1 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4 when --with-gxx-include-dir is also specified.
5 * configure: Regenerate.
7 2018-05-09 Jim Wilson <jimw@sifive.com>
10 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
11 * config/riscv/t-withmultilib: New.
12 * config/riscv/withmultilib.h: New.
13 * doc/install.texi: Document RISC-V --with-multilib-list support.
15 2018-05-09 Richard Biener <rguenther@suse.de>
17 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19 (vect_bb_vectorization_profitable_p): Adjust. Compute
20 actual scalar cost using the cost vector and the add_stmt_cost
23 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
25 PR rtl-optimization/85645
26 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
27 in the REG_CFA_REGISTER note for LR, don't leave it empty.
29 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
31 PR rtl-optimization/85645
32 * shrink-wrap.c (spread_components): Return a boolean saying if
34 (try_shrink_wrapping_separate): Iterate spread_components until
35 nothing changes anymore.
37 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
39 PR rtl-optimization/85645
40 * regrename.c (build_def_use): Also kill the chains that include the
41 destination of a REG_CFA_REGISTER note.
43 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
45 PR rtl-optimization/85645
46 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
47 insn that has a REG_CFA_REGISTER note.
49 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
51 * cfgexpand.c (expand_clobber): New function.
52 (expand_gimple_stmt_1): Use it.
53 * tree-vect-stmts.c (vect_clobber_variable): New function,
55 (vectorizable_simd_clone_call): ...here.
56 (vectorizable_store): Emit a clobber either side of an
57 IFN_STORE_LANES sequence.
58 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
60 2018-05-09 Tom de Vries <tom@codesourcery.com>
63 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
64 (define_insn "trap_if_false"): Add exit after trap.
66 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
68 PR rtl-optimization/85638
69 * bb-reorder.c: Include common/common-target.h.
70 (create_forwarder_block): New function extracted from...
71 (fix_up_crossing_landing_pad): ...here. Rename into...
72 (dw2_fix_up_crossing_landing_pad): ...this.
73 (sjlj_fix_up_crossing_landing_pad): New function.
74 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
75 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
76 from both partitions and exit the loop after one iteration.
78 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
81 * doc/extend.texi (PowerPC Built-in Functions): Rename this
83 (Basic PowerPC Built-in Functions): The new name of the
84 subsection previously known as "PowerPC Built-in Functions".
85 (Basic PowerPC Built-in Functions Available on all Configurations):
87 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
89 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
91 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
93 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
96 2018-05-08 Jim Wilson <jimw@sifive.com>
98 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
99 (LD_EMUL_SUFFIX): New.
102 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
104 * doc/extend.texi (PowerPC Built-in Functions): Rename this
106 (Basic PowerPC Built-in Functions): The new name of the
107 subsection previously known as "PowerPC Built-in Functions".
108 (Basic PowerPC Built-in Functions Available on all Configurations):
110 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
112 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
114 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
116 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
119 2018-05-08 Jakub Jelinek <jakub@redhat.com>
122 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
123 after cmpelim optimization.
125 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
127 * config.gcc: Support "goldmont".
128 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
129 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
131 * config/i386/i386.c (m_GOLDMONT): Define.
132 (processor_target_table): Add "goldmont".
133 (PTA_GOLDMONT): Define.
134 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
135 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
136 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
137 (fold_builtin_cpu): Add "goldmont".
138 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
139 (ix86_option_override_internal): Add "goldmont".
140 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
141 (processor_type): Add PROCESSOR_GOLDMONT.
142 * config/i386/i386.md: Add CPU "glm".
143 * config/i386/glm.md: New file.
144 * config/i386/x86-tune.def: Add m_GOLDMONT.
145 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
147 2018-05-08 Jakub Jelinek <jakub@redhat.com>
150 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
152 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
153 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
154 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
157 * config/i386/i386.c (ix86_fold_builtin): Handle
158 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
161 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
162 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
164 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
167 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
168 (check_arch): Likewise.
169 (check_fpu): Return the result rather than printing it.
170 (end arch): Fix operator precedence.
172 (END): Print the result from check_fpu.
174 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
175 Alan Hayward <alan.hayward@arm.com>
176 David Sherwood <david.sherwood@arm.com>
178 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
179 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
180 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
181 (*fcmuo<mode>_and): New patterns.
183 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
185 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
186 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
187 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
188 (cmp_op, sve_imm_con): New code attributes.
189 (SVE_COND_INT_CMP, imm_con): Delete.
190 (cmp_op): Remove above unspecs from int attribute.
191 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
193 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
194 comparison-specific unspecs.
195 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
196 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
197 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
198 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
199 (*vec_fcm<cmp_op><mode>): Rename to...
200 (*fcm<cmp_op><mode>): ...this and adjust likewise.
201 (*vec_fcmuo<mode>): Rename to...
202 (*fcmuo<mode>): ...this and adjust likewise.
203 (*pred_fcm<cmp_op><mode>): New pattern.
204 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
205 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
207 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
209 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
210 (aarch64_emit_sve_predicated_cond): New function.
211 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
212 (aarch64_emit_unspec_cond_or): Replace with...
213 (aarch64_emit_sve_or_conds): ...this new function. Use
214 aarch64_emit_sve_ptrue_op for the individual comparisons and
215 aarch64_emit_binop to OR them together.
216 (aarch64_emit_inverted_unspec_cond): Replace with...
217 (aarch64_emit_sve_inverted_cond): ...this new function. Use
218 aarch64_emit_sve_ptrue_op for the comparison and
219 aarch64_emit_unop to invert the result.
220 (aarch64_expand_sve_vec_cmp_float): Update after the above
221 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
223 2018-05-07 Nathan Sidwell <nathan@acm.org>
225 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
226 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
227 (Backwards Compatibility): Likewise.
229 2018-05-07 Luis Machado <luis.machado@linaro.org>
233 2018-05-07 Luis Machado <luis.machado@linaro.org>
235 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
236 <prefetch_dynamic_strides>: New const bool field.
237 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
238 prefetch_dynamic_strides.
239 (exynosm1_prefetch_tune): Likewise.
240 (thunderxt88_prefetch_tune): Likewise.
241 (thunderx_prefetch_tune): Likewise.
242 (thunderx2t99_prefetch_tune): Likewise.
243 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
244 (aarch64_override_options_internal): Update to set
245 PARAM_PREFETCH_DYNAMIC_STRIDES.
246 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
247 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
248 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
249 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
250 prefetch-dynamic-strides setting.
252 2018-05-07 Luis Machado <luis.machado@linaro.org>
254 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
255 <minimum_stride>: New const int field.
256 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
257 minimum_stride field.
258 (exynosm1_prefetch_tune): Likewise.
259 (thunderxt88_prefetch_tune): Likewise.
260 (thunderx_prefetch_tune): Likewise.
261 (thunderx2t99_prefetch_tune): Likewise.
262 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
263 (aarch64_override_options_internal): Update to set
264 PARAM_PREFETCH_MINIMUM_STRIDE.
265 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
266 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
267 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
268 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
269 stride is constant and is below the minimum stride threshold.
271 2018-05-07 Luis Machado <luis.machado@linaro.org>
273 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
276 2018-05-07 Luis Machado <luis.machado@linaro.org>
278 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
279 <prefetch_dynamic_strides>: New const bool field.
280 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
281 prefetch_dynamic_strides.
282 (exynosm1_prefetch_tune): Likewise.
283 (thunderxt88_prefetch_tune): Likewise.
284 (thunderx_prefetch_tune): Likewise.
285 (thunderx2t99_prefetch_tune): Likewise.
286 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
287 (aarch64_override_options_internal): Update to set
288 PARAM_PREFETCH_DYNAMIC_STRIDES.
289 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
290 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
291 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
292 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
293 prefetch-dynamic-strides setting.
295 2018-05-07 Luis Machado <luis.machado@linaro.org>
297 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
298 <minimum_stride>: New const int field.
299 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
300 minimum_stride field.
301 (exynosm1_prefetch_tune): Likewise.
302 (thunderxt88_prefetch_tune): Likewise.
303 (thunderx_prefetch_tune): Likewise.
304 (thunderx2t99_prefetch_tune): Likewise.
305 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
306 (aarch64_override_options_internal): Update to set
307 PARAM_PREFETCH_MINIMUM_STRIDE.
308 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
309 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
310 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
311 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
312 stride is constant and is below the minimum stride threshold.
314 2018-05-06 Jakub Jelinek <jakub@redhat.com>
317 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
318 the type is addressable. Don't force op into register if it has
321 2018-05-05 Roland McGrath <mcgrathr@google.com>
324 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
325 any section for which we don't know a specific type it should have,
326 regardless of name. Previously this was done only for the exact
327 names ".init_array", ".fini_array", and ".preinit_array".
328 (default_elf_asm_named_section): Add comment about
329 relationship with default_section_type_flags and SECTION_NOTYPE.
330 (get_section): Don't consider it a type conflict if one side has
331 SECTION_NOTYPE and the other doesn't, as long as neither has the
332 SECTION_BSS et al used in the default_section_type_flags logic.
334 2018-05-05 Tom de Vries <tom@codesourcery.com>
337 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
338 (workaround_barsyncs): New function.
339 (nvptx_reorg): Use workaround_barsyncs.
340 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
341 (define_expand "nvptx_membar_cta"): New define_expand.
342 (define_insn "*nvptx_membar_cta"): New insn.
344 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
346 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
347 To improve optimization opportunities.
348 * builtin-types.def: The new needed builtin types for the above.
350 2018-05-04 Richard Biener <rguenther@suse.de>
352 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
353 * gimple-ssa-store-merging.c
354 (imm_store_chain_info::output_merged_store): Remove redundant create,
355 release split_store vector contents on failure.
356 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
357 scalar stmt vector on cache hit.
359 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
361 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
363 * config.gcc (powerpc-xilinx-eabi*): Remove.
364 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
366 (fusion_addis_mem_combo_load): Ditto.
367 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
369 (rs6000_cpu_cpp_builtins): Ditto.
370 * config/rs6000/rs6000-linux.c
371 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
372 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
373 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
375 (rs6000_setup_reg_addr_masks): Ditto.
376 (rs6000_init_hard_regno_mode_ok): Ditto.
377 (rs6000_option_override_internal): Ditto.
378 (legitimate_lo_sum_address_p): Ditto.
379 (rs6000_legitimize_address): Ditto.
380 (rs6000_legitimize_reload_address): Ditto.
381 (rs6000_legitimate_address_p): Ditto.
382 (abi_v4_pass_in_fpr): Ditto.
383 (setup_incoming_varargs): Ditto.
384 (rs6000_gimplify_va_arg): Ditto.
385 (rs6000_split_multireg_move): Ditto.
386 (rs6000_savres_strategy): Ditto.
387 (rs6000_emit_prologue_components): Ditto.
388 (rs6000_emit_epilogue_components): Ditto.
389 (rs6000_emit_prologue): Ditto.
390 (rs6000_emit_epilogue): Ditto.
391 (rs6000_elf_file_end): Ditto.
392 (rs6000_function_value): Ditto.
393 (rs6000_libcall_value): Ditto.
394 * config/rs6000/rs6000.h: Ditto.
395 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
396 (TARGET_MINMAX): ... this. New.
397 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
398 * config/rs6000/rs6000.md: Remove Xilinx FP support.
399 (*movsi_internal1_single): Delete.
400 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
401 mfpu=, mxilinx-fpu): Delete.
402 * config/rs6000/singlefp.h: Delete.
403 * config/rs6000/sysv4.h: Remove Xilinx FP support.
404 * config/rs6000/t-rs6000: Ditto.
405 * config/rs6000/t-xilinx: Delete.
406 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
407 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
408 (VStype_simple): Delete.
409 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
410 * config/rs6000/xfpu.h: Delete.
411 * config/rs6000/xfpu.md: Delete.
412 * config/rs6000/xilinx.h: Delete.
413 * config/rs6000/xilinx.opt: Delete.
414 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
415 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
417 2018-05-04 Tom de Vries <tom@codesourcery.com>
420 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
423 2018-05-04 Richard Biener <rguenther@suse.de>
426 * tree-complex.c (update_complex_assignment): We are always in SSA form.
427 (expand_complex_div_wide): Likewise.
428 (expand_complex_operations_1): Likewise.
429 (expand_complex_libcall): Preserve EH info of the original stmt.
430 (tree_lower_complex): Handle removed blocks.
431 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
432 on complex multiplication and division libcall builtins.
434 2018-05-04 Richard Biener <rguenther@suse.de>
437 * fold-const.c (negate_expr_p): Restrict negation of operand
438 zero of a division to when we know that can happen without
440 (fold_negate_expr_1): Likewise.
442 2018-05-04 Jakub Jelinek <jakub@redhat.com>
445 * real.h (real_nextafter): Declare.
446 * real.c (real_nextafter): New function.
447 * fold-const-call.c (fold_const_nextafter): New function.
448 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
450 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
451 even when arg1_mode is different from arg0_mode.
453 2018-05-03 Nathan Sidwell <nathan@acm.org>
455 * doc/extend.texi (Deprecated Features): Remove
457 (Backwards Compatibility): Likewise.
458 * doc/invoke.texi (C++ Language Options): Likewise.
459 (C++ Dialect Options): Likewise.
461 2018-05-03 Jakub Jelinek <jakub@redhat.com>
464 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
465 _mm512_mask_mullox_epi64): New intrinsics.
467 2018-05-03 Tom de Vries <tom@codesourcery.com>
470 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
471 dump files): Add offload-tree.
473 2018-05-03 Richard Biener <rguenther@suse.de>
475 PR tree-optimization/85615
476 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
477 to loops not nested in BBs loop father to avoid creating multi-entry
480 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
482 PR tree-optimization/70291
483 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
484 arguments. Change return type to tree. Emit libcall as a new
485 statement rather than replacing existing one when inplace_p is true.
486 (expand_complex_multiplication_components): New function.
487 (expand_complex_multiplication): Expand floating-point complex
488 multiplication using the above.
489 (expand_complex_division): Rename inner_type parameter to type.
490 Update expand_complex_libcall call-site.
491 (expand_complex_operations_1): Update expand_complex_multiplication
492 and expand_complex_division call-sites.
494 2018-05-02 Jakub Jelinek <jakub@redhat.com>
497 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
498 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
499 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
500 the highest significant bit of the shift count mask is clear. In
501 check whether and[sq]i3 is needed verify that all significant bits
502 of the shift count other than the highest are set.
504 2018-05-02 Tom de Vries <tom@codesourcery.com>
507 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
508 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
509 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
510 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
511 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
512 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
513 __builtin_goacc_parlevel_size.
515 2018-05-02 Richard Biener <rguenther@suse.de>
517 PR tree-optimization/85597
518 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
519 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
528 2018-05-02 Tom de Vries <tom@codesourcery.com>
531 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
532 dump files): Add wpa-ipa.
534 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
536 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
537 powerpc*-*-linux*paired* target.
538 * config/rs6000/750cl.h: Delete.
539 * config/rs6000/paired.h: Delete.
540 * config/rs6000/paired.md: Delete.
541 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
543 * config/rs6000/rs6000-builtin.def: Remove paired float support.
544 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
545 comment. Remove paired float support.
546 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
547 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
549 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
550 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
552 * config/rs6000/rs6000.c: Remove paired float support.
553 (paired_expand_vector_init, paired_expand_vector_move,
554 paired_emit_vector_compare, paired_emit_vector_cond_expr,
555 (paired_expand_lv_builtin, paired_expand_stv_builtin,
556 paired_expand_builtin, paired_expand_predicate_builtin,
557 paired_init_builtins): Delete.
558 * config/rs6000/rs6000.h: Remove paired float support.
559 * config/rs6000/rs6000.md: Remove paired float support.
560 (move_from_CR_ov_bit): Delete.
561 * config/rs6000/rs6000.opt (mpaired): Delete.
562 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
563 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
565 2018-05-02 Richard Biener <rguenther@suse.de>
568 * gimplify.c (gimplify_save_expr): When in SSA form allow
569 SAVE_EXPRs to compute to SSA vars.
571 2018-05-02 Jakub Jelinek <jakub@redhat.com>
574 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
575 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
576 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
577 clobber operands[2], instead use a new pseudo. Formatting fixes.
579 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
581 PR tree-optimization/85586
582 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
583 exit early for statements in the same group if the accesses are
586 2018-05-02 Tom de Vries <tom@codesourcery.com>
589 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
592 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
594 PR tree-optimization/85143
595 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
597 2018-05-01 Tom de Vries <tom@codesourcery.com>
600 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
601 not found" error message.
603 2018-05-01 Tom de Vries <tom@codesourcery.com>
606 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
607 * vec.c (test_ordered_remove_if): New function.
608 (vec_c_tests): Call test_ordered_remove_if.
609 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
610 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
611 * tree-vect-patterns.c (vect_pattern_recog_1): Use
612 VEC_ORDERED_REMOVE_IF.
614 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
616 PR tree-optimization/82665
617 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
618 pointer subtraction where arguments come from a memchr call.
620 2018-05-01 Jakub Jelinek <jakub@redhat.com>
622 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
623 --push-state --as-needed and --pop-state instead of --as-needed and
624 --no-as-needed if ld supports it.
625 * configure: Regenerated.
628 * doc/install.texi2html: Replace _002d with - and _002a with * in
629 generated html files using sed.
631 2018-04-30 David Malcolm <dmalcolm@redhat.com>
634 * gcc-rich-location.c (blank_line_before_p): New function.
635 (use_new_line): New function.
636 (gcc_rich_location::add_fixit_insert_formatted): New function.
637 * gcc-rich-location.h
638 (gcc_rich_location::add_fixit_insert_formatted): New function.
640 2018-04-30 David Malcolm <dmalcolm@redhat.com>
642 * selftest.c (assert_streq): Rename "expected" and "actual" to
643 "val1" and "val2". Extend NULL-handling to cover both inputs
644 symmetrically, while still requiring both to be non-NULL for a pass.
645 * selftest.h (assert_streq): Rename "expected" and "actual" to
647 (ASSERT_EQ): Likewise.
648 (ASSERT_EQ_AT): Likewise.
649 (ASSERT_KNOWN_EQ): Likewise.
650 (ASSERT_KNOWN_EQ_AT): Likewise.
651 (ASSERT_NE): Likewise.
652 (ASSERT_MAYBE_NE): Likewise.
653 (ASSERT_MAYBE_NE_AT): Likewise.
654 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
655 the assertion to pass.
656 (ASSERT_STREQ_AT): Likewise.
658 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
660 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
661 interaction with -pie.
663 2018-04-30 David Malcolm <dmalcolm@redhat.com>
665 * selftest.h: Fix alphabetization of per-source-file selftest
668 2018-04-30 Jason Merrill <jason@redhat.com>
670 PR c++/61982 - dead stores to destroyed objects.
671 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
674 2018-04-30 Jason Merrill <jason@redhat.com>
676 * tree.c (build_clobber): New.
677 * tree.h: Declare it.
678 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
680 2018-04-30 David Malcolm <dmalcolm@redhat.com>
682 * diagnostic-show-locus.c (layout::layout): Update for
683 location_get_source_line returning a char_span.
684 (struct char_span): Move to input.h.
685 (struct correction): Update for fields in char_span becoming
687 (struct source_line): Update for location_get_source_line
688 returning a char_span.
689 (layout::print_line): Likewise.
690 * edit-context.c (edited_file::print_content): Likewise.
691 (edited_file::print_diff_hunk): Likewise.
692 (edited_file::print_run_of_changed_lines): Likewise.
693 (edited_file::get_num_lines): Likewise.
694 (edited_line::edited_line): Likewise.
695 * final.c (asm_show_source): Likewise.
696 * input.c (location_get_source_line): Convert return type
697 from const char * to char_span, losing the final "line_len"
699 (dump_location_info): Update for the above.
700 (get_substring_ranges_for_loc): Likewise. Use a char_span
701 when handling the literal within the line.
702 (test_reading_source_line): Update for location_get_source_line
703 returning a char_span.
704 * input.h (class char_span): Move here from
705 diagnostic-show-locus.c, converting from a struct to a class.
706 Make data members private.
707 (char_span::operator bool): New.
708 (char_span::length): New.
709 (char_span::get_buffer): New.
710 (char_span::operator[]): New.
711 (char_span::subspan): Make const.
712 (char_span::xstrdup): New.
713 (location_get_source_line): Convert return type from const char *
714 to char_span, losing the final "line_size" param.
716 2018-04-30 Jan Hubicka <jh@suse.cz>
718 * lto-wrapper.c (ltrans_priorities): New static var.
720 (run_gcc): Read priorities and if doing parallel build order
721 the Makefile by them.
723 2018-04-30 David Malcolm <dmalcolm@redhat.com>
725 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
727 2018-04-30 Richard Biener <rguenther@suse.de>
729 * tree-cfg.c (verify_address): Remove base argument, add
730 flag whether to check TREE_ADDRESSABLE and do that.
731 (verify_expr): Remove.
732 (verify_types_in_gimple_reference): Add pieces from verify_expr.
733 (verify_gimple_assign_single): Likewise.
734 (verify_gimple_switch): Likewise.
735 (verify_expr_location_1): Dereference tp once. Add (disabled)
736 piece from verify_expr.
737 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
739 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
741 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
743 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
745 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
746 (small_data_pattern): Likewise.
747 (arc_rewrite_small_data): Likewise.
748 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
749 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
750 (get_symbol_alignment): New function.
751 (legitimate_small_data_address_p): Likewise.
752 (legitimate_scaled_address): Update, call
753 legitimate_small_data_address_p.
754 (output_sdata): New static variable.
755 (arc_print_operand): Update how we handle small data operands.
756 (arc_print_operand_address): Likewise.
757 (arc_legitimate_address_p): Update, use
758 legitimate_small_data_address_p.
759 (arc_rewrite_small_data_p): Remove.
760 (arc_rewrite_small_data_1): Likewise.
761 (arc_rewrite_small_data): Likewise.
762 (small_data_pattern): Likewise.
763 (compact_sda_memory_operand): Update to use
764 legitimate_small_data_address_p and get_symbol_alignment.
765 (prepare_move_operands): Don't rewite sdata pattern.
766 (prepare_extend_operands): Remove.
767 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
769 (zero_extendqisi2): Likewise.
770 (zero_extendhisi2): Likewise.
771 (extendqihi2): Likewise.
772 (extendqisi2): Likewise.
773 (extendhisi2): Likewise.
777 * config/arc/constraints.md (Usd): Change it to memory constraint.
779 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
781 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
782 as source of std instructions.
783 * config/arc/arc.md (movsi_insn): Update pattern predicate to
784 allow 6-bit constants as source for store instructions.
785 (movdi_insn): Update instruction pattern to allow 6-bit constants
786 as source for store instructions.
788 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
790 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
792 2018-04-30 Nathan Sidwell <nathan@acm.org>
793 Sandra Loosemore <sandra@codesourcery.com>
795 * dumpfile.c (dump_open): Allow '-' for stdout.
796 * doc/invoke.texi (Developer Options): Document dump filename
797 determination early. Document stdin/stdout selection.
799 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
801 Microblaze Target: PIC data text relative
803 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
804 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
806 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
807 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
808 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
809 New addressing mode for data-text relative position indepenedent code.
810 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
811 'ADDRESS_SYMBOLIC_TXT_REL'.
812 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
813 (microblaze_legitimate_pic_operand): Exclude function calls from
814 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
815 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
817 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
818 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
819 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
820 for 'address + offset'.
821 (microblaze_expand_prologue): Add new function prologue call for
823 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
824 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
825 table in case of TARGET_PIC_DATA_TEXT_REL.
826 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
827 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
828 Add new macros 'UNSPEC_TEXT',
829 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
830 + exclude function calls from 'UNSPEC_PLT' in case of data text
832 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
833 new target hook for generating address diff vector tables in case of
835 * doc/tm.texi : Regenerate.
836 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
837 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
838 of addr diff vector generation.
839 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
840 target hook definition.
841 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
842 Add default function for generate_pic_addr_diff_vec -> flag_pic.
843 * doc/invoke.texi (Add new pic option): Add new microblaze pic
844 option for data text relative.
846 2018-04-30 Richard Biener <rguenther@suse.de>
848 * tree-chrec.h (evolution_function_is_constant_p): Remove
850 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
852 2018-04-30 Richard Biener <rguenther@suse.de>
855 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
857 2018-04-30 Richard Biener <rguenther@suse.de>
859 PR tree-optimization/28364
860 PR tree-optimization/85275
861 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
862 copying first exit test.
864 2018-04-28 Mark Wielaard <mark@klomp.org>
866 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
868 (dwarf_AT): Handle DW_AT_addr_base.
869 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
871 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
874 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
875 (*ashl<dwi>3_doubleword_mask_1): Ditto.
876 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
877 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
879 2018-04-28 Richard Biener <rguenther@suse.de>
881 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
882 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
883 to reflect use. Only add interesting stmts.
885 2018-04-27 Martin Jambor <mjambor@suse.cz>
888 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
889 the jump function allows for passing through aggregate values.
891 2018-04-27 David Malcolm <dmalcolm@redhat.com>
893 * input.h (in_system_header_at): Convert from macro to inline
895 (from_macro_expansion_at): Likewise.
896 (from_macro_definition_at): Likewise.
898 2018-04-27 Jeff Law <law@redhat.com>
900 * config.gcc: Mark tile* targets as deprecated/obsolete.
902 2018-04-27 Richard Biener <rguenther@suse.de>
904 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
907 2018-04-27 Richard Biener <rguenther@suse.de>
909 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
911 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
913 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
914 with Yd constraint. Set "preferred_for_speed" attribute from
915 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
917 (*movdi_internal): Ditto.
918 (movti_interunit splitters): Remove
919 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
920 (movdi_interunit splitters): Ditto.
921 * config/i386/constraints.md (Ye): Remove.
922 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
924 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
927 * config/aarch64/constraints.md (Usg): Limit to 31.
930 2018-04-27 Jakub Jelinek <jakub@redhat.com>
932 PR tree-optimization/85529
933 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
934 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
935 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
936 zero extension or masking of the MSB bit.
937 (optimize_range_tests): Add FIRST_BB argument, pass it through
938 to optimize_range_tests_var_bound.
939 (maybe_optimize_range_tests, reassociate_bb): Adjust
940 optimize_range_tests callers.
942 2018-04-26 Richard Biener <rguenther@suse.de>
943 Jakub Jelinek <jakub@redhat.com>
945 * cgraph.h (symbol_table): Just declare debug method here.
946 * symtab.c (symbol_table::debug): Define.
948 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
950 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
952 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
954 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
955 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
956 (*movdi_internal): Substitute Yi and Yj constraint with x
957 and Ym and Yn constraint with y constraint. Update "isa"
958 attribute and set "preferred_for_speed" attribute from
959 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
960 (*movsi_internal): Ditto.
961 (*movdf_internal): Ditto.
962 (*movsf_internal): Ditto.
963 (*zero_extendsidi2): Ditto.
964 * config/i386/sse.md (vec_set<mode>_0): Ditto.
965 (sse2_loadld): Ditto.
966 (*vec_extract<ssevecmodelower>_0): Ditto.
967 (*vec_extractv4si_0_zext_sse4): Ditto.
968 (vec_concatv2di): Ditto.
969 (*vec_dup<mode>): Ditto.
970 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
971 * config/i386/constraints.md (Yi): Remove.
976 2018-04-26 Nathan Sidwell <nathan@acm.org>
978 * dumpfile.c (dump_open): New.
979 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
980 (dump_finish): Detect stdio/stderr by value not name.
982 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
984 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
986 2018-04-26 Tom de Vries <tom@codesourcery.com>
989 * config/nvptx/nvptx.c (verify_neutering_jumps)
990 (verify_neutering_labels): New function
991 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
993 2018-04-26 Tom de Vries <tom@codesourcery.com>
996 * config/nvptx/nvptx.c (needs_neutering_p): New function.
997 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1000 2018-04-26 Richard Biener <rguenther@suse.de>
1001 Tom de Vries <tom@codesourcery.com>
1004 * lto-streamer-out.c (output_function): Fixup loops if required to match
1005 discovery done in the reader.
1007 2018-04-26 Richard Biener <rguenther@suse.de>
1009 PR tree-optimization/85116
1010 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1011 have a loop exit from the single latch predecessor. Remove
1012 case of header with just condition.
1013 (ch_base::copy_headers): Exclude infinite loops from any
1015 (pass_ch::execute): Record exits.
1017 2018-04-26 Richard Biener <rguenther@suse.de>
1019 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1020 prologue cost vector and pass it to vect_get_load_cost.
1021 (vect_get_peeling_costs_all_drs): Likewise.
1022 (vect_peeling_hash_get_lowest_cost): Likewise.
1023 (vect_enhance_data_refs_alignment): Likewise.
1025 2018-04-26 Richard Biener <rguenther@suse.de>
1028 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1029 checking of integer<->pointer conversions.
1030 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1031 sign-/zero-extending pointer types.
1032 (expand_omp_for_static_chunk): Likewise.
1034 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1035 Jean Lee <xiaoyur347@gmail.com>
1037 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1038 (TARGET_ASAN_SHADOW_OFFSET): Define.
1039 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1040 true for -fsanitize=address.
1042 2018-04-25 Mark Wielaard <mark@klomp.org>
1044 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1047 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1049 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1050 than "alu", remove explicit "memory" and "imm_disp" attributes.
1051 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1054 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1055 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1056 gen_lowpart_no_emit.
1058 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1061 * config/i386/i386.c (ix86_expand_builtin): Change memory
1062 operand to XI, extend p0 to Pmode.
1063 * config/i386/i386.md: Change unspec volatile and operand
1064 1 mode to XI, change operand 0 mode to P.
1066 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1068 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1069 GET_MODE_MASK before any checking.
1070 (nds32_can_use_bset_p): Likewise.
1071 (nds32_can_use_btgl_p): Likewise.
1073 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1075 * config/nds32/nds32-doubleword.md: New define_split pattern for
1076 illegal register number.
1078 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1080 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1082 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1084 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1086 2018-04-25 Richard Biener <rguenther@suse.de>
1088 * lto-streamer.h (LTO_major_version): Bump to 8.
1090 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1092 * BASE-VER: Set to 9.0.0.
1094 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1096 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1097 in __abskf2 and __powikf2.
1099 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1102 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1103 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1104 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1105 Use the above on operand 2. Reindent.
1106 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1108 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1111 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1113 * config/i386/i386.opt (mcet): Removed.
1114 * doc/install.texi: Remove -mcet documentation.
1115 * doc/invoke.texi: Likewise.
1117 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1120 * doc/install.texi: Remove -mcet from bootstrap-cet.
1122 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1125 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1126 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1130 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1131 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1132 containing a CONST_VECTOR.
1134 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1136 * doc/install.texi: Update newlib dependency for nvptx.
1138 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1141 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1142 instead of INTVAL when shifting x left.
1144 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1146 PR tree-optimization/85478
1147 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1148 vect_grouped_store_supported for single element vectors.
1150 2018-04-24 Richard Biener <rguenther@suse.de>
1153 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1154 load cost increase to the case of non-constant step.
1156 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1159 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1160 destination if any_malformed_asm.
1162 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1165 * expr.c (store_field): In the bitfield case, if the value comes from
1166 a function call and is returned in registers by means of a PARALLEL,
1167 do not change the mode of the temporary unless BLKmode and VOIDmode.
1169 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1171 PR rtl-optimization/85423
1172 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1173 dependencies to debug insns when the previous insn is non-debug.
1175 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1177 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1178 enums into a single definition.
1179 (fls): Fix predicates and printing.
1182 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1184 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1185 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1186 and short u6 immediate.
1187 (check_if_valid_sleep_operand): Remove.
1188 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1190 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1192 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1193 flag_always_save_lp condition.
1194 * config/nds32/nds32.opt (malways-save-lp): New option.
1196 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1198 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1199 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1200 * config/nds32/nds32.h
1201 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1202 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1204 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1206 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1207 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1209 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1210 Chung-Ju Wu <jasonwucj@gmail.com>
1212 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1214 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1215 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1217 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1219 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1221 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1223 * config/nds32/nds32-protos.h (nds32_data_alignment,
1224 nds32_local_alignment): Declare.
1225 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1226 nds32_local_alignment): New functions.
1227 (TARGET_CONSTANT_ALIGNMENT): Define.
1228 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1230 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1232 * config/nds32/nds32.c
1233 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1234 (TARGET_MODES_TIEABLE_P): Likewise.
1236 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1238 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1241 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1242 Chung-Ju Wu <jasonwucj@gmail.com>
1244 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1245 for unaligned access.
1246 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1248 * config/nds32/nds32-intrinsic.md: Likewise.
1249 * config/nds32/nds32_intrinsic.h: Likewise.
1250 * config/nds32/nds32.h (nds32_builtins): Likewise.
1251 * config/nds32/nds32.opt (munaligned-access): New option.
1252 * config/nds32/nds32.c (nds32_asm_file_start): Display
1253 flag_unaligned_access status.
1255 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1257 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1258 -mno-relax is present.
1259 * config/riscv/linux.h (LINK_SPEC): Ditto.
1261 2018-04-20 Martin Sebor <msebor@redhat.com>
1264 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1266 (gimple_fold_builtin_stxcpy_chk): Same.
1267 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1269 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1272 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1273 __powikf2 when long double is IEEE 128-bit.
1275 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1277 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1278 step to make sure stack always aligned.
1280 2018-04-20 Carl Love <cel@us.ibm.com>
1283 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1284 size check for arg0.
1286 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1287 Tom de Vries <tom@codesourcery.com>
1290 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1291 Emit insns for calls too.
1292 (nvptx_find_par): Always look for worker-level predecessor insn.
1293 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1295 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1296 (nvptx_process_pars): Propagate frames for calls.
1298 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1301 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1303 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1304 (ix86_handle_option): Don't handle OPT_mibt.
1305 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1307 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1309 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1310 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1311 (ix86_target_macros): Define __CET__ with flag_cf_protection
1312 for -fcf-protection.
1313 * config/i386/i386.c (isa2_opts): Remove -mibt.
1314 * config/i386/i386.h (TARGET_IBT): Removed.
1315 (TARGET_IBT_P): Likewise.
1316 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1317 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1318 * config/i386/i386.opt (mcet): Update help message.
1321 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1322 -mcet as an alias for -mshstk.
1324 2018-04-20 Richard Biener <rguenther@suse.de>
1327 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1328 complexity by forcing a single use of the multiply operand.
1330 2018-04-20 Martin Jambor <mjambor@suse.cz>
1333 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1334 recursion dependency to only apply to non-clones.
1336 2018-04-20 Martin Jambor <mjambor@suse.cz>
1339 * ipa-cp.c (create_specialized_node): Check that clones of
1340 self-recursive edges exist during IPA-CP.
1342 2018-04-19 Toon Moene <toon@moene.org>
1344 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1347 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1349 PR tree-optimization/85467
1350 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1351 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1352 VECTOR_CST element to type.
1354 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1357 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1358 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1359 (builtin_longjmp): Likewise.
1360 (save_stack_nonlocal): New pattern.
1361 (restore_stack_nonlocal): Likewise.
1363 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1366 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1367 Replace ASM_OUTPUT_LABEL with fprintf.
1369 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1372 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1373 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1374 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1375 define __IBT__ and __SHSTK__ for -fcf-protection.
1376 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1378 (ix86_trampoline_init): Likewise.
1379 (x86_output_mi_thunk): Likewise.
1380 (ix86_notrack_prefixed_insn_p): Likewise.
1381 (ix86_option_override_internal): Don't disallow -fcf-protection.
1382 * config/i386/i386.md (rdssp<mode>): Also enable for
1384 (incssp<mode>): Likewise.
1385 (nop_endbr): Likewise.
1386 * config/i386/i386.opt (mcet): Change help message to built-in
1390 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1391 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1392 enable CET built-in functions.
1394 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1396 * common/config/i386/i386-common.c
1397 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1398 OPTION_MASK_ISA_MOVDIRI_UNSET,
1399 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1400 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1401 * config.gcc (movdirintrin.h): New header.
1402 * config/i386/cpuid.h (bit_MOVDIRI,
1403 bit_MOVDIR64B): New bits.
1404 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1406 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1407 (VOID, PVOID, PCVOID)): New function types.
1408 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1409 __builtin_ia32_directstoreu_u64,
1410 __builtin_ia32_movdir64b): New builtins.
1411 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1412 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1414 (ix86_valid_target_attribute_inner_p): Ditto.
1415 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1416 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1417 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1418 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1419 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1420 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1421 (movdiri<mode>, movdir64b_<mode>): New.
1422 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1423 * config/i386/immintrin.h: Include movdirintrin.h.
1424 * config/i386/movdirintrin.h: New file.
1425 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1427 2018-04-19 Richard Biener <rguenther@suse.de>
1430 * cfg.c (clear_bb_flags): When loop state says we have
1431 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1433 2018-04-19 Richard Biener <rguenther@suse.de>
1435 PR tree-optimization/84737
1436 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1437 copying restrict info.
1438 (vect_setup_realignment): Use it.
1439 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1440 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1441 the first DR to all generated stores.
1442 (vectorizable_load): Likewise for loads.
1444 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1446 PR tree-optimization/85446
1447 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1448 the integral and pointer types to have the same precision.
1450 * doc/install.texi: Document --disable-cet being the default and
1453 2018-04-18 Martin Liska <mliska@suse.cz>
1455 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1458 2018-04-18 Martin Liska <mliska@suse.cz>
1461 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1464 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1465 arguments if they are comparable.
1467 2018-04-18 Martin Liska <mliska@suse.cz>
1470 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1473 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1476 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1479 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1480 ENDBR after calling __morestack.
1482 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1485 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1486 by using gcc_base_ver to generate a gcc_driver_version, and use
1487 it when generating GCC_DRIVER_NAME.
1488 * configure: Regenerate.
1490 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1493 * config.gcc: Obsolete powerpc*-*-*spe*.
1495 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1498 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1499 (stabstr_D): Change type of unum from unsigned int to
1500 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1503 2018-04-17 Jim Wilson <jimw@sifive.com>
1506 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1507 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1508 Set arg_pointer_offset after using pretend_args_size.
1510 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1512 PR rtl-optimization/85431
1513 * dse.c (record_store): Ignore zero width stores.
1516 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1517 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1518 __builtin_stack_restore rather than after it.
1519 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1520 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1521 argument instead of virtual_dynamic_stack_rtx.
1523 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1525 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1527 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1528 Add note to error message to explain internal mapping of overloaded
1529 built-in function name to non-overloaded built-in function name.
1530 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1533 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1536 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1537 where the inputs overlap with the output.
1539 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1541 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1542 (=v, v) alternative and explicit "memory" attribute.
1543 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1544 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1546 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1547 "sselog1" type instead of "sselog".
1548 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1549 "sselog". Remove explicit "memory" attribute.
1550 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1551 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1553 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1554 "sselog1" type instead of "sselog". Remove explicit "memory"
1556 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1557 use "sselog1" type instead of "sselog". Remove explicit "memory"
1559 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1560 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1562 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1563 "sselog1" type instead of "sselog". Remove explicit "memory"
1565 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1566 use "sselog1" type instead of "sselog". Remove explicit "memory"
1570 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1573 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1576 2018-04-17 Martin Jambor <mjambor@suse.cz>
1579 * ipa-cp.c (create_specialized_node): Call
1580 expand_all_artificial_thunks if necessary.
1582 2018-04-17 Martin Liska <mliska@suse.cz>
1585 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1586 in message, remote space in between '_G' and '('.
1588 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1591 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1592 avx512f_vmcmp<mode>3<round_saeonly_name>,
1593 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1594 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1595 avx512f_rndscale<mode><round_saeonly_name>,
1596 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1597 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1598 Use %<iptr>2 instead of %2 for -masm=intel.
1599 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1600 avx512f_vcvttss2usi<round_saeonly_name>,
1601 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1603 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1604 avx512f_vcvttsd2usi<round_saeonly_name>,
1605 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1606 Use %q1 instead of %1 for -masm=intel.
1607 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1608 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1609 of %3 for -masm=intel.
1610 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1612 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1614 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1616 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1617 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1619 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1621 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1622 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1623 %g1 and one with %0 and %1.
1624 (avx512er_vmrcp28<mode><round_saeonly_name>,
1625 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1627 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1628 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1629 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1630 of %0 and %{%4%} for -masm=intel.
1631 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1632 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1633 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1634 order of %0 and %{%5%}%{z%} for -masm=intel.
1636 2018-04-17 Jan Hubicka <jh@suse.cz>
1639 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1641 2018-04-17 Martin Liska <mliska@suse.cz>
1644 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1645 for target_clone error message. Make default implementation
1646 clone to be a local declaration.
1647 (separate_attrs): Add new argument and check for an empty
1649 (expand_target_clones): Handle it.
