CONCAT.
11137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11138 Alan Hayward <alan.hayward@arm.com>
11139 David Sherwood <david.sherwood@arm.com>
11141 * machmode.h (mode_precision): Change from unsigned short to
11143 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11145 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11146 or if measurement_type is not polynomial.
11147 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11148 in which the mode is already known to be a scalar_int_mode.
11149 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11150 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11152 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11153 for GET_MODE_PRECISION.
11154 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11155 for GET_MODE_PRECISION.
11156 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11158 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11159 (expand_field_assignment, make_extraction): Likewise.
11160 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11161 (get_last_value): Likewise.
11162 * convert.c (convert_to_integer_1): Likewise.
11163 * cse.c (cse_insn): Likewise.
11164 * expr.c (expand_expr_real_1): Likewise.
11165 * lra-constraints.c (simplify_operand_subreg): Likewise.
11166 * optabs-query.c (can_atomic_load_p): Likewise.
11167 * optabs.c (expand_atomic_load): Likewise.
11168 (expand_atomic_store): Likewise.
11169 * ree.c (combine_reaching_defs): Likewise.
11170 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11171 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11172 * tree.h (type_has_mode_precision_p): Likewise.
11173 * ubsan.c (instrument_si_overflow): Likewise.
11175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11176 Alan Hayward <alan.hayward@arm.com>
11177 David Sherwood <david.sherwood@arm.com>
11179 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11180 polynomial numbers of units.
11181 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11182 (valid_vector_subparts_p): New function.
11183 (build_vector_type): Remove temporary shim and take the number
11184 of units as a poly_uint64 rather than an int.
11185 (build_opaque_vector_type): Take the number of units as a
11186 poly_uint64 rather than an int.
11187 * tree.c (build_vector_from_ctor): Handle polynomial
11188 TYPE_VECTOR_SUBPARTS.
11189 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11190 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11191 (build_vector_from_val): If the number of units is variable,
11192 use build_vec_duplicate_cst for constant operands and
11193 VEC_DUPLICATE_EXPR otherwise.
11194 (make_vector_type): Remove temporary is_constant ().
11195 (build_vector_type, build_opaque_vector_type): Take the number of
11196 units as a poly_uint64 rather than an int.
11197 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11199 * cfgexpand.c (expand_debug_expr): Likewise.
11200 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11201 (store_constructor, expand_expr_real_1): Likewise.
11202 (const_scalar_mask_from_tree): Likewise.
11203 * fold-const-call.c (fold_const_reduction): Likewise.
11204 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11205 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11206 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11207 (fold_relational_const): Likewise.
11208 (native_interpret_vector): Likewise. Change the size from an
11209 int to an unsigned int.
11210 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11211 TYPE_VECTOR_SUBPARTS.
11212 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11213 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11214 duplicating a non-constant operand into a variable-length vector.
11215 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11216 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11217 * ipa-icf.c (sem_variable::equals): Likewise.
11218 * match.pd: Likewise.
11219 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11220 * print-tree.c (print_node): Likewise.
11221 * stor-layout.c (layout_type): Likewise.
11222 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11223 * tree-cfg.c (verify_gimple_comparison): Likewise.
11224 (verify_gimple_assign_binary): Likewise.
11225 (verify_gimple_assign_ternary): Likewise.
11226 (verify_gimple_assign_single): Likewise.
11227 * tree-pretty-print.c (dump_generic_node): Likewise.
11228 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11229 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11230 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11231 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11232 (vect_shift_permute_load_chain): Likewise.
11233 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11234 (expand_vector_condition, optimize_vector_constructor): Likewise.
11235 (lower_vec_perm, get_compute_type): Likewise.
11236 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11237 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11238 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11239 (vect_recog_mask_conversion_pattern): Likewise.
11240 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11241 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11242 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11243 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11244 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11245 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11246 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11247 (supportable_widening_operation): Likewise.
11248 (supportable_narrowing_operation): Likewise.
11249 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11251 * varasm.c (output_constant): Likewise.
11253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11254 Alan Hayward <alan.hayward@arm.com>
11255 David Sherwood <david.sherwood@arm.com>
11257 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11258 so that both the length == 3 and length != 3 cases set up their
11259 own permute vectors. Add comments explaining why we know the
11260 number of elements is constant.
11261 (vect_permute_load_chain): Likewise.
11263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11264 Alan Hayward <alan.hayward@arm.com>
11265 David Sherwood <david.sherwood@arm.com>
11267 * machmode.h (mode_nunits): Change from unsigned char to
11269 (ONLY_FIXED_SIZE_MODES): New macro.
11270 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11271 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11272 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11274 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11275 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11276 or if measurement_type is not polynomial.
11277 * genmodes.c (ZERO_COEFFS): New macro.
11278 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11280 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11281 Use ZERO_COEFFS when emitting initializers.
11282 * data-streamer.h (bp_pack_poly_value): New function.
11283 (bp_unpack_poly_value): Likewise.
11284 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11285 for GET_MODE_NUNITS.
11286 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11287 for GET_MODE_NUNITS.
11288 * tree.c (make_vector_type): Remove temporary shim and make
11289 the real function take the number of units as a poly_uint64
11290 rather than an int.
11291 (build_vector_type_for_mode): Handle polynomial nunits.
