1 2018-05-08 Jim Wilson <jimw@sifive.com>
3 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11 (Basic PowerPC Built-in Functions): The new name of the
12 subsection previously known as "PowerPC Built-in Functions".
13 (Basic PowerPC Built-in Functions Available on all Configurations):
15 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
21 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
24 2018-05-08 Jakub Jelinek <jakub@redhat.com>
27 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
28 after cmpelim optimization.
30 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
32 * config.gcc: Support "goldmont".
33 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
34 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
36 * config/i386/i386.c (m_GOLDMONT): Define.
37 (processor_target_table): Add "goldmont".
38 (PTA_GOLDMONT): Define.
39 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
40 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
41 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
42 (fold_builtin_cpu): Add "goldmont".
43 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
44 (ix86_option_override_internal): Add "goldmont".
45 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
46 (processor_type): Add PROCESSOR_GOLDMONT.
47 * config/i386/i386.md: Add CPU "glm".
48 * config/i386/glm.md: New file.
49 * config/i386/x86-tune.def: Add m_GOLDMONT.
50 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
52 2018-05-08 Jakub Jelinek <jakub@redhat.com>
55 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
57 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
58 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
59 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
62 * config/i386/i386.c (ix86_fold_builtin): Handle
63 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
66 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
67 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
69 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
72 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
73 (check_arch): Likewise.
74 (check_fpu): Return the result rather than printing it.
75 (end arch): Fix operator precedence.
77 (END): Print the result from check_fpu.
79 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
80 Alan Hayward <alan.hayward@arm.com>
81 David Sherwood <david.sherwood@arm.com>
83 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
84 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
85 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
86 (*fcmuo<mode>_and): New patterns.
88 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
90 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
91 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
92 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
93 (cmp_op, sve_imm_con): New code attributes.
94 (SVE_COND_INT_CMP, imm_con): Delete.
95 (cmp_op): Remove above unspecs from int attribute.
96 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
98 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
99 comparison-specific unspecs.
100 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
101 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
102 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
103 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
104 (*vec_fcm<cmp_op><mode>): Rename to...
105 (*fcm<cmp_op><mode>): ...this and adjust likewise.
106 (*vec_fcmuo<mode>): Rename to...
107 (*fcmuo<mode>): ...this and adjust likewise.
108 (*pred_fcm<cmp_op><mode>): New pattern.
109 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
110 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
112 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
114 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
115 (aarch64_emit_sve_predicated_cond): New function.
116 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
117 (aarch64_emit_unspec_cond_or): Replace with...
118 (aarch64_emit_sve_or_conds): ...this new function. Use
119 aarch64_emit_sve_ptrue_op for the individual comparisons and
120 aarch64_emit_binop to OR them together.
121 (aarch64_emit_inverted_unspec_cond): Replace with...
122 (aarch64_emit_sve_inverted_cond): ...this new function. Use
123 aarch64_emit_sve_ptrue_op for the comparison and
124 aarch64_emit_unop to invert the result.
125 (aarch64_expand_sve_vec_cmp_float): Update after the above
126 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
128 2018-05-07 Nathan Sidwell <nathan@acm.org>
130 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
131 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
132 (Backwards Compatibility): Likewise.
134 2018-05-07 Luis Machado <luis.machado@linaro.org>
138 2018-05-07 Luis Machado <luis.machado@linaro.org>
140 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
141 <prefetch_dynamic_strides>: New const bool field.
142 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
143 prefetch_dynamic_strides.
144 (exynosm1_prefetch_tune): Likewise.
145 (thunderxt88_prefetch_tune): Likewise.
146 (thunderx_prefetch_tune): Likewise.
147 (thunderx2t99_prefetch_tune): Likewise.
148 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
149 (aarch64_override_options_internal): Update to set
150 PARAM_PREFETCH_DYNAMIC_STRIDES.
151 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
152 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
153 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
154 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
155 prefetch-dynamic-strides setting.
157 2018-05-07 Luis Machado <luis.machado@linaro.org>
159 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
160 <minimum_stride>: New const int field.