1650 (ipa_target_clone): Make redirection just for target_clones
1653 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1654 Tom de Vries <tom@codesourcery.com>
1657 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1658 tiled basic blocks without omp continue statements.
1660 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1663 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1664 vec_extract expression as having side effects to make sure it gets
1667 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1670 * config/i386/i386.c (get_builtin_code_for_version): Check
1673 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1676 * gcc/config.gcc: Support "skylake".
1677 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1679 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1680 (processor_target_table): Add "skylake".
1681 (ix86_option_override_internal): Add "skylake".
1682 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1683 PROCESSOR_CANNONLAKE.
1684 (get_builtin_code_for_version): Fix priority for
1685 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1686 PROCESSOR_SKYLAKE-AVX512.
1687 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1688 (processor_type): Add PROCESSOR_SKYLAKE.
1690 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1691 Jason Merrill <jason@redhat.com>
1694 * convert.c (convert_to_integer_1): Use direct recursion for
1695 enumeral types and types with a precision less than the number
1696 of bits in their mode.
1698 2018-04-16 Julia Koval <julia.koval@intel.com>
1701 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1702 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1704 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1707 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1708 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1709 and -mno-direct-move.
1711 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1714 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1715 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1716 Ensure negative shifts result in {0}.
1718 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1720 PR rtl-optimization/79916
1721 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1722 regs (if any) to define how to gnerate SD moves when LRA is in
1725 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1727 PR rtl-optimization/85393
1728 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1729 * except.c (expand_dw2_landing_pad_for_region): Make static.
1730 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1731 a label and unconditional jump to old_bb, rather than
1732 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1735 PR rtl-optimization/85376
1736 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1737 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1738 instead of a specific value.
1740 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1741 Bin Cheng <bin.cheng@arm.com>
1743 PR tree-optimization/82965
1744 PR tree-optimization/83991
1745 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1746 by_profile_only parameter.
1747 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1748 information if the loop was predicted to iterate too many times.
1749 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1751 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1754 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1757 2018-04-13 Martin Liska <mliska@suse.cz>
1758 Jakub Jelinek <jakub@redhat.com>
1761 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1762 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1763 * builtins.c (expand_builtin_memory_copy_args): Use
1764 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1765 handle dest_addr == pc_rtx.
1767 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1770 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1771 asked to not generate direct moves.
1772 (fix_trunc<mode>si2_stfiwx): Similar.
1773 (fix_trunc<mode>si2_internal): Similar.
1775 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1778 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1779 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1780 lookup if dest in some wider mode is known to be const0_rtx and
1781 if so, record permanent equivalence for it to be ZERO_EXTEND of
1782 the narrower mode destination.
1784 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1786 * lto-streamer-out.c (output_function): Revert 259346.
1787 * omp-expand.c (expand_oacc_for): Likewise.
1789 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1791 PR rtl-optimization/85354
1792 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1793 * sel-sched.c (sel_global_init): ... here.
1795 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1798 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1799 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1800 mode for PE-COFF targets.
1801 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1802 (i386_pe_asm_lto_end): Likewise.
1803 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1804 (TARGET_ASM_LTO_END): Likewise.
1805 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1806 (i386_pe_asm_lto_start): New function.
1807 (i386_pe_asm_lto_end): Likewise.
1809 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1810 Richard Biener <rguenther@suse.de>
1813 * lto-streamer-out.c (output_function): Fix CFG loop state before
1815 * omp-expand.c (expand_oacc_for): Handle calls to internal
1816 functions like regular functions.
1818 2018-04-12 Richard Biener <rguenther@suse.de>
1821 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1822 for the early LTO debug to properly generate references to it
1823 during DIE emission. Do not re-use that for the skeleton for
1825 (dwarf2out_early_finish): Likewise.
1827 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1830 * config/i386/sse.md
1831 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1832 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1833 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1834 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1835 and output is a reg, avoid creating invalid lowpart subreg, but
1836 instead split into a 512-bit move. Don't split if not AVX512VL,
1837 input is xmm16+ reg and output is a mem.
1838 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1839 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1840 xmm16+ reg and output is a mem.
1842 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1844 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1845 also for flag_dwarf2_cfi_asm.
1847 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1849 PR rtl-optimization/85342
1850 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1851 a bool scalar var inside of the loop instead. Don't try to update
1852 recog_data.operand after failed apply_change_group.
1854 2018-04-12 Tom de Vries <tom@codesourcery.com>
1857 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1858 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1859 array with flexible array member as array without given dimension.
1860 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1861 argument for undefined param to true.
1863 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1866 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1867 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1868 from PowerPC section.
1869 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1870 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1871 help text that is too long.
1872 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1873 help text that is too long.
1874 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1875 help text that is too long.
1877 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1879 * config/alpha/alpha.md (stack_probe_internal): Rename
1880 from "probe_stack". Update all callers.
1882 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1884 PR rtl-optimization/84566
1885 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1886 sched_macro_fuse_insns.
1888 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1891 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1892 (compute_block_dependences): ... from here.
1894 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1896 PR tree-optimization/85331
1897 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1898 from int to HOST_WIDE_INT.
1900 2018-04-11 Martin Jambor <mjambor@suse.cz>
1903 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1904 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1905 not the same as the source val.
1906 (cgraph_edge_brings_value_p): New parameter.
1907 (gather_edges_for_value): Pass destination value to
1908 cgraph_edge_brings_value_p.
1909 (perhaps_add_new_callers): Likewise.
1910 (get_info_about_necessary_edges): Likewise and exclude values brought
1911 only by self-recursive edges.
1912 (create_specialized_node): Redirect only clones of self-calling edges.
1913 (+self_recursive_pass_through_p): New function.
1914 (find_more_scalar_values_for_callers_subset): Use it.
1915 (find_aggregate_values_for_callers_subset): Likewise.
1916 (known_aggs_to_agg_replacement_list): Removed.
1917 (decide_whether_version_node): Re-calculate known constants for all
1918 remaining context clones.
1920 2018-04-11 Richard Biener <rguenther@suse.de>
1923 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1924 from early DWARF output.
1925 (dwarf2out_early_finish): Output line info unconditionally into
1926 early DWARF and add reference to it.
1928 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1931 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1932 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1933 other than V2DFmode using iptr mode attribute.
1934 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1936 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1938 PR rtl-optimization/84659
1939 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1941 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1944 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1946 (output_loc_list): Pass address of a dummy size variable even in the
1947 locview handling loop.
1948 (index_location_lists): Add comment on why skip_loc_list_entry can't
1951 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1954 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1957 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1960 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1961 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1962 and -mstring-compare-inline-limit.
1964 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1967 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1968 for stack clash protection in a register whenever we need it to be in
1971 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1973 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1974 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1976 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1979 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1981 (mlong-double-): Ditto.
1982 * config/rs6000/sysv4.opt (msdata=): Ditto.
1983 (mtls-size=): Ditto.
1985 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1987 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1988 erroneous entries for
1989 "vector int vec_ldl (int, long int *)", and
1990 "vector unsigned int vec_ldl (int, unsigned long int *)".
1991 Add comments and entries for
1992 "vector bool char vec_ldl (int, bool char *)",
1993 "vector bool short vec_ldl (int, bool short *)",
1994 "vector bool int vec_ldl (int, bool int *)",
1995 "vector bool long long vec_ldl (int, bool long long *)",
1996 "vector pixel vec_ldl (int, pixel *)",
1997 "vector long long vec_ldl (int, long long *)",
1998 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1999 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2000 type tree bool_long_long_type_node and correct definition of
2001 bool_V2DI_type_node to make reference to this new type tree.
2002 (rs6000_mangle_type): Replace erroneous reference to
2003 bool_long_type_node with bool_long_long_type_node.
2004 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2005 comments to emphasize sign distinctions for char and int types and
2006 replace RS6000_BTI_bool_long constant with
2007 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2008 use of RS6000_BTI_pixel.
2009 (bool_long_type_node): Remove this macro definition.
2010 (bool_long_long_type_node): New macro definition
2012 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2014 PR rtl-optimization/85300
2015 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2016 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2017 simplify_unary_operation fails.
2019 2018-04-10 Martin Liska <mliska@suse.cz>
2021 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2022 cgraph_edge and ipa_ref.
2024 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2028 * config/i386/sse.md
2029 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2030 computation of the VEC_MERGE selector from mask.
2031 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2032 Fix decoding of the VEC_MERGE selector into mask.
2034 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2036 PR tree-optimization/85286
2037 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2039 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2041 * final.c (final_1): Set insn_last_address as well as
2042 insn_current_address.
2044 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2047 * explow.c (emit_stack_probe): Call validize_mem on memory location
2048 before passing it to gen_probe_stack. Create address operand and
2049 legitimize it for the probe_stack_address case.
2051 2018-04-09 Jan Hubicka <jh@suse.cz>
2054 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2055 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2056 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2058 (free_lang_data): Rebuild type inheritance graph.
2060 2018-04-09 Martin Sebor <msebor@redhat.com>
2062 * invoke.texi (-finline-small-functions): Mention other optimization
2064 (-findirect-inlining, -fpartial-inlining): Same.
2065 (-finline-functions-called-once): Same.
2066 (-freorder-blocks-and-partition): Same.
2068 2018-04-09 Jan Hubicka <jh@suse.cz>
2071 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2072 jumps; choose last target that matches the criteria (i.e.
2073 no partition changes for non-crossing jumps).
2074 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2075 support for redirecting crossing jumps to non-crossing.
2077 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2079 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2080 also for naked functions.
2082 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2084 * config/arc/arc.md (add_shift): New pattern.
2085 (add_shift2): Likewise.
2086 (sub_shift): Likewise.
2087 (sub_shift_cmp0_noout): Likewise.
2088 (compare_si_ashiftsi): Likewise.
2089 (xbfu_cmp0_noout): New combine pattern.
2090 (xbfu_cmp0"): Likewise.
2091 (movsi_set_cc_insn): Place the predicable variant first.
2092 (commutative_binary_cmp0_noout): Remove clobber.
2093 (commutative_binary_cmp0): New pattern.
2094 (noncommutative_binary_cmp0): Likewise.
2095 (noncommutative_binary_cmp0_noout): Likewise.
2096 (noncommutative_binary_comparison_result_used): Removed.
2097 (rsub_cmp0): New pattern.
2098 (rsub_cmp0_noout): Likewise.
2099 (extzvsi): Changed, keep only meaningful variants.
2100 (SQH, SEZ): New iterators.
2101 (SQH_postfix): New mode attribute.
2102 (SEZ_prefix): New code attribute.
2103 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2104 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2105 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2107 (noncommutative_operator): Check the availability of barrel
2110 2018-04-09 Richard Biener <rguenther@suse.de>
2112 PR tree-optimization/85284
2113 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2114 Only use the niter constraining form of simple_iv when the exit
2117 2018-04-09 Tom de Vries <tom@codesourcery.com>
2120 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2121 (define_expand "*memory_barrier"): New define_expand.
2122 (define_insn "memory_barrier"): New insn.
2124 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2126 PR rtl-optimization/80463
2127 PR rtl-optimization/83972
2128 PR rtl-optimization/83480
2130 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2131 correct producer for the insn.
2132 (tidy_control_flow): Fixup seqnos in case of debug insns.
2134 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2136 PR rtl-optimization/83913
2138 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2139 different sched-times when merging exprs.
2141 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2143 PR rtl-optimization/83962
2145 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2146 tidy_fallthru_edge and tidy_control_flow.
2148 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2150 PR rtl-optimization/83530
2152 * sel-sched.c (force_next_insn): New global variable.
2153 (remove_insn_for_debug): When force_next_insn is true, also leave only
2154 next insn in the ready list.
2155 (sel_sched_region): When the region wasn't scheduled, make another pass
2156 over it with force_next_insn set to 1.
2158 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2160 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2162 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2163 for interrupt control.
2164 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2165 functions for interrupt control.
2166 * config/nds32/nds32-intrinsic.md: Likewise.
2167 * config/nds32/nds32_intrinsic.h: Likewise.
2168 * config/nds32/nds32.h (nds32_builtins): Likewise.
2170 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2172 * config/nds32/nds32.c (nds32_init_machine_status,
2173 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2174 strict_aligned_p field.
2175 (nds32_expand_to_rtl_hook): New function.
2176 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2177 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2179 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2180 Chung-Ju Wu <jasonwucj@gmail.com>
2182 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2183 * config/nds32/nds32-n7.md: New file.
2184 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2185 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2187 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2188 * config/nds32/nds32.md (pipeline_model): Add n7.
2189 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2190 * config/nds32/pipelines.md: Include n7 settings.
2192 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2193 Chung-Ju Wu <jasonwucj@gmail.com>
2195 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2196 * config/nds32/nds32-e8.md: New file.
2197 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2198 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2200 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2201 * config/nds32/nds32.md (pipeline_model): Add e8.
2202 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2203 * config/nds32/pipelines.md: Include e8 settings.
2205 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2206 Chung-Ju Wu <jasonwucj@gmail.com>
2208 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2209 * config/nds32/nds32-n8.md: New file.
2210 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2211 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2213 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2214 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2215 * config/nds32/nds32.md (pipeline_model): Add n8.
2216 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2217 * config/nds32/pipelines.md: Include n8 settings.
2219 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2220 Chung-Ju Wu <jasonwucj@gmail.com>
2222 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2223 * config/nds32/nds32-n9-2r1w.md: New file.
2224 * config/nds32/nds32-n9-3r2w.md: New file.
2225 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2226 nds32_register_ports): New or modify for cpu n9.
2227 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2229 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2230 * config/nds32/nds32-utils.c: New file.
2231 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2232 TARGET_MUL_SLOW): Define.
2233 * config/nds32/nds32.md (pipeline_model): New attribute.
2234 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2235 New options that support cpu n9.
2236 * config/nds32/pipelines.md: Include n9 settings.
2237 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2239 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2241 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2242 information if necessary.
2243 (output_cond_branch_compare_zero): Likewise.
2244 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2245 (nds32_target_alignment): Refine for alignment.
2246 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2247 (FUNCTION_BOUNDARY): Modify.
2248 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2250 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2252 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2254 * config/nds32/constants.md (unspec_volatile_element): Add values for
2255 TLB operation and data prefetch.
2256 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2257 functions for TLB operation and data prefetch.
2258 * config/nds32/nds32-intrinsic.md: Likewise.
2259 * config/nds32/nds32_intrinsic.h: Likewise.
2260 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2261 (nds32_print_operand): Likewise.
2262 * config/nds32/nds32.h (nds32_builtins): Likewise.
2264 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2265 Andrew Pinski <pinsika@gcc.gnu.org>
2268 * match.pd: Use constant_boolean_node of correct type instead of
2269 boolean_true_node or boolean_false_node for simplifying
2270 pointer comparisons to zero.
2272 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2274 PR tree-optimization/80021
2275 * tree.c (verify_type_variant): Make error call in verify_variant_match
2276 translatable and remove final full stop.
2278 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2280 * config/nds32/constants.md (unspec_volatile_element): Add
2281 UNSPEC_VOLATILE_EH_RETURN.
2282 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2283 nds32_output_stack_pop): Support dwarf exception handling process.
2284 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2285 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2286 exception handling process.
2287 (nds32_compute_stack_frame): Likewise.
2288 (nds32_return_addr_rtx): Likewise.
2289 (nds32_initial_elimination_offset): Likewise.
2290 (nds32_expand_prologue): Likewise.
2291 (nds32_expand_epilogue): Likewise.
2292 (nds32_dynamic_chain_address): New function.
2293 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2295 (DYNAMIC_CHAIN_ADDRESS): Define.
2296 (EH_RETURN_DATA_REGNO): Define.
2297 (EH_RETURN_STACKADJ_RTX): Define.
2298 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2299 patterns for dwarf exception handling.
2301 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2303 * config/nds32/nds32.h: Clean up obsolete macros.
2305 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2307 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2308 Add enum values for particular instructions.
2309 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2310 particular intrinsic functions.
2311 * config/nds32/nds32-intrinsic.md: Likewise.
2312 * config/nds32/nds32_intrinsic.h: Likewise.
2313 * config/nds32/nds32.h (nds32_builtins): Likewise.
2314 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2315 (btst, ave): New patterns for particular instructions.
2317 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2319 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2320 Add enum values for atomic load/store and memory sync.
2321 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2323 * config/nds32/nds32-intrinsic.md: Likewise.
2324 * config/nds32/nds32_intrinsic.h: Likewise.
2325 * config/nds32/nds32.h (nds32_builtins): Likewise.
2327 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2329 PR tree-optimization/85257
2330 * fold-const.c (native_encode_vector): If not all elts could fit
2331 and off is -1, return 0 rather than offset.
2332 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2333 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2334 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2335 adjust buffer in native_interpret_expr call.
2337 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2339 * config/nds32/constants.md (unspec_volatile_element): Add cache
2340 control enum values.
2341 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2342 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2343 * config/nds32/nds32.c (nds32_cctl_names): New.
2344 (nds32_print_operand): Handle cache control register names.
2345 * config/nds32/nds32.h (nds32_builtins): New enum values.
2346 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2348 * config/nds32/nds32.md (type): Add mmu.
2349 * config/nds32/pipelines.md (simple_insn): Add mmu.
2351 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2353 * config/nds32/nds32.md (type): Remove call.
2354 * config/nds32/pipelines.md (simple_insn): Likewise.
2356 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2358 * config/nds32/constants.md (unspec_volatile_element): Add
2359 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2360 UNSPEC_VOLATILE_FMFCFG.
2361 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2362 description for fmfcfg and fmfcsr.
2363 (bdesc_1arg): Add fmtcsr.
2364 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2365 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2366 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2367 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2368 unspec_fmfcfg): New patterns.
2369 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2370 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2371 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2372 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2373 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2374 __nds32__fmfcfg): Define.
2376 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2378 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2379 intrinsic register names.
2380 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2381 intrinsic register enum values and macros.
2383 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2385 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2386 for load/store addressing form.
2387 (nds32_print_operand_address): Likewise.
2389 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2392 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2393 based on LABEL_REF. Remove useless assertion.
2394 (pic_address_needs_scratch): Fix formatting.
2395 (sparc_legitimize_pic_address): Minor tweaks.
2396 (sparc_delegitimize_address): Adjust assertion accordingly.
2397 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2398 into symbolic_operand.
2399 (movsi_high_pic_label_ref): Likewise.
2400 (movsi_lo_sum_pic_label_ref): Likewise.
2401 (movdi_pic_label_ref): Likewise.
2402 (movdi_high_pic_label_ref): Likewise.
2403 (movdi_lo_sum_pic_label_ref): Likewise.
2405 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2407 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2408 custom LIB_SPEC setup.
2410 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2411 Kito Cheng <kito.cheng@gmail.com>
2413 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2414 * config/riscv/freebsd.h: New.
2416 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2418 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2419 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2422 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2423 Kito Cheng <kito.cheng@gmail.com>
2425 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2426 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2427 * config/nds32/nds32-protos.h (nds32_output_call,
2428 nds32_output_return): Declare.
2429 * config/nds32/nds32.md: Refine all the call and return patterns.
2431 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2434 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2435 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2437 PR rtl-optimization/84872
2438 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2439 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2442 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2444 * expr.c (copy_blkmode_to_reg): Revert 254862.
2445 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2447 2018-04-06 Richard Biener <rguenther@suse.de>
2450 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2451 after seeing a component reference with an adjacent field. Treat
2452 refs to arrays at struct end of external decls similar to
2453 refs to unconstrained commons.
2455 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2458 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2459 look through SAVE_EXPRs with non-side-effects argument. Adjust
2461 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2464 2018-04-06 Richard Biener <rguenther@suse.de>
2467 * alias.c (find_base_term): New wrapper around find_base_term
2468 unwinding CSELIB_VAL_PTR changes.
2469 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2472 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2474 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2476 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2477 constant definitions.
2478 ("nop"): lr 0,0 -> nopr r0
2479 ("nop_lr0", "nop_lr1"): New insn definitions.
2481 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2483 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2484 NDS32_V3PUSH_AVAILABLE_P macro.
2486 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2487 Chung-Ju Wu <jasonwucj@gmail.com>
2489 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2490 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2491 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2492 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2493 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2494 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2495 * config/nds32/constraints.md: New constraints and checking for hard
2496 float configuration.
2497 * config/nds32/iterators.md: New mode iterator and attribute for hard
2498 float configuration.
2499 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2501 * config/nds32/nds32-fpu.md: New file.
2502 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2503 deal with hard float code generation.
2504 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2506 (abi_type, float_reg_number): New enum type.
2507 * config/nds32/nds32-predicates.c: New predicates for hard float.
2508 * config/nds32/nds32-protos.h: Declare functions for hard float.
2509 * config/nds32/nds32.c: Implementation for hard float configuration.
2510 * config/nds32/nds32.h: Definitions for hard float configuration.
2511 * config/nds32/nds32.md: Include hard float machine description and
2512 modify patterns for hard float configuration.
2513 * config/nds32/nds32.opt: New options for hard float configuration.
2514 * config/nds32/predicates.md: New predicates for hard float
2517 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2519 * common/config/nds32/nds32-common.c
2520 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2522 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2525 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2526 CONSTRUCTOR_ELT (ctor, ...)->value.
2528 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2531 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2533 2018-04-05 Tom de Vries <tom@codesourcery.com>
2536 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2539 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2540 Kito Cheng <kito.cheng@gmail.com>
2542 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2543 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2544 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2545 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2547 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2548 Kito Cheng <kito.cheng@gmail.com>
2550 * config/nds32/constraints.md (Ufe): New memory constraint.
2551 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2552 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2553 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2555 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2556 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2558 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2560 * config/nds32/nds32.md: Use optimize_size in the condition for
2561 alu-shift instructions.
2563 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2565 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2567 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2569 * config/nds32/nds32.md (negsi2): Refine pattern.
2571 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2572 Chung-Ju Wu <jasonwucj@gmail.com>
2574 * config/nds32/iterators.md (shift_rotate): New code iterator.
2575 (shift): New code attribute.
2576 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2577 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2578 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2579 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2580 bit-wise operations.
2581 (andsi3, *andsi3): Ditto.
2582 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2583 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2584 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2585 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2586 nds32_ior_operand, nds32_xor_operand): New predicates.
2588 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2590 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2591 (addsi3, subsi3): ... this.
2593 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2595 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2597 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2599 * config/nds32/nds32.md: Adjust indention.
2601 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2603 * config/nds32/nds32.md (feature): New attribute.
2605 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2607 * config/nds32/nds32.md (subtype): New attribute.
2609 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2612 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2613 expansion to perform a bitwise AND of the argument followed by a
2614 boolean negation of the result.
2616 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2618 PR rtl-optimization/84878
2619 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2620 the basic block. Assert the use reference is not artificial and that
2621 it has an associated insn.
2623 2018-04-04 Michael Matz <matz@suse.de>
2625 * builtins.c (compute_objsize): Pass correct operand
2626 to array_at_struct_end_p.
2628 2018-04-04 Richard Biener <rguenther@suse.de>
2631 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2632 from contexts for DINFO_LEVEL_TERSE and below.
2634 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2636 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2637 resiter_operand condition.
2638 * config/nds32/nds32.md (*move<mode>): Ditto.
2640 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2641 Monk Chiang <sh.chiang04@gmail.com>
2643 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2645 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2647 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2649 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2650 Kito Cheng <kito.cheng@gmail.com>
2652 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2653 nds32_cond_code_str, output_cond_branch,
2654 output_cond_branch_compare_zero, 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: New functions.
2660 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2661 nds32_expand_cstore, nds32_expand_movcc,
2662 nds32_output_cbranchsi4_equality_zero,
2663 nds32_output_cbranchsi4_equality_reg,
2664 nds32_output_cbranchsi4_equality_reg_or_const_int,
2665 nds32_output_cbranchsi4_greater_less_zero): Declare.
2666 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2667 nds32_rimm11s_operand): New predicates.
2668 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2669 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2672 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2674 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2675 * config/nds32/nds32.md: Ditto.
2676 * config/nds32/pipelines.md: Ditto.
2678 2018-04-04 Richard Biener <rguenther@suse.de>
2680 PR tree-optimization/85168
2681 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2682 propagating abnormals.
2684 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2686 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2688 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2689 Kito Cheng <kito.cheng@gmail.com>
2691 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2692 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2693 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2694 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2695 * config/nds32/nds32.md (sibcall_internal): New.
2696 (sibcall_register): Remove.
2697 (sibcall_immediate): Remove.
2698 (sibcall_value_internal): New.
2699 (sibcall_value_register): Remove.
2700 (sibcall_value_immediate): Remove.
2701 * config/nds32/predicates.md (nds32_general_register_operand): New.
2702 (nds32_call_address_operand): New.
2704 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2706 PR rtl-optimization/85167
2707 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2708 bb_defs if *split_p, instead preinitialize it to NULL.
2710 PR tree-optimization/85156
2711 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2712 evaluating the argument multiple times.
2714 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2716 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2718 (_mm_cvtpd_ps): Likewise.
2719 (_mm_cvttpd_epi32): Likewise.
2720 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2721 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2722 vector, pixel, and bool following altivec.h include.
2724 2018-04-03 Martin Sebor <msebor@redhat.com>
2726 * doc/extend.texi (Common Function Attributes): Clarify.
2727 (const attribute): Likewise.
2728 (pure attribute): Likewise.
2730 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2733 * config/i386/i386.c (ix86_expand_vector_set): Use
2734 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2736 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2738 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2739 instructions when changing rounding bits to preserve precision bits
2740 in the x87 control word.
2742 2018-04-03 Martin Liska <mliska@suse.cz>
2744 PR tree-optimization/82491
2745 * rtl.h (strip_offset_and_add): Replace += suboffset with
2746 poly_uint64 () + suboffset.
2748 2018-03-29 Martin Liska <mliska@suse.cz>
2749 Martin Jambor <mjambor@suse.cz>
2752 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2753 param_type is not an integral or pointer type.
2755 2018-04-03 Richard Biener <rguenther@suse.de>
2757 * sese.h (recompute_all_dominators): Remove.
2759 2018-04-02 Martin Sebor <msebor@redhat.com>
2761 * doc/invoke.texi (-Wrestrict): Fix typos.
2763 2018-04-02 Jim Wilson <jimw@sifive.com>
2765 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2766 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2767 (<optab>di3, <optab>si3_extend): Likewise.
2768 (<optab>si3_mask, <optab>si3_mask_1): New.
2769 (<optab>di3_mask, <optab>di3_mask_1): New.
2770 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2771 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2772 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2774 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2776 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2779 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2781 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2782 (nds32_canonicalize_comparison): New function.
2784 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2785 Kito Cheng <kito.cheng@gmail.com>
2786 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2788 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2789 * config/nds32/constants.md (unspec_volatile_element): Add
2790 UNSPEC_VOLATILE_RELAX_GROUP.
2791 * config/nds32/nds32-relax-opt.c: New file.
2792 * config/nds32/nds32-predicates.c
2793 (nds32_symbol_load_store_p): New function.
2794 * config/nds32/nds32-protos.h
2795 (nds32_symbol_load_store_p): Declare function.
2796 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2797 * config/nds32/nds32.c
2798 (nds32_register_pass): New function to register pass.
2799 (nds32_register_passes): New function to register passes.
2800 * config/nds32/nds32.md (relax_group): New pattern.
2801 * config/nds32/nds32.opt (mrelax-hint): New option.
2802 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2804 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2806 * config/nds32/t-nds32: Modify files dependency.
2808 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2810 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2811 (PROFILE_HOOK): Define its implementation.
2813 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2815 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2816 type and 32-bit size.
2818 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2821 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2822 (V_128_256): New mode iterator.
2823 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2824 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2825 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2827 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2828 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2830 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2833 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2834 NaN inputs correctly.
2836 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2839 * config/rs6000/vsx.md (??r): New mode attribute.
2840 (*vsx_mov<mode>_64bit): Use it.
2841 (*vsx_mov<mode>_32bit): Likewise.
2843 2018-03-30 Martin Sebor <msebor@redhat.com>
2845 PR tree-optimization/84818
2846 * builtins.c (check_access): Use warning_n.
2848 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2851 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2853 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2856 2018-03-30 Julia Koval <julia.koval@intel.com>
2859 * x86-tune.def (movx, partial_reg_dependency): Enable for
2862 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2865 * lra-constraints.c (process_alt_operands): Move setting
2866 this_alternative_matches below.
2868 2018-03-29 Martin Liska <mliska@suse.cz>
2871 * doc/invoke.texi: Document how LTO works with debug info.
2872 Describe auto-load support of binutils. Mention 'x86-64'
2873 as valid option value of -march option.
2875 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2877 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2880 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2881 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2882 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2885 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2888 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2889 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2890 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2891 for RS6000_BTM_POWERPC64.
2892 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2893 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2894 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2899 2018-03-28 Carl Love <cel@us.ibm.com>
2902 2017-09-27 Carl Love <cel@us.ibm.com>
2904 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2905 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2906 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2909 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2911 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2912 instead of __vector bool.
2913 (_mm_max_pu8): Likewise.
2914 (_mm_min_pi16): Likewise.
2916 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2919 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2920 (DIVWEUO): Likewise.
2922 (DIVDEUO): Likewise.
2923 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2924 DIVWEUO and DIVDEUO.
2925 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2926 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2927 (div_extend): Likewise.
2928 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2930 (__builtin_divweuo): Likewise.
2931 (__builtin_divdeo): Likewise.
2932 (__builtin_divdeuo): Likewise.
2934 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2937 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2938 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2940 PR tree-optimization/82004
2941 * gimple-match-head.c (optimize_pow_to_exp): New function.
2942 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2943 Don't fold to exp if optimize_pow_to_exp is false.
2945 2018-03-28 Martin Liska <mliska@suse.cz>
2948 * calls.c (initialize_argument_information): Fix trailing space.
2949 * common.opt: Fix typo and provide better explanation for
2950 -fsanitize-coverage option.
2951 * config/i386/i386.opt: Fix typo.
2953 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2954 Martin Liska <mliska@suse.cz>
2957 * gimplify.c (asan_poison_variable): Don't do the check for
2958 gimplify_omp_ctxp here.
2959 (gimplify_decl_expr): Do it here.
2960 (gimplify_target_expr): Likewise.
2962 2018-03-28 Martin Liska <mliska@suse.cz>
2965 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2966 chkp_type_bounds_count if MPX is not enabled.
2968 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2970 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2972 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2975 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2976 function to create the function decl for complex long double
2977 multiply and divide for -mabi=ieeelongdouble.
2978 (init_float128_ieee): Call it.
2980 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2983 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2984 -fcf-protection=branch -mibt.
2985 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2987 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2990 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2992 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2995 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2996 extern array declarations.
2998 2018-03-27 Richard Biener <rguenther@suse.de>
3001 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3002 explicit single_use checks.
3004 2018-03-27 Richard Biener <rguenther@suse.de>
3006 PR tree-optimization/85082
3007 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3010 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3012 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3013 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3014 Turn on fasynchronous-unwind-tables and funwind-tables.
3016 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3019 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3020 (*bmi_blsr_<mode>_ccz): Ditto.
3022 2018-03-26 Tom de Vries <tom@codesourcery.com>
3024 PR tree-optimization/85063
3025 * omp-general.c (offloading_function_p): New function. Factor out
3027 * omp-offload.c (pass_omp_target_link::gate): ... here.
3028 * omp-general.h (offloading_function_p): Declare.
3029 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3030 with attribute omp declare target for offloading functions.
3032 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3034 PR tree-optimization/84005
3035 * tree-data-ref.h (get_base_for_alignment): Declare.
3036 * tree-data-ref.c (get_base_for_alignment_1): New function.
3037 (get_base_for_alignment): Likewise.
3038 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3039 get_base_for_alignment to find a suitable base object, instead
3040 of always using drb->base_address.
3042 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3045 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3046 known size by default.
3048 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3051 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3052 and non BLKmode operands.
3054 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3057 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3058 Clean up attributes.
3060 2018-03-23 Richard Biener <rguenther@suse.de>
3063 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3064 we are going to emit early debug for LTO.
3066 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3069 * function.c (match_asm_constraints_1): Don't optimize if input
3070 doesn't satisfy general_operand predicate for output's mode.
3073 * alias.c (write_dependence_p): Don't require for x_canonicalized
3074 non-VOIDmode if x has VOIDmode.
3077 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3078 just don't try to optimize it rather than assert it never happens.
3080 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3082 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3083 macro expansions for definition of ST_INTERNAL_<mode> and
3084 LD_INTERNAL_<mode> builtins.
3085 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3087 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3089 (altivec_expand_st_builtin): Likewise.
3090 (altivec_expand_builtin): Remove calls to deleted functions.
3091 (rs6000_address_for_altivec): Delete this function.
3092 * config/rs6000/vector.md: Remove expands for
3093 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3095 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3098 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3099 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3100 re-computing once computed.
3101 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3102 (arm_init_machine_status): Initialize
3103 machine->static_chain_stack_bytes.
3105 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3108 * doc/extend.texi: Add four new prototypes for vec_ld.
3109 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3110 definitions for more logical presentation.
3111 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3112 entries for V1TI variants of __builtin_altivec_ld builtin.
3113 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3114 handling of V1TI variant of LVX icode pattern.
3115 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3116 (rs6000_gimple_fold_builtin): Likewise.
3117 (altivec_init_builtins): Add code to define
3118 __builtin_altivec_lvx_v1ti function.
3120 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3123 * function.c (match_asm_constraints_1): Don't do the optimization
3124 if input isn't a REG, SUBREG, MEM or constant.
3126 2018-03-22 Tom de Vries <tom@codesourcery.com>
3128 PR tree-optimization/84956
3129 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3130 bb_has_abnormal_pred.
3132 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3135 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3136 DECL_INITIAL (decl) to decl at the end.
3137 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3140 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3142 * doc/extend.texi (__builtin_tgmath): Document when complex
3143 integer types are treated as _Complex _Float64.
3145 2018-03-21 Tom de Vries <tom@codesourcery.com>
3147 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3149 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3151 PR tree-optimization/84960
3152 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3153 if it is ENTRY block, move them into single succ of ENTRY in that case.
3155 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3157 PR tree-optimization/84811
3158 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3159 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3160 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3161 (known_subrange_p): Update accordingly. Cast each value involved
3162 in the size comparison, rather than casting the result of the
3165 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3167 PR tree-optimization/84982
3168 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3169 by flipping the least significant bit rather than all bits from
3170 bitpos to bitpos + bitsize - 1.
3172 2018-03-21 Nathan Sidwell <nathan@acm.org>
3174 * doc/extend.texi (Deprecated Features): Remove mention of
3175 long-deleted deprecations.
3177 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3180 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3181 * configure: Regenerate.
3183 2018-03-21 Tom de Vries <tom@codesourcery.com>
3185 PR tree-optimization/83126
3186 * tree-parloops.c (num_phis): New function.
3187 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3189 2018-03-21 Nathan Sidwell <nathan@acm.org>
3191 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3192 mention anon-struct/union members and trailing attributes.
3194 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3196 PR tree-optimization/84969
3197 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3198 builtin memset partitions if they set different rhs values.
3200 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3202 PR rtl-optimization/84989
3203 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3204 VEC_DUPLICATE with scalar result mode.
3206 2018-03-21 Martin Liska <mliska@suse.cz>
3209 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3210 not intended return statement.
3212 2018-03-21 Martin Liska <mliska@suse.cz>
3215 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3216 (chkp_find_bound_slots_1): Limit number of iterations.
3218 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3221 * Minor grammar fixes for x86 options.
3223 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3226 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3227 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3229 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3232 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3233 (altivec_lvx_<mode>_1op): Likewise.
3234 (altivec_stvx_<mode>_2op): Likewise.
3235 (altivec_stvx_<mode>_1op): Likewise.
3236 (altivec_lvx_<VM2:mode>): New define_expand.
3237 (altivec_stvx_<VM2:mode>): Likewise.
3238 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3239 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3240 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3241 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3242 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3243 (rs6000_gen_lvx): Likewise.
3244 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3245 (altivec_expand_stv_builtin): Likewise.
3246 (altivec_expand_builtin): Likewise.
3247 * config/rs6000/vector.md: Likewise.
3249 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3252 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3255 2018-03-20 Richard Biener <rguenther@suse.de>
3258 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3259 sign-conversions as zero, fall back to standard scalar_stmt
3262 2018-03-20 Martin Liska <mliska@suse.cz>
3265 * predict.c (rebuild_frequencies): Handle case when we have
3266 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3268 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3271 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3272 flag_section_anchors.
3273 * varasm.c (use_blocks_for_decl_p): Remove hack for
3274 dw2_force_const_mem.
3277 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3279 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3280 be created, use lowpart_subreg of operands[0] rather than operands[0]
3282 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3283 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3284 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3285 and n constraint instead of aarch64_shift_imm_di and Usd.
3286 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3287 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3289 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3292 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3293 to favor GPR over NEON registers.
3294 (<shift>di3_neon): Likewise.
3296 2018-03-20 Tom de Vries <tom@codesourcery.com>
3299 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3300 (nvptx_process_pars): Emit bar.sync asap and alap.
3302 2018-03-20 Tom de Vries <tom@codesourcery.com>
3305 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3306 seen_label if seen_label is already set.
3308 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3311 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3312 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3313 Use 1U instead of 1. Formatting fixes.
3316 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3317 instead of TREE_TYPE (s1) for the return value.
3319 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3321 PR tree-optimization/84946
3322 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3323 bitsize + bitsize in poly_uint64 rather than poly_int64.
3326 * dwarf2asm.c: Include fold-const.c.
3327 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3328 of decl rather than decl itself.
3330 PR rtl-optimization/84643
3331 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3333 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3336 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3337 calling assemble_variable.
3339 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3342 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3343 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3345 2018-03-19 Jim Wilson <jimw@sifive.com>
3348 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3349 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3350 (riscv_first_stack_step): Likewise.
3351 (riscv_option_override): Use STACK_BOUNDARY instead of
3353 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3355 (BIGGEST_ALIGNMENT): Set to 128.
3356 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3357 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3360 2018-03-19 Richard Biener <rguenther@suse.de>
3362 PR tree-optimization/84933
3363 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3364 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3366 2018-03-19 Richard Biener <rguenther@suse.de>
3368 PR tree-optimization/84859
3369 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3370 (cond_if_else_store_replacement): Perform sinking operation on
3371 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3372 Generalize what a BB with a single eligible store is.
3374 2018-03-19 Richard Biener <rguenther@suse.de>
3376 PR tree-optimization/84929
3377 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3378 chrec_is_positive against non-chrec arg.
3380 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3383 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3385 2018-03-18 Martin Liska <mliska@suse.cz>
3387 PR rtl-optimization/84635
3388 * regrename.c (build_def_use): Use matches_mode only when
3391 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3393 PR tree-optimization/84913
3394 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3395 vectorize chains of COND_EXPRs.
3397 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3399 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3401 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3403 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3405 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3407 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3409 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3410 Kito Cheng <kito.cheng@gmail.com>
3412 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3413 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3414 (nds32_adjust_reg_alloc_order): New function.
3415 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3417 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3419 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3420 nds32_print_operand, nds32_print_operand_address): Use
3421 HOST_WIDE_INT_PRINT_DEC instead.
3423 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3425 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3427 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3430 * config/i386/i386.c (initial_ix86_tune_features,
3431 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3433 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3434 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3435 rather than 1u << ix86_tune. Formatting fix.
3436 (ix86_option_override_internal): Change ix86_arch_mask from
3437 unsigned int to unsigned HOST_WIDE_INT, initialize to
3438 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3439 (ix86_function_specific_restore): Likewise.
3441 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3444 * postreload.c (reload_combine_recognize_pattern): Perform
3445 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3446 truncate_int_for_mode the result for the destination's mode.
3449 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3450 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3452 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3455 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3458 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3459 regno_allocno_class_array and sorted_pseudos.
3460 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3461 insns where regno is used.
3463 2018-03-16 Martin Liska <mliska@suse.cz>
3466 * multiple_target.c (create_dispatcher_calls): Redirect
3467 reference in the symbol table.
3469 2018-03-16 Martin Liska <mliska@suse.cz>
3472 * multiple_target.c (create_dispatcher_calls): Redirect also
3475 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3479 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3480 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3481 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3483 2018-03-16 Julia Koval <julia.koval@intel.com>
3485 * doc/invoke.texi (Skylake Server): Add CLWB.
3486 Cannonlake): Remove CLWB.
3488 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3490 PR tree-optimization/84841
3491 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3493 (FLOAT_ONE_CONST_TYPE): Define.
3494 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3495 (sort_by_operand_rank): Put entries with higher constant_type last
3496 rather than first to match comments.
3498 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3500 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3503 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3506 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3507 check and use address of target always.
3509 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3512 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3513 (indirect_thunk_bnd_needed): Likewise.
3514 (indirect_thunks_used): Likewise.
3515 (indirect_thunks_bnd_used): Likewise.
3516 (indirect_return_needed): New.