11292 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11293 * emit-rtl.c (const_vec_series_p_1): Likewise.
11294 (gen_rtx_CONST_VECTOR): Likewise.
11295 * fold-const.c (test_vec_duplicate_folding): Likewise.
11296 * genrecog.c (validate_pattern): Likewise.
11297 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11298 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11299 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11300 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11301 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11302 * rtlanal.c (subreg_get_info): Likewise.
11303 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11304 (vect_grouped_load_supported): Likewise.
11305 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11306 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11307 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11308 (simplify_const_unary_operation, simplify_binary_operation_1)
11309 (simplify_const_binary_operation, simplify_ternary_operation)
11310 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11311 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11312 instead of CONST_VECTOR_NUNITS.
11313 * varasm.c (output_constant_pool_2): Likewise.
11314 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11315 explicit-encoded elements in the XVEC for variable-length vectors.
11317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11319 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11322 Alan Hayward <alan.hayward@arm.com>
11323 David Sherwood <david.sherwood@arm.com>
11325 * coretypes.h (fixed_size_mode): Declare.
11326 (fixed_size_mode_pod): New typedef.
11327 * builtins.h (target_builtins::x_apply_args_mode)
11328 (target_builtins::x_apply_result_mode): Change type to
11329 fixed_size_mode_pod.
11330 * builtins.c (apply_args_size, apply_result_size, result_vector)
11331 (expand_builtin_apply_args_1, expand_builtin_apply)
11332 (expand_builtin_return): Update accordingly.
11334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11336 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11337 * cselib.c (cselib_hash_rtx): Likewise.
11338 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11339 CONST_VECTOR encoding.
11341 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11342 Jeff Law <law@redhat.com>
11345 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11346 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11347 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11348 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11351 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11352 explicitly probe *sp in a noreturn function if there were any callee
11353 register saves or frame pointer is needed.
11355 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11358 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11359 BLKmode for ternary, binary or unary expressions.
11362 * var-tracking.c (delete_vta_debug_insn): New inline function.
11363 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11364 insns from get_insns () to NULL instead of each bb separately.
11365 Use delete_vta_debug_insn. No longer static.
11366 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11367 delete_vta_debug_insns callers.
11368 * rtl.h (delete_vta_debug_insns): Declare.
11369 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11370 instead of variable_tracking_main.
11372 2018-01-03 Martin Sebor <msebor@redhat.com>
11374 PR tree-optimization/83603
11375 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11376 arguments past the endof the argument list in functions declared
11377 without a prototype.
11378 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11379 Avoid checking when arguments are null.
11381 2018-01-03 Martin Sebor <msebor@redhat.com>
11384 * doc/extend.texi (attribute const): Fix a typo.
11385 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11386 issuing -Wsuggest-attribute for void functions.
11388 2018-01-03 Martin Sebor <msebor@redhat.com>
11390 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11391 offset_int::from instead of wide_int::to_shwi.
11392 (maybe_diag_overlap): Remove assertion.
11393 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11394 * gimple-ssa-sprintf.c (format_directive): Same.
11395 (parse_directive): Same.
11396 (sprintf_dom_walker::compute_format_length): Same.
11397 (try_substitute_return_value): Same.
11399 2018-01-03 Jeff Law <law@redhat.com>
11401 PR middle-end/83654
11402 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11403 non-constant residual for zero at runtime and avoid probing in
11404 that case. Reorganize code for trailing problem to mirror handling
11407 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11409 PR tree-optimization/83501
11410 * tree-ssa-strlen.c (get_string_cst): New.
11411 (handle_char_store): Call get_string_cst.
11413 2018-01-03 Martin Liska <mliska@suse.cz>
11415 PR tree-optimization/83593
11416 * tree-ssa-strlen.c: Include tree-cfg.h.
11417 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11418 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11419 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11421 (strlen_dom_walker::before_dom_children): Call
11422 gimple_purge_dead_eh_edges. Dump tranformation with details
11424 (strlen_dom_walker::before_dom_children): Update call by adding
11425 new argument cleanup_eh.
11426 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11428 2018-01-03 Martin Liska <mliska@suse.cz>
11431 * cif-code.def (VARIADIC_THUNK): New enum value.
11432 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11435 2018-01-03 Jan Beulich <jbeulich@suse.com>
11437 * sse.md (mov<mode>_internal): Tighten condition for when to use
11438 vmovdqu<ssescalarsize> for TI and OI modes.
11440 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11442 Update copyright years.
11444 2018-01-03 Martin Liska <mliska@suse.cz>
11447 * ipa-visibility.c (function_and_variable_visibility): Skip
11448 functions with noipa attribure.
11450 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11452 * gcc.c (process_command): Update copyright notice dates.
11453 * gcov-dump.c (print_version): Ditto.
11454 * gcov.c (print_version): Ditto.
11455 * gcov-tool.c (print_version): Ditto.
11456 * gengtype.c (create_file): Ditto.
11457 * doc/cpp.texi: Bump @copying's copyright year.
11458 * doc/cppinternals.texi: Ditto.
11459 * doc/gcc.texi: Ditto.
11460 * doc/gccint.texi: Ditto.
11461 * doc/gcov.texi: Ditto.
11462 * doc/install.texi: Ditto.
11463 * doc/invoke.texi: Ditto.