3517 (indirect_return_bnd_needed): Likewise.
3518 (output_indirect_thunk_function): Add a bool argument for
3520 (output_indirect_thunk_function): Don't generate alias for
3521 function return thunk.
3522 (ix86_code_end): Call output_indirect_thunk_function to generate
3523 function return thunks.
3524 (ix86_output_function_return): Set indirect_return_bnd_needed
3525 and indirect_return_needed instead of indirect_thunk_bnd_needed
3526 and indirect_thunk_needed.
3528 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3530 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3531 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3532 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3534 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3535 Paul Hua <paul.hua.gm@gmail.com>
3538 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3540 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3542 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3543 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3546 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3549 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3550 instead of GET_MODE_SIZE when comparing Units.
3552 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3555 * varasm.c (hash_section): Return an unchangeble hash value
3556 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3557 Return !aarch64_can_use_per_function_literal_pools_p ().
3559 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3562 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3563 rather than address of cmode as last argument to prepare_cmp_insn.
3565 2018-03-15 Julia Koval <julia.koval@intel.com>
3567 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3568 F_AVX512VNNI, F_AVX512BITALG): New.
3570 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3573 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3574 insn for floating-point loads and stores.
3576 2018-03-14 Carl Love <cel@us.ibm.com>
3578 * config/rs6000/rs6000-c.c: Add macro definitions for
3579 ALTIVEC_BUILTIN_VEC_PERMXOR.
3580 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3581 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3582 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3583 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3585 * config/doc/extend.texi: Add prototypes for vec_permxor.
3587 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3590 * diagnostic-show-locus.c (class layout_point): Convert m_line
3591 from int to linenum_type.
3592 (line_span::comparator): Use linenum "compare" function when
3593 comparing line numbers.
3594 (test_line_span): New function.
3595 (layout_range::contains_point): Convert param "row" from int to
3597 (layout_range::intersects_line_p): Likewise.
3598 (layout::will_show_line_p): Likewise.
3599 (layout::print_source_line): Likewise.
3600 (layout::should_print_annotation_line_p): Likewise.
3601 (layout::print_annotation_line): Likewise.
3602 (layout::print_leading_fixits): Likewise.
3603 (layout::annotation_line_showed_range_p): Likewise.
3604 (struct line_corrections): Likewise for field m_row.
3605 (line_corrections::line_corrections): Likewise for param "row".
3606 (layout::print_trailing_fixits): Likewise.
3607 (layout::get_state_at_point): Likewise.
3608 (layout::get_x_bound_for_row): Likewise.
3609 (layout::print_line): Likewise.
3610 (diagnostic_show_locus): Likewise for locals "last_line" and
3612 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3613 * input.c (selftest::test_linenum_comparisons): New function.
3614 (selftest::input_c_tests): Call it.
3615 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3616 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3617 * selftest.h (ASSERT_GT): New macro.
3618 (ASSERT_GT_AT): New macro.
3619 (ASSERT_LT): New macro.
3620 (ASSERT_LT_AT): New macro.
3622 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3624 PR rtl-optimization/84780
3625 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3627 2018-03-14 Martin Liska <mliska@suse.cz>
3629 * tree.c (record_node_allocation_statistics): Use
3630 get_stats_node_kind.
3631 (get_stats_node_kind): New function extracted from
3632 record_node_allocation_statistics.
3633 (free_node): Use get_stats_node_kind.
3635 2018-03-14 Richard Biener <rguenther@suse.de>
3637 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3638 that the value-set of ANTIC_IN doesn't grow.
3641 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3643 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3644 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3646 2018-03-14 Julia Koval <julia.koval@intel.com>
3648 * config.gcc (icelake-client, icelake-server): New.
3650 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3651 (initial_ix86_arch_features): Ditto.
3652 (PTA_SKYLAKE): Add SGX.
3653 (PTA_ICELAKE): Remove.
3654 (PTA_ICELAKE_CLIENT): New.
3655 (PTA_ICELAKE_SERVER): New.
3656 (ix86_option_override_internal): Split up icelake on icelake client and
3658 (get_builtin_code_for_version): Ditto.
3659 (fold_builtin_cpu): Ditto.
3660 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3661 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3662 * config/i386/i386.h (processor_type): Ditto.
3663 * doc/invoke.texi: Ditto.
3665 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3668 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3669 INTEGER_CST offset, add it together with bitpos / 8 and
3670 sign extend based on POINTER_SIZE.
3674 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3677 * config/i386/constraints.md (Yc): New register constraint.
3678 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3679 Use Yc constraint for alternative 2 of operand 0. Remove
3680 preferred_for_speed attribute.
3682 2018-03-14 Richard Biener <rguenther@suse.de>
3684 PR tree-optimization/84830
3685 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3686 with the old one to avoid oscillations.
3688 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3691 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3693 (assign_by_spills): Return a flag of reload assignment failure.
3694 Do not process the reload assignment failures. Do not spill other
3695 reload pseudos if they has the same reg class. Update n if
3697 (lra_assign): Add a return arg. Set up from the result of
3698 assign_by_spills call.
3699 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3700 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3701 usage_insns if it is not NULL.
3702 (spill_hard_reg_in_range): New function.
3703 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3704 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3705 function prototypes.
3706 (lra_assign): Change prototype.
3707 * lra.c (lra): Add code to deal with fails by splitting hard reg
3710 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3712 * config/riscv/riscv.opt (mrelax): New option.
3713 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3714 "norelax" when riscv_mrelax is disabled.
3715 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3717 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3720 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3721 reassociation for int modes.
3723 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3725 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3726 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3728 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3729 * config/aarch64/aarch64-sve.md
3730 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3731 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3732 (*extend<mode><Vwide>2): Rename to...
3733 (aarch64_sve_extend<mode><Vwide>2): ...this.
3734 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3735 renaming the old pattern to...
3736 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3738 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3739 define_expand, renaming the old pattern to...
3740 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3741 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3742 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3743 account when deciding which SVE instruction the optab should use.
3744 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3746 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3748 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3749 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3750 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3751 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3752 (tlsdesc_small_<mode>): Turn a define_expand and use
3753 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3754 (tlsdesc_small_advsimd_<mode>): ...this.
3755 (tlsdesc_small_sve_<mode>): New pattern.
3757 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3759 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3760 (UNSPEC_UMUL_HIGHPART): New constants.
3761 (MUL_HIGHPART): New int iteraor.
3762 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3763 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3765 (*<su>mul<mode>3_highpart): New define_insn.
3767 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3770 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3773 2018-03-13 Martin Liska <mliska@suse.cz>
3776 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3778 (sem_item_optimizer::~sem_item_optimizer): Release it.
3779 (sem_item_optimizer::merge_classes): Register variable aliases.
3780 (sem_item_optimizer::fixup_pt_set): New function.
3781 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3782 * ipa-icf.h: Declare new variables and functions.
3784 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3787 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3788 integer_pow2p@2 and test integer_pow2p in condition.
3789 (A < 0 ? C : 0): Similarly for @1.
3792 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3793 characters starting at p contain '\0' character, don't look beyond
3797 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3798 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3801 * reg-stack.c (change_stack): Change update_end var from int to
3802 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3803 also call set_block_for_insn on the newly added insns and rescan.
3806 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3807 on the last operand.
3810 * tree.c (stabilize_reference_1): Return save_expr (e) for
3811 STATEMENT_LIST even if it doesn't have side-effects.
3813 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3815 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3817 2018-03-12 Renlin Li <renlin.li@arm.com>
3819 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3820 aarch64_output_scalar_simd_mov_immediate.
3822 2018-03-12 Martin Sebor <msebor@redhat.com>
3824 PR tree-optimization/83456
3825 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3826 for perfectly overlapping calls to memcpy.
3827 (gimple_fold_builtin_memory_chk): Same.
3828 (gimple_fold_builtin_strcpy): Handle no-warning.
3829 (gimple_fold_builtin_stxcpy_chk): Same.
3830 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3832 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3834 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3835 parameter. Use it for SFmode.
3836 (rs6000_function_arg_advance_1): Adjust.
3837 (rs6000_function_arg): Adjust.
3838 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3840 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3842 PR rtl-optimization/84169
3843 PR rtl-optimization/84780
3844 * combine.c (can_combine_p): Check for a 2-insn combination whether
3845 the destination register is used between the two insns, too.
3847 2018-03-12 Richard Biener <rguenther@suse.de>
3849 PR tree-optimization/84803
3850 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3851 for refs DR analysis didn't process.
3853 2018-03-12 Richard Biener <rguenther@suse.de>
3855 PR tree-optimization/84777
3856 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3857 force-vectorize loops ignore whether we are optimizing for size.
3859 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3861 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3862 (TARGET_MD_ASM_ADJUST): Define.
3864 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3865 Kito Cheng <kito.cheng@gmail.com>
3866 Chung-Ju Wu <jasonwucj@gmail.com>
3868 * config/nds32/nds32.c (nds32_compute_stack_frame,
3869 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3870 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3871 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3872 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3873 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3874 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3875 * config/nds32/nds32.md (prologue, epilogue): Use macro
3876 NDS32_V3PUSH_AVAILABLE_P to do checking.
3878 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3881 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3882 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3883 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3884 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3885 addition of most attributes on !orig_type_die or the attribute not
3886 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3888 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3889 Chung-Ju Wu <jasonwucj@gmail.com>
3891 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3893 * config/nds32/nds32.opt (mvh): New option.
3895 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3896 Chung-Ju Wu <jasonwucj@gmail.com>
3898 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3900 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3901 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3904 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3905 Chung-Ju Wu <jasonwucj@gmail.com>
3907 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3909 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3910 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3912 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3913 Kito Cheng <kito.cheng@gmail.com>
3914 Chung-Ju Wu <jasonwucj@gmail.com>
3916 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3917 UNSPEC_FFMISM and UNSPEC_FLMISM.
3918 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3919 for ffb, ffmism and flmism.
3920 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3921 (unspec_ffmism): Ditto.
3922 (unspec_flmism): Ditto.
3923 (nds32_expand_builtin_impl): Check if string extension is available.
3924 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3925 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3927 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3930 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3933 * lra-assigns.c (assign_by_spills): Return a flag of reload
3934 assignment failure. Do not process the reload assignment
3935 failures. Do not spill other reload pseudos if they has the same
3937 (lra_assign): Add a return arg. Set up from the result of
3938 assign_by_spills call.
3939 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3940 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3941 usage_insns if it is not NULL.
3942 (spill_hard_reg_in_range): New function.
3943 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3944 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3945 function prototypes.
3946 (lra_assign): Change prototype.
3947 * lra.c (lra): Add code to deal with fails by splitting hard reg
3950 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3953 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3955 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3958 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3959 (dw_val_node): Add val_symbolic_view.
3960 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3961 (symview_upper_bound): New.
3962 (new_line_info_table): Initialize symviews_since_reset.
3963 (dwarf2out_source_line): Count symviews_since_reset and set
3964 symview_upper_bound.
3965 (dw_val_equal_p): Handle symview.
3966 (add_AT_symview): New.
3967 (print_dw_val): Handle symview.
3968 (attr_checksum, attr_checksum_ordered): Likewise.
3969 (same_dw_val_p, size_of_die): Likewise.
3970 (value_format, output_die): Likewise.
3971 (add_high_low_attributes): Use add_AT_symview for entry_view.
3972 (dwarf2out_finish): Reset symview_upper_bound, clear
3975 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3978 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3979 Add strict argument and use it.
3980 (rs6000_split_multireg_move): Update for new strict argument.
3981 (mem_operand_gpr): Disallow all non-offsettable addresses.
3982 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3984 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3987 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3988 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3989 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3992 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3993 decl, use remap_type if we want to use the type.
3995 2018-03-09 Martin Sebor <msebor@redhat.com>
3997 PR tree-optimization/84526
3998 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4000 (builtin_access::generic_overlap): Be prepared to handle non-array
4003 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4005 PR rtl-optimization/84682
4006 * lra-constraints.c (process_address_1): Check is_address flag
4007 for address constraints.
4008 (process_alt_operands): Likewise.
4009 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4010 preprocess_constraints.
4011 * recog.h (preprocess_constraints): Add oploc parameter.
4013 * recog.c (preprocess_constraints): Test address_operand for
4014 CT_ADDRESS constraints.
4016 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4019 * lra-assigns.c (assign_by_spills): Return a flag of reload
4020 assignment failure. Do not process the reload assignment
4021 failures. Do not spill other reload pseudos if they has the same
4023 (lra_assign): Add a return arg. Set up from the result of
4024 assign_by_spills call.
4025 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4026 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4027 usage_insns if it is not NULL.
4028 (spill_hard_reg_in_range): New function.
4029 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4030 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4031 function prototypes.
4032 (lra_assign): Change prototype.
4033 * lra.c (lra): Add code to deal with fails by splitting hard reg
4036 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4039 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4040 Accept complain bool parameter. Only emit errors if it is true.
4041 (arm_parse_cpu_option_name): Likewise.
4042 (arm_target_thumb_only): Adjust callers of the above.
4043 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4044 prototype to take a default true bool parameter.
4045 (arm_parse_arch_option_name): Likewise.
4047 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4048 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4052 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4053 * configure: Regenerate.
4054 * configure.ac ("linker --version-script option"): New.
4055 ("linker soname option"): New.
4057 2018-03-09 Richard Biener <rguenther@suse.de>
4059 PR tree-optimization/84775
4060 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4061 immediate uses of predicate stmts and mark them modified.
4064 PR tree-optimization/84178
4065 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4067 (version_loop_for_if_conversion): Delay update_ssa call.
4068 (tree_if_conversion): Delay update_ssa until after predicate
4071 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4074 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4075 when the function accesses prior frames.
4077 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4080 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4081 gen_llsym, otherwise call maybe_gen_llsym.
4084 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4085 has ',' character inside of it.
4087 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4090 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4091 as clobbering CC_REGNUM.
4093 2018-03-08 Richard Biener <rguenther@suse.de>
4096 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4097 (follow_copies_to_constant): Do not follow SSA names registered
4100 2018-03-08 Richard Biener <rguenther@suse.de>
4102 PR tree-optimization/84178
4103 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4105 (version_loop_for_if_conversion): Delay update_ssa call.
4106 (tree_if_conversion): Delay update_ssa until after predicate
4109 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4111 PR tree-optimization/84178
4112 * tree-if-conv.c (release_bb_predicate): Remove the
4113 the assertion that the stmts have NULL use_ops.
4114 Discard the statements, asserting that they haven't
4115 yet been added to a BB.
4117 2018-03-08 Richard Biener <rguenther@suse.de>
4119 PR tree-optimization/84746
4120 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4121 (phi_translate): Pass in destination ANTIC_OUT set.
4122 (phi_translate_1): Likewise. For a simplified result lookup
4123 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4124 (phi_translate_set): Adjust.
4125 (do_pre_regular_insertion): Likewise.
4126 (do_pre_partial_partial_insertion): Likewise.
4128 2018-03-08 Martin Liska <mliska@suse.cz>
4130 PR gcov-profile/84735
4131 * doc/gcov.texi: Document usage of profile files.
4132 * gcov-io.h: Document changes in the format.
4134 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4138 * dwarf2out.c (struct dw_line_info_table): Update comments for
4140 (FORCE_RESET_NEXT_VIEW): New.
4141 (FORCE_RESETTING_VIEW_P): New.
4142 (RESETTING_VIEW_P): Check for -1 too.
4143 (ZERO_VIEW_P): Likewise.
4144 (new_line_info_table): Force-reset next view.
4145 (dwarf2out_begin_function): Likewise.
4146 (dwarf2out_source_line): Simplify zero_view_p initialization.
4147 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4148 view directly. Omit view when omitting .loc at line 0.
4150 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4152 PR tree-optimization/84740
4153 * tree-switch-conversion.c (process_switch): Call build_constructors
4154 only if info.phi_count is non-zero.
4156 PR tree-optimization/84739
4157 * tree-tailcall.c (find_tail_calls): Check call arguments against
4158 DECL_ARGUMENTS (current_function_decl) rather than
4159 DECL_ARGUMENTS (func) when checking for tail recursion.
4161 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4163 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4164 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4165 Volker Reichelt's entry and add entries for people that perform
4166 GCC fuzzy testing and report numerous bugs.
4168 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4171 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4172 readonly data in sdata, if that is disabled.
4173 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4174 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4175 -mreadonly-in-sdata option.
4177 2018-03-07 Martin Sebor <msebor@redhat.com>
4179 PR tree-optimization/84468
4180 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4181 basic block when looking for nul assignment.
4183 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4186 * except.h (output_function_exception_table): Adjust prototype.
4187 * except.c (output_function_exception_table): Remove FNNAME parameter
4188 and add SECTION parameter. Ouput one part of the table at a time.
4189 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4190 the first part of the exception table and emit unwind directives.
4191 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4192 (i386_pe_seh_cold_init): Likewise.
4193 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4194 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4195 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4196 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4197 * config/i386/winnt.c: Include except.h.
4198 (struct seh_frame_state): Add reg_offset, after_prologue and
4199 in_cold_section fields.
4200 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4201 (i386_pe_seh_cold_init): New function.
4202 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4203 to seh->in_cold_section.
4204 (seh_emit_push): Record the offset of the push.
4205 (seh_emit_save): Record the offet of the save.
4206 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4207 Test seh->after_prologue to disregard the epilogue.
4208 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4209 (i386_pe_end_cold_function): New function.
4211 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4214 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4215 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4218 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4220 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4221 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4224 * multiple_target.c: Include tree-inline.h and intl.h.
4225 (expand_target_clones): Diagnose and fail if node->definition and
4226 !tree_versionable_function_p (node->decl).
4228 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4230 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4232 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4233 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4234 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4236 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4239 * combine.c (try_combine): Use reg_or_subregno instead of handling
4240 just paradoxical SUBREGs and REGs.
4242 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4244 * config/arc/arc.c (arc_finalize_pic): Remove function.
4245 (arc_must_save_register): We use single base PIC register, remove
4246 checks to save/restore the PIC register.
4247 (arc_expand_prologue): Likewise.
4248 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4250 (arc_verify_short): Likewise.
4251 (arc_attr_type): Likewise.
4252 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4253 (walk_stores): Likewise.
4254 (arc_address_cost): Make it static.
4255 (arc_verify_short): Likewise.
4256 (branch_dest): Likewise.
4257 (arc_attr_type): Likewise.
4258 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4259 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4260 (arc_final_prescan_insn): Remove inserting the nops due to
4261 hardware hazards. It is done in reorg step.
4262 (insn_length_variant_t): Remove.
4263 (insn_length_parameters_t): Likewise.
4264 (arc_insn_length_parameters): Likewise.
4265 (arc_get_insn_variants): Likewise.
4266 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4268 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4271 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4274 PR tree-optimization/84687
4275 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4277 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4279 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4281 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4282 Rename to ppc_speculation_barrier.
4283 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4284 __builtin_ppc_speculation_barrier.
4286 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4289 * combine.c (combine_simplify_rtx): Don't try to simplify if
4290 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4293 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4295 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4296 to 32 bytes when compiling for POWER9.
4298 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4301 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4302 regparm >= 3 with no arg reg available also for calls with
4303 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4306 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4308 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4310 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4313 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4315 2018-03-05 Richard Biener <rguenther@suse.de>
4317 PR tree-optimization/84486
4318 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4319 When inserting a __builtin_assume_aligned call set the LHS
4320 SSA name alignment info accordingly.
4322 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4324 PR tree-optimization/84114
4325 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4326 Avoid reassociation of FLOAT_MODE addition.
4328 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4330 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4331 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4332 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4333 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4334 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4335 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4336 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4338 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4339 __builtin_ia32_wbinvd): New builtins.
4340 (SPECIAL_ARGS2): New.
4341 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4342 (SPECIAL_ARGS2): New.
4343 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4344 (ix86_valid_target_attribute_inner_p): Ditto.
4345 (ix86_init_mmx_sse_builtins): Add special_args2.
4346 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4347 TARGET_WBNOINVD_P): New.
4348 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4349 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4350 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4351 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4352 * config/i386/pconfigintrin.h: New file.
4353 * config/i386/wbnoinvdintrin.h: Ditto.
4354 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4355 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4357 2018-03-05 Richard Biener <rguenther@suse.de>
4359 PR tree-optimization/84670
4360 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4362 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4363 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4364 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4365 doesn't grow if all successors (recursively) were visited at least
4368 2018-03-05 Richard Biener <rguenther@suse.de>
4370 PR tree-optimization/84650
4371 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4372 if executed in the loop pipeline.
4374 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4376 * doc/configfiles.texi (Configuration Files): Move info about
4377 conditionalizing $target-protos.h to...
4378 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4379 differs from $target-protos.h.
4381 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4382 Chung-Ju Wu <jasonwucj@gmail.com>
4384 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4385 * config/nds32/nds32-multiple.md (setmemsi): Define.
4386 * config/nds32/nds32-memory-manipulation.c
4387 (nds32_gen_dup_4_byte_to_word_value): New.
4388 (emit_setmem_word_loop): New.
4389 (emit_setmem_byte_loop): New.
4390 (nds32_expand_setmem_loop): New.
4391 (nds32_expand_setmem_loop_v3m): New.
4392 (nds32_expand_setmem_unroll): New.
4393 (nds32_expand_setmem): New.
4395 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4396 Chung-Ju Wu <jasonwucj@gmail.com>
4398 * config/nds32/nds32-memory-manipulation.c
4399 (nds32_emit_load_store): New.
4400 (nds32_emit_post_inc_load_store): New.
4401 (nds32_emit_mem_move): New.
4402 (nds32_emit_mem_move_block): New.
4403 (nds32_expand_movmemsi_loop_unknown_size): New.
4404 (nds32_expand_movmemsi_loop_known_size): New.
4405 (nds32_expand_movmemsi_loop): New.
4406 (nds32_expand_movmemsi_unroll): New.
4407 (nds32_expand_movmemqi): Rename ...
4408 (nds32_expand_movmemsi): ... to this.
4409 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4410 (movmemsi): ... to this.
4411 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4412 (nds32_expand_movmemsi): ... to this.
4414 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4415 Monk Chiang <sh.chiang04@gmail.com>
4416 Chung-Ju Wu <jasonwucj@gmail.com>
4418 * config/nds32/nds32-protos.h
4419 (nds32_expand_load_multiple): New arguments.
4420 (nds32_expand_store_multiple): Ditto.
4421 (nds32_valid_multiple_load_store): Rename ...
4422 (nds32_valid_multiple_load_store_p): ... to this.
4423 * config/nds32/nds32-memory-manipulation.c
4424 (nds32_expand_load_multiple): Refine implementation.
4425 (nds32_expand_store_multiple): Ditto.
4426 * config/nds32/nds32-multiple.md
4427 (load_multiple): Update nds32_expand_load_multiple interface.
4428 (store_multiple): Update nds32_expand_store_multiple interface.
4429 * config/nds32/nds32-predicates.c
4430 (nds32_valid_multiple_load_store): Rename ...
4431 (nds32_valid_multiple_load_store_p): ... to this and refine
4433 * config/nds32/predicates.md
4434 (nds32_load_multiple_and_update_address_operation): New predicate.
4435 (nds32_store_multiple_and_update_address_operation): New predicate.
4437 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4438 Chung-Ju Wu <jasonwucj@gmail.com>
4440 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4441 (combo): New attribute.
4442 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4444 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4446 * config/nds32/nds32.opt: Change -mcmodel= default value.
4448 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4449 Monk Chiang <sh.chiang04@gmail.com>
4450 Chung-Ju Wu <jasonwucj@gmail.com>
4452 * config/nds32/constants.md (unspec_element): New enum.
4453 * config/nds32/constraints.md (Umw): New constraint.
4454 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4455 * config/nds32/nds32-intrinsic.md: Likewise.
4456 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4457 (nds32_valid_smw_lwm_base_p): New.
4458 (nds32_output_smw_single_word): New.
4459 (nds32_output_lmw_single_word): New.
4460 (nds32_expand_unaligned_load): New.
4461 (nds32_expand_unaligned_store): New.
4462 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4463 (nds32_output_smw_single_word): Declare.
4464 (nds32_output_lmw_single_word): Declare.
4465 (nds32_expand_unaligned_load): Declare.
4466 (nds32_expand_unaligned_store): Declare.
4467 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4468 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4469 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4470 NDS32_BUILTIN_UASTORE_DW.
4471 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4474 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4475 Kito Cheng <kito.cheng@gmail.com>
4476 Chung-Ju Wu <jasonwucj@gmail.com>
4478 * config/nds32/nds32-intrinsic.c
4479 (nds32_expand_builtin_null_ftype_reg): Delete.
4480 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4481 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4482 (nds32_read_argument): New.
4483 (nds32_legitimize_target): Ditto.
4484 (nds32_legitimize_argument): Ditto.
4485 (nds32_check_constant_argument): Ditto.
4486 (nds32_expand_unop_builtin): Ditto.
4487 (nds32_expand_unopimm_builtin): Ditto.
4488 (nds32_expand_binop_builtin): Ditto.
4489 (nds32_builtin_decl_impl): Ditto.
4490 (builtin_description): Ditto.
4491 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4492 (nds32_init_builtins_impl): Ditto.
4493 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4494 (nds32_builtin_decl): New.
4495 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4496 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4498 2018-03-02 Jeff Law <law@redhat.com>
4500 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4501 (redundant_insn, fill_simple_delay_slots): Likewise.
4502 (fill_slots_from_thread): Likewise.
4503 * resource.c (mark_referenced_resources): Likewise.
4504 (mark_set_resources, find_dead_or_set_registers): Likewise.
4506 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4508 * substring-locations.h (format_warning_va): Formatting fix for
4510 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4512 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4513 * substring-locations.c: Include intl.h.
4514 (format_warning_va): Turned into small wrapper around
4515 format_warning_n_va, renamed to ...
4516 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4517 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4519 (format_warning_at_substring_n): New function.
4520 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4521 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4522 format_warning_at_substring with just a shorter name instead of
4523 const function pointer.
4524 (fmtwarn_n): New function.
4525 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4526 appropriate, get rid of all the fmtstr temporaries, move conditionals
4527 with G_() wrapped string literals directly into fmtwarn arguments,
4528 cast dir.len to (int), formatting fixes.
4530 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4532 * doc/invoke.texi: Remove "Cilk Plus" references.
4534 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4535 Richard Biener <rguenther@suse.de>
4538 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4539 for error or warning attributes if CALL_FROM_THUNK_P is set.
4542 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4545 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4546 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4549 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4550 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4552 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4553 instead of -1U in last predictors element's probability member.
4555 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4558 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4559 arguments if they are comparable.
4561 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4563 PR tree-optimization/84634
4564 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4565 masks and masked_loop_p with a single loop_masks, making sure it's
4566 null for bb vectorization.
4568 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4570 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4571 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4572 loop->force_vectorize to check whether there is no alias.
4574 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4577 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4579 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4581 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4582 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4583 instead of a loop around prev_real_insn.
4584 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4588 * config/i386/i386.c (ix86_print_operand): Use conditional
4589 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4592 2018-03-02 Richard Biener <rguenther@suse.de>
4594 PR tree-optimization/84427
4595 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4596 (bitmap_set_subtract_values): Rewrite to handle multiple
4599 (prune_clobbered_mems): Likewise.
4600 (phi_translate): Take edge instead of pred/phiblock.
4601 (phi_translate_1): Likewise.
4602 (phi_translate_set): Likewise. Insert all translated
4603 exprs for a value into the set, keeping possibly multiple
4604 expressions per value.
4605 (compute_antic_aux): Adjust for phi_translate changes.
4606 When intersecting union the expressions and prune those
4607 not in the final value set, keeping possibly multiple
4608 expressions per value. Do not use value-insertion
4609 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4610 all expressions. Add verification that the value-sets
4611 only shrink during iteration.
4612 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4613 (do_pre_regular_insertion): Likewise.
4614 (do_pre_partial_partial_insertion): Likewise.
4616 2018-03-02 Richard Biener <rguenther@suse.de>
4619 * config/darwin.c (saved_debug_info_level): New static global.
4620 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4621 (darwin_asm_lto_end): Restore debug info generation settings.
4623 2018-03-01 Martin Liska <mliska@suse.cz>
4626 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4629 2018-03-01 Richard Biener <rguenther@suse.de>
4632 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4633 type annotation with LTO when debug was disabled at compile-time.
4635 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4637 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4639 (mips_final_postscan_insn): Likewise.
4641 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4643 PR rtl-optimization/84528
4644 * alias.c (init_alias_target): Add commentary.
4645 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4646 a unique base value if the frame pointer is not eliminated
4647 to the stack pointer.
4649 2018-03-01 Tom de Vries <tom@codesourcery.com>
4651 PR rtl-optimization/83327
4652 * lra-int.h (hard_regs_spilled_into): Declare.
4653 * lra.c (hard_regs_spilled_into): Define.
4654 (init_reg_info): Init hard_regs_spilled_into.
4655 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4656 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4657 (process_bb_lives): Handle hard_regs_spilled_into.
4658 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4659 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4661 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4663 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4664 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4665 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4666 * config/rs6000/aix72.h: New file.
4668 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4670 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4671 instead of warning_at with conditional singular and plural messages
4675 * stor-layout.c (update_alignment_for_field): For
4676 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4677 && !DECL_PACKED (field), do the alignment update, just use
4678 only desired_align instead of MAX (type_align, desired_align)
4680 (place_field): Don't do known_align < desired_align handling
4681 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4682 is non-NULL, instead do it after rli->prev_field handling and
4683 only if not within a bitfield word. For DECL_PACKED (field)
4684 use type_align of BITS_PER_UNIT.
4686 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4688 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4689 superfluous parentheses and trailing spaces.
4691 2018-02-28 Richard Biener <rguenther@suse.de>
4693 PR tree-optimization/84584
4694 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4695 SCoPs with fake exit edge.
4697 2018-02-28 Martin Liska <mliska@suse.cz>
4700 * timevar.c (timer::print): Fix format to properly print 100%
4703 2018-02-28 Richard Biener <rguenther@suse.de>
4706 * genmatch.c (capture_info::walk_match): Do not mark
4707 captured expressions without operands as expr_p given
4708 they act more like predicates and should be subject to
4709 "lost tail" side-effect preserving.
4711 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4713 PR rtl-optimization/81611
4714 * auto-inc-dec.c (attempt_change): Move dead note from
4715 mem_insn if it's the next use of regno
4716 (find_address): Take address use of reg holding
4717 non-incremented value. Add parm to limit search to the named
4719 (merge_in_block): Attempt to use a mem insn that is the next
4720 use of the original regno.
4722 2018-02-27 Martin Sebor <msebor@redhat.com>
4725 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4726 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4728 2018-02-27 Martin Sebor <msebor@redhat.com>
4730 PR translation/84207
4731 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4732 n argument to unsigned HOST_WIDE_INT.
4733 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4734 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4735 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4736 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4738 2018-02-27 Richard Biener <rguenther@suse.de>
4740 PR tree-optimization/84512
4741 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4742 Do not use the estimate returned from record_stmt_cost for
4743 the scalar iteration cost but sum properly using add_stmt_cost.
4745 2018-02-27 Richard Biener <rguenther@suse.de>
4747 PR tree-optimization/84466
4748 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4749 Adjust last change to less strictly validate use operands.
4751 2018-02-27 Martin Liska <mliska@suse.cz>
4753 PR gcov-profile/84548
4754 * gcov.c (process_file): Allow partial overlap and consider it
4755 also as group functions.
4756 (output_lines): Properly calculate range of lines for a group.
4758 2018-02-27 Martin Liska <mliska@suse.cz>
4760 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4761 'ggc' suffixes. Change first column width.
4762 (timer::print): Fix formatting of the column.
4764 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4766 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4767 preserve inline entry blocks for the sake of debug inline
4768 entry point markers alone.
4769 (remove_unused_locals): Suggest in comments a better place to
4770 force the preservation of inline entry blocks that are
4771 otherwise unused, but do not preserve them.
4773 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4775 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4777 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4780 * config/i386/constraints.md (Bs): Replace
4781 ix86_indirect_branch_register with
4782 TARGET_INDIRECT_BRANCH_REGISTER.
4784 * config/i386/i386.md (indirect_jump): Likewise.
4785 (tablejump): Likewise.
4786 (*sibcall_memory): Likewise.
4787 (*sibcall_value_memory): Likewise.
4788 Peepholes of indirect call and jump via memory: Likewise.
4789 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4790 (*sibcall_value_GOT_32): Likewise.
4791 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4792 (GOT_memory_operand): Likewise.
4793 (call_insn_operand): Likewise.
4794 (sibcall_insn_operand): Likewise.
4795 (GOT32_symbol_operand): Likewise.
4796 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4798 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4800 PR rtl-optimization/83496
4801 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4802 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4803 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4804 redundant insn, if any.
4805 (relax_delay_slots): Likewise.
4806 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4808 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4810 PR tree-optimization/83965
4811 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4812 that grouped statements are part of a reduction chain. Return
4813 true if the statement is not marked as a reduction itself but
4815 (vect_recog_dot_prod_pattern): Don't check whether the statement
4816 is part of a group here.
4817 (vect_recog_sad_pattern): Likewise.
4818 (vect_recog_widen_sum_pattern): Likewise.
4820 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4823 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4825 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4828 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4830 (ix86_output_indirect_function_return): New prototype.
4831 (ix86_split_simple_return_pop_internal): Likewise.
4832 * config/i386/i386.c (indirect_return_via_cx): New.
4833 (indirect_return_via_cx_bnd): Likewise.
4834 (indirect_thunk_name): Handle return va CX_REG.
4835 (output_indirect_thunk_function): Create alias for
4836 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4837 (ix86_output_indirect_jmp): Remove the bool argument.
4838 (ix86_output_indirect_function_return): New function.
4839 (ix86_split_simple_return_pop_internal): Likewise.
4840 * config/i386/i386.md (*indirect_jump): Don't pass false
4841 to ix86_output_indirect_jmp.
4842 (*tablejump_1): Likewise.
4843 (simple_return_pop_internal): Change it to define_insn_and_split.
4844 Call ix86_split_simple_return_pop_internal to split it for
4846 (simple_return_indirect_internal): Call
4847 ix86_output_indirect_function_return instead of
4848 ix86_output_indirect_jmp.
4850 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4853 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4854 memset and value initialization afterwards.
4856 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4858 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4860 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4863 * common/config/aarch64/aarch64-common.c
4864 (aarch_option_optimization_table[]): Switch
4865 off fomit-frame-pointer
4867 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4868 Chung-Ju Wu <jasonwucj@gmail.com>
4870 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4872 (store_multiple): Ditto.
4874 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4876 * config.gcc: Add --with-cpu support for nds32 target.
4877 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4878 * config/nds32/nds32.opt: Add -mcpu= option.
4880 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4882 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4883 isel=yes): Warn for these deprecated options.
4885 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4887 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4888 ISA_2_5_MASKS_EMBEDDED.
4890 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4892 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4893 p->max as pointers rather than using iterative_hash_expr.
4895 2018-02-23 Carl Love <cel@us.ibm.com>
4897 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4898 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4900 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4901 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4902 P8V_BUILTIN_VEC_VUNSIGNED2.
4904 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4907 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4908 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4910 * lra-constraints.c (curr_insn_transform): Set up
4911 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4913 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4914 * lra-eliminations.c (spill_pseudos): Ditto.
4915 (process_insn_for_elimination): Ditto.
4916 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4917 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4918 LRA_NON_CLOBBERED_ALT.
4920 2018-02-22 Martin Sebor <msebor@redhat.com>
4922 PR tree-optimization/84480
4923 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4924 to maybe_diag_stxncpy_trunc. Call it.
4925 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4926 from gimple_fold_builtin_strcpy. Print inlining stack.
4927 (handle_builtin_stxncpy): Print inlining stack.
4928 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4930 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4933 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4934 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4935 and -fcheck-pointer-bounds are used together.
4936 (indirect_thunk_prefix): New enum.
4937 (indirect_thunk_need_prefix): New function.
4938 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4939 "_nt" instead of "_bnd" for NOTRACK prefix.
4940 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4941 (output_indirect_thunk_function): Likewise.
4943 (ix86_code_end): Update output_indirect_thunk_function calls.
4944 (ix86_output_indirect_branch_via_reg): Replace
4945 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4946 (ix86_output_indirect_branch_via_push): Likewise.
4947 (ix86_output_function_return): Likewise.
4948 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4949 incompatible with -fcf-protection=branch and
4950 -fcheck-pointer-bounds.
4952 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4955 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4956 Change gcc_assert call to output_operand_lossage.
4958 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4960 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4962 2018-02-22 DJ Delorie <dj@redhat.com>
4963 Sebastian Perta <sebastian.perta@renesas.com>
4964 Oleg Endo <olegendo@gcc.gnu.org>
4966 * config/rx/rx.c (rx_rtx_costs): New function.
4967 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4969 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4971 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4973 2018-02-22 Martin Liska <mliska@suse.cz>
4976 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4977 Add "native" as a possible value.
4979 2018-02-22 Martin Liska <mliska@suse.cz>
4982 * config/i386/i386.c (ix86_option_override_internal):
4983 Add "native" as a possible value for -march and -mtune.
4985 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4988 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4989 to all type variants.
4991 PR tree-optimization/84503
4992 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4993 width as info->bitpos + info->bitsize - start.
4994 (merged_store_group::merge_overlapping): Simplify width computation.
4995 (check_no_overlap): New function.
4996 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4997 start + width and last_order of the group, fail if check_no_overlap
4999 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5000 to group if check_no_overlap fails.
5002 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5004 * config/rs6000/altivec.md: Delete contraint arguments to
5005 define_expand, define_split, and define_peephole2, and in
5006 define_insn_and_split if always unused.
5007 * config/rs6000/darwin.md: Ditto.
5008 * config/rs6000/dfp.md: Ditto.
5009 * config/rs6000/rs6000.md: Ditto.
5010 * config/rs6000/sync.md: Ditto.
5011 * config/rs6000/vector.md: Ditto.
5012 * config/rs6000/vsx.md: Ditto.
5014 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5016 * config/rs6000/altivec.md: Write output control strings as braced
5017 blocks instead of double-quoted strings.
5018 * config/rs6000/darwin.md: Ditto.
5019 * config/rs6000/rs6000.md: Ditto.
5020 * config/rs6000/vector.md: Ditto.
5021 * config/rs6000/vsx.md: Ditto.
5023 2018-02-21 Jason Merrill <jason@redhat.com>
5025 PR c++/84314 - ICE with templates and fastcall attribute.
5026 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5028 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5030 * ipa-cp.c (determine_versionability): Fix comment typos.
5032 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5035 * ipa-cp.c (determine_versionability): Do not version functions caling
5038 2018-02-21 Martin Liska <mliska@suse.cz>
5041 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5042 Add "native" as a possible value.
5043 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5044 the macro when native cpu detection is available.
5046 2018-02-21 Martin Liska <mliska@suse.cz>
5049 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5050 Add "native" as a possible value.
5051 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5052 when native cpu detection is available.
5054 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5055 Martin Sebor <msebor@redhat.com>
5057 PR tree-optimization/84478
5058 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5060 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5061 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5062 support which is conservatively correct, for 2 only stay conservative
5063 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5064 argument to the 2 argument get_range_strlen, adjust 6 arg
5065 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5067 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5068 (gimple_fold_builtin_strlen): Pass true as last argument to
5071 2018-02-20 Martin Sebor <msebor@redhat.com>
5074 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5075 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5076 (builtin_memref::builtin_memref): Factor out parts into
5077 set_base_and_offset and call it.
5079 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5082 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5083 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5084 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5085 search at the associated MODE_INT.
5087 2018-02-20 Jeff Law <law@redhat.com>
5090 PR tree-optimization/81592
5092 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5093 for range data rather than using global data.
5094 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5095 range data rather than using global data.
5096 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5097 pass it to children as needed.
5098 (struct directive::fmtresult): Similarly.
5099 (struct directive::set_width): Similarly.
5100 (struct directive::set_precision): Similarly.
5101 (format_integer, format_directive, parse_directive): Similarly.
5102 (format_none): Accept unnamed vr_values parameter.
5103 (format_percent, format_floating, format_character): Similarly.
5104 (format_string, format_plain): Similarly.
5105 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5106 the EVRP range analyzer for range data rather than using global data.
5107 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5108 gimple-ssa-evrp-analyze.h
5109 (class sprintf_dom_walker): Add after_dom_children member function.
5110 Add evrp_range_analyzer member.
5111 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5112 range analyzer as needed.
5113 (sprintf_dom_walker::after_dom_children): New member function.
5114 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5116 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5117 (evrp_range_analyzer::pop_to_marker): Likewise.
5119 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5121 PR tree-optimization/84419
5122 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5123 with the required type if its current type is compatible but
5126 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5129 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5130 after vectorization.
5132 2018-02-20 Martin Liska <mliska@suse.cz>
5135 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5136 possible values if we don't have a hint.
5138 2018-02-20 Martin Liska <mliska@suse.cz>
5142 * final.c (shorten_branches): Build align_tab array with one
5144 * opts.c (finish_options): Add alignment option limit check.
5145 (MAX_CODE_ALIGN): Likewise.
5146 (MAX_CODE_ALIGN_VALUE): Likewise.
5147 * doc/invoke.texi: Document maximum allowed option value for
5148 all -falign-* options.
5150 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5153 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5154 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5155 * var-tracking.c (emit_note_insn_var_location): Remove all references
5156 to NOTE_INSN_CALL_ARG_LOCATION.
5157 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5158 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5159 Use copy_rtx_if_shared.
5160 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5161 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5162 (dwarf2out_var_location): Remove handling of
5163 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5165 * final.c (final_scan_insn): Remove all references to
5166 NOTE_INSN_CALL_ARG_LOCATION.
5167 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5168 before dumping final insns.
5169 * except.c (emit_note_eh_region_end): Remove all references to
5170 NOTE_INSN_CALL_ARG_LOCATION.
5171 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5172 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5173 * config/arc/arc.c (hwloop_optimize): Likewise.
5174 * config/arm/arm.c (create_fix_barrier): Likewise.
5175 * config/s390/s390.c (s390_chunkify_start): Likewise.
5176 * config/sh/sh.c (find_barrier): Likewise.
5177 * config/i386/i386.c (rest_of_insert_endbranch,
5178 ix86_seh_fixup_eh_fallthru): Likewise.
5179 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5180 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5181 * config/frv/frv.c (frv_function_prologue): Likewise.
5182 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5184 (note_outside_basic_block_p): Remove all references to
5185 NOTE_INSN_CALL_ARG_LOCATION.
5186 * gengtype.c (adjust_field_rtx_def): Likewise.
5187 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5189 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5190 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5193 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5196 PR tree-optimization/84452
5197 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5198 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5201 2018-02-19 Martin Liska <mliska@suse.cz>
5204 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5206 2018-02-19 Martin Liska <mliska@suse.cz>
5207 Richard Sandiford <richard.sandiford@linaro.org>
5209 PR tree-optimization/82491
5210 * gimple-fold.c (get_base_constructor): Make earlier bail out
5213 2018-02-19 Carl Love <cel@us.ibm.com>
5215 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5216 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5218 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5219 P8V_BUILTIN_VEC_NEG.
5221 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5223 * config/rl78/rl78.md (movdf): New define expand.
5225 2018-02-19 Martin Liska <mliska@suse.cz>
5228 * doc/invoke.texi: Fix typo.
5229 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5231 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5233 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5234 handle rs6000_single_float and rs6000_double_float specially for
5237 2018-02-16 Jeff Law <law@redhat.com>
5239 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5242 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5245 * ipa-inline.c (inline_small_functions): Fix a typo.
5247 2018-02-16 Nathan Sidwell <nathan@acm.org>
5249 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5251 2018-02-16 Carl Love <cel@us.ibm.com>
5253 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5254 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5255 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5256 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5257 expansion to P8V_BUILTIN_VEC_FLOAT2.
5259 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5261 PR rtl-optimization/70023
5262 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5263 src_regno into account.
5265 2018-02-16 Carl Love <cel@us.ibm.com>
5267 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5268 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5269 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5270 * config/rs6000/rs6000.c: Remove case statements for
5271 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5272 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5273 and P9V_BUILTIN_VEC_VINSERT4B.
5274 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5275 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5276 * config/rs6000/vsx.md:
5277 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5280 2018-02-16 Carl Love <cel@us.ibm.com>
5282 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5284 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5286 * config/rs6000/rs6000-c.c: Add the definitions for
5287 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5288 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5289 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5290 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5291 definition for insert4b and define insn *insert3b_internal.
5292 * doc/extend.texi: Add documentation for vec_extract4b.
5294 2018-02-16 Nathan Sidwell <nathan@acm.org>
5296 * doc/extend.texi (Backwards Compatibility): Mention friend
5297 injection. Note for-scope is deprecated.
5298 * doc/invoke.texi (-ffriend-injection): Deprecate.
5300 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5302 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5303 that moved to I2, also allow destinations that are a paradoxical
5304 subreg (instead of a normal reg).
5306 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5309 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5312 2018-02-16 Richard Biener <rguenther@suse.de>
5314 PR tree-optimization/84037
5315 PR tree-optimization/84016
5317 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5318 Adjust vec_construct for the fact we need additional higher latency
5319 128bit inserts for AVX256 and AVX512 vector builds.
5320 (ix86_add_stmt_cost): Scale vector construction cost for
5323 2018-02-16 Richard Biener <rguenther@suse.de>
5325 PR tree-optimization/84417
5326 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5327 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5328 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5330 2018-02-16 Martin Liska <mliska@suse.cz>
5333 * internal-fn.def (ASAN_CHECK): Set proper flags.
5334 (ASAN_MARK): Likewise.
5336 2018-02-16 Julia Koval <julia.koval@intel.com>
5338 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5339 from PTA_CANNONLAKE.
5341 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5344 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5345 Use ++iter rather than iter++ for std::list iterators.
5346 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5347 defer deleting them until all nodes in the forest are processed. Do
5348 free even leaf nodes. Change to_process into auto_vec.
5351 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5352 * vec.h (vec_default_construct): Use memset instead of placement new
5353 if BROKEN_VALUE_INITIALIZATION is defined.
5354 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5355 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5358 PR rtl-optimization/83723
5359 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5360 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5361 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5363 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5365 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5367 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5369 PR rtl-optimization/81443
5370 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5371 from inner REGs to paradoxical SUBREGs.
5373 2018-02-16 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84399
5376 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5377 For operands we can analyze at their definition make sure we can
5378 analyze them at each use as well.
5380 2018-02-16 Richard Biener <rguenther@suse.de>
5382 PR tree-optimization/84190
5383 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5384 volatile accesses if the decl isn't volatile.
5386 2018-02-15 Jason Merrill <jason@redhat.com>
5388 PR c++/84314 - ICE with templates and fastcall attribute.
5389 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5390 TYPE_CANONICAL on an existing type.
5392 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5394 PR tree-optimization/84383
5395 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5396 dstoff nor call operand_equal_p if dstbase is NULL.
5398 PR tree-optimization/84334
5399 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5400 also a CONSTANT_CLASS_P, punt.
5402 2018-02-14 Jim Wilson <jimw@sifive.com>
5404 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5405 first SMALL_OPERAND check. New local min_second_step. Move assert
5406 to where locals are set. Add TARGET_RVC support.
5407 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5409 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5411 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5413 2018-02-14 Martin Sebor <msebor@redhat.com>
5415 PR tree-optimization/83698
5416 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5417 arrays constrain the offset range to their bounds.
5418 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5419 (builtin_access::overlap): Avoid setting the size of overlap if it's
5421 (maybe_diag_overlap): Also consider arrays when deciding what values
5422 of offsets to include in diagnostics.
5424 2018-02-14 Martin Sebor <msebor@redhat.com>
5427 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5428 that correspond to the kind of a declaration.
5430 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5433 * config/pa/pa.md: Load address of PIC label using the linkage table
5434 if the label is nonlocal.
5436 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5438 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5439 warning message if user requests -maltivec=be.
5440 * doc/invoke.texi: Document deprecation of -maltivec=be.
5442 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5445 * config/rs6000/rs6000-c.c: Update definitions for
5446 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5447 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5449 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5452 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5453 add _get_ssp intrinsics. Remove argument from
5454 __builtin_ia32_rdssp[d|q].
5455 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5456 * config/i386/i386-builtin.def: Remove argument from
5457 __builtin_ia32_rdssp[d|q].
5458 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5459 ix86_expand_special_args_builtin for _rdssp[d|q].
5460 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5461 Clear register before usage.
5462 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5463 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5465 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5467 PR tree-optimization/84357
5468 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5469 operand 1 of an ARRAY_REF too.
5471 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5474 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5475 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5477 (set_of_reg): New struct.
5478 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5479 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5480 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5482 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5483 Split into bitclr, bitset, bitinvert patterns if appropriate.
5484 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5485 use rx_fuse_in_memory_bitop.
5486 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5487 to named insn, correct maximum insn length.
5489 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5492 * machmode.def: Define a complex mode for PARTIAL_INT.
5493 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5495 * doc/rtl.texi: Document CSPImode.
5496 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5498 (msp430_hard_regno_nregs_with_padding): Likewise.
5500 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5503 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5505 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5507 PR rtl-optimization/84169
5508 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5509 we generated a parallel as new i3 and we split that to new i2 and i3
5510 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5511 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5512 those to i2, not i1. Partially rewrite this scan code.
5514 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5517 * stor-layout.c (place_field): For variable length fields, adjust
5518 offset_align afterwards not just based on the field's alignment,
5519 but also on the size.
5522 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5523 of exps and logs in the use_exp2 case.
5525 2018-02-13 Jeff Law <law@redhat.com>
5527 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5530 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5533 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5537 * common.opt (gas-loc-support, gas-locview-support): New.
5538 (ginline-points, ginternal-reset-location-views): New.
5539 * doc/invoke.texi: Document them. Use @itemx where intended.
5540 (gvariable-location-views): Adjust.
5541 * target.def (reset_location_view): New.
5542 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5543 (TARGET_RESET_LOCATION_VIEW): New.
5544 * doc/tm.texi: Rebuilt.
5545 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5546 (dwarf2out_default_as_locview_support): New.
5547 (output_asm_line_debug_info): Use option variables.
5548 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5549 (output_loc_list): Likewise.
5550 (add_high_low_attributes): Check option variables.
5551 Don't output entry view attribute in strict mode.
5552 (gen_inlined_subroutine_die): Check option variables.
5553 (dwarf2out_inline_entry): Likewise.
5554 (init_sections_and_labels): Likewise.
5555 (dwarf2out_early_finish): Likewise.
5556 (maybe_reset_location_view): New, from...
5557 (dwarf2out_var_location): ... here. Call it.
5558 * debug.h (dwarf2out_default_as_loc_support): Declare.
5559 (dwarf2out_default_as_locview_support): Declare.
5560 * hooks.c (hook_int_rtx_insn_0): New.
5561 * hooks.h (hook_int_rtx_insn_0): Declare.
5562 * toplev.c (process_options): Take -gas-loc-support and
5563 -gas-locview-support from dwarf2out. Enable
5564 -gvariable-location-views by default only with locview
5565 assembler support. Enable -ginternal-reset-location-views by
5566 default only if the target defines the corresponding hook.
5567 Enable -ginline-points by default if location views are
5568 enabled; force it disabled if statement frontiers are
5570 * tree-inline.c (expand_call_inline): Check option variables.
5571 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5573 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5575 PR tree-optimization/84321
5576 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5577 handling. Also check whether the anti-range contains any values
5578 that satisfy the mask; switch to a VR_RANGE if not.
5580 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5583 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5585 2018-02-13 Martin Jambor <mjambor@suse.cz>
5588 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5589 of call statements, also set location of a load to a temporary.
5591 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5593 * config/rl78/rl78.c (add_vector_labels): New function.
5594 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5595 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5596 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5597 which checks that no arguments are passed.
5598 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5599 * doc/extend.texi: Documentation for the new attribute.
5601 2018-02-13 Andreas Schwab <schwab@suse.de>
5603 * config/riscv/linux.h (CPP_SPEC): Define.
5605 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5608 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5609 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5610 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5611 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5612 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5613 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5614 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5615 temporarily for AES and PCLMUL builtins.
5617 PR tree-optimization/84339
5618 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5619 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5623 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5624 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5625 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5627 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5629 * omp-simd-clone.h: New file.
5630 * omp-simd-clone.c: Include omp-simd-clone.h.
5631 (expand_simd_clones): No longer static.
5632 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5633 cgraph.h and omp-simd-clone.h.
5634 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5635 (vect_recog_widen_shift_pattern): Formatting fix.
5636 (vect_pattern_recog_1): Don't check optab for calls.
5639 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5640 operands[2] into a REG before using gen_lowpart on it.
5642 2018-02-12 Jeff Law <law@redhat.com>
5645 * config/sh/sh.c (find_barrier): Consider a sibling call
5648 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5649 successfully back substituting a reg.
5651 2018-02-12 Richard Biener <rguenther@suse.de>
5653 PR tree-optimization/84037
5654 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5655 parameter, move visited init to caller.
5656 (vect_slp_analyze_operations): Separate cost from validity
5657 check, initialize visited once for all instances.
5658 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5660 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5662 (vect_model_promotion_demotion_cost): Likewise.
5663 (vectorizable_bswap): Guard cost modeling with !slp_node
5664 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5666 (vectorizable_call): Likewise.
5667 (vectorizable_conversion): Likewise.
5668 (vectorizable_assignment): Likewise.
5669 (vectorizable_shift): Likewise.
5670 (vectorizable_operation): Likewise.
5671 (vectorizable_store): Likewise.
5672 (vectorizable_load): Likewise.
5673 (vectorizable_condition): Likewise.
5674 (vectorizable_comparison): Likewise.
5676 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5679 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5680 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5681 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5683 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5686 * params.def (inline-min-speedup): Increase from 8 to 15.
5687 (max-inline-insns-auto): Decrease from 40 to 30.
5688 * ipa-split.c (consider_split): Add some buffer for function to
5689 be considered inlining candidate.
5690 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5693 2018-02-12 Richard Biener <rguenther@suse.de>
5695 PR tree-optimization/84037
5696 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5697 matched stmts if we cannot swap the non-matched ones.
5699 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5701 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5702 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5703 _mm_maskz_scalef_round_ss): New intrinsics.
5704 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5705 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5706 __builtin_ia32_scalefss_round): Remove.
5707 (__builtin_ia32_scalefsd_mask_round,
5708 __builtin_ia32_scalefss_mask_round): New intrinsics.
5709 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5710 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5711 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5712 "<round_constraint>")): Changed to ...
5713 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5714 "<round_scalar_constraint>")): ... this.
5715 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5716 %0, %1, %2<round_op3>}"): Changed to ...
5717 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5718 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5719 %2<round_scalar_mask_op3>}"): ... this.
5720 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5722 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5724 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5725 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5726 (_mm_maskz_sqrt_round_ss): New intrinsics.
5727 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5728 (__builtin_ia32_sqrtsd_mask_round)
5729 (__builtin_ia32_sqrtss_mask_round): New builtins.
5730 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5731 (__builtin_ia32_sqrtss_round): Remove.
5732 (__builtin_ia32_sqrtsd_mask_round)
5733 (__builtin_ia32_sqrtss_mask_round): New builtins.
5734 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5735 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5736 ((match_operand:VF_128 1 "vector_operand"
5737 "xBm,<round_constraint>")): Changed to ...
5738 ((match_operand:VF_128 1 "vector_operand"
5739 "xBm,<round_scalar_constraint>")): ... this.
5740 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5741 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5742 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5743 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5744 %<iptr>1<round_scalar_mask_op3>}): ... this.
5745 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5746 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5748 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5751 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5752 Cast vec_cmpeq result to correct type.
5753 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5754 Cast vec_cmpgt result to correct type.
5756 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5758 * final.c (final_scan_insn_1): Renamed from...
5759 (final_scan_insn): ... this. New wrapper, to recover
5760 seen from the outermost call in recursive ones.
5761 * config/sparc/sparc.c (output_return): Drop seen from call.
5762 (output_sibcall): Likewise.
5763 * config/visium/visium.c (output_branch): Likewise.
5765 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5767 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5770 2018-02-10 Alan Modra <amodra@gmail.com>
5773 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5774 Specify LR as an input.
5776 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5779 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5780 remove_member_access_dummy_vars): New functions.
5781 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5782 lower_omp_1, execute_lower_omp): Use them.
5784 PR rtl-optimization/84308
5785 * shrink-wrap.c (spread_components): Release todo vector.
5787 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5789 PR rtl-optimization/57193
5790 * ira-color.c (struct allocno_color_data): Add member
5791 conflict_allocno_hard_prefs.
5792 (update_conflict_allocno_hard_prefs): New.
5793 (bucket_allocno_compare_func): Add a preference based on
5794 conflict_allocno_hard_prefs.
5795 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5796 (color_allocnos): Remove a dead code. Initiate
5797 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5799 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5802 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5803 constraint from =wa to wa. Avoid a subreg on the output operand,
5804 instead use a pseudo and subreg it in a move.
5805 (p9_xxbrd_<mode>): Changed to ...
5806 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5807 (p9_xxbrd_v2df): New expander.
5808 (p9_xxbrw_<mode>): Changed to ...
5809 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5810 (p9_xxbrw_v4sf): New expander.
5812 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5814 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5816 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5819 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5820 multiply in 32-bit mode.
5821 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5822 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5825 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5827 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5828 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5829 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5830 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5832 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5835 * dwarf2out.c (is_trivial_indirect_ref): New function.
5836 (dwarf2out_late_global_decl): Do not generate a location
5837 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5838 and that are not defined in the current unit.
5840 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5842 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5843 instead of a libcall for UNORDERED.
5845 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5848 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5849 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5851 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5854 * config/s390/s390.c (s390_set_current_function): Invoke
5855 s390_indirect_branch_settings also if fndecl didn't change.
5857 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5859 * config/rs6000/rs6000.md (blockage): Set length to zero.
5861 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5863 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5865 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5868 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5869 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5873 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5874 PARALLEL incoming that failed vt_get_decl_and_offset check.
5877 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5878 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5880 (get_variable_section): For decls in named .bss* sections pass true as
5881 second argument to bss_initializer_p.
5883 2018-02-09 Marek Polacek <polacek@redhat.com>
5884 Jakub Jelinek <jakub@redhat.com>
5887 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5888 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5889 Sync some changes from cxx_fold_indirect_ref.
5891 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5893 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5895 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5896 (BLOCK_INLINE_ENTRY_LABEL): New.
5897 (dwarf2out_var_location): Disregard inline entry markers.
5898 (inline_entry_data): New struct.
5899 (inline_entry_data_hasher): New hashtable type.
5900 (inline_entry_data_hasher::hash): New.
5901 (inline_entry_data_hasher::equal): New.
5902 (inline_entry_data_table): New variable.
5903 (add_high_low_attributes): Add DW_AT_entry_pc and
5904 DW_AT_GNU_entry_view attributes if a pending entry is found
5905 in inline_entry_data_table. Add old entry_pc attribute only
5906 if debug nonbinding markers are disabled.
5907 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5908 markers are enabled.
5909 (block_within_block_p, dwarf2out_inline_entry): New.
5910 (dwarf2out_finish): Check that no entries remained in
5911 inline_entry_data_table.
5912 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5913 (final_scan_insn, notice_source_line): Likewise.
5914 (rest_of_clean_state): Skip inline entry markers.
5915 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5917 * gimple.c (gimple_build_debug_inline_entry): New.
5918 * gimple.h (enum gimple_debug_subcode): Add
5919 GIMPLE_DEBUG_INLINE_ENTRY.
5920 (gimple_build_debug_inline_entry): Declare.
5921 (gimple_debug_inline_entry_p): New.
5922 (gimple_debug_nonbind_marker_p): Adjust.
5923 * insn-notes.def (INLINE_ENTRY): New.
5924 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5925 inline entry marker notes.
5926 (print_insn): Likewise.
5927 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5928 (INSN_DEBUG_MARKER_KIND): Likewise.
5929 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5930 * tree-inline.c (expand_call_inline): Build and insert
5931 debug_inline_entry stmt.
5932 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5933 inline entry blocks early, if nonbind markers are enabled.
5934 (dump_scope_block): Dump fragment info.
5935 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5936 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5937 (gimple_build_debug_inline_entry): New.
5938 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5939 Enable/disable inline entry points too.
5940 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5941 (DEBUG_INSN): Describe inline entry markers.
5943 * common.opt (gvariable-location-views): New.
5944 (gvariable-location-views=incompat5): New.
5945 * config.in: Rebuilt.
5946 * configure: Rebuilt.
5947 * configure.ac: Test assembler for view support.
5948 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5949 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5950 * dwarf2out.c (var_loc_view): New typedef.
5951 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5952 (dwarf2out_locviews_in_attribute): New.
5953 (dwarf2out_locviews_in_loclist): New.
5954 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5955 (enum dw_line_info_opcode): Add LI_adv_address.
5956 (struct dw_line_info_table): Add view.
5957 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5958 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5959 (zero_view_p): New variable.
5960 (ZERO_VIEW_P): New macro.
5961 (output_asm_line_debug_info): New.
5962 (struct var_loc_node): Add view.
5963 (add_AT_view_list, AT_loc_list): New.
5964 (add_var_loc_to_decl): Add view param. Test it against last.
5965 (new_loc_list): Add view params. Record them.
5966 (AT_loc_list_ptr): Handle loc and view lists.
5967 (view_list_to_loc_list_val_node): New.
5968 (print_dw_val): Handle dw_val_class_view_list.
5969 (size_of_die): Likewise.
5970 (value_format): Likewise.
5971 (loc_list_has_views): New.
5972 (gen_llsym): Set vl_symbol too.
5973 (maybe_gen_llsym, skip_loc_list_entry): New.
5974 (dwarf2out_maybe_output_loclist_view_pair): New.
5975 (output_loc_list): Output view list or entries too.
5976 (output_view_list_offset): New.
5977 (output_die): Handle dw_val_class_view_list.
5978 (output_dwarf_version): New.
5979 (output_compilation_unit_header): Use it.
5980 (output_skeleton_debug_sections): Likewise.
5981 (output_rnglists, output_line_info): Likewise.
5982 (output_pubnames, output_aranges): Update version comments.
5983 (output_one_line_info_table): Output view numbers in asm comments.
5984 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5985 Call maybe_gen_llsym.
5986 (loc_list_from_tree_1): Adjust.
5987 (add_AT_location_description): Create view list attribute if
5988 needed, check it's absent otherwise.
5989 (convert_cfa_to_fb_loc_list): Adjust.
5990 (maybe_emit_file): Call output_asm_line_debug_info for test.
5991 (dwarf2out_var_location): Reset views as needed. Precompute
5992 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5993 attribute. Set view.
5994 (new_line_info_table): Reset next view.
5995 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5996 (dwarf2out_source_line): Likewise. Output view resets and labels to
5997 the assembler, or select appropriate line info opcodes.
5998 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5999 (optimize_string_length): Catch it. Adjust.
6000 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6001 dw_val_class_view_list, and remove it if no longer needed.
6002 (hash_loc_list): Hash view numbers.
6003 (loc_list_hasher::equal): Compare them.
6004 (optimize_location_lists): Check whether a view list symbol is
6005 needed, and whether the locview attribute is present, and
6006 whether they match. Remove the locview attribute if no longer
6008 (index_location_lists): Call skip_loc_list_entry for test.
6009 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6010 Use output_dwarf_version.
6011 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6012 (struct dw_val_node): Add val_view_list.
6013 * final.c (SEEN_NEXT_VIEW): New.
6014 (set_next_view_needed): New.
6015 (clear_next_view_needed): New.
6016 (maybe_output_next_view): New.
6017 (final_start_function): Rename to...
6018 (final_start_function_1): ... this. Take pointer to FIRST,
6019 add SEEN parameter. Emit param bindings in the initial view.
6020 (final_start_function): Reintroduce SEEN-less interface.
6021 (final): Rename to...
6022 (final_1): ... this. Take SEEN parameter. Output final pending
6023 next view at the end.
6024 (final): Reintroduce seen-less interface.
6025 (final_scan_insn): Output pending next view before switching
6026 sections or ending a block. Mark the next view as needed when
6027 outputting variable locations. Notify debug backend of section
6028 changes, and of location view changes.
6029 (rest_of_handle_final): Adjust.
6030 * toplev.c (process_options): Autodetect value for debug variable
6031 location views option. Warn on incompat5 without -gdwarf-5.
6032 * doc/invoke.texi (gvariable-location-views): New.
6033 (gvariable-location-views=incompat5): New.
6034 (gno-variable-location-views): New.
6036 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6038 PR tree-optimization/84136
6039 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6040 that the result of find_edge is non-NULL.
6042 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6045 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6046 storing integer register in SImode. Fix cost of 256 and 512
6047 byte aligned SSE register store.
6049 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6051 * config/i386/i386.c (ix86_multiplication_cost): Fix
6052 multiplication cost for TARGET_AVX512DQ.
6054 2018-02-08 Marek Polacek <polacek@redhat.com>
6056 PR tree-optimization/84238
6057 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6060 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6062 PR tree-optimization/84265
6063 * tree-vect-stmts.c (vectorizable_store): Don't treat
6064 VMAT_CONTIGUOUS accesses as grouped.
6065 (vectorizable_load): Likewise.
6067 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6069 PR tree-optimization/81635
6070 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6071 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6072 (test_round_for_mask): New functions.
6073 (wide_int_cc_tests): Call test_round_for_mask.
6074 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6075 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6076 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6077 range returned by get_range_info.
6079 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6082 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6083 * symtab.c: Include builtins.h
6084 (symtab_node::output_to_lto_symbol_table_p): Move here
6085 from lto-streamer-out.c:output_symbol_p.
6086 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6087 (output_symbol_p): Move all logic to symtab.c
6088 (produce_symtab): Update.
6090 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6092 * config/s390/s390-opts.h (enum indirect_branch): Define.
6093 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6094 (s390_indirect_branch_via_thunk)
6095 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6096 (enum s390_indirect_branch_type): Define.
6097 * config/s390/s390.c (struct s390_frame_layout, struct
6098 machine_function): Remove.
6099 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6100 (indirect_branch_table_label_no, indirect_branch_table_name):
6102 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6103 (enum s390_indirect_branch_option): Define.
6104 (s390_return_addr_from_memory): New function.
6105 (s390_handle_string_attribute): New function.
6106 (s390_attribute_table): Add new attribute handler.
6107 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6108 (s390_indirect_branch_via_thunk): New function.
6109 (s390_indirect_branch_via_inline_thunk): New function.
6110 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6111 sibling call optimization for non z10 compiles.
6112 (s390_emit_call): Force indirect branch target to be a single
6113 register. Add r1 clobber for non-z10 compiles.
6114 (s390_emit_epilogue): Emit return jump via return_use expander.
6115 (s390_reorg): Handle JUMP_INSNs as execute targets.
6116 (s390_option_override_internal): Perform validity checks for the
6117 new command line options.
6118 (s390_indirect_branch_attrvalue): New function.
6119 (s390_indirect_branch_settings): New function.
6120 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6121 (s390_output_indirect_thunk_function): New function.
6122 (s390_code_end): Implement target hook.
6123 (s390_case_values_threshold): Implement target hook.
6124 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6126 * config/s390/s390.h (struct s390_frame_layout)
6127 (struct machine_function): Move here from s390.c.
6128 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6129 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6130 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6131 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6132 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6133 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6134 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6135 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6136 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6137 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6138 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6139 (mnemonic attribute): Add values which aren't recognized
6141 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6142 pattern for branch conversion. Fix mnemonic attribute.
6143 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6144 indirect branch via thunk if requested.
6145 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6146 ("*indirect_jump"): Disable for branch conversion using out of
6148 ("indirect_jump_via_thunk<mode>_z10")
6149 ("indirect_jump_via_thunk<mode>")
6150 ("indirect_jump_via_inlinethunk<mode>_z10")
6151 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6152 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6153 ("casesi_jump_via_inlinethunk<mode>_z10")
6154 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6155 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6156 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6157 ("*indirect2_jump"): Disable for branch conversion.
6158 ("casesi_jump"): Turn into expander and expand patterns for branch
6160 ("return_use"): New expander.
6161 ("*return"): Emit return via thunk and rename it to ...
6162 ("*return<mode>"): ... this one.
6163 * config/s390/s390.opt: Add new options and and enum for the
6166 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6168 * lra-constraints.c (match_reload): Unconditionally use
6169 gen_lowpart_SUBREG, rather than selecting between that
6170 and equivalent gen_rtx_SUBREG code.
6172 2018-02-08 Richard Biener <rguenther@suse.de>
6174 PR tree-optimization/84233
6175 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6176 changed flag instead of boguously re-using phi_inserted.
6178 2018-02-08 Martin Jambor <mjambor@suse.cz>
6180 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6181 static local variables.
6183 2018-02-08 Richard Biener <rguenther@suse.de>
6185 PR tree-optimization/84278
6186 * tree-vect-stmts.c (vectorizable_store): When looking for
6187 smaller vector types to perform grouped strided loads/stores
6188 make sure the mode is supported by the target.
6189 (vectorizable_load): Likewise.
6191 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6193 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6194 Increase LDP/STP opportunities by adding adjacent callee-saves.
6196 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6198 PR rtl-optimization/84068
6199 PR rtl-optimization/83459
6200 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6202 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6204 PR tree-optimization/84224
6205 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6206 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6209 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6212 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6213 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6214 count, remove one USE.
6216 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6218 * doc/install.texi (Configuration): Document the
6219 --with-long-double-format={ibm,ieee} PowerPC configuration
6223 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6224 Convert from define_expand to be define_insn_and_split. Rework
6225 float/double/_Float128 conversions to QI/HI/SImode to work with
6226 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6227 conversions to QI/HImode types did a store and then a load to
6228 truncate the value. For conversions to VSX registers, don't split
6229 the insn, instead emit the code directly. Use the code iterator
6230 any_fix to combine signed and unsigned conversions.
6231 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6232 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6233 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6234 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6235 (fix_<mode>di2_hw): Likewise.
6236 (fixuns_<mode>di2_hw): Likewise.
6237 (fix_<mode>si2_hw): Likewise.
6238 (fixuns_<mode>si2_hw): Likewise.
6239 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6240 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6241 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6242 fix<uns>_trunc<SFDF:mode>si2_p8.
6243 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6245 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6246 (fix<uns>_<mode>_mem): Likewise.
6247 (fctiw<u>z_<mode>_mem): Likewise.
6248 (fix<uns>_<mode>_mem): Likewise.
6249 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6250 the register allocator from doing a direct move to the GPRs to do
6251 a store, and instead use the ISA 3.0 store byte/half-word from
6252 vector register instruction. For IEEE 128-bit floating point,
6253 also optimize stores of 32-bit ints.
6254 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6256 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6258 * genextract.c (push_pathstr_operand): New function to support
6260 (walk_rtx): Call push_pathstr_operand.
6261 (print_path): Support [a-zA-Z].
6263 2018-02-07 Richard Biener <rguenther@suse.de>
6265 PR tree-optimization/84037
6266 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6267 (cse_and_gimplify_to_preheader): Declare.
6268 (vect_get_place_in_interleaving_chain): Likewise.
6269 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6271 (_loop_vec_info::~_loop_vec_info): Delete it.
6272 (cse_and_gimplify_to_preheader): New function.
6273 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6274 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6275 (vectorizable_load): Likewise. For grouped stores always base
6276 the IV on the first element.
6277 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6278 condition before gimplifying.
6280 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6282 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6283 *DIV_EXPR and *MOD_EXPR.
6285 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6288 * config/i386/i386.c (ix86_option_override_internal): Mask out
6289 the CF_SET bit when checking -fcf-protection.
6291 2018-02-07 Tom de Vries <tom@codesourcery.com>
6294 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6297 2018-02-07 Richard Biener <rguenther@suse.de>
6299 PR tree-optimization/84204
6300 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6303 PR tree-optimization/84205
6304 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6305 special-case isl_ast_op_zdiv_r.
6307 PR tree-optimization/84223
6308 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6309 Only add conditions from within the region.
6310 (gather_bbs::after_dom_children): Adjust.
6312 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6315 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6316 * config/avr/avr.md: Only post-reload split REG-REG moves if
6317 either register is GENERAL_REG_P.
6319 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6321 PR tree-optimization/84235
6322 * tree-ssa-scopedtables.c
6323 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6324 if the subtraction is performed in floating point type where NaNs are
6325 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6326 build 1. Formatting fix.
6328 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6331 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6332 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6333 and skip it regardless of bb boundaries. Use CALL_P macro,
6334 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6335 unnecessarily, formatting fix.
6337 2018-02-06 Michael Collison <michael.collison@arm.com>
6339 * config/arm/thumb2.md:
6340 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6341 (*thumb_mov_notscc): Ditto.
6343 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6346 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6347 unsigned_fix, not "s".
6349 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6351 * configure.ac (gcc_fn_eh_frame_ro): New function.
6352 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6353 correct .eh_frame permissions.
6354 * configure: Regenerate.
6356 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6358 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6361 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6363 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6364 Display warning message for -mno-speculate-indirect-jumps.
6366 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6368 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6370 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6372 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6374 PR tree-optimization/84225
6375 * tree-eh.c (find_trapping_overflow): Only call
6376 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6378 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6381 * config/i386/i386.c: Reimplement the check of possible options
6382 -mibt/-mshstk conbination. Change error messages.
6383 * doc/invoke.texi: Fix a typo: remove extra '='.
6385 2018-02-06 Marek Polacek <polacek@redhat.com>
6387 PR tree-optimization/84228
6388 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6390 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6393 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6394 emitted arch directives.
6395 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6396 __ARM_FEATURE_COPROC before changing architectures.
6398 2018-02-06 Richard Biener <rguenther@suse.de>
6400 * config/i386/i386.c (print_reg): Fix typo.
6401 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6403 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6405 * configure: Regenerate.
6407 2018-02-05 Martin Sebor <msebor@redhat.com>
6409 PR tree-optimization/83369
6410 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6413 2018-02-05 Martin Liska <mliska@suse.cz>
6415 * doc/invoke.texi: Cherry-pick upstream r323995.
6417 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6419 * ira.c (ira_init_register_move_cost): Adjust comment.
6421 2018-02-05 Martin Liska <mliska@suse.cz>
6423 PR gcov-profile/84137
6424 * doc/gcov.texi: Fix typo in documentation.
6426 2018-02-05 Martin Liska <mliska@suse.cz>
6428 PR gcov-profile/83879
6429 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6430 using dlopen functionality.
6432 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6434 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6435 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6436 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6437 _mm_maskz_range_round_ss): New intrinsics.
6438 (__builtin_ia32_rangesd128_round)
6439 (__builtin_ia32_rangess128_round): Remove.
6440 (__builtin_ia32_rangesd128_mask_round,
6441 __builtin_ia32_rangess128_mask_round): New builtins.
6442 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6443 __builtin_ia32_rangess128_round): Remove.
6444 (__builtin_ia32_rangesd128_mask_round,
6445 __builtin_ia32_rangess128_mask_round): New builtins.
6446 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6447 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6448 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6449 "<round_saeonly_constraint>")): Changed to ...
6450 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6451 "<round_saeonly_scalar_constraint>")): ... this.
6452 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6453 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6454 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6455 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6456 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6458 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6460 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6462 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6463 Remove all values except native, 8540 and 8548.
6465 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6467 * config/i386/i386.c (ix86_output_function_return): Pass
6468 INVALID_REGNUM, instead of -1, as invalid register number to
6469 indirect_thunk_name and output_indirect_thunk.
6471 2018-02-02 Julia Koval <julia.koval@intel.com>
6473 * config.gcc: Add -march=icelake.
6474 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6475 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6476 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6477 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6478 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6479 (processor_target_table): Add icelake.
6480 (ix86_option_override_internal): Handle new PTAs.
6481 (get_builtin_code_for_version): Handle icelake.
6482 (M_INTEL_COREI7_ICELAKE): New.
6483 (fold_builtin_cpu): Handle icelake.
6484 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6485 * doc/invoke.texi: Add -march=icelake.
6487 2018-02-02 Julia Koval <julia.koval@intel.com>
6489 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6490 to wide_int_bitmask.
6491 * wide-int-bitmask.h: New.
6493 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6496 * config/i386/i386.md: Replace Pmode with word_mode in
6497 builtin_setjmp_setup and builtin_longjmp to support x32.
6499 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6503 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6505 (rs6000_supported_cpu_names): New static variable.
6506 (linux_cpu_translation_table): Likewise.
6507 (elf_platform) <cpu>: Define new static variable and use it.
6508 Translate kernel AT_PLATFORM name to canonical name if needed.
6509 Error if platform name is unknown.
6511 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6514 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6516 2018-02-01 Jeff Law <law@redhat.com>
6519 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6520 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6521 the scratch if RELEASE_VIA_POP is false.
6522 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6523 If we have to save a temporary register, decrement SIZE appropriately.
6524 Pass new arguments to release_scratch_register_on_entry.
6525 (ix86_adjust_stack_and_probe): Likewise.
6526 (ix86_emit_probe_stack_range): Pass new arguments to
6527 release_scratch_register_on_entry.
6529 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6531 PR rtl-optimization/84157
6532 * combine.c (change_zero_ext): Use REG_P predicate in
6533 front of HARD_REGISTER_P predicate.
6535 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6537 * config/avr/avr.c (avr_option_override): Move disabling of
6538 -fdelete-null-pointer-checks to...
6539 * common/config/avr/avr-common.c (avr_option_optimization_table):
6542 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6544 PR tree-optimization/81635
6545 * tree-data-ref.c (split_constant_offset_1): For types that
6546 wrap on overflow, try to use range info to prove that wrapping
6549 2018-02-01 Renlin Li <renlin.li@arm.com>
6552 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6554 (aarch64_register_move_cost): Likewise.
6555 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6557 (REG_CLASS_NAMES): Likewise.
6558 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6559 TAILCALL_ADDR_REGS. Remove IP registers.
6560 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6562 2018-02-01 Richard Biener <rguenther@suse.de>
6564 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6565 for specifying RPO order and allow NULL for that.
6566 * domwalk.c (dom_walker::dom_walker): Likewise.
6567 (dom_walker::walk): Handle NULL RPO order.
6568 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6570 (rewrite_update_dom_walker): Likewise.
6571 (mark_def_dom_walker): Likewise.
6573 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6575 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6576 (aarch64_maybe_expand_sve_subreg_move): Declare.
6577 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6578 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6580 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6581 that are semantically a reverse operation.
6582 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6583 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6584 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6586 (aarch64_can_change_mode_class): For big-endian, forbid changes
6587 between two SVE modes if they have different element sizes.
6589 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6591 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6592 the TImode handling for big-endian targets.
6594 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6596 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6597 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6599 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6600 Remove BSWAP handing for big-endian targets and use the form of
6601 LD1RQ appropariate for the mode.
6603 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6605 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6606 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6609 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6612 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6613 check for operands that need to go through aarch64_sve_reload_be.
6615 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6617 PR tree-optimization/81661
6618 PR tree-optimization/84117
6619 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6620 * tree-eh.c: Include gimplify.h.
6621 (find_trapping_overflow, replace_trapping_overflow,
6622 rewrite_to_non_trapping_overflow): New functions.
6623 * tree-vect-loop.c: Include tree-eh.h.
6624 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6625 * tree-data-ref.c: Include tree-eh.h.
6626 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6628 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6630 PR rtl-optimization/84123
6631 * combine.c (change_zero_ext): Check if hard register satisfies
6632 can_change_dest_mode before calling gen_lowpart_SUBREG.
6634 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6637 * ira.c (ira_init_register_move_cost): Remove assert.
6639 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6641 PR rtl-optimization/84071
6642 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6643 * doc/tm.texi: Regenerate.
6645 2018-01-31 Richard Biener <rguenther@suse.de>
6647 PR tree-optimization/84132
6648 * tree-data-ref.c (analyze_miv_subscript): Properly
6649 check whether evolution_function_is_affine_multivariate_p
6650 before calling gcd_of_steps_may_divide_p.
6652 2018-01-31 Julia Koval <julia.koval@intel.com>
6655 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6656 * config/i386/i386.md (rdpid_rex64) New.
6657 (rdpid): Make 32bit only.
6659 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6662 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6663 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6665 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6668 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6670 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6672 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6674 PR rtl-optimization/84071
6675 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6676 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6678 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6680 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6681 (arc_attribute_table): Add 'aux' attribute.
6682 (arc_in_small_data_p): Consider aux like variables.
6683 (arc_is_aux_reg_p): New function.
6684 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6685 (arc_get_aux_arg): New function.
6686 (prepare_move_operands): Handle aux-register access.
6687 (arc_handle_aux_attribute): New function.
6688 * doc/extend.texi (ARC Variable attributes): Add subsection.
6690 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6692 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6693 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6694 (arc_attribute_table): Add 'uncached' attribute.
6695 (arc_print_operand): Print '.di' flag for uncached memory
6697 (arc_in_small_data_p): Do not consider for small data the uncached
6699 (arc_is_uncached_mem_p): New function.
6700 * config/arc/predicates.md (compact_store_memory_operand): Check
6701 for uncached memory accesses.
6702 (nonvol_nonimm_operand): Likewise.
6703 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6705 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6708 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6709 falign-loops=): Add Optimization flag.
6711 2018-01-30 Jeff Law <law@redhat.com>
6714 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6715 INT_REGISTERS_SAVED. Check it prior to calling
6716 get_scratch_register_on_entry.