11465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11467 * vector-builder.h (vector_builder::m_full_nelts): Change from
11468 unsigned int to poly_uint64.
11469 (vector_builder::full_nelts): Update prototype accordingly.
11470 (vector_builder::new_vector): Likewise.
11471 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11472 (vector_builder::operator ==): Likewise.
11473 (vector_builder::finalize): Likewise.
11474 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11475 Take the number of elements as a poly_uint64 rather than an
11477 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11478 from unsigned int to poly_uint64.
11479 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11480 (vec_perm_indices::new_vector): Likewise.
11481 (vec_perm_indices::length): Likewise.
11482 (vec_perm_indices::nelts_per_input): Likewise.
11483 (vec_perm_indices::input_nelts): Likewise.
11484 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11485 number of elements per input as a poly_uint64 rather than an
11486 unsigned int. Use the original encoding for variable-length
11487 vectors, rather than clamping each individual element.
11488 For the second and subsequent elements in each pattern,
11489 clamp the step and base before clamping their sum.
11490 (vec_perm_indices::series_p): Handle polynomial element counts.
11491 (vec_perm_indices::all_in_range_p): Likewise.
11492 (vec_perm_indices_to_tree): Likewise.
11493 (vec_perm_indices_to_rtx): Likewise.
11494 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11495 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11496 (tree_vector_builder::new_binary_operation): Handle polynomial
11497 element counts. Return false if we need to know the number
11498 of elements at compile time.
11499 * fold-const.c (fold_vec_perm): Punt if the number of elements
11500 isn't known at compile time.
11502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11504 * vec-perm-indices.h (vec_perm_builder): Change element type
11505 from HOST_WIDE_INT to poly_int64.
11506 (vec_perm_indices::element_type): Update accordingly.
11507 (vec_perm_indices::clamp): Handle polynomial element_types.
11508 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11509 (vec_perm_indices::all_in_range_p): Likewise.
11510 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11512 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11513 polynomial vec_perm_indices element types.
11514 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11515 * fold-const.c (fold_vec_perm): Likewise.
11516 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11517 * tree-vect-generic.c (lower_vec_perm): Likewise.
11518 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11519 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11520 element type to HOST_WIDE_INT.
11522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11523 Alan Hayward <alan.hayward@arm.com>
11524 David Sherwood <david.sherwood@arm.com>
11526 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11527 rather than an int. Use plus_constant.
11528 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11529 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11532 Alan Hayward <alan.hayward@arm.com>
11533 David Sherwood <david.sherwood@arm.com>
11535 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11536 a HOST_WIDE_INT to a poly_int64.
11538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11539 Alan Hayward <alan.hayward@arm.com>
11540 David Sherwood <david.sherwood@arm.com>
11542 * calls.c (load_register_parameters): Cope with polynomial
11543 mode sizes. Require a constant size for BLKmode parameters
11544 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11545 forces a parameter to be padded at the lsb end in order to
11546 fill a complete number of words, require the parameter size
11547 to be ordered wrt UNITS_PER_WORD.
11549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11550 Alan Hayward <alan.hayward@arm.com>
11551 David Sherwood <david.sherwood@arm.com>
11553 * reload1.c (spill_stack_slot_width): Change element type
11554 from unsigned int to poly_uint64_pod.
11555 (alter_reg): Treat mode sizes as polynomial.
11557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11558 Alan Hayward <alan.hayward@arm.com>
11559 David Sherwood <david.sherwood@arm.com>
11561 * reload.c (complex_word_subreg_p): New function.
11562 (reload_inner_reg_of_subreg, push_reload): Use it.
11564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11565 Alan Hayward <alan.hayward@arm.com>
11566 David Sherwood <david.sherwood@arm.com>
11568 * lra-constraints.c (process_alt_operands): Reject matched
11569 operands whose sizes aren't ordered.
11570 (match_reload): Refer to this check here.
11572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11573 Alan Hayward <alan.hayward@arm.com>
11574 David Sherwood <david.sherwood@arm.com>
11576 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11577 that the mode size is in the set {1, 2, 4, 8, 16}.
11579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11580 Alan Hayward <alan.hayward@arm.com>
11581 David Sherwood <david.sherwood@arm.com>
11583 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11584 Use plus_constant instead of gen_rtx_PLUS.
11586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11587 Alan Hayward <alan.hayward@arm.com>
11588 David Sherwood <david.sherwood@arm.com>
11590 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11591 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11592 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11593 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11594 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11595 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11596 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11597 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11598 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11599 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11601 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11602 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11603 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11604 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11605 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11606 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11607 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11608 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11609 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11610 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11612 * expr.c (emit_move_resolve_push): Treat the input and result
11613 of PUSH_ROUNDING as a poly_int64.
11614 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11615 (emit_push_insn): Likewise.
11616 * lra-eliminations.c (mark_not_eliminable): Likewise.
11617 * recog.c (push_operand): Likewise.
11618 * reload1.c (elimination_effects): Likewise.
11619 * rtlanal.c (nonzero_bits1): Likewise.
11620 * calls.c (store_one_arg): Likewise. Require the padding to be
11621 known at compile time.
11623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11624 Alan Hayward <alan.hayward@arm.com>
11625 David Sherwood <david.sherwood@arm.com>
11627 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11628 Use plus_constant instead of gen_rtx_PLUS.