6717 (ix86_adjust_stack_and_probe): Similarly.
6718 (ix86_emit_probe_stack_range): Similarly.
6719 (ix86_expand_prologue): Corresponding changes.
6721 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6724 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6725 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6727 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6730 * lra-constraints.c (curr_insn_transform): Process AND in the
6733 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6735 PR rtl-optimization/83986
6736 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6737 dependence against last_pending_memory_flush in addition to
6740 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6742 PR tree-optimization/81611
6743 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6746 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6749 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6752 2018-01-30 Richard Biener <rguenther@suse.de>
6753 Jakub Jelinek <jakub@redhat.com>
6755 PR tree-optimization/84111
6756 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6757 inner loops added during recursion, as they don't have up-to-date
6760 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6763 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6764 (can_inline_edge_by_limits_p): ... here.
6765 (can_early_inline_edge_p, check_callers,
6766 update_caller_keys, update_callee_keys, recursive_inlining,
6767 add_new_edges_to_heap, speculation_useful_p,
6768 inline_small_functions,
6769 inline_small_functions, flatten_function,
6770 inline_to_all_callers_1): Update.
6772 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6774 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6777 2018-01-30 Richard Biener <rguenther@suse.de>
6779 PR tree-optimization/83008
6780 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6781 invariant and constant vector uses in stmts when they need
6784 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6787 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6788 * configure: Regenerate.
6790 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6792 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6794 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6795 Use gen_rtx_REG rather than gen_lowpart.
6797 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6799 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6800 rather than 0 when creating partial subregs.
6802 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6804 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6807 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6810 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6811 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6812 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6813 flags. This restores the settings used before the 2017-07-24.
6814 Turning off pre increment/decrement/modify allows IVOPTS to
6815 optimize DF/SF loops where the index is an int.
6817 2018-01-29 Richard Biener <rguenther@suse.de>
6818 Kelvin Nilsen <kelvin@gcc.gnu.org>
6821 * tree-vect-stmts.c (vectorizable_call): Don't call
6822 targetm.vectorize_builtin_md_vectorized_function if callee is
6825 2018-01-22 Carl Love <cel@us.ibm.com>
6827 * doc/extend.tex: Fix typo in second arg in
6828 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6830 2018-01-29 Richard Biener <rguenther@suse.de>
6832 PR tree-optimization/84086
6833 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6834 (flush_ssaname_freelist): When SSA names were released reset
6835 the SCEV hash table.
6837 2018-01-29 Richard Biener <rguenther@suse.de>
6839 PR tree-optimization/84057
6840 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6841 removed paths when removing edges.
6843 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6845 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6846 -mfunction-return=@var{choice}.
6848 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6851 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6853 (layout::print_source_line): Likewise.
6854 (test_get_line_width_without_trailing_whitespace): Add test cases.
6856 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6859 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6862 2018-01-26 Jim Wilson <jimw@sifive.com>
6864 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6866 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6869 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6871 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6872 and CMP + SUB-immediate -> SUBS.
6874 2018-01-26 Martin Sebor <msebor@redhat.com>
6876 PR tree-optimization/83896
6877 * tree-ssa-strlen.c (get_string_len): Rename...
6878 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6879 Avoid assuming length is constant.
6880 (handle_char_store): Use HOST_WIDE_INT for string length.
6882 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6885 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6886 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6888 2018-01-26 Richard Biener <rguenther@suse.de>
6890 PR rtl-optimization/84003
6891 * dse.c (record_store): Only record redundant stores when
6892 the earlier store aliases at least all accesses the later one does.
6894 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6896 PR rtl-optimization/83985
6897 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6898 REG_CFA_RESTORE insns.
6899 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6900 REG_CFA_RESTORE insns here.
6903 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6904 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6906 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6908 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6909 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6910 (arc_init): Likewise.
6911 (arc_override_options): Likewise.
6912 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6914 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6916 * config/arc/arc.h (TARGET_DBNZ): Define.
6917 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6918 properly set the tune attribute.
6919 (dbnz): Use TARGET_DBNZ guard.
6920 * config/arc/arc.opt (mtune): Add core3 option.
6922 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6924 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6925 recognize new pic like addresses.
6926 (arc_delegitimize_address): Clean up.
6928 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6930 * config/arc/arc-arches.def: Option mrf16 valid for all
6932 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6933 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6934 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6935 * config/arc/arc-tables.opt: Regenerate.
6936 * config/arc/arc.c (arc_conditional_register_usage): Handle
6937 reduced register file case.
6938 (arc_file_start): Set must have build attributes.
6939 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6941 * config/arc/arc.opt (mrf16): Add new option.
6942 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6943 * config/arc/genmultilib.awk: Handle new mrf16 option.
6944 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6945 * config/arc/t-multilib: Regenerate.
6946 * doc/invoke.texi (ARC Options): Document mrf16 option.
6948 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6950 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6951 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6952 (arc_attribute_table): Add 'secure_call' attribute.
6953 (arc_print_operand): Print secure call operand.
6954 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6956 (arc_is_secure_call_p): New function. * config/arc/arc.md
6957 (call_i): Add support for sjli instruction.
6958 (call_value_i): Likewise.
6959 * config/arc/constraints.md (Csc): New constraint.
6961 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6962 John Eric Martin <John.Martin@emmicro-us.com>
6964 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6965 * config/arc/arc.c (_arc_jli_section): New struct.
6966 (arc_jli_section): New type.
6967 (rc_jli_sections): New static variable.
6968 (arc_handle_jli_attribute): New function.
6969 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6970 (arc_file_end): New function.
6971 (TARGET_ASM_FILE_END): Define.
6972 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6973 (arc_add_jli_section): New function.
6974 (jli_call_scan): Likewise.
6975 (arc_reorg): Call jli_call_scan.
6976 (arc_output_addsi): Remove 'S' from printing asm operand.
6977 (arc_is_jli_call_p): New function.
6978 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6980 (movhi_insn): Likewise.
6981 (movsi_insn): Likewise.
6982 (movsi_set_cc_insn): Likewise.
6983 (loadqi_update): Likewise.
6984 (load_zeroextendqisi_update): Likewise.
6985 (load_signextendqisi_update): Likewise.
6986 (loadhi_update): Likewise.
6987 (load_zeroextendhisi_update): Likewise.
6988 (load_signextendhisi_update): Likewise.
6989 (loadsi_update): Likewise.
6990 (loadsf_update): Likewise.
6991 (movsicc_insn): Likewise.
6992 (bset_insn): Likewise.
6993 (bxor_insn): Likewise.
6994 (bclr_insn): Likewise.
6995 (bmsk_insn): Likewise.
6996 (bicsi3_insn): Likewise.
6997 (cmpsi_cc_c_insn): Likewise.
6998 (movsi_ne): Likewise.
6999 (movsi_cond_exec): Likewise.
7000 (clrsbsi2): Likewise.
7011 (call_i): Remove 'S' asm letter, add jli instruction.
7012 (call_value_i): Likewise.
7013 * config/arc/arc.op (mjli-always): New option.
7014 * config/arc/constraints.md (Cji): New constraint.
7015 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7017 (subsf3_fpx): Likewise.
7018 (mulsf3_fpx): Likewise.
7019 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7021 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7023 * doc/invoke.texi (ARC): Document mjli-always option.
7025 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7027 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7028 avoid addition with 0 and use incw and decw where possible.
7030 2018-01-26 Richard Biener <rguenther@suse.de>
7032 PR tree-optimization/81082
7033 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7034 association if it requires casting to unsigned.
7035 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7036 from fold_plusminus_mult_expr to catch important cases late when
7037 range info is available.
7039 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7041 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7042 * configure.ac (hidden_linkonce): New test.
7043 * configure: Regenerate.
7044 * config.in: Regenerate.
7046 2018-01-26 Julia Koval <julia.koval@intel.com>
7048 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7049 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7050 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7051 _mm_mask_bitshuffle_epi64_mask): Fix type.
7052 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7053 USI_FTYPE_V4DI_V4DI_USI): Remove.
7054 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7055 __builtin_ia32_vpshufbitqmb256_mask,
7056 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7057 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7058 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7060 2018-01-26 Alan Modra <amodra@gmail.com>
7063 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7064 UNSPEC_VBPERMQ. Sort other unspecs.
7066 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7068 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7070 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7073 * predict.c (drop_profile): Do not push/pop cfun; update also
7075 (handle_missing_profiles): Fix logic looking for zero profiles.
7077 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7080 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7081 on functions with #pragma omp declare simd or functions with simd
7083 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7084 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7085 Remove trailing \n from warning_at calls.
7087 2018-01-25 Tom de Vries <tom@codesourcery.com>
7090 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7091 for neutered workers.
7093 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7096 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7097 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7099 2018-01-24 Jeff Law <law@redhat.com>
7102 * i386.c (get_probe_interval): Move to earlier point.
7103 (ix86_compute_frame_layout): If -fstack-clash-protection and
7104 the frame is larger than the probe interval, then use pushes
7105 to save registers rather than reg->mem moves.
7106 (ix86_expand_prologue): Remove conditional for int_registers_saved
7109 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7112 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7113 min/max for never referenced object.
7115 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7118 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7120 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7121 attributes from DECL_ATTRIBUTES (decl) without affecting
7122 DECL_ATTRIBUTES (current_function_decl).
7123 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7124 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7126 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7128 PR tree-optimization/83979
7129 * fold-const.c (fold_comparison): Use constant_boolean_node
7130 instead of boolean_{true,false}_node.
7132 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7134 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7137 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7139 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7140 Simplify the clause that sets the length attribute.
7141 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7142 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7143 clause that sets the length attribute.
7144 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7146 2018-01-24 Tom de Vries <tom@codesourcery.com>
7149 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7150 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7151 Add strict parameter.
7152 (prevent_branch_around_nothing): Insert dummy insn between branch to
7153 label and label with no ptx insn inbetween.
7154 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7156 2018-01-24 Tom de Vries <tom@codesourcery.com>
7159 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7160 for neutered threads in warp.
7161 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7163 2018-01-24 Richard Biener <rguenther@suse.de>
7165 PR tree-optimization/83176
7166 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7169 2018-01-24 Richard Biener <rguenther@suse.de>
7171 PR tree-optimization/82819
7172 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7173 code generating pluses that are no-ops in the target precision.
7175 2018-01-24 Richard Biener <rguenther@suse.de>
7178 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7180 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7182 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7183 to merge probabilities.
7184 * predict.c (probably_never_executed): Also mark as cold functions
7185 with global 0 profile and guessed local profile.
7186 * profile-count.c (profile_probability::combine_with_count): New
7188 * profile-count.h (profile_probability::operator*,
7189 profile_probability::operator*=, profile_probability::operator/,
7190 profile_probability::operator/=): Reduce precision to adjusted
7191 and set value to guessed on contradictory divisions.
7192 (profile_probability::combine_with_freq): Remove.
7193 (profile_probability::combine_wiht_count): Declare.
7194 (profile_count::force_nonzero):: Set to adjusted.
7195 (profile_count::probability_in):: Set quality to adjusted.
7196 * tree-ssa-tail-merge.c (replace_block_by): Use
7199 2018-01-23 Andrew Waterman <andrew@sifive.com>
7200 Jim Wilson <jimw@sifive.com>
7202 * config/riscv/riscv.c (riscv_stack_boundary): New.
7203 (riscv_option_override): Set riscv_stack_boundary. Handle
7204 riscv_preferred_stack_boundary_arg.
7205 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7206 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7207 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7208 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7209 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7210 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7212 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7215 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7216 of struct ix86_frame.
7217 (ix86_expand_epilogue): Likewise. Add a local variable for
7218 the reg_save_offset field in struct ix86_frame.
7220 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7222 PR tree-optimization/82604
7223 * tree-loop-distribution.c (enum partition_kind): New enum item
7224 PKIND_PARTIAL_MEMSET.
7225 (partition_builtin_p): Support above new enum item.
7226 (generate_code_for_partition): Ditto.
7227 (compute_access_range): Differentiate cases that equality can be
7228 proven at all loops, the innermost loops or no loops.
7229 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7230 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7231 (finalize_partitions, distribute_loop): Don't fuse partition of
7232 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7233 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7236 2018-01-23 Martin Liska <mliska@suse.cz>
7238 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7239 order to ignore the predictor.
7240 (PRED_POLYMORPHIC_CALL): Likewise.
7241 (PRED_RECURSIVE_CALL): Likewise.
7243 2018-01-23 Martin Liska <mliska@suse.cz>
7245 * tree-profile.c (tree_profiling): Print function header to
7246 aware reader which function we are working on.
7247 * value-prof.c (gimple_find_values_to_profile): Do not print
7248 not interesting value histograms.
7250 2018-01-23 Martin Liska <mliska@suse.cz>
7252 * profile-count.h (enum profile_quality): Add
7253 profile_uninitialized as the first value. Do not number values
7254 as they are zero based.
7255 (profile_count::verify): Update sanity check.
7256 (profile_probability::verify): Likewise.
7258 2018-01-23 Nathan Sidwell <nathan@acm.org>
7260 * doc/invoke.texi (ffor-scope): Deprecate.
7262 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7264 PR tree-optimization/83510
7265 * domwalk.c (set_all_edges_as_executable): New function.
7266 (dom_walker::dom_walker): Convert bool param
7267 "skip_unreachable_blocks" to enum reachability. Move setup of
7268 edge flags to set_all_edges_as_executable and only do it when
7269 reachability is REACHABLE_BLOCKS.
7270 * domwalk.h (enum dom_walker::reachability): New enum.
7271 (dom_walker::dom_walker): Convert bool param
7272 "skip_unreachable_blocks" to enum reachability.
7273 (set_all_edges_as_executable): New decl.
7274 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7275 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7277 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7278 but converting true to REACHABLE_BLOCKS.
7279 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7281 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7282 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7283 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7285 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7286 if check_all_array_refs will be called.
7288 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7290 * tree.c (selftest::test_location_wrappers): Add more test
7293 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7295 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7296 (selftest::test_bit_in_range): Likewise.
7298 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7301 * doc/sourcebuild.texi (vect_float): Say that the selector
7302 only describes the situation when -funsafe-math-optimizations is on.
7303 (vect_float_strict): Document.
7305 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7307 PR tree-optimization/83965
7308 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7309 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7310 instead of checking only for a reduction.
7311 (vect_recog_widen_sum_pattern): Likewise.
7313 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7315 * predict.c (probably_never_executed): Only use precise profile info.
7316 (compute_function_frequency): Skip after inlining hack since we now
7317 have quality checking.
7319 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7321 * profile-count.h (profile_probability::very_unlikely,
7322 profile_probability::unlikely, profile_probability::even): Set
7323 precision to guessed.
7325 2018-01-23 Richard Biener <rguenther@suse.de>
7327 PR tree-optimization/83963
7328 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7329 Properly terminate dominator walk when crossing the exit edge not
7330 when visiting its source block.
7332 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7335 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7336 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7338 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7340 PR tree-optimization/83957
7341 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7342 semicolon after for body surrounded by braces.
7344 PR tree-optimization/83081
7345 * profile-count.h (profile_probability::split): New method.
7346 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7347 Use profile_probability::split.
7348 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7349 when splitting a single conditional jump into 2.
7351 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7353 PR tree-optimization/69452
7354 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7357 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7359 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7360 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7361 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7363 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7365 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7366 * config/rl78/rl78.md (movdi): New define_expand.
7367 * config/rl78/rl78.c (rl78_split_movdi): New function.
7369 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7372 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7374 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7375 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7376 128-bit to produce an UNSPEC move to get the double word with the
7377 signbit and then a shift directly to do signbit.
7378 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7379 implementation with a new version that just does either a direct
7380 move or a regular move. Move memory interface to separate insns.
7381 Move insns so they are next to the expander.
7382 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7383 with signbit move. Split big and little endian case.
7384 (signbit<mode>2_dm_mem_le): Likewise.
7385 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7386 (signbit<mode>2_dm2): Likewise.
7388 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7390 * config/rl78/rl78.md (anddi3): New define_expand.
7392 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7394 * config/rl78/rl78.md (umindi3): New define_expand.
7396 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7398 * config/rl78/rl78.md (smindi3): New define_expand.
7400 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7402 * config/rl78/rl78.md (smaxdi3): New define_expand.
7404 2018-01-22 Carl Love <cel@us.ibm.com>
7406 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7407 LVX_V1TI): Add macro expansion.
7408 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7409 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7410 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7411 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7412 Change check to determine if the instruction is a byte reversing
7413 entry. Fix typo in comment.
7414 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7415 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7416 Add def_builtin calls for new builtins.
7417 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7418 Add define_insn expansion.
7420 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7422 * config/rl78/rl78.md (umaxdi3): New define_expand.
7424 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7426 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7427 for non-QImode registers.
7429 2018-01-22 Richard Biener <rguenther@suse.de>
7431 PR tree-optimization/83963
7432 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7433 including the loop exit block.
7434 (scop_detection::merge_sese): Likewise.
7435 (scop_detection::add_scop): Do it here instead.
7437 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7439 * doc/sourcebuild.texi (arm_softfloat): Document.
7441 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7444 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7445 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7446 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7448 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7449 David Edelsohn <dje.gcc@gmail.com>
7452 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7453 Change "crset eq" to "crset 2".
7454 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7455 (*call_indirect_aix<mode>_nospec): Likewise.
7456 (*call_value_indirect_aix<mode>_nospec): Likewise.
7457 (*call_indirect_elfv2<mode>_nospec): Likewise.
7458 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7459 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7460 change assembly output from . to $.
7461 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7462 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7463 (*tablejump<mode>_internal1_nospec): Likewise.
7465 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7468 * config/sh/sh_optimize_sett_clrt.cc:
7469 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7471 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7473 PR tree-optimization/83940
7474 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7475 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7476 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7477 use it to pass back the definition type.
7478 (vect_check_store_rhs): Likewise rhs_dt_out.
7479 (vect_build_gather_load_calls): Add a mask_dt argument and use
7480 it instead of a call to vect_is_simple_use.
7481 (vectorizable_store): Update calls to vect_check_load_store_mask
7482 and vect_check_store_rhs. Use the dt returned by the latter instead
7483 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7484 instead of calls to vect_is_simple_use. Pass the scalar rather
7485 than the vector operand to vect_is_simple_use when handling
7486 second and subsequent copies of an rhs value.
7487 (vectorizable_load): Update calls to vect_check_load_store_mask
7488 and vect_build_gather_load_calls. Use the cached mask_dt and
7489 gs_info.offset_dt instead of calls to vect_is_simple_use.
7491 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7494 * tree-emutls.c: Include gimplify.h.
7495 (lower_emutls_2): New function.
7496 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7497 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7498 it before further processing.
7501 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7502 UINTVAL (trueop1) instead of INTVAL (op1).
7504 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7508 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7509 INCOMING_FRAME_SP_OFFSET if not defined.
7510 (scan_trace): Add ENTRY argument. If true and
7511 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7512 emit a note to adjust the CFA offset.
7513 (create_cfi_notes): Adjust scan_trace callers.
7514 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7515 INCOMING_FRAME_SP_OFFSET in the CIE.
7516 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7517 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7519 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7520 * doc/tm.texi: Regenerated.
7522 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7524 PR rtl-optimization/83147
7525 * lra-constraints.c (remove_inheritance_pseudos): Use
7526 lra_substitute_pseudo_within_insn.
7528 2018-01-19 Tom de Vries <tom@codesourcery.com>
7529 Cesar Philippidis <cesar@codesourcery.com>
7532 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7534 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7537 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7538 spaces for function labels.
7540 2018-01-19 Martin Liska <mliska@suse.cz>
7542 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7543 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7544 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7545 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7546 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7547 (PRED_CONST_RETURN): Change from 69 to 65.
7548 (PRED_NULL_RETURN): Change from 91 to 71.
7549 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7550 (PRED_LOOP_GUARD): Change from 66 to 73.
7552 2018-01-19 Martin Liska <mliska@suse.cz>
7554 * predict.c (predict_insn_def): Add new assert.
7555 (struct branch_predictor): Change type to signed integer.
7556 (test_prediction_value_range): Amend test to cover
7558 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7559 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7560 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7561 (PRED_LOOP_IV_COMPARE): Likewise.
7562 * predict.h (PROB_UNINITIALIZED): Define new constant.
7564 2018-01-19 Martin Liska <mliska@suse.cz>
7566 * predict.c (dump_prediction): Add new format for
7567 analyze_brprob.py script which is enabled with -details
7569 * profile-count.h (precise_p): New function.
7571 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7573 PR tree-optimization/83922
7574 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7575 there are no statements that need masking.
7576 (vect_active_double_reduction_p): New function.
7577 (vect_analyze_loop_operations): Use it when handling phis that
7578 are not in the loop header.
7580 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7582 PR tree-optimization/83914
7583 * tree-vect-loop.c (vectorizable_induction): Don't convert
7584 init_expr or apply the peeling adjustment for inductions
7585 that are nested within the vectorized loop.
7587 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7589 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7592 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7596 * function.h (gimplify_parameters): Add gimple_seq * argument.
7597 * function.c: Include gimple.h and options.h.
7598 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7599 for the added local temporaries if needed.
7600 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7601 if there are any parameter cleanups, wrap whole body into a
7602 try/finally with the cleanups.
7604 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7607 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7608 Use GET_MODE_CLASS for scalar floating point.
7610 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7614 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7615 Fix call of call_cgraph_insertion_hooks.
7617 2018-01-18 Martin Sebor <msebor@redhat.com>
7619 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7621 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7624 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7627 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7630 * common.opt: (-ffile-prefix-map): New option.
7631 * opts.c (common_handle_option): Defer it.
7632 * opts-global.c (handle_common_deferred_options): Handle it.
7633 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7634 * file-prefix-map.h: New file.
7635 (remap_debug_filename, add_debug_prefix_map): ...here.
7636 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7637 * final.c (debug_prefix_map, add_debug_prefix_map
7638 remap_debug_filename): Move to...
7639 * file-prefix-map.c: New file.
7640 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7641 generalize, get rid of alloca(), use strrchr() instead of strchr().
7642 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7643 Implement in terms of add_prefix_map().
7644 (remap_macro_filename, remap_debug_filename): Implement in term of
7646 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7647 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7648 * dbxout.c: Include file-prefix-map.h.
7649 * varasm.c: Likewise.
7650 * vmsdbgout.c: Likewise.
7651 * xcoffout.c: Likewise.
7652 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7653 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7654 * doc/invoke.texi (-ffile-prefix-map): Document.
7655 (-fdebug-prefix-map): Update description.
7657 2018-01-18 Martin Liska <mliska@suse.cz>
7659 * config/i386/i386.c (indirect_thunk_name): Document that also
7661 (output_indirect_thunk): Document why both instructions
7662 (pause and lfence) are generated.
7664 2018-01-18 Richard Biener <rguenther@suse.de>
7666 PR tree-optimization/83887
7667 * graphite-scop-detection.c
7668 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7669 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7670 (scop_detection::merge_sese): Re-implement with a flood-fill
7671 algorithm that properly finds a SESE region if it exists.
7673 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7676 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7677 pointer_diff optimizations use view_convert instead of convert.
7679 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7681 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7682 Generate different code for -mno-speculate-indirect-jumps.
7683 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7684 (*call_indirect_aix<mode>): Disable for
7685 -mno-speculate-indirect-jumps.
7686 (*call_indirect_aix<mode>_nospec): New define_insn.
7687 (*call_value_indirect_aix<mode>): Disable for
7688 -mno-speculate-indirect-jumps.
7689 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7690 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7691 -mno-speculate-indirect-jumps.
7692 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7694 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7696 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7697 long double type, set the flags for noting the default long double
7698 type, even if we don't pass or return a long double type.
7700 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7703 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7706 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7708 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7709 support for merge[hl].
7710 (fold_mergehl_helper): New helper function.
7711 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7712 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7713 (altivec_vmrglw_direct): Add xxmrglw insn.
7715 2018-01-17 Andrew Waterman <andrew@sifive.com>
7717 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7718 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7720 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7723 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7724 call the lto_location_cache before reading the
7725 DECL_SOURCE_LOCATION of the types.
7727 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7728 Richard Sandiford <richard.sandiford@linaro.org>
7730 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7731 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7732 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7733 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7734 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7736 * config/aarch64/constraints.md (aarch64_movti_operand):
7738 * config/aarch64/predicates.md (Uti): Add new constraint.
7740 2018-01-17 Carl Love <cel@us.ibm.com>
7742 * config/rs6000/vsx.md (define_expand xl_len_r,
7743 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7744 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7746 (define_expand, define_insn): Move the shift left from the
7747 define_insn to the define_expand for lxvl and stxvl instructions.
7748 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7749 and XL_LEN_R definitions to PURE.
7751 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7753 * config/i386/i386.c (indirect_thunk_name): Declare regno
7754 as unsigned int. Compare regno with INVALID_REGNUM.
7755 (output_indirect_thunk): Ditto.
7756 (output_indirect_thunk_function): Ditto.
7757 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7758 in the call to output_indirect_thunk_function.
7760 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7763 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7764 rather than the size of inner_type to determine the stack slot size
7765 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7767 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7770 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7773 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7775 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7776 endian Linux systems to optionally enable multilibs for selecting
7777 the long double type if the user configured an explicit type.
7778 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7779 have no long double multilibs if not defined.
7780 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7781 warn if the user used -mabi={ieee,ibm}longdouble and we built
7782 multilibs for long double.
7783 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7784 appropriate multilib option.
7785 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7787 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7788 for building long double multilibs.
7789 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7791 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7793 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7796 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7798 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7801 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7804 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7807 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7809 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7810 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7812 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7814 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7815 different rtl trees depending on TARGET_64BIT.
7816 (rs6000_gen_lvx): Likewise.
7818 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7820 * config/visium/visium.md (nop): Tweak comment.
7821 (hazard_nop): Likewise.
7823 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7825 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7826 -mspeculate-indirect-jumps.
7827 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7828 for -mno-speculate-indirect-jumps.
7829 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7830 (*call_value_indirect_elfv2<mode>): Disable for
7831 -mno-speculate-indirect-jumps.
7832 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7833 (indirect_jump): Emit different RTL for
7834 -mno-speculate-indirect-jumps.
7835 (*indirect_jump<mode>): Disable for
7836 -mno-speculate-indirect-jumps.
7837 (*indirect_jump<mode>_nospec): New define_insn.
7838 (tablejump): Emit different RTL for
7839 -mno-speculate-indirect-jumps.
7840 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7841 (tablejumpsi_nospec): New define_expand.
7842 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7843 (tablejumpdi_nospec): New define_expand.
7844 (*tablejump<mode>_internal1): Disable for
7845 -mno-speculate-indirect-jumps.
7846 (*tablejump<mode>_internal1_nospec): New define_insn.
7847 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7850 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7852 * caller-save.c (insert_save): Drop unnecessary parameter. All
7855 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7856 Richard Biener <rguenth@suse.de>
7859 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7860 return early, inline manually is_gimple_sizepos. Make sure if we
7861 call gimplify_expr we don't end up with a gimple constant.
7862 * tree.c (variably_modified_type_p): Don't return true for
7863 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7864 * gimplify.h (is_gimple_sizepos): Remove.
7866 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7868 PR tree-optimization/83857
7869 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7870 vectorizable_live_operation for pure SLP statements.
7871 (vectorizable_live_operation): Handle PHIs.
7873 2018-01-16 Richard Biener <rguenther@suse.de>
7875 PR tree-optimization/83867
7876 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7877 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7879 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7882 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7883 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7884 If off is not INTEGER_CST, issue a may not be aligned warning
7885 rather than isn't aligned. Use isn%'t rather than isn't.
7886 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7888 <case MULT_EXPR>: Improve the case when bottom and one of the
7889 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7890 operand, in that case check if the other operand is multiple of
7891 bottom divided by the INTEGER_CST operand.
7893 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7896 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7897 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7898 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7899 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7900 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7901 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7902 (pa_function_arg_size): New function.
7904 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7906 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7907 in a separate statement.
7909 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7911 PR tree-optimization/83847
7912 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7913 group gathers and scatters.
7915 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7917 PR rtl-optimization/86620
7918 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7920 PR rtl-optimization/83213
7921 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7922 to last if both are JUMP_INSNs.
7924 PR tree-optimization/83843
7925 * gimple-ssa-store-merging.c
7926 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7927 store_immediate_info for bswap/nop orig_stores.
7929 2018-01-15 Andrew Waterman <andrew@sifive.com>
7931 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7933 <UDIV>: Increase cost if !TARGET_DIV.
7935 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7937 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7938 (define_attr "cr_logical_3op"): New.
7939 (cceq_ior_compare): Adjust.
7940 (cceq_ior_compare_complement): Adjust.
7941 (*cceq_rev_compare): Adjust.
7942 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7943 (is_cracked_insn): Adjust.
7944 (insn_must_be_first_in_group): Adjust.
7945 * config/rs6000/40x.md: Adjust.
7946 * config/rs6000/440.md: Adjust.
7947 * config/rs6000/476.md: Adjust.
7948 * config/rs6000/601.md: Adjust.
7949 * config/rs6000/603.md: Adjust.
7950 * config/rs6000/6xx.md: Adjust.
7951 * config/rs6000/7450.md: Adjust.
7952 * config/rs6000/7xx.md: Adjust.
7953 * config/rs6000/8540.md: Adjust.
7954 * config/rs6000/cell.md: Adjust.
7955 * config/rs6000/e300c2c3.md: Adjust.
7956 * config/rs6000/e500mc.md: Adjust.
7957 * config/rs6000/e500mc64.md: Adjust.
7958 * config/rs6000/e5500.md: Adjust.
7959 * config/rs6000/e6500.md: Adjust.
7960 * config/rs6000/mpc.md: Adjust.
7961 * config/rs6000/power4.md: Adjust.
7962 * config/rs6000/power5.md: Adjust.
7963 * config/rs6000/power6.md: Adjust.
7964 * config/rs6000/power7.md: Adjust.
7965 * config/rs6000/power8.md: Adjust.
7966 * config/rs6000/power9.md: Adjust.
7967 * config/rs6000/rs64.md: Adjust.
7968 * config/rs6000/titan.md: Adjust.
7970 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7972 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7973 ix86_indirect_branch_register logic.
7975 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7977 * config/i386/constraints.md (Bs): Update
7978 ix86_indirect_branch_register check. Don't check
7979 ix86_indirect_branch_register with GOT_memory_operand.
7981 * config/i386/predicates.md (GOT_memory_operand): Don't check
7982 ix86_indirect_branch_register here.
7983 (GOT32_symbol_operand): Likewise.
7985 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7987 * config/i386/predicates.md (constant_call_address_operand):
7988 Rewrite ix86_indirect_branch_register logic.
7989 (sibcall_insn_operand): Likewise.
7991 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7993 * config/i386/constraints.md (Bs): Replace
7994 ix86_indirect_branch_thunk_register with
7995 ix86_indirect_branch_register.
7997 * config/i386/i386.md (indirect_jump): Likewise.
7998 (tablejump): Likewise.
7999 (*sibcall_memory): Likewise.
8000 (*sibcall_value_memory): Likewise.
8001 Peepholes of indirect call and jump via memory: Likewise.
8002 * config/i386/i386.opt: Likewise.
8003 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8004 (GOT_memory_operand): Likewise.
8005 (call_insn_operand): Likewise.
8006 (sibcall_insn_operand): Likewise.
8007 (GOT32_symbol_operand): Likewise.
8009 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8012 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8013 type rather than type addr's type points to.
8014 (expand_omp_atomic_mutex): Likewise.
8015 (expand_omp_atomic): Likewise.
8017 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8020 * config/i386/i386.c (output_indirect_thunk_function): Use
8021 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8022 for __x86_return_thunk.
8024 2018-01-15 Richard Biener <rguenther@suse.de>
8027 * expmed.c (extract_bit_field_1): Fix typo.
8029 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8032 * config/arm/iterators.md (VF): New mode iterator.
8033 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8034 Remove integer-related logic from pattern.
8035 (neon_vabd<mode>_3): Likewise.
8037 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8040 * common.opt (fstrict-overflow): No longer an alias.
8041 (fwrapv-pointer): New option.
8042 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8043 also for pointer types based on flag_wrapv_pointer.
8044 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8045 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8046 opts->x_flag_wrapv got set.
8047 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8048 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8049 POINTER_TYPE_OVERFLOW_UNDEFINED.
8050 * match.pd: Likewise in address comparison pattern.
8051 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8053 2018-01-15 Richard Biener <rguenther@suse.de>
8056 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8057 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8058 Reset type names to their identifier if their TYPE_DECL doesn't
8059 have linkage (and thus is used for ODR and devirt).
8060 (save_debug_info_for_decl): Remove.
8061 (save_debug_info_for_type): Likewise.
8062 (add_tree_to_fld_list): Adjust.
8063 * tree-pretty-print.c (dump_generic_node): Make dumping of
8064 type names more robust.
8066 2018-01-15 Richard Biener <rguenther@suse.de>
8068 * BASE-VER: Bump to 8.0.1.
8070 2018-01-14 Martin Sebor <msebor@redhat.com>
8073 * builtins.c (check_access): Avoid warning when the no-warning bit
8076 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8078 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8079 * ira-color (allocno_hard_regs_compare): Likewise.
8081 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8084 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8085 Use .pushsection/.popsection.
8087 2018-01-14 Martin Sebor <msebor@redhat.com>
8090 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8092 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8094 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8095 entry from extra_headers.
8096 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8097 extra_headers, make the list bitwise identical to the i?86-*-* one.
8099 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8101 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8102 -mcmodel=large with -mindirect-branch=thunk,
8103 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8104 -mfunction-return=thunk-extern.
8105 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8106 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8107 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8109 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8111 * config/i386/i386.c (print_reg): Print the name of the full
8112 integer register without '%'.
8113 (ix86_print_operand): Handle 'V'.
8114 * doc/extend.texi: Document 'V' modifier.
8116 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8118 * config/i386/constraints.md (Bs): Disallow memory operand for
8119 -mindirect-branch-register.
8121 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8122 (GOT_memory_operand): Likewise.
8123 (call_insn_operand): Likewise.
8124 (sibcall_insn_operand): Likewise.
8125 (GOT32_symbol_operand): Likewise.
8126 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8127 for -mindirect-branch-register.
8128 (tablejump): Likewise.
8129 (*sibcall_memory): Likewise.
8130 (*sibcall_value_memory): Likewise.
8131 Disallow peepholes of indirect call and jump via memory for
8132 -mindirect-branch-register.
8133 (*call_pop): Replace m with Bw.
8134 (*call_value_pop): Likewise.
8135 (*sibcall_pop_memory): Replace m with Bs.
8136 * config/i386/i386.opt (mindirect-branch-register): New option.
8137 * doc/invoke.texi: Document -mindirect-branch-register option.
8139 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8141 * config/i386/i386-protos.h (ix86_output_function_return): New.
8142 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8143 set function_return_type.
8144 (indirect_thunk_name): Add ret_p to indicate thunk for function
8146 (output_indirect_thunk_function): Pass false to
8147 indirect_thunk_name.
8148 (ix86_output_indirect_branch_via_reg): Likewise.
8149 (ix86_output_indirect_branch_via_push): Likewise.
8150 (output_indirect_thunk_function): Create alias for function
8151 return thunk if regno < 0.
8152 (ix86_output_function_return): New function.
8153 (ix86_handle_fndecl_attribute): Handle function_return.
8154 (ix86_attribute_table): Add function_return.
8155 * config/i386/i386.h (machine_function): Add
8156 function_return_type.
8157 * config/i386/i386.md (simple_return_internal): Use
8158 ix86_output_function_return.
8159 (simple_return_internal_long): Likewise.
8160 * config/i386/i386.opt (mfunction-return=): New option.
8161 (indirect_branch): Mention -mfunction-return=.
8162 * doc/extend.texi: Document function_return function attribute.
8163 * doc/invoke.texi: Document -mfunction-return= option.
8165 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8167 * config/i386/i386-opts.h (indirect_branch): New.
8168 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8169 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8170 with local indirect jump when converting indirect call and jump.
8171 (ix86_set_indirect_branch_type): New.
8172 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8173 (indirectlabelno): New.
8174 (indirect_thunk_needed): Likewise.
8175 (indirect_thunk_bnd_needed): Likewise.
8176 (indirect_thunks_used): Likewise.
8177 (indirect_thunks_bnd_used): Likewise.
8178 (INDIRECT_LABEL): Likewise.
8179 (indirect_thunk_name): Likewise.
8180 (output_indirect_thunk): Likewise.
8181 (output_indirect_thunk_function): Likewise.
8182 (ix86_output_indirect_branch_via_reg): Likewise.
8183 (ix86_output_indirect_branch_via_push): Likewise.
8184 (ix86_output_indirect_branch): Likewise.
8185 (ix86_output_indirect_jmp): Likewise.
8186 (ix86_code_end): Call output_indirect_thunk_function if needed.
8187 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8189 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8190 (ix86_attribute_table): Add indirect_branch.
8191 * config/i386/i386.h (machine_function): Add indirect_branch_type
8192 and has_local_indirect_jump.
8193 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8195 (tablejump): Likewise.
8196 (*indirect_jump): Use ix86_output_indirect_jmp.
8197 (*tablejump_1): Likewise.
8198 (simple_return_indirect_internal): Likewise.
8199 * config/i386/i386.opt (mindirect-branch=): New option.
8200 (indirect_branch): New.
8203 (thunk-inline): Likewise.
8204 (thunk-extern): Likewise.
8205 * doc/extend.texi: Document indirect_branch function attribute.
8206 * doc/invoke.texi: Document -mindirect-branch= option.
8208 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8211 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8213 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8215 * ipa-inline.c (want_inline_small_function_p): Return false if
8216 inlining has already failed with CIF_FINAL_ERROR.
8217 (update_caller_keys): Call want_inline_small_function_p before
8219 (update_callee_keys): Likewise.
8221 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8223 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8225 (rs6000_quadword_masked_address_p): Likewise.
8226 (quad_aligned_load_p): Likewise.
8227 (quad_aligned_store_p): Likewise.
8228 (const_load_sequence_p): Add comment to describe the outer-most loop.
8229 (mimic_memory_attributes_and_flags): New function.
8230 (rs6000_gen_stvx): Likewise.
8231 (replace_swapped_aligned_store): Likewise.
8232 (rs6000_gen_lvx): Likewise.
8233 (replace_swapped_aligned_load): Likewise.
8234 (replace_swapped_load_constant): Capitalize argument name in
8235 comment describing this function.
8236 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8237 and stores that access quad-word aligned addresses and replace
8238 with stvx or lvx instructions when appropriate.
8239 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8240 New function prototype.
8241 (rs6000_quadword_masked_address_p): Likewise.
8242 (rs6000_gen_lvx): Likewise.
8243 (rs6000_gen_stvx): Likewise.
8244 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8245 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8246 when memory address is aligned.
8247 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8248 this split to select lvx instruction when memory address is aligned.
8249 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8250 instruction when memory address is aligned.
8251 (*vsx_le_perm_load_v16qi): Likewise.
8252 (four unnamed splitters): Modify to select the stvx instruction
8253 when memory is aligned.
8255 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8257 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8258 which appears in the queue multiple times.
8260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8261 Alan Hayward <alan.hayward@arm.com>
8262 David Sherwood <david.sherwood@arm.com>
8264 * tree-vectorizer.h (vec_lower_bound): New structure.
8265 (_loop_vec_info): Add check_nonzero and lower_bounds.
8266 (LOOP_VINFO_CHECK_NONZERO): New macro.
8267 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8268 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8269 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8270 fields. Make seg_len the distance travelled, not including the
8272 (dr_direction_indicator): Declare.
8273 (dr_zero_step_indicator): Likewise.
8274 (dr_known_forward_stride_p): Likewise.
8275 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8277 (runtime_alias_check_p): Allow runtime alias checks with
8279 (operator ==): Compare access_size and align.
8280 (prune_runtime_alias_test_list): Rework for new distinction between
8281 the access_size and seg_len.
8282 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8284 (get_segment_min_max): New function.
8285 (create_intersect_range_checks): Use it.
8286 (dr_step_indicator): New function.
8287 (dr_direction_indicator): Likewise.
8288 (dr_zero_step_indicator): Likewise.
8289 (dr_known_forward_stride_p): Likewise.
8290 * tree-loop-distribution.c (data_ref_segment_size): Return
8291 DR_STEP * (niters - 1).
8292 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8294 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8295 (vect_preserves_scalar_order_p): New function, split out from...
8296 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8297 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8298 (vect_vfa_access_size): New function.
8299 (vect_vfa_align): Likewise.
8300 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8301 (dump_lower_bound): New function.
8302 (vect_check_lower_bound): Likewise.
8303 (vect_small_gap_p): Likewise.
8304 (vectorizable_with_step_bound_p): Likewise.