11630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11631 Alan Hayward <alan.hayward@arm.com>
11632 David Sherwood <david.sherwood@arm.com>
11634 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11635 rather than an int.
11637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11638 Alan Hayward <alan.hayward@arm.com>
11639 David Sherwood <david.sherwood@arm.com>
11641 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11642 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11643 via stack temporaries. Treat the mode size as polynomial too.
11645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11646 Alan Hayward <alan.hayward@arm.com>
11647 David Sherwood <david.sherwood@arm.com>
11649 * expr.c (expand_expr_real_2): When handling conversions involving
11650 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11651 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11652 as a poly_uint64 too.
11654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11655 Alan Hayward <alan.hayward@arm.com>
11656 David Sherwood <david.sherwood@arm.com>
11658 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11661 Alan Hayward <alan.hayward@arm.com>
11662 David Sherwood <david.sherwood@arm.com>
11664 * combine.c (can_change_dest_mode): Handle polynomial
11665 REGMODE_NATURAL_SIZE.
11666 * expmed.c (store_bit_field_1): Likewise.
11667 * expr.c (store_constructor): Likewise.
11668 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11669 and polynomial REGMODE_NATURAL_SIZE.
11670 (gen_lowpart_common): Likewise.
11671 * reginfo.c (record_subregs_of_mode): Likewise.
11672 * rtlanal.c (read_modify_subreg_p): Likewise.
11674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11675 Alan Hayward <alan.hayward@arm.com>
11676 David Sherwood <david.sherwood@arm.com>
11678 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11679 numbers of elements.
11681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11682 Alan Hayward <alan.hayward@arm.com>
11683 David Sherwood <david.sherwood@arm.com>
11685 * match.pd: Cope with polynomial numbers of vector elements.
11687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11688 Alan Hayward <alan.hayward@arm.com>
11689 David Sherwood <david.sherwood@arm.com>
11691 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11692 in a POINTER_PLUS_EXPR.
11694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11695 Alan Hayward <alan.hayward@arm.com>
11696 David Sherwood <david.sherwood@arm.com>
11698 * omp-simd-clone.c (simd_clone_subparts): New function.
11699 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11700 (ipa_simd_modify_function_body): Likewise.
11702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11703 Alan Hayward <alan.hayward@arm.com>
11704 David Sherwood <david.sherwood@arm.com>
11706 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11707 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11708 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11709 (expand_vector_condition, vector_element): Likewise.
11710 (subparts_gt): New function.
11711 (get_compute_type): Use subparts_gt.
11712 (count_type_subparts): Delete.
11713 (expand_vector_operations_1): Use subparts_gt instead of
11714 count_type_subparts.
11716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11717 Alan Hayward <alan.hayward@arm.com>
11718 David Sherwood <david.sherwood@arm.com>
11720 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11721 (vect_compile_time_alias): ...this new function. Do the calculation
11722 on poly_ints rather than trees.
11723 (vect_prune_runtime_alias_test_list): Update call accordingly.
11725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11726 Alan Hayward <alan.hayward@arm.com>
11727 David Sherwood <david.sherwood@arm.com>
11729 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11731 (vect_schedule_slp_instance): Likewise.
11733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11734 Alan Hayward <alan.hayward@arm.com>
11735 David Sherwood <david.sherwood@arm.com>
11737 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11738 constant and extern definitions for variable-length vectors.
11739 (vect_get_constant_vectors): Note that the number of units
11740 is known to be constant.
11742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11743 Alan Hayward <alan.hayward@arm.com>
11744 David Sherwood <david.sherwood@arm.com>
11746 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11747 of units as polynomial. Choose between WIDE and NARROW based
11750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11751 Alan Hayward <alan.hayward@arm.com>
11752 David Sherwood <david.sherwood@arm.com>
11754 * tree-vect-stmts.c (simd_clone_subparts): New function.
11755 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11758 Alan Hayward <alan.hayward@arm.com>
11759 David Sherwood <david.sherwood@arm.com>
11761 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11762 vectors as polynomial. Use build_index_vector for
11763 IFN_GOMP_SIMD_LANE.
11765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11766 Alan Hayward <alan.hayward@arm.com>
11767 David Sherwood <david.sherwood@arm.com>
11769 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11770 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11771 for variable-length vectors.
11772 (vectorizable_mask_load_store): Treat the number of units as
11773 polynomial, asserting that it is constant if the condition has
11774 already been enforced.
11775 (vectorizable_store, vectorizable_load): Likewise.
11777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11778 Alan Hayward <alan.hayward@arm.com>
11779 David Sherwood <david.sherwood@arm.com>
11781 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11782 of units as polynomial. Punt if we can't tell at compile time
11783 which vector contains the final result.
11785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11786 Alan Hayward <alan.hayward@arm.com>
11787 David Sherwood <david.sherwood@arm.com>
11789 * tree-vect-loop.c (vectorizable_induction): Treat the number
11790 of units as polynomial. Punt on SLP inductions. Use an integer
11791 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11792 cast of such a series for variable-length floating-point
11795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11796 Alan Hayward <alan.hayward@arm.com>
11797 David Sherwood <david.sherwood@arm.com>
11799 * tree.h (build_index_vector): Declare.
11800 * tree.c (build_index_vector): New function.
11801 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11802 of units as polynomial, forcibly converting it to a constant if
11803 vectorizable_reduction has already enforced the condition.
11804 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11805 to create a {1,2,3,...} vector.
11806 (vectorizable_reduction): Treat the number of units as polynomial.
11807 Choose vectype_in based on the largest scalar element size rather
11808 than the smallest number of units. Enforce the restrictions
11811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11812 Alan Hayward <alan.hayward@arm.com>
11813 David Sherwood <david.sherwood@arm.com>
11815 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11816 number of units as polynomial.
11818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11819 Alan Hayward <alan.hayward@arm.com>
11820 David Sherwood <david.sherwood@arm.com>
11822 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11823 * target.def (autovectorize_vector_sizes): Return the vector sizes
11824 by pointer, using vector_sizes rather than a bitmask.
11825 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11826 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11827 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11829 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11830 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11831 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11832 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11833 * omp-general.c (omp_max_vf): Likewise.
11834 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11835 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11836 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11837 * tree-vect-slp.c (vect_slp_bb): Likewise.
11838 * doc/tm.texi: Regenerate.
11839 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11841 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11842 the vector size as a poly_uint64 rather than an unsigned int.
11843 (current_vector_size): Change from an unsigned int to a poly_uint64.
11844 (get_vectype_for_scalar_type): Update accordingly.
11845 * tree.h (build_truth_vector_type): Take the size and number of
11846 units as a poly_uint64 rather than an unsigned int.
11847 (build_vector_type): Add a temporary overload that takes
11848 the number of units as a poly_uint64 rather than an unsigned int.
11849 * tree.c (make_vector_type): Likewise.
11850 (build_truth_vector_type): Take the number of units as a poly_uint64
11851 rather than an unsigned int.
11853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11854 Alan Hayward <alan.hayward@arm.com>
11855 David Sherwood <david.sherwood@arm.com>
11857 * target.def (get_mask_mode): Take the number of units and length
11858 as poly_uint64s rather than unsigned ints.
11859 * targhooks.h (default_get_mask_mode): Update accordingly.
11860 * targhooks.c (default_get_mask_mode): Likewise.
11861 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11862 * doc/tm.texi: Regenerate.
11864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11865 Alan Hayward <alan.hayward@arm.com>
11866 David Sherwood <david.sherwood@arm.com>
11868 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11869 * omp-general.c (omp_max_vf): Likewise.
11870 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11871 (expand_omp_simd): Handle polynomial safelen.
11872 * omp-low.c (omplow_simd_context): Add a default constructor.
11873 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11874 (lower_rec_simd_input_clauses): Update accordingly.
11875 (lower_rec_input_clauses): Likewise.
11877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11878 Alan Hayward <alan.hayward@arm.com>
11879 David Sherwood <david.sherwood@arm.com>
11881 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11882 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11883 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11884 (vect_analyze_slp_cost): Likewise.
11885 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11886 (vect_model_load_cost): Likewise.
11888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11889 Alan Hayward <alan.hayward@arm.com>
11890 David Sherwood <david.sherwood@arm.com>
11892 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11893 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11894 from an unsigned int * to a poly_uint64_pod *.
11895 (calculate_unrolling_factor): New function.
11896 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11899 Alan Hayward <alan.hayward@arm.com>
11900 David Sherwood <david.sherwood@arm.com>
11902 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11903 from an unsigned int to a poly_uint64.
11904 (_loop_vec_info::slp_unrolling_factor): Likewise.
11905 (_loop_vec_info::vectorization_factor): Change from an int
11907 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11908 (vect_get_num_vectors): New function.
11909 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11910 (vect_get_num_copies): Use vect_get_num_vectors.
11911 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11912 to an unsigned int *.
11913 (vect_analyze_data_refs): Change min_vf from an int * to a
11915 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11916 than an unsigned HOST_WIDE_INT.
11917 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11918 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11919 to an unsigned int *.
11920 (vect_analyze_data_ref_dependences): Likewise.
11921 (vect_compute_data_ref_alignment): Handle polynomial vf.
11922 (vect_enhance_data_refs_alignment): Likewise.
11923 (vect_prune_runtime_alias_test_list): Likewise.
11924 (vect_shift_permute_load_chain): Likewise.
11925 (vect_supportable_dr_alignment): Likewise.
11926 (dependence_distance_ge_vf): Take the vectorization factor as a
11927 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11928 (vect_analyze_data_refs): Change min_vf from an int * to a
11930 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11931 vfm1 as a poly_uint64 rather than an int. Make the same change
11932 for the returned bound_scalar.
11933 (vect_gen_vector_loop_niters): Handle polynomial vf.
11934 (vect_do_peeling): Likewise. Update call to
11935 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11936 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11938 * tree-vect-loop.c (vect_determine_vectorization_factor)
11939 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11940 (vect_get_known_peeling_cost): Likewise.
11941 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11942 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11943 (vect_transform_loop): Likewise. Use the lowest possible VF when
11944 updating the upper bounds of the loop.
11945 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11946 rather than an int.
11947 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11948 polynomial unroll factors.
11949 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11950 (vect_make_slp_decision): Likewise.