8305 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8306 depencies if the vectorization factor is 1. Convert the checks
8307 for nonzero steps into checks on the bounds of DR_STEP. Try using
8308 a bunds check for variable steps if the minimum required step is
8309 relatively small. Update calls to the dr_with_seg_len
8310 constructor and to vect_compile_time_alias.
8311 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8313 (vect_loop_versioning): Call it.
8314 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8316 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8319 Alan Hayward <alan.hayward@arm.com>
8320 David Sherwood <david.sherwood@arm.com>
8322 * doc/sourcebuild.texi (vect_scatter_store): Document.
8323 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8325 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8327 * genopinit.c (main): Add supports_vec_scatter_store and
8328 supports_vec_scatter_store_cached to target_optabs.
8329 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8330 IFN_MASK_SCATTER_STORE.
8331 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8333 * internal-fn.h (internal_store_fn_p): Declare.
8334 (internal_fn_stored_value_index): Likewise.
8335 * internal-fn.c (scatter_store_direct): New macro.
8336 (expand_scatter_store_optab_fn): New function.
8337 (direct_scatter_store_optab_supported_p): New macro.
8338 (internal_store_fn_p): New function.
8339 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8340 IFN_MASK_SCATTER_STORE.
8341 (internal_fn_mask_index): Likewise.
8342 (internal_fn_stored_value_index): New function.
8343 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8345 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8346 * optabs-query.c (supports_vec_scatter_store_p): New function.
8347 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8348 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8349 true for scatter stores.
8350 (vect_gather_scatter_fn_p): Handle scatter stores too.
8351 (vect_check_gather_scatter): Consider using scatter stores if
8352 supports_vec_scatter_store_p.
8353 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8355 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8356 internal_fn_stored_value_index.
8357 (check_load_store_masking): Handle scatter stores too.
8358 (vect_get_store_rhs): Make public.
8359 (vectorizable_call): Use internal_store_fn_p.
8360 (vectorizable_store): Handle scatter store internal functions.
8361 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8362 when deciding whether the end of the group has been reached.
8363 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8364 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8365 (mask_scatter_store<mode>): New insns.
8367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8368 Alan Hayward <alan.hayward@arm.com>
8369 David Sherwood <david.sherwood@arm.com>
8371 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8372 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8373 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8375 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8376 Use vect_truncate_gather_scatter_offset if we can't treat the
8377 operation as a normal gather load or scatter store.
8378 (get_group_load_store_type): Take the gather_scatter_info
8379 as argument. Try using a gather load or scatter store for
8380 single-element groups.
8381 (get_load_store_type): Update calls to get_group_load_store_type
8382 and vect_use_strided_gather_scatters_p.
8384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8385 Alan Hayward <alan.hayward@arm.com>
8386 David Sherwood <david.sherwood@arm.com>
8388 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8389 optional tree argument.
8390 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8392 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8393 but continue to use the current value as a fallback.
8394 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8395 to compare the updates.
8396 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8397 (get_load_store_type): Use it when handling a strided access.
8398 (vect_get_strided_load_store_ops): New function.
8399 (vect_get_data_ptr_increment): Likewise.
8400 (vectorizable_load): Handle strided gather loads. Always pass
8401 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8404 Alan Hayward <alan.hayward@arm.com>
8405 David Sherwood <david.sherwood@arm.com>
8407 * doc/md.texi (gather_load@var{m}): Document.
8408 (mask_gather_load@var{m}): Likewise.
8409 * genopinit.c (main): Add supports_vec_gather_load and
8410 supports_vec_gather_load_cached to target_optabs.
8411 * optabs-tree.c (init_tree_optimization_optabs): Use
8412 ggc_cleared_alloc to allocate target_optabs.
8413 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8414 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8416 * internal-fn.h (internal_load_fn_p): Declare.
8417 (internal_gather_scatter_fn_p): Likewise.
8418 (internal_fn_mask_index): Likewise.
8419 (internal_gather_scatter_fn_supported_p): Likewise.
8420 * internal-fn.c (gather_load_direct): New macro.
8421 (expand_gather_load_optab_fn): New function.
8422 (direct_gather_load_optab_supported_p): New macro.
8423 (direct_internal_fn_optab): New function.
8424 (internal_load_fn_p): Likewise.
8425 (internal_gather_scatter_fn_p): Likewise.
8426 (internal_fn_mask_index): Likewise.
8427 (internal_gather_scatter_fn_supported_p): Likewise.
8428 * optabs-query.c (supports_at_least_one_mode_p): New function.
8429 (supports_vec_gather_load_p): Likewise.
8430 * optabs-query.h (supports_vec_gather_load_p): Declare.
8431 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8432 and memory_type field.
8433 (NUM_PATTERNS): Bump to 15.
8434 * tree-vect-data-refs.c: Include internal-fn.h.
8435 (vect_gather_scatter_fn_p): New function.
8436 (vect_describe_gather_scatter_call): Likewise.
8437 (vect_check_gather_scatter): Try using internal functions for
8438 gather loads. Recognize existing calls to a gather load function.
8439 (vect_analyze_data_refs): Consider using gather loads if
8440 supports_vec_gather_load_p.
8441 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8442 (vect_get_gather_scatter_offset_type): Likewise.
8443 (vect_convert_mask_for_vectype): Likewise.
8444 (vect_add_conversion_to_patterm): Likewise.
8445 (vect_try_gather_scatter_pattern): Likewise.
8446 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8447 (vect_vect_recog_func_ptrs): Add it.
8448 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8449 internal_fn_mask_index and internal_gather_scatter_fn_p.
8450 (check_load_store_masking): Take the gather_scatter_info as an
8451 argument and handle gather loads.
8452 (vect_get_gather_scatter_ops): New function.
8453 (vectorizable_call): Check internal_load_fn_p.
8454 (vectorizable_load): Likewise. Handle gather load internal
8456 (vectorizable_store): Update call to check_load_store_masking.
8457 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8458 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8459 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8460 (aarch64_gather_scale_operand_d): New predicates.
8461 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8462 (mask_gather_load<mode>): New insns.
8464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8465 Alan Hayward <alan.hayward@arm.com>
8466 David Sherwood <david.sherwood@arm.com>
8468 * optabs.def (fold_left_plus_optab): New optab.
8469 * doc/md.texi (fold_left_plus_@var{m}): Document.
8470 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8471 * internal-fn.c (fold_left_direct): Define.
8472 (expand_fold_left_optab_fn): Likewise.
8473 (direct_fold_left_optab_supported_p): Likewise.
8474 * fold-const-call.c (fold_const_fold_left): New function.
8475 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8476 * tree-parloops.c (valid_reduction_p): New function.
8477 (gather_scalar_reductions): Use it.
8478 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8479 (vect_finish_replace_stmt): Declare.
8480 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8481 (needs_fold_left_reduction_p): New function, split out from...
8482 (vect_is_simple_reduction): ...here. Accept reductions that
8483 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8484 (vect_force_simple_reduction): Also store the reduction type in
8485 the assignment's STMT_VINFO_REDUC_TYPE.
8486 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8487 (merge_with_identity): New function.
8488 (vect_expand_fold_left): Likewise.
8489 (vectorize_fold_left_reduction): Likewise.
8490 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8491 scalar phi in place for it. Check for target support and reject
8492 cases that would reassociate the operation. Defer the transform
8493 phase to vectorize_fold_left_reduction.
8494 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8495 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8496 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8500 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8501 call to ifc_temp_var.
8503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8504 Alan Hayward <alan.hayward@arm.com>
8505 David Sherwood <david.sherwood@arm.com>
8507 * target.def (legitimize_address_displacement): Take the original
8508 offset as a poly_int.
8509 * targhooks.h (default_legitimize_address_displacement): Update
8511 * targhooks.c (default_legitimize_address_displacement): Likewise.
8512 * doc/tm.texi: Regenerate.
8513 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8514 as an argument, moving assert of ad->disp == ad->disp_term to...
8515 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8516 Try calling targetm.legitimize_address_displacement before expanding
8517 the address rather than afterwards, and adjust for the new interface.
8518 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8519 Match the new hook interface. Handle SVE addresses.
8520 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8525 * Makefile.in (OBJS): Add early-remat.o.
8526 * target.def (select_early_remat_modes): New hook.
8527 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8528 * doc/tm.texi: Regenerate.
8529 * targhooks.h (default_select_early_remat_modes): Declare.
8530 * targhooks.c (default_select_early_remat_modes): New function.
8531 * timevar.def (TV_EARLY_REMAT): New timevar.
8532 * passes.def (pass_early_remat): New pass.
8533 * tree-pass.h (make_pass_early_remat): Declare.
8534 * early-remat.c: New file.
8535 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8537 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8540 Alan Hayward <alan.hayward@arm.com>
8541 David Sherwood <david.sherwood@arm.com>
8543 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8544 vfm1 with a bound_epilog parameter.
8545 (vect_do_peeling): Update calls accordingly, and move the prologue
8546 call earlier in the function. Treat the base bound_epilog as 0 for
8547 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8548 this base when peeling for gaps.
8549 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8550 with fully-masked loops.
8551 (vect_estimate_min_profitable_iters): Handle the single peeled
8552 iteration in that case.
8554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8555 Alan Hayward <alan.hayward@arm.com>
8556 David Sherwood <david.sherwood@arm.com>
8558 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8559 single-element interleaving even if the size is not a power of 2.
8560 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8561 accesses for single-element interleaving if the group size is
8564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8565 Alan Hayward <alan.hayward@arm.com>
8566 David Sherwood <david.sherwood@arm.com>
8568 * doc/md.texi (fold_extract_last_@var{m}): Document.
8569 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8570 * optabs.def (fold_extract_last_optab): New optab.
8571 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8572 * internal-fn.c (fold_extract_direct): New macro.
8573 (expand_fold_extract_optab_fn): Likewise.
8574 (direct_fold_extract_optab_supported_p): Likewise.
8575 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8576 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8577 EXTRACT_LAST_REDUCTION.
8578 (get_initial_def_for_reduction): Do not create an initial vector
8579 for EXTRACT_LAST_REDUCTION reductions.
8580 (vectorizable_reduction): Leave the scalar phi in place for
8581 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8582 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8583 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8584 transform phase to vectorizable_condition.
8585 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8587 (vect_finish_stmt_generation): ...here.
8588 (vect_finish_replace_stmt): New function.
8589 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8590 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8592 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8595 Alan Hayward <alan.hayward@arm.com>
8596 David Sherwood <david.sherwood@arm.com>
8598 * doc/md.texi (extract_last_@var{m}): Document.
8599 * optabs.def (extract_last_optab): New optab.
8600 * internal-fn.def (EXTRACT_LAST): New internal function.
8601 * internal-fn.c (cond_unary_direct): New macro.
8602 (expand_cond_unary_optab_fn): Likewise.
8603 (direct_cond_unary_optab_supported_p): Likewise.
8604 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8605 loops using EXTRACT_LAST.
8606 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8607 (extract_last_<mode>): ...this optab.
8608 (vec_extract<mode><Vel>): Update accordingly.
8610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8611 Alan Hayward <alan.hayward@arm.com>
8612 David Sherwood <david.sherwood@arm.com>
8614 * target.def (empty_mask_is_expensive): New hook.
8615 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8616 * doc/tm.texi: Regenerate.
8617 * targhooks.h (default_empty_mask_is_expensive): Declare.
8618 * targhooks.c (default_empty_mask_is_expensive): New function.
8619 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8620 if the target says that empty masks are expensive.
8621 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8623 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8626 Alan Hayward <alan.hayward@arm.com>
8627 David Sherwood <david.sherwood@arm.com>
8629 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8630 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8631 (vect_use_loop_mask_for_alignment_p): New function.
8632 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8633 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8634 niters_skip argument. Make sure that the first niters_skip elements
8635 of the first iteration are inactive.
8636 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8637 Update call to vect_set_loop_masks_directly.
8638 (get_misalign_in_elems): New function, split out from...
8639 (vect_gen_prolog_loop_niters): ...here.
8640 (vect_update_init_of_dr): Take a code argument that specifies whether
8641 the adjustment should be added or subtracted.
8642 (vect_update_init_of_drs): Likewise.
8643 (vect_prepare_for_masked_peels): New function.
8644 (vect_do_peeling): Skip prologue peeling if we're using a mask
8645 instead. Update call to vect_update_inits_of_drs.
8646 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8648 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8649 alignment. Do not include the number of peeled iterations in
8650 the minimum threshold in that case.
8651 (vectorizable_induction): Adjust the start value down by
8652 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8653 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8654 Take the number of skipped iterations into account when calculating
8656 * tree-vect-stmts.c (vect_gen_while_not): New function.
8658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8659 Alan Hayward <alan.hayward@arm.com>
8660 David Sherwood <david.sherwood@arm.com>
8662 * doc/sourcebuild.texi (vect_fully_masked): Document.
8663 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8665 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8667 (vect_analyze_loop_2): ...here. Don't check the vectorization
8668 factor against the number of loop iterations if the loop is
8671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8672 Alan Hayward <alan.hayward@arm.com>
8673 David Sherwood <david.sherwood@arm.com>
8675 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8676 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8677 (dump_groups): Update accordingly.
8678 (iv_use::mem_type): New member variable.
8679 (address_p): New function.
8680 (record_use): Add a mem_type argument and initialize the new
8682 (record_group_use): Add a mem_type argument. Use address_p.
8683 Remove obsolete null checks of base_object. Update call to record_use.
8684 (find_interesting_uses_op): Update call to record_group_use.
8685 (find_interesting_uses_cond): Likewise.
8686 (find_interesting_uses_address): Likewise.
8687 (get_mem_type_for_internal_fn): New function.
8688 (find_address_like_use): Likewise.
8689 (find_interesting_uses_stmt): Try find_address_like_use before
8690 calling find_interesting_uses_op.
8691 (addr_offset_valid_p): Use the iv mem_type field as the type
8692 of the addressed memory.
8693 (add_autoinc_candidates): Likewise.
8694 (get_address_cost): Likewise.
8695 (split_small_address_groups_p): Use address_p.
8696 (split_address_groups): Likewise.
8697 (add_iv_candidate_for_use): Likewise.
8698 (autoinc_possible_for_pair): Likewise.
8699 (rewrite_groups): Likewise.
8700 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8701 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8702 (get_alias_ptr_type_for_ptr_address): New function.
8703 (rewrite_use_address): Rewrite address uses in calls that were
8704 identified by find_address_like_use.
8706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8707 Alan Hayward <alan.hayward@arm.com>
8708 David Sherwood <david.sherwood@arm.com>
8710 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8712 * gimple-expr.h (is_gimple_addressable: Likewise.
8713 * gimple-expr.c (is_gimple_address): Likewise.
8714 * internal-fn.c (expand_call_mem_ref): New function.
8715 (expand_mask_load_optab_fn): Use it.
8716 (expand_mask_store_optab_fn): Likewise.
8718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8719 Alan Hayward <alan.hayward@arm.com>
8720 David Sherwood <david.sherwood@arm.com>
8722 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8723 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8724 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8725 (cond_umax@var{mode}): Document.
8726 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8727 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8728 (cond_umin_optab, cond_umax_optab): New optabs.
8729 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8730 (COND_IOR, COND_XOR): New internal functions.
8731 * internal-fn.h (get_conditional_internal_fn): Declare.
8732 * internal-fn.c (cond_binary_direct): New macro.
8733 (expand_cond_binary_optab_fn): Likewise.
8734 (direct_cond_binary_optab_supported_p): Likewise.
8735 (get_conditional_internal_fn): New function.
8736 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8737 Cope with reduction statements that are vectorized as calls rather
8739 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8740 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8741 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8742 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8743 (UNSPEC_COND_EOR): New unspecs.
8744 (optab): Add mappings for them.
8745 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8746 (sve_int_op, sve_fp_op): New int attributes.
8748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8749 Alan Hayward <alan.hayward@arm.com>
8750 David Sherwood <david.sherwood@arm.com>
8752 * optabs.def (while_ult_optab): New optab.
8753 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8754 * internal-fn.def (WHILE_ULT): New internal function.
8755 * internal-fn.h (direct_internal_fn_supported_p): New override
8756 that takes two types as argument.
8757 * internal-fn.c (while_direct): New macro.
8758 (expand_while_optab_fn): New function.
8759 (convert_optab_supported_p): Likewise.
8760 (direct_while_optab_supported_p): New macro.
8761 * wide-int.h (wi::udiv_ceil): New function.
8762 * tree-vectorizer.h (rgroup_masks): New structure.
8763 (vec_loop_masks): New typedef.
8764 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8766 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8767 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8768 (vect_max_vf): New function.
8769 (slpeel_make_loop_iterate_ntimes): Delete.
8770 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8771 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8772 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8773 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8774 internal-fn.h, stor-layout.h and optabs-query.h.
8775 (vect_set_loop_mask): New function.
8776 (add_preheader_seq): Likewise.
8777 (add_header_seq): Likewise.
8778 (interleave_supported_p): Likewise.
8779 (vect_maybe_permute_loop_masks): Likewise.
8780 (vect_set_loop_masks_directly): Likewise.
8781 (vect_set_loop_condition_masked): Likewise.
8782 (vect_set_loop_condition_unmasked): New function, split out from
8783 slpeel_make_loop_iterate_ntimes.
8784 (slpeel_make_loop_iterate_ntimes): Rename to..
8785 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8786 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8787 (vect_do_peeling): Update call accordingly.
8788 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8790 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8791 mask_compare_type, can_fully_mask_p and fully_masked_p.
8792 (release_vec_loop_masks): New function.
8793 (_loop_vec_info): Use it to free the loop masks.
8794 (can_produce_all_loop_masks_p): New function.
8795 (vect_get_max_nscalars_per_iter): Likewise.
8796 (vect_verify_full_masking): Likewise.
8797 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8798 retries, and free the mask rgroups before retrying. Check loop-wide
8799 reasons for disallowing fully-masked loops. Make the final decision
8800 about whether use a fully-masked loop or not.
8801 (vect_estimate_min_profitable_iters): Do not assume that peeling
8802 for the number of iterations will be needed for fully-masked loops.
8803 (vectorizable_reduction): Disable fully-masked loops.
8804 (vectorizable_live_operation): Likewise.
8805 (vect_halve_mask_nunits): New function.
8806 (vect_double_mask_nunits): Likewise.
8807 (vect_record_loop_mask): Likewise.
8808 (vect_get_loop_mask): Likewise.
8809 (vect_transform_loop): Handle the case in which the final loop
8810 iteration might handle a partial vector. Call vect_set_loop_condition
8811 instead of slpeel_make_loop_iterate_ntimes.
8812 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8813 (check_load_store_masking): New function.
8814 (prepare_load_store_mask): Likewise.
8815 (vectorizable_store): Handle fully-masked loops.
8816 (vectorizable_load): Likewise.
8817 (supportable_widening_operation): Use vect_halve_mask_nunits for
8819 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8820 (vect_gen_while): New function.
8821 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8822 (aarch64_uqdec<mode>): New insn.
8824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8825 Alan Hayward <alan.hayward@arm.com>
8826 David Sherwood <david.sherwood@arm.com>
8828 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8829 (reduc_xor_scal_optab): New optabs.
8830 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8831 (reduc_xor_scal_@var{m}): Document.
8832 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8833 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8835 * fold-const-call.c (fold_const_call): Handle them.
8836 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8837 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8838 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8839 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8840 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8841 (UNSPEC_XORV): New unspecs.
8842 (optab): Add entries for them.
8843 (BITWISEV): New int iterator.
8844 (bit_reduc_op): New int attributes.
8846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8847 Alan Hayward <alan.hayward@arm.com>
8848 David Sherwood <david.sherwood@arm.com>
8850 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8851 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8852 * optabs.def (vec_shl_insert_optab): New optab.
8853 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8854 (duplicate_and_interleave): Likewise.
8855 * tree-vect-loop.c: Include internal-fn.h.
8856 (neutral_op_for_slp_reduction): New function, split out from
8857 get_initial_defs_for_reduction.
8858 (get_initial_def_for_reduction): Handle option 2 for variable-length
8859 vectors by loading the neutral value into a vector and then shifting
8860 the initial value into element 0.
8861 (get_initial_defs_for_reduction): Replace the code argument with
8862 the neutral value calculated by neutral_op_for_slp_reduction.
8863 Use gimple_build_vector for constant-length vectors.
8864 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8865 but the first group_size elements have a neutral value.
8866 Use duplicate_and_interleave otherwise.
8867 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8868 Update call to get_initial_defs_for_reduction. Handle SLP
8869 reductions for variable-length vectors by creating one vector
8870 result for each scalar result, with the elements associated
8871 with other scalar results stubbed out with the neutral value.
8872 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8873 Require IFN_VEC_SHL_INSERT for double reductions on
8874 variable-length vectors, or SLP reductions that have
8875 a neutral value. Require can_duplicate_and_interleave_p
8876 support for variable-length unchained SLP reductions if there
8877 is no neutral value, such as for MIN/MAX reductions. Also require
8878 the number of vector elements to be a multiple of the number of
8879 SLP statements when doing variable-length unchained SLP reductions.
8880 Update call to vect_create_epilog_for_reduction.
8881 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8882 and remove initial values.
8883 (duplicate_and_interleave): Make public.
8884 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8885 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8888 Alan Hayward <alan.hayward@arm.com>
8889 David Sherwood <david.sherwood@arm.com>
8891 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8892 (can_duplicate_and_interleave_p): New function.
8893 (vect_get_and_check_slp_defs): Take the vector of statements
8894 rather than just the current one. Remove excess parentheses.
8895 Restriction rejectinon of vect_constant_def and vect_external_def
8896 for variable-length vectors to boolean types, or types for which
8897 can_duplicate_and_interleave_p is false.
8898 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8899 (duplicate_and_interleave): New function.
8900 (vect_get_constant_vectors): Use gimple_build_vector for
8901 constant-length vectors and suitable variable-length constant
8902 vectors. Use duplicate_and_interleave for other variable-length
8903 vectors. Don't defer the update when inserting new statements.
8905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8906 Alan Hayward <alan.hayward@arm.com>
8907 David Sherwood <david.sherwood@arm.com>
8909 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8910 min_profitable_iters doesn't go negative.
8912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8913 Alan Hayward <alan.hayward@arm.com>
8914 David Sherwood <david.sherwood@arm.com>
8916 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8917 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8918 * optabs.def (vec_mask_load_lanes_optab): New optab.
8919 (vec_mask_store_lanes_optab): Likewise.
8920 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8921 (MASK_STORE_LANES): Likewise.
8922 * internal-fn.c (mask_load_lanes_direct): New macro.
8923 (mask_store_lanes_direct): Likewise.
8924 (expand_mask_load_optab_fn): Handle masked operations.
8925 (expand_mask_load_lanes_optab_fn): New macro.
8926 (expand_mask_store_optab_fn): Handle masked operations.
8927 (expand_mask_store_lanes_optab_fn): New macro.
8928 (direct_mask_load_lanes_optab_supported_p): Likewise.
8929 (direct_mask_store_lanes_optab_supported_p): Likewise.
8930 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8932 (vect_load_lanes_supported): Likewise.
8933 * tree-vect-data-refs.c (strip_conversion): New function.
8934 (can_group_stmts_p): Likewise.
8935 (vect_analyze_data_ref_accesses): Use it instead of checking
8936 for a pair of assignments.
8937 (vect_store_lanes_supported): Take a masked_p parameter.
8938 (vect_load_lanes_supported): Likewise.
8939 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8940 vect_store_lanes_supported and vect_load_lanes_supported.
8941 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8942 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8943 parameter. Don't allow gaps for masked accesses.
8944 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8945 and vect_load_lanes_supported.
8946 (get_load_store_type): Take a masked_p parameter and update
8947 call to get_group_load_store_type.
8948 (vectorizable_store): Update call to get_load_store_type.
8949 Handle IFN_MASK_STORE_LANES.
8950 (vectorizable_load): Update call to get_load_store_type.
8951 Handle IFN_MASK_LOAD_LANES.
8953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8954 Alan Hayward <alan.hayward@arm.com>
8955 David Sherwood <david.sherwood@arm.com>
8957 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8959 * config/aarch64/aarch64-protos.h
8960 (aarch64_sve_struct_memory_operand_p): Declare.
8961 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8962 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8963 (VPRED, vpred): Handle SVE structure modes.
8964 * config/aarch64/constraints.md (Utx): New constraint.
8965 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8966 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8967 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8968 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8969 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8970 structure modes. Split into pieces after RA.
8971 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8972 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8974 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8975 SVE structure modes.
8976 (aarch64_classify_address): Likewise.
8977 (sizetochar): Move earlier in file.
8978 (aarch64_print_operand): Handle SVE register lists.
8979 (aarch64_array_mode): New function.
8980 (aarch64_sve_struct_memory_operand_p): Likewise.
8981 (TARGET_ARRAY_MODE): Redefine.
8983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8984 Alan Hayward <alan.hayward@arm.com>
8985 David Sherwood <david.sherwood@arm.com>
8987 * target.def (array_mode): New target hook.
8988 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8989 * doc/tm.texi: Regenerate.
8990 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8991 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8992 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8994 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8997 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8998 Alan Hayward <alan.hayward@arm.com>
8999 David Sherwood <david.sherwood@arm.com>
9001 * fold-const.c (fold_binary_loc): Check the argument types
9002 rather than the result type when testing for a vector operation.
9004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9006 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9007 * doc/tm.texi: Regenerate.
9009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9014 (sve): Document new AArch64 extension.
9015 * doc/md.texi (w): Extend the description of the AArch64
9016 constraint to include SVE vectors.
9017 (Upl, Upa): Document new AArch64 predicate constraints.
9018 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9020 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9021 (msve-vector-bits=): New option.
9022 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9023 SVE when these are disabled.
9024 (sve): New extension.
9025 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9026 modes. Adjust their number of units based on aarch64_sve_vg.
9027 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9028 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9029 aarch64_addr_query_type.
9030 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9031 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9032 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9033 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9034 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9035 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9036 (aarch64_simd_imm_zero_p): Delete.
9037 (aarch64_check_zero_based_sve_index_immediate): Declare.
9038 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9039 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9040 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9041 (aarch64_sve_float_mul_immediate_p): Likewise.
9042 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9044 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9045 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9046 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9047 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9048 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9049 (aarch64_regmode_natural_size): Likewise.
9050 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9051 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9053 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9054 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9055 for VG and the SVE predicate registers.
9056 (V_ALIASES): Add a "z"-prefixed alias.
9057 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9058 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9059 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9060 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9061 (REG_CLASS_NAMES): Add entries for them.
9062 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9063 and the predicate registers.
9064 (aarch64_sve_vg): Declare.
9065 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9066 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9067 (REGMODE_NATURAL_SIZE): Define.
9068 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9070 * config/aarch64/aarch64.c: Include cfgrtl.h.
9071 (simd_immediate_info): Add a constructor for series vectors,
9072 and an associated step field.
9073 (aarch64_sve_vg): New variable.
9074 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9075 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9076 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9077 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9078 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9079 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9080 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9081 (aarch64_get_mask_mode): New functions.
9082 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9083 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9084 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9085 predicate modes and predicate registers. Explicitly restrict
9086 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9087 to store a vector mode if it is recognized by
9088 aarch64_classify_vector_mode.
9089 (aarch64_regmode_natural_size): New function.
9090 (aarch64_hard_regno_caller_save_mode): Return the original mode
9092 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9093 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9094 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9095 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9097 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9098 does not overlap dest if the function is frame-related. Handle
9100 (aarch64_split_add_offset): New function.
9101 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9102 them aarch64_add_offset.
9103 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9104 and update call to aarch64_sub_sp.
9105 (aarch64_add_cfa_expression): New function.
9106 (aarch64_expand_prologue): Pass extra temporary registers to the
9107 functions above. Handle the case in which we need to emit new
9108 DW_CFA_expressions for registers that were originally saved
9109 relative to the stack pointer, but now have to be expressed
9110 relative to the frame pointer.
9111 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9113 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9114 IP0 and IP1 values for SVE frames.
9115 (aarch64_expand_vec_series): New function.
9116 (aarch64_expand_sve_widened_duplicate): Likewise.
9117 (aarch64_expand_sve_const_vector): Likewise.
9118 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9119 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9120 into the register, rather than emitting a SET directly.
9121 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9122 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9123 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9124 (offset_9bit_signed_scaled_p): New functions.
9125 (aarch64_replicate_bitmask_imm): New function.
9126 (aarch64_bitmask_imm): Use it.
9127 (aarch64_cannot_force_const_mem): Reject expressions involving
9128 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9129 (aarch64_classify_index): Handle SVE indices, by requiring
9130 a plain register index with a scale that matches the element size.
9131 (aarch64_classify_address): Handle SVE addresses. Assert that
9132 the mode of the address is VOIDmode or an integer mode.
9133 Update call to aarch64_classify_symbol.
9134 (aarch64_classify_symbolic_expression): Update call to
9135 aarch64_classify_symbol.
9136 (aarch64_const_vec_all_in_range_p): New function.
9137 (aarch64_print_vector_float_operand): Likewise.
9138 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9139 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9140 and the FP immediates 1.0 and 0.5.
9141 (aarch64_print_address_internal): Handle SVE addresses.
9142 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9143 (aarch64_regno_regclass): Handle predicate registers.
9144 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9146 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9147 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9148 (aarch64_convert_sve_vector_bits): New function.
9149 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9150 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9152 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9153 Handle SVE vector and predicate modes. Accept VL-based constants
9154 that need only one temporary register, and VL offsets that require
9155 no temporary registers.
9156 (aarch64_conditional_register_usage): Mark the predicate registers
9157 as fixed if SVE isn't available.
9158 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9159 Return true for SVE vector and predicate modes.
9160 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9161 rather than an unsigned int. Handle SVE modes.
9162 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9164 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9166 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9167 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9168 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9169 (aarch64_sve_float_mul_immediate_p): New functions.
9170 (aarch64_sve_valid_immediate): New function.
9171 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9172 Explicitly reject structure modes. Check for INDEX constants.
9173 Handle PTRUE and PFALSE constants.
9174 (aarch64_check_zero_based_sve_index_immediate): New function.
9175 (aarch64_simd_imm_zero_p): Delete.
9176 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9177 vector modes. Accept constants in the range of CNT[BHWD].
9178 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9179 ask for an Advanced SIMD mode.
9180 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9181 (aarch64_simd_vector_alignment): Handle SVE predicates.
9182 (aarch64_vectorize_preferred_vector_alignment): New function.
9183 (aarch64_simd_vector_alignment_reachable): Use it instead of
9185 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9186 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9188 (MAX_VECT_LEN): Delete.
9189 (expand_vec_perm_d): Add a vec_flags field.
9190 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9191 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9192 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9194 (aarch64_evpc_rev): Rename to...
9195 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9196 (aarch64_evpc_rev_global): New function.
9197 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9198 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9200 (aarch64_evpc_sve_tbl): New function.
9201 (aarch64_expand_vec_perm_const_1): Update after rename of
9202 aarch64_evpc_rev. Handle SVE permutes too, trying
9203 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9204 than aarch64_evpc_tbl.
9205 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9206 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9207 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9208 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9209 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9210 (aarch64_expand_sve_vcond): New functions.
9211 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9212 of aarch64_vector_mode_p.
9213 (aarch64_dwarf_poly_indeterminate_value): New function.
9214 (aarch64_compute_pressure_classes): Likewise.
9215 (aarch64_can_change_mode_class): Likewise.
9216 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9217 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9218 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9219 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9220 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9221 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9222 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9223 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9225 (Dn, Dl, Dr): Accept const as well as const_vector.
9226 (Dz): Likewise. Compare against CONST0_RTX.
9227 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9228 of "vector" where appropriate.
9229 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9230 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9231 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9232 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9233 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9234 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9235 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9236 (v_int_equiv): Extend to SVE modes.
9237 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9239 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9240 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9241 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9242 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9243 (SVE_COND_FP_CMP): New int iterators.
9244 (perm_hilo): Handle the new unpack unspecs.
9245 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9247 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9248 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9249 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9250 (aarch64_equality_operator, aarch64_constant_vector_operand)
9251 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9252 (aarch64_sve_nonimmediate_operand): Likewise.
9253 (aarch64_sve_general_operand): Likewise.
9254 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9255 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9256 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9257 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9258 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9259 (aarch64_sve_float_arith_immediate): Likewise.
9260 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9261 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9262 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9263 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9264 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9265 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9266 (aarch64_sve_float_arith_operand): Likewise.
9267 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9268 (aarch64_sve_float_mul_operand): Likewise.
9269 (aarch64_sve_vec_perm_operand): Likewise.
9270 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9271 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9272 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9273 as well as const_vector.
9274 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9275 in file. Use CONST0_RTX and CONSTM1_RTX.
9276 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9277 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9278 Use aarch64_simd_imm_zero.
9279 * config/aarch64/aarch64-sve.md: New file.
9280 * config/aarch64/aarch64.md: Include it.
9281 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9282 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9283 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9284 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9285 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9286 (sve): New attribute.
9287 (enabled): Disable instructions with the sve attribute unless
9289 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9290 aarch64_expand_mov_immediate.
9291 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9292 CNT[BHSD] immediates.
9293 (movti): Split CONST_POLY_INT moves into two halves.
9294 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9295 Split additions that need a temporary here if the destination
9296 is the stack pointer.
9297 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9298 (*add<mode>3_poly_1): New instruction.
9299 (set_clobber_cc): New expander.
9301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9303 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9304 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9305 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9306 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9307 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9308 Change innermode from fixed_mode_size to machine_mode.
9309 (simplify_subreg): Update call accordingly. Handle a constant-sized
9310 subreg of a variable-length CONST_VECTOR.
9312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9313 Alan Hayward <alan.hayward@arm.com>
9314 David Sherwood <david.sherwood@arm.com>
9316 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9317 (add_offset_to_base): New function, split out from...
9318 (create_mem_ref): ...here. When handling a scale other than 1,
9319 check first whether the address is valid without the offset.
9320 Add it into the base if so, leaving the index and scale as-is.
9322 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9325 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9326 fold_for_warn before checking if arg2 is INTEGER_CST.
9328 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9330 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9331 (store_multiple_operation): Delete.
9332 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9333 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9334 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9335 guarded by TARGET_STRING.
9336 (rs6000_output_load_multiple): Delete.
9337 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9338 OPTION_MASK_STRING / TARGET_STRING handling.
9339 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9340 (const rs6000_opt_masks) <"string">: Change mask to 0.
9341 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9342 (MASK_STRING): Delete.
9343 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9345 (load_multiple): Delete.
9352 (store_multiple): Delete.
9359 (movmemsi_8reg): Delete.
9360 (corresponding unnamed define_insn): Delete.
9361 (movmemsi_6reg): Delete.
9362 (corresponding unnamed define_insn): Delete.
9363 (movmemsi_4reg): Delete.
9364 (corresponding unnamed define_insn): Delete.
9365 (movmemsi_2reg): Delete.
9366 (corresponding unnamed define_insn): Delete.
9367 (movmemsi_1reg): Delete.
9368 (corresponding unnamed define_insn): Delete.
9369 * config/rs6000/rs6000.opt (mno-string): New.
9370 (mstring): Replace by deprecation warning stub.
9371 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9373 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9375 * regrename.c (regrename_do_replace): If replacing the same
9376 reg multiple times, try to reuse last created gen_raw_REG.
9379 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9380 main to workaround a bug in GDB.
9382 2018-01-12 Tom de Vries <tom@codesourcery.com>
9385 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9387 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9389 PR rtl-optimization/80481
9390 * ira-color.c (get_cap_member): New function.
9391 (allocnos_conflict_by_live_ranges_p): Use it.
9392 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9393 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9395 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9398 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9399 (*saddl_se_1): Ditto.
9401 (*ssubl_se_1): Ditto.
9403 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9405 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9406 rather than wi::to_widest for DR_INITs.
9407 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9408 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9409 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9411 (vect_analyze_group_access_1): Note that here.
9413 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9415 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9416 polynomial type sizes.
9418 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9420 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9421 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9422 (gimple_add_tmp_var): Likewise.
9424 2018-01-12 Martin Liska <mliska@suse.cz>
9426 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9427 (gimple_alloc_sizes): Likewise.
9428 (dump_gimple_statistics): Use PRIu64 in printf format.
9429 * gimple.h: Change uint64_t to int.
9431 2018-01-12 Martin Liska <mliska@suse.cz>
9433 * tree-core.h: Use uint64_t instead of int.
9434 * tree.c (tree_node_counts): Likewise.
9435 (tree_node_sizes): Likewise.
9436 (dump_tree_statistics): Use PRIu64 in printf format.
9438 2018-01-12 Martin Liska <mliska@suse.cz>
9440 * Makefile.in: As qsort_chk is implemented in vec.c, add
9441 vec.o to linkage of gencfn-macros.
9442 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9443 passing the info to record_node_allocation_statistics.
9444 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9446 * ggc-common.c (struct ggc_usage): Add operator== and use
9447 it in operator< and compare function.
9448 * mem-stats.h (struct mem_usage): Likewise.
9449 * vec.c (struct vec_usage): Remove operator< and compare
9450 function. Can be simply inherited.
9452 2018-01-12 Martin Jambor <mjambor@suse.cz>
9455 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9456 * tree-ssa-math-opts.c: Include domwalk.h.
9457 (convert_mult_to_fma_1): New function.
9458 (fma_transformation_info): New type.
9459 (fma_deferring_state): Likewise.
9460 (cancel_fma_deferring): New function.
9461 (result_of_phi): Likewise.
9462 (last_fma_candidate_feeds_initial_phi): Likewise.
9463 (convert_mult_to_fma): Added deferring logic, split actual
9464 transformation to convert_mult_to_fma_1.
9465 (math_opts_dom_walker): New type.
9466 (math_opts_dom_walker::after_dom_children): New method, body moved
9467 here from pass_optimize_widening_mul::execute, added deferring logic
9469 (pass_optimize_widening_mul::execute): Moved most of code to
9470 math_opts_dom_walker::after_dom_children.
9471 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9472 * config/i386/i386.c (ix86_option_override_internal): Added
9473 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9475 2018-01-12 Richard Biener <rguenther@suse.de>
9478 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9479 inline instance vars.
9481 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9484 * config/rx/rx.c (rx_is_restricted_memory_address):
9487 2018-01-12 Richard Biener <rguenther@suse.de>
9489 PR tree-optimization/80846
9490 * target.def (split_reduction): New target hook.
9491 * targhooks.c (default_split_reduction): New function.
9492 * targhooks.h (default_split_reduction): Declare.
9493 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9494 target requests first reduce vectors by combining low and high
9496 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9497 (get_vectype_for_scalar_type_and_size): Export.
9498 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9499 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9500 * doc/tm.texi: Regenerate.
9501 * config/i386/i386.c (ix86_split_reduction): Implement
9502 TARGET_VECTORIZE_SPLIT_REDUCTION.
9504 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9507 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9508 in PIC mode except for TARGET_VXWORKS_RTP.
9509 * config/sparc/sparc.c: Include cfgrtl.h.
9510 (TARGET_INIT_PIC_REG): Define.
9511 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9512 (sparc_pic_register_p): New predicate.
9513 (sparc_legitimate_address_p): Use it.
9514 (sparc_legitimize_pic_address): Likewise.
9515 (sparc_delegitimize_address): Likewise.
9516 (sparc_mode_dependent_address_p): Likewise.
9517 (gen_load_pcrel_sym): Remove 4th parameter.
9518 (load_got_register): Adjust call to above. Remove obsolete stuff.
9519 (sparc_expand_prologue): Do not call load_got_register here.
9520 (sparc_flat_expand_prologue): Likewise.
9521 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9522 (sparc_use_pseudo_pic_reg): New function.
9523 (sparc_init_pic_reg): Likewise.
9524 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9525 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9527 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9529 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9530 Add item for branch_cost.
9532 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9534 PR rtl-optimization/83565
9535 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9536 not extend the result to a larger mode for rotate operations.
9537 (num_sign_bit_copies1): Likewise.
9539 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9542 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9544 Use values-Xc.o for -pedantic.
9545 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9547 2018-01-12 Martin Liska <mliska@suse.cz>
9550 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9552 (possible_polymorphic_call_targets): Use it.
9553 (ipa_devirt): Likewise.
9555 2018-01-12 Martin Liska <mliska@suse.cz>
9557 * profile-count.h (enum profile_quality): Use 0 as invalid
9558 enum value of profile_quality.
9560 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9562 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9563 -mext-string options.
9565 2018-01-12 Richard Biener <rguenther@suse.de>
9567 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9568 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9569 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9571 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9573 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9575 * configure.ac (--with-long-double-format): Add support for the
9576 configuration option to change the default long double format on
9578 * config.gcc (powerpc*-linux*-*): Likewise.
9579 * configure: Regenerate.
9580 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9581 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9582 used without modification.
9584 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9586 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9587 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9588 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9589 MISC_BUILTIN_SPEC_BARRIER.