11951 (vect_supported_load_permutation_p): Likewise, and polynomial
11953 (vect_analyze_slp_cost): Handle polynomial vf.
11954 (vect_slp_analyze_node_operations): Likewise.
11955 (vect_slp_analyze_bb_1): Likewise.
11956 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11957 than an unsigned HOST_WIDE_INT.
11958 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11959 (vectorizable_load): Handle polynomial vf.
11960 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11962 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11965 Alan Hayward <alan.hayward@arm.com>
11966 David Sherwood <david.sherwood@arm.com>
11968 * match.pd: Handle bit operations involving three constants
11969 and try to fold one pair.
11971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11973 * tree-vect-loop-manip.c: Include gimple-fold.h.
11974 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11975 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11976 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11977 Add a path that uses a step of VF instead of 1, but disable it
11979 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11980 and niters_no_overflow parameters. Update calls to
11981 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11982 Create a new SSA name if the latter choses to use a ste other
11983 than zero, and return it via niters_vector_mult_vf_var.
11984 * tree-vect-loop.c (vect_transform_loop): Update calls to
11985 vect_do_peeling, vect_gen_vector_loop_niters and
11986 slpeel_make_loop_iterate_ntimes.
11987 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11988 (vect_gen_vector_loop_niters): Update declarations after above changes.
11990 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11992 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11993 128-bit round to integer instructions.
11994 (ceil<mode>2): Likewise.
11995 (btrunc<mode>2): Likewise.
11996 (round<mode>2): Likewise.
11998 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12000 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12001 unaligned VSX load/store on P8/P9.
12002 (expand_block_clear): Allow the use of unaligned VSX
12003 load/store on P8/P9.
12005 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12007 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12009 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12010 swap associated with both a load and a store.
12012 2018-01-02 Andrew Waterman <andrew@sifive.com>
12014 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12015 * config/riscv/riscv.md (clear_cache): Use it.
12017 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12019 * web.c: Remove out-of-date comment.
12021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12023 * expr.c (fixup_args_size_notes): Check that any existing
12024 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12025 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12026 (emit_single_push_insn): ...here.
12028 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12030 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12031 (const_vector_encoded_nelts): New function.
12032 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12033 (const_vector_int_elt, const_vector_elt): Declare.
12034 * emit-rtl.c (const_vector_int_elt_1): New function.
12035 (const_vector_elt): Likewise.
12036 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12037 of CONST_VECTOR_ELT.
12039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12041 * expr.c: Include rtx-vector-builder.h.
12042 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12043 directly on the tree encoding.
12044 (const_vector_from_tree): Likewise.
12045 * optabs.c: Include rtx-vector-builder.h.
12046 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12047 sequence of "u" values.
12048 * vec-perm-indices.c: Include rtx-vector-builder.h.
12049 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12050 directly on the vec_perm_indices encoding.
12052 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12054 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12055 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12056 * rtx-vector-builder.h: New file.
12057 * rtx-vector-builder.c: Likewise.
12058 * rtl.h (rtx_def::u2): Add a const_vector field.
12059 (CONST_VECTOR_NPATTERNS): New macro.
12060 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12061 (CONST_VECTOR_DUPLICATE_P): Likewise.
12062 (CONST_VECTOR_STEPPED_P): Likewise.
12063 (CONST_VECTOR_ENCODED_ELT): Likewise.
12064 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12065 (unwrap_const_vec_duplicate): Likewise.
12066 (const_vec_series_p): Check for a non-duplicated vector encoding.
12067 Say that the function only returns true for integer vectors.
12068 * emit-rtl.c: Include rtx-vector-builder.h.
12069 (gen_const_vec_duplicate_1): Delete.
12070 (gen_const_vector): Call gen_const_vec_duplicate instead of
12071 gen_const_vec_duplicate_1.
12072 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12073 (gen_const_vec_duplicate): Use rtx_vector_builder.
12074 (gen_const_vec_series): Likewise.
12075 (gen_rtx_CONST_VECTOR): Likewise.
12076 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12077 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12078 Build a new vector rather than modifying a CONST_VECTOR in-place.
12079 (handle_special_swappables): Update call accordingly.
12080 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12081 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12082 Build a new vector rather than modifying a CONST_VECTOR in-place.
12083 (handle_special_swappables): Update call accordingly.
12085 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12087 * simplify-rtx.c (simplify_const_binary_operation): Use
12088 CONST_VECTOR_ELT instead of XVECEXP.
12090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12092 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12093 the selector elements to be different from the data elements
12094 if the selector is a VECTOR_CST.
12095 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12096 ssizetype for the selector.
12098 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12100 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12101 before testing each element individually.
12102 * tree-vect-generic.c (lower_vec_perm): Likewise.
12104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12106 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12107 * selftest-run-tests.c (selftest::run_tests): Call it.
12108 * vector-builder.h (vector_builder::operator ==): New function.
12109 (vector_builder::operator !=): Likewise.
12110 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12111 (vec_perm_indices::all_from_input_p): New function.
12112 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12113 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12114 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12115 instead of reading the VECTOR_CST directly. Detect whether both
12116 vector inputs are the same before constructing the vec_perm_indices,
12117 and update the number of inputs argument accordingly. Use the
12118 utility functions added above. Only construct sel2 if we need to.
12120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12122 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12123 the broadcast of the low byte.