9590 (rs6000_init_builtins): Likewise.
9591 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9593 (speculation_barrier): New define_insn.
9594 * doc/extend.texi: Document __builtin_speculation_barrier.
9596 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9599 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9600 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9601 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9603 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9604 integral modes instead of "ss" and "sd".
9605 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9606 vectors with 32-bit and 64-bit elements.
9607 (vecdupssescalarmodesuffix): New mode attribute.
9608 (vec_dup<mode>): Use it.
9610 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9613 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9614 frame if argument is passed on stack.
9616 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9619 * ree.c (combine_reaching_defs): Optimize also
9620 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9621 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9623 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9626 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9628 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9631 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9632 after they are computed.
9634 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9636 PR tree-optimization/83695
9637 * gimple-loop-linterchange.cc
9638 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9639 reset cached scev information after interchange.
9640 (pass_linterchange::execute): Remove call to scev_reset_htab.
9642 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9644 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9645 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9646 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9647 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9648 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9649 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9650 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9651 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9652 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9653 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9654 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9655 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9656 (V_lane_reg): Likewise.
9657 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9659 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9660 (vfmal_lane_low<mode>_intrinsic,
9661 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9662 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9663 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9664 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9665 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9666 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9668 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9670 * config/arm/arm-cpus.in (fp16fml): New feature.
9671 (ALL_SIMD): Add fp16fml.
9672 (armv8.2-a): Add fp16fml as an option.
9673 (armv8.3-a): Likewise.
9674 (armv8.4-a): Add fp16fml as part of fp16.
9675 * config/arm/arm.h (TARGET_FP16FML): Define.
9676 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9678 * config/arm/arm-modes.def (V2HF): Define.
9679 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9680 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9681 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9682 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9683 vfmsl_low, vfmsl_high): New set of builtins.
9684 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9685 (vfml_op): New code attribute.
9686 (VFMLHALVES): New int iterator.
9687 (VFML, VFMLSEL): New mode attributes.
9688 (V_reg): Define mapping for V2HF.
9689 (V_hi, V_lo): New mode attributes.
9690 (VF_constraint): Likewise.
9691 (vfml_half, vfml_half_selector): New int attributes.
9692 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9694 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9695 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9697 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9698 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9699 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9700 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9702 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9703 Document new effective target and option set.
9705 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9707 * config/arm/arm-cpus.in (armv8_4): New feature.
9708 (ARMv8_4a): New fgroup.
9709 (armv8.4-a): New arch.
9710 * config/arm/arm-tables.opt: Regenerate.
9711 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9712 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9713 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9714 Add matching rules for -march=armv8.4-a and extensions.
9715 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9717 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9720 * config/rx/rx.md (BW): New mode attribute.
9721 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9723 2018-01-11 Richard Biener <rguenther@suse.de>
9725 PR tree-optimization/83435
9726 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9727 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9728 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9730 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9731 Alan Hayward <alan.hayward@arm.com>
9732 David Sherwood <david.sherwood@arm.com>
9734 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9736 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9737 (aarch64_print_address_internal): Use it to check for a zero offset.
9739 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9740 Alan Hayward <alan.hayward@arm.com>
9741 David Sherwood <david.sherwood@arm.com>
9743 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9744 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9745 Return a poly_int64 rather than a HOST_WIDE_INT.
9746 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9747 rather than a HOST_WIDE_INT.
9748 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9749 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9750 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9751 final_offset from HOST_WIDE_INT to poly_int64.
9752 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9753 to_constant when getting the number of units in an Advanced SIMD
9755 (aarch64_builtin_vectorized_function): Check for a constant number
9757 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9759 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9760 attribute instead of GET_MODE_NUNITS.
9761 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9762 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9763 GET_MODE_SIZE for fixed-size registers.
9764 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9765 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9766 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9767 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9768 (aarch64_print_operand, aarch64_print_address_internal)
9769 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9770 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9771 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9772 Handle polynomial GET_MODE_SIZE.
9773 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9774 wider than SImode without modification.
9775 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9776 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9777 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9778 passing and returning SVE modes.
9779 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9780 rather than GEN_INT.
9781 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9782 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9783 (aarch64_allocate_and_probe_stack_space): Likewise.
9784 (aarch64_layout_frame): Cope with polynomial offsets.
9785 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9786 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9788 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9789 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9790 poly_int64 rather than a HOST_WIDE_INT.
9791 (aarch64_get_separate_components, aarch64_process_components)
9792 (aarch64_expand_prologue, aarch64_expand_epilogue)
9793 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9794 (aarch64_anchor_offset): New function, split out from...
9795 (aarch64_legitimize_address): ...here.
9796 (aarch64_builtin_vectorization_cost): Handle polynomial
9797 TYPE_VECTOR_SUBPARTS.
9798 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9800 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9801 number of elements from the PARALLEL rather than the mode.
9802 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9803 rather than GET_MODE_BITSIZE.
9804 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9805 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9806 (aarch64_expand_vec_perm_const_1): Handle polynomial
9807 d->perm.length () and d->perm elements.
9808 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9809 Apply to_constant to d->perm elements.
9810 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9811 polynomial CONST_VECTOR_NUNITS.
9812 (aarch64_move_pointer): Take amount as a poly_int64 rather
9814 (aarch64_progress_pointer): Avoid temporary variable.
9815 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9816 the mode attribute instead of GET_MODE.
9818 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9819 Alan Hayward <alan.hayward@arm.com>
9820 David Sherwood <david.sherwood@arm.com>
9822 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9823 x exists before using it.
9824 (aarch64_add_constant_internal): Rename to...
9825 (aarch64_add_offset_1): ...this. Replace regnum with separate
9826 src and dest rtxes. Handle the case in which they're different,
9827 including when the offset is zero. Replace scratchreg with an rtx.
9828 Use 2 additions if there is no spare register into which we can
9829 move a 16-bit constant.
9830 (aarch64_add_constant): Delete.
9831 (aarch64_add_offset): Replace reg with separate src and dest
9832 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9833 Use aarch64_add_offset_1.
9834 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9835 an rtx rather than an int. Take the delta as a poly_int64
9836 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9837 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9838 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9839 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9840 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9842 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9843 aarch64_add_constant.
9845 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9847 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9848 Use scalar_float_mode.
9850 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9852 * config/aarch64/aarch64-simd.md
9853 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9854 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9855 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9856 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9857 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9858 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9859 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9860 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9861 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9862 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9864 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9867 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9868 targ_options->x_arm_arch_string is non NULL.
9870 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9872 * config/aarch64/aarch64.h
9873 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9875 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9878 * expmed.c (emit_store_flag_force): Swap if const op0
9879 and change VOIDmode to mode of op0.
9881 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9883 PR rtl-optimization/83761
9884 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9885 than bytes to mode_for_size.
9887 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9890 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9891 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9894 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9897 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9898 when in layout mode.
9899 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9900 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9903 2018-01-10 Michael Collison <michael.collison@arm.com>
9905 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9906 * config/aarch64/aarch64-option-extension.def: Add
9907 AARCH64_OPT_EXTENSION of 'fp16fml'.
9908 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9909 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9910 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9911 * config/aarch64/constraints.md (Ui7): New constraint.
9912 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9913 (VFMLA_SEL_W): Ditto.
9916 (VFMLA16_LOW): New int iterator.
9917 (VFMLA16_HIGH): Ditto.
9918 (UNSPEC_FMLAL): New unspec.
9919 (UNSPEC_FMLSL): Ditto.
9920 (UNSPEC_FMLAL2): Ditto.
9921 (UNSPEC_FMLSL2): Ditto.
9922 (f16mac): New code attribute.
9923 * config/aarch64/aarch64-simd-builtins.def
9924 (aarch64_fmlal_lowv2sf): Ditto.
9925 (aarch64_fmlsl_lowv2sf): Ditto.
9926 (aarch64_fmlalq_lowv4sf): Ditto.
9927 (aarch64_fmlslq_lowv4sf): Ditto.
9928 (aarch64_fmlal_highv2sf): Ditto.
9929 (aarch64_fmlsl_highv2sf): Ditto.
9930 (aarch64_fmlalq_highv4sf): Ditto.
9931 (aarch64_fmlslq_highv4sf): Ditto.
9932 (aarch64_fmlal_lane_lowv2sf): Ditto.
9933 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9934 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9935 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9936 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9937 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9938 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9939 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9940 (aarch64_fmlal_lane_highv2sf): Ditto.
9941 (aarch64_fmlsl_lane_highv2sf): Ditto.
9942 (aarch64_fmlal_laneq_highv2sf): Ditto.
9943 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9944 (aarch64_fmlalq_lane_highv4sf): Ditto.
9945 (aarch64_fmlsl_lane_highv4sf): Ditto.
9946 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9947 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9948 * config/aarch64/aarch64-simd.md:
9949 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9950 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9951 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9952 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9953 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9954 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9955 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9956 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9957 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9958 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9959 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9960 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9961 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9962 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9963 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9964 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9965 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9966 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9967 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9968 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9969 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9970 (vfmlsl_low_u32): Ditto.
9971 (vfmlalq_low_u32): Ditto.
9972 (vfmlslq_low_u32): Ditto.
9973 (vfmlal_high_u32): Ditto.
9974 (vfmlsl_high_u32): Ditto.
9975 (vfmlalq_high_u32): Ditto.
9976 (vfmlslq_high_u32): Ditto.
9977 (vfmlal_lane_low_u32): Ditto.
9978 (vfmlsl_lane_low_u32): Ditto.
9979 (vfmlal_laneq_low_u32): Ditto.
9980 (vfmlsl_laneq_low_u32): Ditto.
9981 (vfmlalq_lane_low_u32): Ditto.
9982 (vfmlslq_lane_low_u32): Ditto.
9983 (vfmlalq_laneq_low_u32): Ditto.
9984 (vfmlslq_laneq_low_u32): Ditto.
9985 (vfmlal_lane_high_u32): Ditto.
9986 (vfmlsl_lane_high_u32): Ditto.
9987 (vfmlal_laneq_high_u32): Ditto.
9988 (vfmlsl_laneq_high_u32): Ditto.
9989 (vfmlalq_lane_high_u32): Ditto.
9990 (vfmlslq_lane_high_u32): Ditto.
9991 (vfmlalq_laneq_high_u32): Ditto.
9992 (vfmlslq_laneq_high_u32): Ditto.
9993 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9994 (AARCH64_FL_FOR_ARCH8_4): New.
9995 (AARCH64_ISA_F16FML): New ISA flag.
9996 (TARGET_F16FML): New feature flag for fp16fml.
9997 (doc/invoke.texi): Document new fp16fml option.
9999 2018-01-10 Michael Collison <michael.collison@arm.com>
10001 * config/aarch64/aarch64-builtins.c:
10002 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10003 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10004 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10005 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10006 (AARCH64_ISA_SHA3): New ISA flag.
10007 (TARGET_SHA3): New feature flag for sha3.
10008 * config/aarch64/iterators.md (sha512_op): New int attribute.
10009 (CRYPTO_SHA512): New int iterator.
10010 (UNSPEC_SHA512H): New unspec.
10011 (UNSPEC_SHA512H2): Ditto.
10012 (UNSPEC_SHA512SU0): Ditto.
10013 (UNSPEC_SHA512SU1): Ditto.
10014 * config/aarch64/aarch64-simd-builtins.def
10015 (aarch64_crypto_sha512hqv2di): New builtin.
10016 (aarch64_crypto_sha512h2qv2di): Ditto.
10017 (aarch64_crypto_sha512su0qv2di): Ditto.
10018 (aarch64_crypto_sha512su1qv2di): Ditto.
10019 (aarch64_eor3qv8hi): Ditto.
10020 (aarch64_rax1qv2di): Ditto.
10021 (aarch64_xarqv2di): Ditto.
10022 (aarch64_bcaxqv8hi): Ditto.
10023 * config/aarch64/aarch64-simd.md:
10024 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10025 (aarch64_crypto_sha512su0qv2di): Ditto.
10026 (aarch64_crypto_sha512su1qv2di): Ditto.
10027 (aarch64_eor3qv8hi): Ditto.
10028 (aarch64_rax1qv2di): Ditto.
10029 (aarch64_xarqv2di): Ditto.
10030 (aarch64_bcaxqv8hi): Ditto.
10031 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10032 (vsha512h2q_u64): Ditto.
10033 (vsha512su0q_u64): Ditto.
10034 (vsha512su1q_u64): Ditto.
10035 (veor3q_u16): Ditto.
10036 (vrax1q_u64): Ditto.
10037 (vxarq_u64): Ditto.
10038 (vbcaxq_u16): Ditto.
10039 * config/arm/types.md (crypto_sha512): New type attribute.
10040 (crypto_sha3): Ditto.
10041 (doc/invoke.texi): Document new sha3 option.
10043 2018-01-10 Michael Collison <michael.collison@arm.com>
10045 * config/aarch64/aarch64-builtins.c:
10046 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10047 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10048 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10049 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10050 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10051 (AARCH64_ISA_SM4): New ISA flag.
10052 (TARGET_SM4): New feature flag for sm4.
10053 * config/aarch64/aarch64-simd-builtins.def
10054 (aarch64_sm3ss1qv4si): Ditto.
10055 (aarch64_sm3tt1aq4si): Ditto.
10056 (aarch64_sm3tt1bq4si): Ditto.
10057 (aarch64_sm3tt2aq4si): Ditto.
10058 (aarch64_sm3tt2bq4si): Ditto.
10059 (aarch64_sm3partw1qv4si): Ditto.
10060 (aarch64_sm3partw2qv4si): Ditto.
10061 (aarch64_sm4eqv4si): Ditto.
10062 (aarch64_sm4ekeyqv4si): Ditto.
10063 * config/aarch64/aarch64-simd.md:
10064 (aarch64_sm3ss1qv4si): Ditto.
10065 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10066 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10067 (aarch64_sm4eqv4si): Ditto.
10068 (aarch64_sm4ekeyqv4si): Ditto.
10069 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10070 (sm3part_op): Ditto.
10071 (CRYPTO_SM3TT): Ditto.
10072 (CRYPTO_SM3PART): Ditto.
10073 (UNSPEC_SM3SS1): New unspec.
10074 (UNSPEC_SM3TT1A): Ditto.
10075 (UNSPEC_SM3TT1B): Ditto.
10076 (UNSPEC_SM3TT2A): Ditto.
10077 (UNSPEC_SM3TT2B): Ditto.
10078 (UNSPEC_SM3PARTW1): Ditto.
10079 (UNSPEC_SM3PARTW2): Ditto.
10080 (UNSPEC_SM4E): Ditto.
10081 (UNSPEC_SM4EKEY): Ditto.
10082 * config/aarch64/constraints.md (Ui2): New constraint.
10083 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10084 * config/arm/types.md (crypto_sm3): New type attribute.
10085 (crypto_sm4): Ditto.
10086 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10087 (vsm3tt1aq_u32): Ditto.
10088 (vsm3tt1bq_u32): Ditto.
10089 (vsm3tt2aq_u32): Ditto.
10090 (vsm3tt2bq_u32): Ditto.
10091 (vsm3partw1q_u32): Ditto.
10092 (vsm3partw2q_u32): Ditto.
10093 (vsm4eq_u32): Ditto.
10094 (vsm4ekeyq_u32): Ditto.
10095 (doc/invoke.texi): Document new sm4 option.
10097 2018-01-10 Michael Collison <michael.collison@arm.com>
10099 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10100 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10101 (AARCH64_FL_FOR_ARCH8_4): New.
10102 (AARCH64_FL_V8_4): New flag.
10103 (doc/invoke.texi): Document new armv8.4-a option.
10105 2018-01-10 Michael Collison <michael.collison@arm.com>
10107 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10108 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10109 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10110 * config/aarch64/aarch64-option-extension.def: Add
10111 AARCH64_OPT_EXTENSION of 'sha2'.
10112 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10113 (crypto): Disable sha2 and aes if crypto disabled.
10114 (crypto): Enable aes and sha2 if enabled.
10115 (simd): Disable sha2 and aes if simd disabled.
10116 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10118 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10119 (TARGET_SHA2): New feature flag for sha2.
10120 (TARGET_AES): New feature flag for aes.
10121 * config/aarch64/aarch64-simd.md:
10122 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10123 conditional on TARGET_AES.
10124 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10125 (aarch64_crypto_sha1hsi): Make pattern conditional
10127 (aarch64_crypto_sha1hv4si): Ditto.
10128 (aarch64_be_crypto_sha1hv4si): Ditto.
10129 (aarch64_crypto_sha1su1v4si): Ditto.
10130 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10131 (aarch64_crypto_sha1su0v4si): Ditto.
10132 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10133 (aarch64_crypto_sha256su0v4si): Ditto.
10134 (aarch64_crypto_sha256su1v4si): Ditto.
10135 (doc/invoke.texi): Document new aes and sha2 options.
10137 2018-01-10 Martin Sebor <msebor@redhat.com>
10139 PR tree-optimization/83781
10140 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10143 2018-01-11 Martin Sebor <msebor@gmail.com>
10144 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10146 PR tree-optimization/83501
10147 PR tree-optimization/81703
10149 * tree-ssa-strlen.c (get_string_cst): Rename...
10150 (get_string_len): ...to this. Handle global constants.
10151 (handle_char_store): Adjust.
10153 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10154 Jim Wilson <jimw@sifive.com>
10156 * config/riscv/riscv-protos.h (riscv_output_return): New.
10157 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10158 (riscv_attribute_table, riscv_output_return),
10159 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10160 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10161 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10163 (riscv_expand_prologue): Add early return for naked function.
10164 (riscv_expand_epilogue): Likewise.
10165 (riscv_function_ok_for_sibcall): Return false for naked function.
10166 (riscv_set_current_function): New.
10167 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10168 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10169 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10170 * doc/extend.texi (RISC-V Function Attributes): New.
10172 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10174 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10175 check for 128-bit long double before checking TCmode.
10176 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10177 128-bit long doubles before checking TFmode or TCmode.
10178 (FLOAT128_IBM_P): Likewise.
10180 2018-01-10 Martin Sebor <msebor@redhat.com>
10182 PR tree-optimization/83671
10183 * builtins.c (c_strlen): Unconditionally return zero for the empty
10185 Use -Warray-bounds for warnings.
10186 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10187 for non-constant array indices with COMPONENT_REF, arrays of
10188 arrays, and pointers to arrays.
10189 (gimple_fold_builtin_strlen): Determine and set length range for
10190 non-constant character arrays.
10192 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10194 PR middle-end/81897
10195 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10198 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10200 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10202 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10205 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10206 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10207 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10208 indexed_or_indirect_operand predicate.
10209 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10210 (*vsx_le_perm_load_v8hi): Likewise.
10211 (*vsx_le_perm_load_v16qi): Likewise.
10212 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10213 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10214 (*vsx_le_perm_store_v8hi): Likewise.
10215 (*vsx_le_perm_store_v16qi): Likewise.
10216 (eight unnamed splitters): Likewise.
10218 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10220 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10221 * config/rs6000/emmintrin.h: Likewise.
10222 * config/rs6000/mmintrin.h: Likewise.
10223 * config/rs6000/xmmintrin.h: Likewise.
10225 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10228 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10230 * tree.c (tree_nop_conversion): Return true for location wrapper
10232 (maybe_wrap_with_location): New function.
10233 (selftest::check_strip_nops): New function.
10234 (selftest::test_location_wrappers): New function.
10235 (selftest::tree_c_tests): Call it.
10236 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10237 (maybe_wrap_with_location): New decl.
10238 (EXPR_LOCATION_WRAPPER_P): New macro.
10239 (location_wrapper_p): New inline function.
10240 (tree_strip_any_location_wrapper): New inline function.
10242 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10245 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10246 stack_realign_offset for the largest alignment of stack slot
10248 (ix86_find_max_used_stack_alignment): New function.
10249 (ix86_finalize_stack_frame_flags): Use it. Set
10250 max_used_stack_alignment if we don't realign stack.
10251 * config/i386/i386.h (machine_function): Add
10252 max_used_stack_alignment.
10254 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10256 * config/arm/arm.opt (-mbranch-cost): New option.
10257 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10260 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10263 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10264 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10266 2018-01-10 Richard Biener <rguenther@suse.de>
10269 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10270 early out so it also covers the case where we have a non-NULL
10273 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10275 PR tree-optimization/83753
10276 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10277 for non-strided grouped accesses if the number of elements is 1.
10279 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10282 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10283 * i386.h (TARGET_USE_GATHER): Define.
10284 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10286 2018-01-10 Martin Liska <mliska@suse.cz>
10289 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10290 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10292 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10293 CLEANUP_NO_PARTITIONING is not set.
10295 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10297 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10298 for vectors, as a partial revert of r254296.
10299 * rtl.h (const_vec_p): Delete.
10300 (const_vec_duplicate_p): Don't test for vector CONSTs.
10301 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10302 * expmed.c (make_tree): Likewise.
10305 * common.md (E, F): Use CONSTANT_P instead of checking for
10307 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10308 checking for CONST_VECTOR.
10310 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10312 PR middle-end/83575
10313 * predict.c (force_edge_cold): Handle in more sane way edges
10314 with no prediction.
10316 2018-01-09 Carl Love <cel@us.ibm.com>
10318 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10320 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10321 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10322 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10323 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10324 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10325 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10326 * config/rs6000/rs6000-protos.h: Add extern defition for
10327 rs6000_generate_float2_double_code.
10328 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10330 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10331 (float2_v2df): Add define_expand.
10333 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10336 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10337 op_mode in the force_to_mode call.
10339 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10341 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10342 instead of checking each element individually.
10343 (aarch64_evpc_uzp): Likewise.
10344 (aarch64_evpc_zip): Likewise.
10345 (aarch64_evpc_ext): Likewise.
10346 (aarch64_evpc_rev): Likewise.
10347 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10348 instead of checking each element individually. Return true without
10350 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10351 whether all selected elements come from the same input, instead of
10352 checking each element individually. Remove calls to gen_rtx_REG,
10353 start_sequence and end_sequence and instead assert that no rtl is
10356 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10358 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10359 order of HIGH and CONST checks.
10361 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10363 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10364 if the destination isn't an SSA_NAME.
10366 2018-01-09 Richard Biener <rguenther@suse.de>
10368 PR tree-optimization/83668
10369 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10371 (canonicalize_loop_form): ... here, renamed from ...
10372 (canonicalize_loop_closed_ssa_form): ... this and amended to
10373 swap successor edges for loop exit blocks to make us use
10374 the RPO order we need for initial schedule generation.
10376 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10378 PR tree-optimization/64811
10379 * match.pd: When optimizing comparisons with Inf, avoid
10380 introducing or losing exceptions from comparisons with NaN.
10382 2018-01-09 Martin Liska <mliska@suse.cz>
10385 * asan.c (shadow_mem_size): Add gcc_assert.
10387 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10389 Don't save registers in main().
10392 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10393 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10394 * config/avr/avr.c (avr_set_current_function): Don't error if
10395 naked, OS_task or OS_main are specified at the same time.
10396 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10398 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10400 * common/config/avr/avr-common.c (avr_option_optimization_table):
10401 Switch on -mmain-is-OS_task for optimizing compilations.
10403 2018-01-09 Richard Biener <rguenther@suse.de>
10405 PR tree-optimization/83572
10406 * graphite.c: Include cfganal.h.
10407 (graphite_transform_loops): Connect infinite loops to exit
10408 and remove fake edges at the end.
10410 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10412 * ipa-inline.c (edge_badness): Revert accidental checkin.
10414 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10417 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10418 symbols; not inline clones.
10420 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10423 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10424 hard registers. Formatting fixes.
10426 PR preprocessor/83722
10427 * gcc.c (try_generate_repro): Pass
10428 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10429 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10432 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10433 Kito Cheng <kito.cheng@gmail.com>
10435 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10436 (riscv_leaf_function_p): Delete.
10437 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10439 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10441 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10443 (do_ifelse): New function.
10444 (do_isel): New function.
10445 (do_sub3): New function.
10446 (do_add3): New function.
10447 (do_load_mask_compare): New function.
10448 (do_overlap_load_compare): New function.
10449 (expand_compare_loop): New function.
10450 (expand_block_compare): Call expand_compare_loop() when appropriate.
10451 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10452 option description.
10453 (-mblock-compare-inline-loop-limit): New option.
10455 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10458 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10459 Reverse order of second and third operands in first alternative.
10460 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10461 of first and second elements in UNSPEC_VPERMR vector.
10462 (altivec_expand_vec_perm_le): Likewise.
10464 2018-01-08 Jeff Law <law@redhat.com>
10466 PR rtl-optimizatin/81308
10467 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10468 (process_switch): If group_case_labels makes a change, then set
10470 (pass_convert_switch::execute): If a switch is converted, then
10471 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10473 PR rtl-optimization/81308
10474 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10477 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10479 PR target/83663 - Revert r255946
10480 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10481 generation for cases where splatting a value is not useful.
10482 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10483 across a vec_duplicate and a paradoxical subreg forming a vector
10484 mode to a vec_concat.
10486 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10488 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10489 -march=armv8.3-a variants.
10490 * config/arm/t-multilib: Likewise.
10491 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10493 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10495 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10497 (cceq_ior_compare_complement): Give it a name so I can use it, and
10498 change boolean_or_operator predicate to boolean_operator so it can
10499 be used to generate a crand.
10500 (eqne): New code iterator.
10501 (bd/bd_neg): New code_attrs.
10502 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10503 a single define_insn.
10504 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10505 decrement (bdnzt/bdnzf/bdzt/bdzf).
10506 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10507 with the new names of the branch decrement patterns, and added the
10508 names of the branch decrement conditional patterns.
10510 2018-01-08 Richard Biener <rguenther@suse.de>
10512 PR tree-optimization/83563
10513 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10516 2018-01-08 Richard Biener <rguenther@suse.de>
10518 PR middle-end/83713
10519 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10521 2018-01-08 Richard Biener <rguenther@suse.de>
10523 PR tree-optimization/83685
10524 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10525 references to abnormals.
10527 2018-01-08 Richard Biener <rguenther@suse.de>
10530 * dwarf2out.c (output_indirect_strings): Handle empty
10531 skeleton_debug_str_hash.
10532 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10534 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10536 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10537 (emit_store_direct): Likewise.
10538 (arc_trampoline_adjust_address): Likewise.
10539 (arc_asm_trampoline_template): New function.
10540 (arc_initialize_trampoline): Use asm_trampoline_template.
10541 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10542 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10543 * config/arc/arc.md (flush_icache): Delete pattern.
10545 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10547 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10548 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10551 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10554 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10555 by not USED_FOR_TARGET.
10556 (make_pass_resolve_sw_modes): Likewise.
10558 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10560 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10563 2018-01-08 Richard Biener <rguenther@suse.de>
10565 PR middle-end/83580
10566 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10568 2018-01-08 Richard Biener <rguenther@suse.de>
10570 PR middle-end/83517
10571 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10573 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10575 PR middle-end/81897
10576 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10577 basic blocks with a small number of successors.
10578 (convert_control_dep_chain_into_preds): Improve handling of
10580 (dump_predicates): Split apart into...
10581 (dump_pred_chain): ...here...
10582 (dump_pred_info): ...and here.
10583 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10584 (can_chain_union_be_invalidated_p): Improve check for invalidation
10586 (uninit_uses_cannot_happen): Avoid unnecessary if
10587 convert_control_dep_chain_into_preds yielded nothing.
10589 2018-01-06 Martin Sebor <msebor@redhat.com>
10591 PR tree-optimization/83640
10592 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10593 subtracting negative offset from size.
10594 (builtin_access::overlap): Adjust offset bounds of the access to fall
10595 within the size of the object if possible.
10597 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10599 PR rtl-optimization/83699
10600 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10601 extract_bit_field_as_subreg to cases in which the extracted
10602 value is also a vector.
10604 * lra-constraints.c (process_alt_operands): Test for the equivalence
10605 substitutions when detecting a possible reload cycle.
10607 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10610 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10611 by default if flag_selective_schedling{,2}. Formatting fixes.
10613 PR rtl-optimization/83682
10614 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10615 if it has non-VECTOR_MODE element mode.
10616 (vec_duplicate_p): Likewise.
10618 PR middle-end/83694
10619 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10620 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10622 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10625 * config/i386/i386-builtin.def
10626 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10627 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10628 Require also OPTION_MASK_ISA_AVX512F in addition to
10629 OPTION_MASK_ISA_GFNI.
10630 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10631 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10632 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10633 to OPTION_MASK_ISA_GFNI.
10634 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10635 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10636 OPTION_MASK_ISA_AVX512BW.
10637 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10638 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10639 addition to OPTION_MASK_ISA_GFNI.
10640 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10641 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10642 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10643 to OPTION_MASK_ISA_GFNI.
10644 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10645 a requirement for all ISAs rather than any of them with a few
10647 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10649 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10650 bitmasks to be enabled with 3 exceptions, instead of requiring any
10651 enabled ISA with lots of exceptions.
10652 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10653 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10654 Change avx512bw in isa attribute to avx512f.
10655 * config/i386/sgxintrin.h: Add license boilerplate.
10656 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10657 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10658 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10659 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10661 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10662 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10663 temporarily sse2 rather than sse if not enabled already.
10666 * config/i386/sse.md (VI248_VLBW): Rename to ...
10667 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10668 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10669 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10670 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10671 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10672 mode iterator instead of VI248_VLBW.
10674 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10676 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10677 (record_modified): Skip clobbers; add debug output.
10678 (param_change_prob): Use sreal frequencies.
10680 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10682 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10683 punt for user-aligned variables.
10685 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10687 * tree-chrec.c (chrec_contains_symbols): Return true for
10690 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10693 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10694 of (x|y) == x for BICS pattern.
10696 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10698 PR tree-optimization/83605
10699 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10700 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10703 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10705 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10706 * config/epiphany/rtems.h: New file.
10708 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10709 Uros Bizjak <ubizjak@gmail.com>
10712 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10713 QIreg_operand instead of register_operand predicate.
10714 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10715 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10716 comments instead of -fmitigate[-_]rop.
10718 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10721 * cgraphunit.c (symbol_table::compile): Switch to text_section
10722 before calling assembly_start debug hook.
10723 * run-rtl-passes.c (run_rtl_passes): Likewise.
10726 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10728 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10729 range_int_cst_p rather than !symbolic_range_p before calling
10730 extract_range_from_multiplicative_op_1.
10732 2018-01-04 Jeff Law <law@redhat.com>
10734 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10735 redundant test in assertion.
10737 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10739 * doc/rtl.texi: Document machine_mode wrapper classes.
10741 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10743 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10744 using tree_to_uhwi.
10746 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10748 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10749 the VEC_PERM_EXPR fold to fail.
10751 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10754 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10755 to switched_sections.
10757 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10760 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10761 test for d.testing.
10763 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10766 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10767 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10769 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10772 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10773 is BLKmode and bitpos not zero or mode change is needed.
10775 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10778 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10781 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10784 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10785 instead of MULT rtx. Update all corresponding splitters.
10786 (*saddl_se): Ditto.
10787 (*ssub<modesuffix>): Ditto.
10788 (*ssubl_se): Ditto.
10789 (*cmp_sadd_di): Update split patterns.
10790 (*cmp_sadd_si): Ditto.
10791 (*cmp_sadd_sidi): Ditto.
10792 (*cmp_ssub_di): Ditto.
10793 (*cmp_ssub_si): Ditto.
10794 (*cmp_ssub_sidi): Ditto.
10795 * config/alpha/predicates.md (const23_operand): New predicate.
10796 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10797 Look for ASHIFT, not MULT inner operand.
10798 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10800 2018-01-04 Martin Liska <mliska@suse.cz>
10802 PR gcov-profile/83669
10803 * gcov.c (output_intermediate_file): Add version to intermediate
10805 * doc/gcov.texi: Document new field 'version' in intermediate
10806 file format. Fix location of '-k' option of gcov command.
10808 2018-01-04 Martin Liska <mliska@suse.cz>
10811 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10813 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10815 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10817 2018-01-03 Martin Sebor <msebor@redhat.com>
10819 PR tree-optimization/83655
10820 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10821 checking calls with invalid arguments.
10823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10825 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10826 (vectorizable_mask_load_store): Delete.
10827 (vectorizable_call): Return false for masked loads and stores.
10828 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10829 instead of gimple_assign_rhs1.
10830 (vectorizable_load): Handle IFN_MASK_LOAD.
10831 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10835 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10837 (vectorizable_mask_load_store): ...here.
10838 (vectorizable_load): ...and here.
10840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10842 * tree-vect-stmts.c (vect_build_all_ones_mask)
10843 (vect_build_zero_merge_argument): New functions, split out from...
10844 (vectorizable_load): ...here.
10846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10848 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10850 (vectorizable_mask_load_store): ...here.
10851 (vectorizable_store): ...and here.
10853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10855 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10857 (vectorizable_mask_load_store): ...here.
10859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10861 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10862 (vect_model_store_cost): Take a vec_load_store_type instead of a
10864 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10865 (vect_model_store_cost): Take a vec_load_store_type instead of a
10867 (vectorizable_mask_load_store): Update accordingly.
10868 (vectorizable_store): Likewise.
10869 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10873 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10874 IFN_MASK_LOAD calls here rather than...
10875 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10878 Alan Hayward <alan.hayward@arm.com>
10879 David Sherwood <david.sherwood@arm.com>
10881 * expmed.c (extract_bit_field_1): For vector extracts,
10882 fall back to extract_bit_field_as_subreg if vec_extract
10885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10886 Alan Hayward <alan.hayward@arm.com>
10887 David Sherwood <david.sherwood@arm.com>
10889 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10890 they are variable or constant sized.
10891 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10892 slots for constant-sized data.
10894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10895 Alan Hayward <alan.hayward@arm.com>
10896 David Sherwood <david.sherwood@arm.com>
10898 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10899 handling COND_EXPRs with boolean comparisons, try to find a better
10900 basis for the mask type than the boolean itself.
10902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10904 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10905 is calculated and how it can be overridden.
10906 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10907 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10909 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10913 Alan Hayward <alan.hayward@arm.com>
10914 David Sherwood <david.sherwood@arm.com>
10916 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10917 Remove the mode argument.
10918 (aarch64_simd_valid_immediate): Remove the mode and inverse
10920 * config/aarch64/iterators.md (bitsize): New iterator.
10921 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10922 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10923 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10924 aarch64_simd_valid_immediate.
10925 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10926 (aarch64_reg_or_bic_imm): Likewise.
10927 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10928 with an insn_type enum and msl with a modifier_type enum.
10929 Replace element_width with a scalar_mode. Change the shift
10930 to unsigned int. Add constructors for scalar_float_mode and
10931 scalar_int_mode elements.
10932 (aarch64_vect_float_const_representable_p): Delete.
10933 (aarch64_can_const_movi_rtx_p)
10934 (aarch64_simd_scalar_immediate_valid_for_move)
10935 (aarch64_simd_make_constant): Update call to
10936 aarch64_simd_valid_immediate.
10937 (aarch64_advsimd_valid_immediate_hs): New function.
10938 (aarch64_advsimd_valid_immediate): Likewise.
10939 (aarch64_simd_valid_immediate): Remove mode and inverse
10940 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10941 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10942 and aarch64_float_const_representable_p on the result.
10943 (aarch64_output_simd_mov_immediate): Remove mode argument.
10944 Update call to aarch64_simd_valid_immediate and use of
10945 simd_immediate_info.
10946 (aarch64_output_scalar_simd_mov_immediate): Update call
10949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10950 Alan Hayward <alan.hayward@arm.com>
10951 David Sherwood <david.sherwood@arm.com>
10953 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10954 (mode_nunits): Likewise CONST_MODE_NUNITS.
10955 * machmode.def (ADJUST_NUNITS): Document.
10956 * genmodes.c (mode_data::need_nunits_adj): New field.
10957 (blank_mode): Update accordingly.
10958 (adj_nunits): New variable.
10959 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10961 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10962 listed in adj_nunits.
10963 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10964 listed in adj_nunits. Don't emit case statements for such modes.
10965 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10966 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10967 nothing if adj_nunits is nonnull.
10968 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10969 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10970 (emit_mode_fbit): Update use of print_maybe_const_decl.
10971 (emit_move_size): Likewise. Treat the array as non-const
10973 (emit_mode_adjustments): Handle adj_nunits.
10975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10977 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10978 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10979 (VECTOR_MODES): Use it.
10980 (make_vector_modes): Take the prefix as an argument.
10982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10983 Alan Hayward <alan.hayward@arm.com>
10984 David Sherwood <david.sherwood@arm.com>
10986 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10987 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10988 for MODE_VECTOR_BOOL.
10989 * machmode.def (VECTOR_BOOL_MODE): Document.
10990 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10991 (make_vector_bool_mode): New function.
10992 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10994 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10995 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10997 * stor-layout.c (int_mode_for_mode): Likewise.
10998 * tree.c (build_vector_type_for_mode): Likewise.
10999 * varasm.c (output_constant_pool_2): Likewise.
11000 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11001 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11002 for MODE_VECTOR_BOOL.
11003 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11004 of mode class checks.
11005 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11006 instead of a list of mode class checks.
11007 (expand_vector_scalar_condition): Likewise.
11008 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11011 Alan Hayward <alan.hayward@arm.com>
11012 David Sherwood <david.sherwood@arm.com>
11014 * machmode.h (mode_size): Change from unsigned short to
11016 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11017 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11018 or if measurement_type is not polynomial.
11019 (fixed_size_mode::includes_p): Check for constant-sized modes.
11020 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11021 return a poly_uint16 rather than an unsigned short.
11022 (emit_mode_size): Change the type of mode_size from unsigned short
11023 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11024 (emit_mode_adjustments): Cope with polynomial vector sizes.
11025 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11027 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11029 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11030 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11031 * caller-save.c (setup_save_areas): Likewise.
11032 (replace_reg_with_saved_mem): Likewise.
11033 * calls.c (emit_library_call_value_1): Likewise.
11034 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11035 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11036 (gen_lowpart_for_combine): Likewise.
11037 * convert.c (convert_to_integer_1): Likewise.
11038 * cse.c (equiv_constant, cse_insn): Likewise.
11039 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11040 (cselib_subst_to_values): Likewise.
11041 * dce.c (word_dce_process_block): Likewise.
11042 * df-problems.c (df_word_lr_mark_ref): Likewise.
11043 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11044 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11045 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11046 (rtl_for_decl_location): Likewise.
11047 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11048 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11049 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11050 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11051 (expand_expr_real_1): Likewise.
11052 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11053 (pad_below): Likewise.
11054 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11055 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11056 * ira.c (get_subreg_tracking_sizes): Likewise.
11057 * ira-build.c (ira_create_allocno_objects): Likewise.
11058 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11059 (ira_sort_regnos_for_alter_reg): Likewise.
11060 * ira-costs.c (record_operand_costs): Likewise.
11061 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11062 (resolve_simple_move): Likewise.
11063 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11064 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11065 (lra_constraints): Likewise.
11066 (CONST_POOL_OK_P): Reject variable-sized modes.
11067 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11068 (add_pseudo_to_slot, lra_spill): Likewise.
11069 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11070 * optabs-query.c (get_best_extraction_insn): Likewise.
11071 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11072 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11073 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11074 * recog.c (offsettable_address_addr_space_p): Likewise.
11075 * regcprop.c (maybe_mode_change): Likewise.
11076 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11077 * regrename.c (build_def_use): Likewise.
11078 * regstat.c (dump_reg_info): Likewise.
11079 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11080 (find_reloads, find_reloads_subreg_address): Likewise.
11081 * reload1.c (eliminate_regs_1): Likewise.
11082 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11083 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11084 (simplify_binary_operation_1, simplify_subreg): Likewise.
11085 * targhooks.c (default_function_arg_padding): Likewise.
11086 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11087 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11088 (verify_gimple_assign_ternary): Likewise.
11089 * tree-inline.c (estimate_move_cost): Likewise.
11090 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11091 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11092 (get_address_cost_ainc): Likewise.
11093 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11094 (vect_supportable_dr_alignment): Likewise.
11095 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11096 (vectorizable_reduction): Likewise.
11097 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11098 (vectorizable_operation, vectorizable_load): Likewise.
11099 * tree.c (build_same_sized_truth_vector_type): Likewise.
11100 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11101 * var-tracking.c (emit_note_insn_var_location): Likewise.
11102 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11103 (ADDR_VEC_ALIGN): Likewise.
11105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11106 Alan Hayward <alan.hayward@arm.com>
11107 David Sherwood <david.sherwood@arm.com>
11109 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11111 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11112 or if measurement_type is polynomial.
11113 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11114 * combine.c (make_extraction): Likewise.
11115 * dse.c (find_shift_sequence): Likewise.
11116 * dwarf2out.c (mem_loc_descriptor): Likewise.
11117 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11118 (extract_bit_field, extract_low_bits): Likewise.