12124 (expand_mult_highpart): Use an explicit encoding for the permutes.
12125 * optabs-query.c (can_mult_highpart_p): Likewise.
12126 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12127 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12128 (vectorizable_bswap): Likewise.
12129 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12130 explicit encoding for the power-of-2 permutes.
12131 (vect_permute_store_chain): Likewise.
12132 (vect_grouped_load_supported): Likewise.
12133 (vect_permute_load_chain): Likewise.
12135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12137 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12138 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12139 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12140 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12141 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12142 (vect_gen_perm_mask_any): Likewise.
12144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12146 * int-vector-builder.h: New file.
12147 * vec-perm-indices.h: Include int-vector-builder.h.
12148 (vec_perm_indices): Redefine as an int_vector_builder.
12149 (auto_vec_perm_indices): Delete.
12150 (vec_perm_builder): Redefine as a stand-alone class.
12151 (vec_perm_indices::vec_perm_indices): New function.
12152 (vec_perm_indices::clamp): Likewise.
12153 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12154 (vec_perm_indices::new_vector): New function.
12155 (vec_perm_indices::new_expanded_vector): Update for new
12156 vec_perm_indices class.
12157 (vec_perm_indices::rotate_inputs): New function.
12158 (vec_perm_indices::all_in_range_p): Operate directly on the
12159 encoded form, without computing elided elements.
12160 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12161 encoding. Update for new vec_perm_indices class.
12162 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12163 the given vec_perm_builder.
12164 (expand_vec_perm_var): Update vec_perm_builder constructor.
12165 (expand_mult_highpart): Use vec_perm_builder instead of
12166 auto_vec_perm_indices.
12167 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12168 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12169 or double series encoding as appropriate.
12170 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12171 vec_perm_indices instead of auto_vec_perm_indices.
12172 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12173 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12174 (vect_permute_store_chain): Likewise.
12175 (vect_grouped_load_supported): Likewise.
12176 (vect_permute_load_chain): Likewise.
12177 (vect_shift_permute_load_chain): Likewise.
12178 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12179 (vect_transform_slp_perm_load): Likewise.
12180 (vect_schedule_slp_instance): Likewise.
12181 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12182 (vectorizable_mask_load_store): Likewise.
12183 (vectorizable_bswap): Likewise.
12184 (vectorizable_store): Likewise.
12185 (vectorizable_load): Likewise.
12186 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12187 vec_perm_indices instead of auto_vec_perm_indices. Use
12188 tree_to_vec_perm_builder to read the vector from a tree.
12189 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12190 vec_perm_builder instead of a vec_perm_indices.
12191 (have_whole_vector_shift): Use vec_perm_builder and
12192 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12193 truncation to calc_vec_perm_mask_for_shift.
12194 (vect_create_epilog_for_reduction): Likewise.
12195 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12196 from auto_vec_perm_indices to vec_perm_indices.
12197 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12198 instead of changing individual elements.
12199 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12200 the vector in d.perm.
12201 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12202 from auto_vec_perm_indices to vec_perm_indices.
12203 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12204 instead of changing individual elements.
12205 (arm_vectorize_vec_perm_const): Use new_vector to install
12206 the vector in d.perm.
12207 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12208 Update vec_perm_builder constructor.
12209 (rs6000_expand_interleave): Likewise.
12210 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12211 (rs6000_expand_interleave): Likewise.
12213 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12215 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12216 to qimode could truncate the indices.
12217 * optabs.c (expand_vec_perm_var): Likewise.
12219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12221 * Makefile.in (OBJS): Add vec-perm-indices.o.
12222 * vec-perm-indices.h: New file.
12223 * vec-perm-indices.c: Likewise.
12224 * target.h (vec_perm_indices): Replace with a forward class
12226 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12227 * optabs.h: Include vec-perm-indices.h.
12228 (expand_vec_perm): Delete.
12229 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12230 (expand_vec_perm_const): Declare.
12231 * target.def (vec_perm_const_ok): Replace with...
12232 (vec_perm_const): ...this new hook.
12233 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12234 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12235 * doc/tm.texi: Regenerate.
12236 * optabs.def (vec_perm_const): Delete.
12237 * doc/md.texi (vec_perm_const): Likewise.
12238 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12239 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12240 expand_vec_perm for constant permutation vectors. Assert that
12241 the mode of variable permutation vectors is the integer equivalent
12242 of the mode that is being permuted.
12243 * optabs-query.h (selector_fits_mode_p): Declare.
12244 * optabs-query.c: Include vec-perm-indices.h.
12245 (selector_fits_mode_p): New function.
12246 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12247 is defined, instead of checking whether the vec_perm_const_optab
12248 exists. Use targetm.vectorize.vec_perm_const instead of
12249 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12250 fit in the vector mode before using a variable permute.
12251 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12252 vec_perm_indices instead of an rtx.
12253 (expand_vec_perm): Replace with...
12254 (expand_vec_perm_const): ...this new function. Take the selector
12255 as a vec_perm_indices rather than an rtx. Also take the mode of
12256 the selector. Update call to shift_amt_for_vec_perm_mask.
12257 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12258 Use vec_perm_indices::new_expanded_vector to expand the original
12259 selector into bytes. Check whether the indices fit in the vector
12260 mode before using a variable permute.
12261 (expand_vec_perm_var): Make global.
12262 (expand_mult_highpart): Use expand_vec_perm_const.
12263 * fold-const.c: Includes vec-perm-indices.h.
12264 * tree-ssa-forwprop.c: Likewise.
12265 * tree-vect-data-refs.c: Likewise.
12266 * tree-vect-generic.c: Likewise.
12267 * tree-vect-loop.c: Likewise.
12268 * tree-vect-slp.c: Likewise.
12269 * tree-vect-stmts.c: Likewise.
12270 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12272 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12273 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12274 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12275 (aarch64_vectorize_vec_perm_const): ...this new function.
12276 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12277 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12278 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12279 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12280 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12281 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12282 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12284 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12285 check for NEON modes.
12286 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12287 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12288 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12289 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12291 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12292 the old VEC_PERM_CONST conditions.
12293 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12294 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12295 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12296 (ia64_vectorize_vec_perm_const_ok): Merge into...
12297 (ia64_vectorize_vec_perm_const): ...this new function.
12298 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12299 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12300 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12301 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12302 * config/mips/mips.c (mips_expand_vec_perm_const)
12303 (mips_vectorize_vec_perm_const_ok): Merge into...
12304 (mips_vectorize_vec_perm_const): ...this new function.
12305 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12306 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12307 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12308 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12309 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12310 (rs6000_expand_vec_perm_const): Delete.
12311 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12313 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12314 (altivec_expand_vec_perm_const_le): Take each operand individually.
12315 Operate on constant selectors rather than rtxes.
12316 (altivec_expand_vec_perm_const): Likewise. Update call to
12317 altivec_expand_vec_perm_const_le.
12318 (rs6000_expand_vec_perm_const): Delete.
12319 (rs6000_vectorize_vec_perm_const_ok): Delete.
12320 (rs6000_vectorize_vec_perm_const): New function.
12321 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12322 an element count and rtx array.
12323 (rs6000_expand_extract_even): Update call accordingly.
12324 (rs6000_expand_interleave): Likewise.
12325 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12326 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12327 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12328 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12329 (rs6000_expand_vec_perm_const): Delete.
12330 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12331 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12332 (altivec_expand_vec_perm_const_le): Take each operand individually.
12333 Operate on constant selectors rather than rtxes.
12334 (altivec_expand_vec_perm_const): Likewise. Update call to
12335 altivec_expand_vec_perm_const_le.
12336 (rs6000_expand_vec_perm_const): Delete.
12337 (rs6000_vectorize_vec_perm_const_ok): Delete.
12338 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12339 reference to the SPE evmerge intructions.
12340 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12341 an element count and rtx array.
12342 (rs6000_expand_extract_even): Update call accordingly.
12343 (rs6000_expand_interleave): Likewise.
12344 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12345 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12347 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12349 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12351 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12352 vector mode and that that mode matches the mode of the data
12354 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12355 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12356 directly using expand_vec_perm_1 when forcing selectors into
12358 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12362 * optabs-query.h (can_vec_perm_p): Delete.
12363 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12364 * optabs-query.c (can_vec_perm_p): Split into...
12365 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12366 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12367 particular selector is valid.
12368 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12369 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12370 (vect_grouped_load_supported): Likewise.
12371 (vect_shift_permute_load_chain): Likewise.
12372 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12373 (vect_transform_slp_perm_load): Likewise.
12374 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12375 (vectorizable_bswap): Likewise.
12376 (vect_gen_perm_mask_checked): Likewise.
12377 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12378 implementations of variable permutation vectors into account
12379 when deciding which selector to use.
12380 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12381 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12382 with a false third argument.
12383 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12384 to test whether the constant selector is valid and can_vec_perm_var_p
12385 to test whether a variable selector is valid.
12387 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12389 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12390 * optabs-query.c (can_vec_perm_p): Likewise.
12391 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12392 instead of vec_perm_indices.
12393 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12394 (vect_gen_perm_mask_checked): Likewise,
12395 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12396 (vect_gen_perm_mask_checked): Likewise,
12398 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12400 * optabs-query.h (qimode_for_vec_perm): Declare.
12401 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12402 (qimode_for_vec_perm): ...this new function.
12403 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12405 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12407 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12408 does not have a conditional at the top.
12410 2018-01-02 Richard Biener <rguenther@suse.de>
12412 * ipa-inline.c (big_speedup_p): Fix expression.
12414 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12417 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12420 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12424 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12425 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12426 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12427 cond_taken_branch_cost 3->4.
12429 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12431 PR tree-optimization/83581
12432 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12433 TODO_cleanup_cfg if any changes have been made.
12435 PR middle-end/83608
12436 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12437 convert_modes if target mode has the right side, but different mode
12440 PR middle-end/83609
12441 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12442 last argument when extracting from CONCAT. If either from_real or
12443 from_imag is NULL, use expansion through memory. If result is not
12444 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12445 the parts directly to inner mode, if even that fails, use expansion
12448 PR middle-end/83623
12449 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12450 check for bswap in mode rather than HImode and use that in expand_unop
12453 Copyright (C) 2018 Free Software Foundation, Inc.
12455 Copying and distribution of this file, with or without modification,
12456 are permitted in any medium without royalty provided the copyright
12457 notice and this notice are preserved.