11119 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11120 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11121 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11122 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11123 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11124 * reload.c (find_reloads): Likewise.
11125 * reload1.c (alter_reg): Likewise.
11126 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11127 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11128 * tree-if-conv.c (predicate_mem_writes): Likewise.
11129 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11130 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11131 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11132 * valtrack.c (dead_debug_insert_temp): Likewise.
11133 * varasm.c (mergeable_constant_section): Likewise.
11134 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11137 Alan Hayward <alan.hayward@arm.com>
11138 David Sherwood <david.sherwood@arm.com>
11140 * expr.c (expand_assignment): Cope with polynomial mode sizes
11141 when assigning to a CONCAT.
11143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11144 Alan Hayward <alan.hayward@arm.com>
11145 David Sherwood <david.sherwood@arm.com>
11147 * machmode.h (mode_precision): Change from unsigned short to
11149 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11151 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11152 or if measurement_type is not polynomial.
11153 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11154 in which the mode is already known to be a scalar_int_mode.
11155 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11156 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11158 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11159 for GET_MODE_PRECISION.
11160 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11161 for GET_MODE_PRECISION.
11162 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11164 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11165 (expand_field_assignment, make_extraction): Likewise.
11166 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11167 (get_last_value): Likewise.
11168 * convert.c (convert_to_integer_1): Likewise.
11169 * cse.c (cse_insn): Likewise.
11170 * expr.c (expand_expr_real_1): Likewise.
11171 * lra-constraints.c (simplify_operand_subreg): Likewise.
11172 * optabs-query.c (can_atomic_load_p): Likewise.
11173 * optabs.c (expand_atomic_load): Likewise.
11174 (expand_atomic_store): Likewise.
11175 * ree.c (combine_reaching_defs): Likewise.
11176 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11177 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11178 * tree.h (type_has_mode_precision_p): Likewise.
11179 * ubsan.c (instrument_si_overflow): Likewise.
11181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11186 polynomial numbers of units.
11187 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11188 (valid_vector_subparts_p): New function.
11189 (build_vector_type): Remove temporary shim and take the number
11190 of units as a poly_uint64 rather than an int.
11191 (build_opaque_vector_type): Take the number of units as a
11192 poly_uint64 rather than an int.
11193 * tree.c (build_vector_from_ctor): Handle polynomial
11194 TYPE_VECTOR_SUBPARTS.
11195 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11196 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11197 (build_vector_from_val): If the number of units is variable,
11198 use build_vec_duplicate_cst for constant operands and
11199 VEC_DUPLICATE_EXPR otherwise.
11200 (make_vector_type): Remove temporary is_constant ().
11201 (build_vector_type, build_opaque_vector_type): Take the number of
11202 units as a poly_uint64 rather than an int.
11203 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11205 * cfgexpand.c (expand_debug_expr): Likewise.
11206 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11207 (store_constructor, expand_expr_real_1): Likewise.
11208 (const_scalar_mask_from_tree): Likewise.
11209 * fold-const-call.c (fold_const_reduction): Likewise.
11210 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11211 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11212 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11213 (fold_relational_const): Likewise.
11214 (native_interpret_vector): Likewise. Change the size from an
11215 int to an unsigned int.
11216 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11217 TYPE_VECTOR_SUBPARTS.
11218 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11219 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11220 duplicating a non-constant operand into a variable-length vector.
11221 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11222 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11223 * ipa-icf.c (sem_variable::equals): Likewise.
11224 * match.pd: Likewise.
11225 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11226 * print-tree.c (print_node): Likewise.
11227 * stor-layout.c (layout_type): Likewise.
11228 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11229 * tree-cfg.c (verify_gimple_comparison): Likewise.
11230 (verify_gimple_assign_binary): Likewise.
11231 (verify_gimple_assign_ternary): Likewise.
11232 (verify_gimple_assign_single): Likewise.
11233 * tree-pretty-print.c (dump_generic_node): Likewise.
11234 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11235 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11236 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11237 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11238 (vect_shift_permute_load_chain): Likewise.
11239 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11240 (expand_vector_condition, optimize_vector_constructor): Likewise.
11241 (lower_vec_perm, get_compute_type): Likewise.
11242 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11243 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11244 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11245 (vect_recog_mask_conversion_pattern): Likewise.
11246 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11247 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11248 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11249 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11250 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11251 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11252 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11253 (supportable_widening_operation): Likewise.
11254 (supportable_narrowing_operation): Likewise.
11255 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11257 * varasm.c (output_constant): Likewise.
11259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11260 Alan Hayward <alan.hayward@arm.com>
11261 David Sherwood <david.sherwood@arm.com>
11263 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11264 so that both the length == 3 and length != 3 cases set up their
11265 own permute vectors. Add comments explaining why we know the
11266 number of elements is constant.
11267 (vect_permute_load_chain): Likewise.
11269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11270 Alan Hayward <alan.hayward@arm.com>
11271 David Sherwood <david.sherwood@arm.com>
11273 * machmode.h (mode_nunits): Change from unsigned char to
11275 (ONLY_FIXED_SIZE_MODES): New macro.
11276 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11277 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11278 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11280 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11281 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11282 or if measurement_type is not polynomial.
11283 * genmodes.c (ZERO_COEFFS): New macro.
11284 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11286 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11287 Use ZERO_COEFFS when emitting initializers.
11288 * data-streamer.h (bp_pack_poly_value): New function.
11289 (bp_unpack_poly_value): Likewise.
11290 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11291 for GET_MODE_NUNITS.
11292 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11293 for GET_MODE_NUNITS.
11294 * tree.c (make_vector_type): Remove temporary shim and make
11295 the real function take the number of units as a poly_uint64
11296 rather than an int.
11297 (build_vector_type_for_mode): Handle polynomial nunits.
11298 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11299 * emit-rtl.c (const_vec_series_p_1): Likewise.
11300 (gen_rtx_CONST_VECTOR): Likewise.
11301 * fold-const.c (test_vec_duplicate_folding): Likewise.
11302 * genrecog.c (validate_pattern): Likewise.
11303 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11304 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11305 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11306 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11307 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11308 * rtlanal.c (subreg_get_info): Likewise.
11309 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11310 (vect_grouped_load_supported): Likewise.
11311 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11312 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11313 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11314 (simplify_const_unary_operation, simplify_binary_operation_1)
11315 (simplify_const_binary_operation, simplify_ternary_operation)
11316 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11317 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11318 instead of CONST_VECTOR_NUNITS.
11319 * varasm.c (output_constant_pool_2): Likewise.
11320 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11321 explicit-encoded elements in the XVEC for variable-length vectors.
11323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11325 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11328 Alan Hayward <alan.hayward@arm.com>
11329 David Sherwood <david.sherwood@arm.com>
11331 * coretypes.h (fixed_size_mode): Declare.
11332 (fixed_size_mode_pod): New typedef.
11333 * builtins.h (target_builtins::x_apply_args_mode)
11334 (target_builtins::x_apply_result_mode): Change type to
11335 fixed_size_mode_pod.
11336 * builtins.c (apply_args_size, apply_result_size, result_vector)
11337 (expand_builtin_apply_args_1, expand_builtin_apply)
11338 (expand_builtin_return): Update accordingly.
11340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11342 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11343 * cselib.c (cselib_hash_rtx): Likewise.
11344 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11345 CONST_VECTOR encoding.
11347 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11348 Jeff Law <law@redhat.com>
11351 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11352 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11353 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11354 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11357 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11358 explicitly probe *sp in a noreturn function if there were any callee
11359 register saves or frame pointer is needed.
11361 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11364 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11365 BLKmode for ternary, binary or unary expressions.
11368 * var-tracking.c (delete_vta_debug_insn): New inline function.
11369 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11370 insns from get_insns () to NULL instead of each bb separately.
11371 Use delete_vta_debug_insn. No longer static.
11372 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11373 delete_vta_debug_insns callers.
11374 * rtl.h (delete_vta_debug_insns): Declare.
11375 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11376 instead of variable_tracking_main.
11378 2018-01-03 Martin Sebor <msebor@redhat.com>
11380 PR tree-optimization/83603
11381 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11382 arguments past the endof the argument list in functions declared
11383 without a prototype.
11384 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11385 Avoid checking when arguments are null.
11387 2018-01-03 Martin Sebor <msebor@redhat.com>
11390 * doc/extend.texi (attribute const): Fix a typo.
11391 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11392 issuing -Wsuggest-attribute for void functions.
11394 2018-01-03 Martin Sebor <msebor@redhat.com>
11396 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11397 offset_int::from instead of wide_int::to_shwi.
11398 (maybe_diag_overlap): Remove assertion.
11399 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11400 * gimple-ssa-sprintf.c (format_directive): Same.
11401 (parse_directive): Same.
11402 (sprintf_dom_walker::compute_format_length): Same.
11403 (try_substitute_return_value): Same.
11405 2018-01-03 Jeff Law <law@redhat.com>
11407 PR middle-end/83654
11408 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11409 non-constant residual for zero at runtime and avoid probing in
11410 that case. Reorganize code for trailing problem to mirror handling
11413 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11415 PR tree-optimization/83501
11416 * tree-ssa-strlen.c (get_string_cst): New.
11417 (handle_char_store): Call get_string_cst.
11419 2018-01-03 Martin Liska <mliska@suse.cz>
11421 PR tree-optimization/83593
11422 * tree-ssa-strlen.c: Include tree-cfg.h.
11423 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11424 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11425 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11427 (strlen_dom_walker::before_dom_children): Call
11428 gimple_purge_dead_eh_edges. Dump tranformation with details
11430 (strlen_dom_walker::before_dom_children): Update call by adding
11431 new argument cleanup_eh.
11432 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11434 2018-01-03 Martin Liska <mliska@suse.cz>
11437 * cif-code.def (VARIADIC_THUNK): New enum value.
11438 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11441 2018-01-03 Jan Beulich <jbeulich@suse.com>
11443 * sse.md (mov<mode>_internal): Tighten condition for when to use
11444 vmovdqu<ssescalarsize> for TI and OI modes.
11446 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11448 Update copyright years.
11450 2018-01-03 Martin Liska <mliska@suse.cz>
11453 * ipa-visibility.c (function_and_variable_visibility): Skip
11454 functions with noipa attribure.
11456 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11458 * gcc.c (process_command): Update copyright notice dates.
11459 * gcov-dump.c (print_version): Ditto.
11460 * gcov.c (print_version): Ditto.
11461 * gcov-tool.c (print_version): Ditto.
11462 * gengtype.c (create_file): Ditto.
11463 * doc/cpp.texi: Bump @copying's copyright year.
11464 * doc/cppinternals.texi: Ditto.
11465 * doc/gcc.texi: Ditto.
11466 * doc/gccint.texi: Ditto.
11467 * doc/gcov.texi: Ditto.
11468 * doc/install.texi: Ditto.
11469 * doc/invoke.texi: Ditto.
11471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11473 * vector-builder.h (vector_builder::m_full_nelts): Change from
11474 unsigned int to poly_uint64.
11475 (vector_builder::full_nelts): Update prototype accordingly.
11476 (vector_builder::new_vector): Likewise.
11477 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11478 (vector_builder::operator ==): Likewise.
11479 (vector_builder::finalize): Likewise.
11480 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11481 Take the number of elements as a poly_uint64 rather than an
11483 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11484 from unsigned int to poly_uint64.
11485 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11486 (vec_perm_indices::new_vector): Likewise.
11487 (vec_perm_indices::length): Likewise.
11488 (vec_perm_indices::nelts_per_input): Likewise.
11489 (vec_perm_indices::input_nelts): Likewise.
11490 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11491 number of elements per input as a poly_uint64 rather than an
11492 unsigned int. Use the original encoding for variable-length
11493 vectors, rather than clamping each individual element.
11494 For the second and subsequent elements in each pattern,
11495 clamp the step and base before clamping their sum.
11496 (vec_perm_indices::series_p): Handle polynomial element counts.
11497 (vec_perm_indices::all_in_range_p): Likewise.
11498 (vec_perm_indices_to_tree): Likewise.
11499 (vec_perm_indices_to_rtx): Likewise.
11500 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11501 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11502 (tree_vector_builder::new_binary_operation): Handle polynomial
11503 element counts. Return false if we need to know the number
11504 of elements at compile time.
11505 * fold-const.c (fold_vec_perm): Punt if the number of elements
11506 isn't known at compile time.
11508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11510 * vec-perm-indices.h (vec_perm_builder): Change element type
11511 from HOST_WIDE_INT to poly_int64.
11512 (vec_perm_indices::element_type): Update accordingly.
11513 (vec_perm_indices::clamp): Handle polynomial element_types.
11514 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11515 (vec_perm_indices::all_in_range_p): Likewise.
11516 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11518 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11519 polynomial vec_perm_indices element types.
11520 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11521 * fold-const.c (fold_vec_perm): Likewise.
11522 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11523 * tree-vect-generic.c (lower_vec_perm): Likewise.
11524 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11525 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11526 element type to HOST_WIDE_INT.
11528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11529 Alan Hayward <alan.hayward@arm.com>
11530 David Sherwood <david.sherwood@arm.com>
11532 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11533 rather than an int. Use plus_constant.
11534 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11535 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11538 Alan Hayward <alan.hayward@arm.com>
11539 David Sherwood <david.sherwood@arm.com>
11541 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11542 a HOST_WIDE_INT to a poly_int64.
11544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11545 Alan Hayward <alan.hayward@arm.com>
11546 David Sherwood <david.sherwood@arm.com>
11548 * calls.c (load_register_parameters): Cope with polynomial
11549 mode sizes. Require a constant size for BLKmode parameters
11550 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11551 forces a parameter to be padded at the lsb end in order to
11552 fill a complete number of words, require the parameter size
11553 to be ordered wrt UNITS_PER_WORD.
11555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11556 Alan Hayward <alan.hayward@arm.com>
11557 David Sherwood <david.sherwood@arm.com>
11559 * reload1.c (spill_stack_slot_width): Change element type
11560 from unsigned int to poly_uint64_pod.
11561 (alter_reg): Treat mode sizes as polynomial.
11563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11564 Alan Hayward <alan.hayward@arm.com>
11565 David Sherwood <david.sherwood@arm.com>
11567 * reload.c (complex_word_subreg_p): New function.
11568 (reload_inner_reg_of_subreg, push_reload): Use it.
11570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11571 Alan Hayward <alan.hayward@arm.com>
11572 David Sherwood <david.sherwood@arm.com>
11574 * lra-constraints.c (process_alt_operands): Reject matched
11575 operands whose sizes aren't ordered.
11576 (match_reload): Refer to this check here.
11578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11579 Alan Hayward <alan.hayward@arm.com>
11580 David Sherwood <david.sherwood@arm.com>
11582 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11583 that the mode size is in the set {1, 2, 4, 8, 16}.
11585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11586 Alan Hayward <alan.hayward@arm.com>
11587 David Sherwood <david.sherwood@arm.com>
11589 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11590 Use plus_constant instead of gen_rtx_PLUS.
11592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11593 Alan Hayward <alan.hayward@arm.com>
11594 David Sherwood <david.sherwood@arm.com>
11596 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11597 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11598 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11599 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11600 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11601 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11602 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11603 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11604 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11605 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11607 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11608 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11609 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11610 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11611 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11612 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11613 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11614 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11615 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11616 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11618 * expr.c (emit_move_resolve_push): Treat the input and result
11619 of PUSH_ROUNDING as a poly_int64.
11620 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11621 (emit_push_insn): Likewise.
11622 * lra-eliminations.c (mark_not_eliminable): Likewise.
11623 * recog.c (push_operand): Likewise.
11624 * reload1.c (elimination_effects): Likewise.
11625 * rtlanal.c (nonzero_bits1): Likewise.
11626 * calls.c (store_one_arg): Likewise. Require the padding to be
11627 known at compile time.
11629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11630 Alan Hayward <alan.hayward@arm.com>
11631 David Sherwood <david.sherwood@arm.com>
11633 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11634 Use plus_constant instead of gen_rtx_PLUS.
11636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11637 Alan Hayward <alan.hayward@arm.com>
11638 David Sherwood <david.sherwood@arm.com>
11640 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11641 rather than an int.
11643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11644 Alan Hayward <alan.hayward@arm.com>
11645 David Sherwood <david.sherwood@arm.com>
11647 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11648 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11649 via stack temporaries. Treat the mode size as polynomial too.
11651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11652 Alan Hayward <alan.hayward@arm.com>
11653 David Sherwood <david.sherwood@arm.com>
11655 * expr.c (expand_expr_real_2): When handling conversions involving
11656 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11657 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11658 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11667 Alan Hayward <alan.hayward@arm.com>
11668 David Sherwood <david.sherwood@arm.com>
11670 * combine.c (can_change_dest_mode): Handle polynomial
11671 REGMODE_NATURAL_SIZE.
11672 * expmed.c (store_bit_field_1): Likewise.
11673 * expr.c (store_constructor): Likewise.
11674 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11675 and polynomial REGMODE_NATURAL_SIZE.
11676 (gen_lowpart_common): Likewise.
11677 * reginfo.c (record_subregs_of_mode): Likewise.
11678 * rtlanal.c (read_modify_subreg_p): Likewise.
11680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11681 Alan Hayward <alan.hayward@arm.com>
11682 David Sherwood <david.sherwood@arm.com>
11684 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11685 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
11693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11694 Alan Hayward <alan.hayward@arm.com>
11695 David Sherwood <david.sherwood@arm.com>
11697 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11698 in a POINTER_PLUS_EXPR.
11700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11701 Alan Hayward <alan.hayward@arm.com>
11702 David Sherwood <david.sherwood@arm.com>
11704 * omp-simd-clone.c (simd_clone_subparts): New function.
11705 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11706 (ipa_simd_modify_function_body): Likewise.
11708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11709 Alan Hayward <alan.hayward@arm.com>
11710 David Sherwood <david.sherwood@arm.com>
11712 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11713 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11714 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11715 (expand_vector_condition, vector_element): Likewise.
11716 (subparts_gt): New function.
11717 (get_compute_type): Use subparts_gt.
11718 (count_type_subparts): Delete.
11719 (expand_vector_operations_1): Use subparts_gt instead of
11720 count_type_subparts.
11722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11723 Alan Hayward <alan.hayward@arm.com>
11724 David Sherwood <david.sherwood@arm.com>
11726 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11727 (vect_compile_time_alias): ...this new function. Do the calculation
11728 on poly_ints rather than trees.
11729 (vect_prune_runtime_alias_test_list): Update call accordingly.
11731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11732 Alan Hayward <alan.hayward@arm.com>
11733 David Sherwood <david.sherwood@arm.com>
11735 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11737 (vect_schedule_slp_instance): Likewise.
11739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11740 Alan Hayward <alan.hayward@arm.com>
11741 David Sherwood <david.sherwood@arm.com>
11743 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11744 constant and extern definitions for variable-length vectors.
11745 (vect_get_constant_vectors): Note that the number of units
11746 is known to be constant.
11748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11749 Alan Hayward <alan.hayward@arm.com>
11750 David Sherwood <david.sherwood@arm.com>
11752 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11753 of units as polynomial. Choose between WIDE and NARROW based
11756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 Alan Hayward <alan.hayward@arm.com>
11758 David Sherwood <david.sherwood@arm.com>
11760 * tree-vect-stmts.c (simd_clone_subparts): New function.
11761 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11764 Alan Hayward <alan.hayward@arm.com>
11765 David Sherwood <david.sherwood@arm.com>
11767 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11768 vectors as polynomial. Use build_index_vector for
11769 IFN_GOMP_SIMD_LANE.
11771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11772 Alan Hayward <alan.hayward@arm.com>
11773 David Sherwood <david.sherwood@arm.com>
11775 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11776 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11777 for variable-length vectors.
11778 (vectorizable_mask_load_store): Treat the number of units as
11779 polynomial, asserting that it is constant if the condition has
11780 already been enforced.
11781 (vectorizable_store, vectorizable_load): Likewise.
11783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11784 Alan Hayward <alan.hayward@arm.com>
11785 David Sherwood <david.sherwood@arm.com>
11787 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11788 of units as polynomial. Punt if we can't tell at compile time
11789 which vector contains the final result.
11791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11792 Alan Hayward <alan.hayward@arm.com>
11793 David Sherwood <david.sherwood@arm.com>
11795 * tree-vect-loop.c (vectorizable_induction): Treat the number
11796 of units as polynomial. Punt on SLP inductions. Use an integer
11797 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11798 cast of such a series for variable-length floating-point
11801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11802 Alan Hayward <alan.hayward@arm.com>
11803 David Sherwood <david.sherwood@arm.com>
11805 * tree.h (build_index_vector): Declare.
11806 * tree.c (build_index_vector): New function.
11807 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11808 of units as polynomial, forcibly converting it to a constant if
11809 vectorizable_reduction has already enforced the condition.
11810 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11811 to create a {1,2,3,...} vector.
11812 (vectorizable_reduction): Treat the number of units as polynomial.
11813 Choose vectype_in based on the largest scalar element size rather
11814 than the smallest number of units. Enforce the restrictions
11817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11818 Alan Hayward <alan.hayward@arm.com>
11819 David Sherwood <david.sherwood@arm.com>
11821 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11822 number of units as polynomial.
11824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11825 Alan Hayward <alan.hayward@arm.com>
11826 David Sherwood <david.sherwood@arm.com>
11828 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11829 * target.def (autovectorize_vector_sizes): Return the vector sizes
11830 by pointer, using vector_sizes rather than a bitmask.
11831 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11832 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11833 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11835 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11836 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11837 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11838 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11839 * omp-general.c (omp_max_vf): Likewise.
11840 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11841 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11842 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11843 * tree-vect-slp.c (vect_slp_bb): Likewise.
11844 * doc/tm.texi: Regenerate.
11845 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11847 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11848 the vector size as a poly_uint64 rather than an unsigned int.
11849 (current_vector_size): Change from an unsigned int to a poly_uint64.
11850 (get_vectype_for_scalar_type): Update accordingly.
11851 * tree.h (build_truth_vector_type): Take the size and number of
11852 units as a poly_uint64 rather than an unsigned int.
11853 (build_vector_type): Add a temporary overload that takes
11854 the number of units as a poly_uint64 rather than an unsigned int.
11855 * tree.c (make_vector_type): Likewise.
11856 (build_truth_vector_type): Take the number of units as a poly_uint64
11857 rather than an unsigned int.
11859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11860 Alan Hayward <alan.hayward@arm.com>
11861 David Sherwood <david.sherwood@arm.com>
11863 * target.def (get_mask_mode): Take the number of units and length
11864 as poly_uint64s rather than unsigned ints.
11865 * targhooks.h (default_get_mask_mode): Update accordingly.
11866 * targhooks.c (default_get_mask_mode): Likewise.
11867 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11868 * doc/tm.texi: Regenerate.
11870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11871 Alan Hayward <alan.hayward@arm.com>
11872 David Sherwood <david.sherwood@arm.com>
11874 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11875 * omp-general.c (omp_max_vf): Likewise.
11876 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11877 (expand_omp_simd): Handle polynomial safelen.
11878 * omp-low.c (omplow_simd_context): Add a default constructor.
11879 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11880 (lower_rec_simd_input_clauses): Update accordingly.
11881 (lower_rec_input_clauses): Likewise.
11883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11884 Alan Hayward <alan.hayward@arm.com>
11885 David Sherwood <david.sherwood@arm.com>
11887 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11888 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11889 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11890 (vect_analyze_slp_cost): Likewise.
11891 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11892 (vect_model_load_cost): Likewise.
11894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11895 Alan Hayward <alan.hayward@arm.com>
11896 David Sherwood <david.sherwood@arm.com>
11898 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11899 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11900 from an unsigned int * to a poly_uint64_pod *.
11901 (calculate_unrolling_factor): New function.
11902 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11905 Alan Hayward <alan.hayward@arm.com>
11906 David Sherwood <david.sherwood@arm.com>
11908 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11909 from an unsigned int to a poly_uint64.
11910 (_loop_vec_info::slp_unrolling_factor): Likewise.
11911 (_loop_vec_info::vectorization_factor): Change from an int
11913 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11914 (vect_get_num_vectors): New function.
11915 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11916 (vect_get_num_copies): Use vect_get_num_vectors.
11917 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11918 to an unsigned int *.
11919 (vect_analyze_data_refs): Change min_vf from an int * to a
11921 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11922 than an unsigned HOST_WIDE_INT.
11923 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11924 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11925 to an unsigned int *.
11926 (vect_analyze_data_ref_dependences): Likewise.
11927 (vect_compute_data_ref_alignment): Handle polynomial vf.
11928 (vect_enhance_data_refs_alignment): Likewise.
11929 (vect_prune_runtime_alias_test_list): Likewise.
11930 (vect_shift_permute_load_chain): Likewise.
11931 (vect_supportable_dr_alignment): Likewise.
11932 (dependence_distance_ge_vf): Take the vectorization factor as a
11933 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11934 (vect_analyze_data_refs): Change min_vf from an int * to a
11936 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11937 vfm1 as a poly_uint64 rather than an int. Make the same change
11938 for the returned bound_scalar.
11939 (vect_gen_vector_loop_niters): Handle polynomial vf.
11940 (vect_do_peeling): Likewise. Update call to
11941 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11942 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11944 * tree-vect-loop.c (vect_determine_vectorization_factor)
11945 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11946 (vect_get_known_peeling_cost): Likewise.
11947 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11948 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11949 (vect_transform_loop): Likewise. Use the lowest possible VF when
11950 updating the upper bounds of the loop.
11951 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11952 rather than an int.
11953 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11954 polynomial unroll factors.
11955 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11956 (vect_make_slp_decision): Likewise.
11957 (vect_supported_load_permutation_p): Likewise, and polynomial
11959 (vect_analyze_slp_cost): Handle polynomial vf.
11960 (vect_slp_analyze_node_operations): Likewise.
11961 (vect_slp_analyze_bb_1): Likewise.
11962 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11963 than an unsigned HOST_WIDE_INT.
11964 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11965 (vectorizable_load): Handle polynomial vf.
11966 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11968 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11971 Alan Hayward <alan.hayward@arm.com>
11972 David Sherwood <david.sherwood@arm.com>
11974 * match.pd: Handle bit operations involving three constants
11975 and try to fold one pair.
11977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11979 * tree-vect-loop-manip.c: Include gimple-fold.h.
11980 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11981 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11982 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11983 Add a path that uses a step of VF instead of 1, but disable it
11985 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11986 and niters_no_overflow parameters. Update calls to
11987 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11988 Create a new SSA name if the latter choses to use a ste other
11989 than zero, and return it via niters_vector_mult_vf_var.
11990 * tree-vect-loop.c (vect_transform_loop): Update calls to
11991 vect_do_peeling, vect_gen_vector_loop_niters and
11992 slpeel_make_loop_iterate_ntimes.
11993 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11994 (vect_gen_vector_loop_niters): Update declarations after above changes.
11996 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11998 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11999 128-bit round to integer instructions.
12000 (ceil<mode>2): Likewise.
12001 (btrunc<mode>2): Likewise.
12002 (round<mode>2): Likewise.
12004 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12006 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12007 unaligned VSX load/store on P8/P9.
12008 (expand_block_clear): Allow the use of unaligned VSX
12009 load/store on P8/P9.
12011 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12013 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12015 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12016 swap associated with both a load and a store.
12018 2018-01-02 Andrew Waterman <andrew@sifive.com>
12020 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12021 * config/riscv/riscv.md (clear_cache): Use it.
12023 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12025 * web.c: Remove out-of-date comment.
12027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12029 * expr.c (fixup_args_size_notes): Check that any existing
12030 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12031 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12032 (emit_single_push_insn): ...here.
12034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12036 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12037 (const_vector_encoded_nelts): New function.
12038 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12039 (const_vector_int_elt, const_vector_elt): Declare.
12040 * emit-rtl.c (const_vector_int_elt_1): New function.
12041 (const_vector_elt): Likewise.
12042 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12043 of CONST_VECTOR_ELT.
12045 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12047 * expr.c: Include rtx-vector-builder.h.
12048 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12049 directly on the tree encoding.
12050 (const_vector_from_tree): Likewise.
12051 * optabs.c: Include rtx-vector-builder.h.
12052 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12053 sequence of "u" values.
12054 * vec-perm-indices.c: Include rtx-vector-builder.h.
12055 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12056 directly on the vec_perm_indices encoding.
12058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12060 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12061 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12062 * rtx-vector-builder.h: New file.
12063 * rtx-vector-builder.c: Likewise.
12064 * rtl.h (rtx_def::u2): Add a const_vector field.
12065 (CONST_VECTOR_NPATTERNS): New macro.
12066 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12067 (CONST_VECTOR_DUPLICATE_P): Likewise.
12068 (CONST_VECTOR_STEPPED_P): Likewise.
12069 (CONST_VECTOR_ENCODED_ELT): Likewise.
12070 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12071 (unwrap_const_vec_duplicate): Likewise.
12072 (const_vec_series_p): Check for a non-duplicated vector encoding.
12073 Say that the function only returns true for integer vectors.
12074 * emit-rtl.c: Include rtx-vector-builder.h.
12075 (gen_const_vec_duplicate_1): Delete.
12076 (gen_const_vector): Call gen_const_vec_duplicate instead of
12077 gen_const_vec_duplicate_1.
12078 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12079 (gen_const_vec_duplicate): Use rtx_vector_builder.
12080 (gen_const_vec_series): Likewise.
12081 (gen_rtx_CONST_VECTOR): Likewise.
12082 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12083 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12084 Build a new vector rather than modifying a CONST_VECTOR in-place.
12085 (handle_special_swappables): Update call accordingly.
12086 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12087 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12088 Build a new vector rather than modifying a CONST_VECTOR in-place.
12089 (handle_special_swappables): Update call accordingly.
12091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12093 * simplify-rtx.c (simplify_const_binary_operation): Use
12094 CONST_VECTOR_ELT instead of XVECEXP.
12096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12098 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12099 the selector elements to be different from the data elements
12100 if the selector is a VECTOR_CST.
12101 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12102 ssizetype for the selector.
12104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12106 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12107 before testing each element individually.
12108 * tree-vect-generic.c (lower_vec_perm): Likewise.
12110 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12112 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12113 * selftest-run-tests.c (selftest::run_tests): Call it.
12114 * vector-builder.h (vector_builder::operator ==): New function.
12115 (vector_builder::operator !=): Likewise.
12116 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12117 (vec_perm_indices::all_from_input_p): New function.
12118 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12119 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12120 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12121 instead of reading the VECTOR_CST directly. Detect whether both
12122 vector inputs are the same before constructing the vec_perm_indices,
12123 and update the number of inputs argument accordingly. Use the
12124 utility functions added above. Only construct sel2 if we need to.
12126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12128 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12129 the broadcast of the low byte.
12130 (expand_mult_highpart): Use an explicit encoding for the permutes.
12131 * optabs-query.c (can_mult_highpart_p): Likewise.
12132 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12133 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12134 (vectorizable_bswap): Likewise.
12135 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12136 explicit encoding for the power-of-2 permutes.
12137 (vect_permute_store_chain): Likewise.
12138 (vect_grouped_load_supported): Likewise.
12139 (vect_permute_load_chain): Likewise.
12141 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12143 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12144 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12145 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12146 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12147 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12148 (vect_gen_perm_mask_any): Likewise.
12150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12152 * int-vector-builder.h: New file.
12153 * vec-perm-indices.h: Include int-vector-builder.h.
12154 (vec_perm_indices): Redefine as an int_vector_builder.
12155 (auto_vec_perm_indices): Delete.
12156 (vec_perm_builder): Redefine as a stand-alone class.
12157 (vec_perm_indices::vec_perm_indices): New function.
12158 (vec_perm_indices::clamp): Likewise.
12159 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12160 (vec_perm_indices::new_vector): New function.
12161 (vec_perm_indices::new_expanded_vector): Update for new
12162 vec_perm_indices class.
12163 (vec_perm_indices::rotate_inputs): New function.
12164 (vec_perm_indices::all_in_range_p): Operate directly on the
12165 encoded form, without computing elided elements.
12166 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12167 encoding. Update for new vec_perm_indices class.
12168 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12169 the given vec_perm_builder.
12170 (expand_vec_perm_var): Update vec_perm_builder constructor.
12171 (expand_mult_highpart): Use vec_perm_builder instead of
12172 auto_vec_perm_indices.
12173 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12174 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12175 or double series encoding as appropriate.
12176 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12177 vec_perm_indices instead of auto_vec_perm_indices.
12178 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12179 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12180 (vect_permute_store_chain): Likewise.
12181 (vect_grouped_load_supported): Likewise.
12182 (vect_permute_load_chain): Likewise.
12183 (vect_shift_permute_load_chain): Likewise.
12184 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12185 (vect_transform_slp_perm_load): Likewise.
12186 (vect_schedule_slp_instance): Likewise.
12187 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12188 (vectorizable_mask_load_store): Likewise.
12189 (vectorizable_bswap): Likewise.
12190 (vectorizable_store): Likewise.
12191 (vectorizable_load): Likewise.
12192 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12193 vec_perm_indices instead of auto_vec_perm_indices. Use
12194 tree_to_vec_perm_builder to read the vector from a tree.
12195 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12196 vec_perm_builder instead of a vec_perm_indices.
12197 (have_whole_vector_shift): Use vec_perm_builder and
12198 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12199 truncation to calc_vec_perm_mask_for_shift.
12200 (vect_create_epilog_for_reduction): Likewise.
12201 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12202 from auto_vec_perm_indices to vec_perm_indices.
12203 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12204 instead of changing individual elements.
12205 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12206 the vector in d.perm.
12207 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12208 from auto_vec_perm_indices to vec_perm_indices.
12209 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12210 instead of changing individual elements.
12211 (arm_vectorize_vec_perm_const): Use new_vector to install
12212 the vector in d.perm.
12213 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12214 Update vec_perm_builder constructor.
12215 (rs6000_expand_interleave): Likewise.
12216 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12217 (rs6000_expand_interleave): Likewise.
12219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12221 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12222 to qimode could truncate the indices.
12223 * optabs.c (expand_vec_perm_var): Likewise.
12225 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12227 * Makefile.in (OBJS): Add vec-perm-indices.o.
12228 * vec-perm-indices.h: New file.
12229 * vec-perm-indices.c: Likewise.
12230 * target.h (vec_perm_indices): Replace with a forward class
12232 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12233 * optabs.h: Include vec-perm-indices.h.
12234 (expand_vec_perm): Delete.
12235 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12236 (expand_vec_perm_const): Declare.
12237 * target.def (vec_perm_const_ok): Replace with...
12238 (vec_perm_const): ...this new hook.
12239 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12240 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12241 * doc/tm.texi: Regenerate.
12242 * optabs.def (vec_perm_const): Delete.
12243 * doc/md.texi (vec_perm_const): Likewise.
12244 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12245 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12246 expand_vec_perm for constant permutation vectors. Assert that
12247 the mode of variable permutation vectors is the integer equivalent
12248 of the mode that is being permuted.
12249 * optabs-query.h (selector_fits_mode_p): Declare.
12250 * optabs-query.c: Include vec-perm-indices.h.
12251 (selector_fits_mode_p): New function.
12252 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12253 is defined, instead of checking whether the vec_perm_const_optab
12254 exists. Use targetm.vectorize.vec_perm_const instead of
12255 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12256 fit in the vector mode before using a variable permute.
12257 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12258 vec_perm_indices instead of an rtx.
12259 (expand_vec_perm): Replace with...
12260 (expand_vec_perm_const): ...this new function. Take the selector
12261 as a vec_perm_indices rather than an rtx. Also take the mode of
12262 the selector. Update call to shift_amt_for_vec_perm_mask.
12263 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12264 Use vec_perm_indices::new_expanded_vector to expand the original
12265 selector into bytes. Check whether the indices fit in the vector
12266 mode before using a variable permute.
12267 (expand_vec_perm_var): Make global.
12268 (expand_mult_highpart): Use expand_vec_perm_const.
12269 * fold-const.c: Includes vec-perm-indices.h.
12270 * tree-ssa-forwprop.c: Likewise.
12271 * tree-vect-data-refs.c: Likewise.
12272 * tree-vect-generic.c: Likewise.
12273 * tree-vect-loop.c: Likewise.
12274 * tree-vect-slp.c: Likewise.
12275 * tree-vect-stmts.c: Likewise.
12276 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12278 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12279 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12280 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12281 (aarch64_vectorize_vec_perm_const): ...this new function.
12282 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12283 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12284 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12285 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12286 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12287 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12288 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12290 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12291 check for NEON modes.
12292 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12293 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12294 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12295 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12297 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12298 the old VEC_PERM_CONST conditions.
12299 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12300 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12301 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12302 (ia64_vectorize_vec_perm_const_ok): Merge into...
12303 (ia64_vectorize_vec_perm_const): ...this new function.
12304 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12305 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12306 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12307 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12308 * config/mips/mips.c (mips_expand_vec_perm_const)
12309 (mips_vectorize_vec_perm_const_ok): Merge into...
12310 (mips_vectorize_vec_perm_const): ...this new function.
12311 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12312 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12313 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12314 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12315 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12316 (rs6000_expand_vec_perm_const): Delete.
12317 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12319 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12320 (altivec_expand_vec_perm_const_le): Take each operand individually.
12321 Operate on constant selectors rather than rtxes.
12322 (altivec_expand_vec_perm_const): Likewise. Update call to
12323 altivec_expand_vec_perm_const_le.
12324 (rs6000_expand_vec_perm_const): Delete.
12325 (rs6000_vectorize_vec_perm_const_ok): Delete.
12326 (rs6000_vectorize_vec_perm_const): New function.
12327 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12328 an element count and rtx array.
12329 (rs6000_expand_extract_even): Update call accordingly.
12330 (rs6000_expand_interleave): Likewise.
12331 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12332 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12333 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12334 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12335 (rs6000_expand_vec_perm_const): Delete.
12336 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12337 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12338 (altivec_expand_vec_perm_const_le): Take each operand individually.
12339 Operate on constant selectors rather than rtxes.
12340 (altivec_expand_vec_perm_const): Likewise. Update call to
12341 altivec_expand_vec_perm_const_le.
12342 (rs6000_expand_vec_perm_const): Delete.
12343 (rs6000_vectorize_vec_perm_const_ok): Delete.
12344 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12345 reference to the SPE evmerge intructions.
12346 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12347 an element count and rtx array.
12348 (rs6000_expand_extract_even): Update call accordingly.
12349 (rs6000_expand_interleave): Likewise.
12350 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12351 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12353 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12357 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12358 vector mode and that that mode matches the mode of the data
12360 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12361 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12362 directly using expand_vec_perm_1 when forcing selectors into
12364 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12368 * optabs-query.h (can_vec_perm_p): Delete.
12369 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12370 * optabs-query.c (can_vec_perm_p): Split into...
12371 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12372 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12373 particular selector is valid.
12374 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12375 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12376 (vect_grouped_load_supported): Likewise.
12377 (vect_shift_permute_load_chain): Likewise.
12378 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12379 (vect_transform_slp_perm_load): Likewise.
12380 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12381 (vectorizable_bswap): Likewise.
12382 (vect_gen_perm_mask_checked): Likewise.
12383 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12384 implementations of variable permutation vectors into account
12385 when deciding which selector to use.
12386 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12387 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12388 with a false third argument.
12389 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12390 to test whether the constant selector is valid and can_vec_perm_var_p
12391 to test whether a variable selector is valid.
12393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12395 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12396 * optabs-query.c (can_vec_perm_p): Likewise.
12397 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12398 instead of vec_perm_indices.
12399 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12400 (vect_gen_perm_mask_checked): Likewise,
12401 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12402 (vect_gen_perm_mask_checked): Likewise,
12404 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12406 * optabs-query.h (qimode_for_vec_perm): Declare.
12407 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12408 (qimode_for_vec_perm): ...this new function.
12409 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12411 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12413 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12414 does not have a conditional at the top.
12416 2018-01-02 Richard Biener <rguenther@suse.de>
12418 * ipa-inline.c (big_speedup_p): Fix expression.
12420 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12423 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12426 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12430 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12431 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12432 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12433 cond_taken_branch_cost 3->4.
12435 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12437 PR tree-optimization/83581
12438 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12439 TODO_cleanup_cfg if any changes have been made.
12441 PR middle-end/83608
12442 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12443 convert_modes if target mode has the right side, but different mode
12446 PR middle-end/83609
12447 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12448 last argument when extracting from CONCAT. If either from_real or
12449 from_imag is NULL, use expansion through memory. If result is not
12450 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12451 the parts directly to inner mode, if even that fails, use expansion
12454 PR middle-end/83623
12455 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12456 check for bswap in mode rather than HImode and use that in expand_unop
12459 Copyright (C) 2018 Free Software Foundation, Inc.
12461 Copying and distribution of this file, with or without modification,
12462 are permitted in any medium without royalty provided the copyright
12463 notice and this notice are preserved.