161 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
162 minimum_stride field.
163 (exynosm1_prefetch_tune): Likewise.
164 (thunderxt88_prefetch_tune): Likewise.
165 (thunderx_prefetch_tune): Likewise.
166 (thunderx2t99_prefetch_tune): Likewise.
167 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
168 (aarch64_override_options_internal): Update to set
169 PARAM_PREFETCH_MINIMUM_STRIDE.
170 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
171 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
172 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
173 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
174 stride is constant and is below the minimum stride threshold.
176 2018-05-07 Luis Machado <luis.machado@linaro.org>
178 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
181 2018-05-07 Luis Machado <luis.machado@linaro.org>
183 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
184 <prefetch_dynamic_strides>: New const bool field.
185 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
186 prefetch_dynamic_strides.
187 (exynosm1_prefetch_tune): Likewise.
188 (thunderxt88_prefetch_tune): Likewise.
189 (thunderx_prefetch_tune): Likewise.
190 (thunderx2t99_prefetch_tune): Likewise.
191 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
192 (aarch64_override_options_internal): Update to set
193 PARAM_PREFETCH_DYNAMIC_STRIDES.
194 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
195 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
196 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
197 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
198 prefetch-dynamic-strides setting.
200 2018-05-07 Luis Machado <luis.machado@linaro.org>
202 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
203 <minimum_stride>: New const int field.
204 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
205 minimum_stride field.
206 (exynosm1_prefetch_tune): Likewise.
207 (thunderxt88_prefetch_tune): Likewise.
208 (thunderx_prefetch_tune): Likewise.
209 (thunderx2t99_prefetch_tune): Likewise.
210 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
211 (aarch64_override_options_internal): Update to set
212 PARAM_PREFETCH_MINIMUM_STRIDE.
213 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
214 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
215 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
216 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
217 stride is constant and is below the minimum stride threshold.
219 2018-05-06 Jakub Jelinek <jakub@redhat.com>
222 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
223 the type is addressable. Don't force op into register if it has
226 2018-05-05 Roland McGrath <mcgrathr@google.com>
229 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
230 any section for which we don't know a specific type it should have,
231 regardless of name. Previously this was done only for the exact
232 names ".init_array", ".fini_array", and ".preinit_array".
233 (default_elf_asm_named_section): Add comment about
234 relationship with default_section_type_flags and SECTION_NOTYPE.
235 (get_section): Don't consider it a type conflict if one side has
236 SECTION_NOTYPE and the other doesn't, as long as neither has the
237 SECTION_BSS et al used in the default_section_type_flags logic.
239 2018-05-05 Tom de Vries <tom@codesourcery.com>
242 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
243 (workaround_barsyncs): New function.
244 (nvptx_reorg): Use workaround_barsyncs.
245 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
246 (define_expand "nvptx_membar_cta"): New define_expand.
247 (define_insn "*nvptx_membar_cta"): New insn.
249 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
251 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
252 To improve optimization opportunities.
253 * builtin-types.def: The new needed builtin types for the above.
255 2018-05-04 Richard Biener <rguenther@suse.de>
257 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
258 * gimple-ssa-store-merging.c
259 (imm_store_chain_info::output_merged_store): Remove redundant create,
260 release split_store vector contents on failure.
261 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
262 scalar stmt vector on cache hit.
264 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
266 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
268 * config.gcc (powerpc-xilinx-eabi*): Remove.
269 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
271 (fusion_addis_mem_combo_load): Ditto.
272 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
274 (rs6000_cpu_cpp_builtins): Ditto.
275 * config/rs6000/rs6000-linux.c
276 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
277 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
278 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
280 (rs6000_setup_reg_addr_masks): Ditto.
281 (rs6000_init_hard_regno_mode_ok): Ditto.
282 (rs6000_option_override_internal): Ditto.
283 (legitimate_lo_sum_address_p): Ditto.
284 (rs6000_legitimize_address): Ditto.
285 (rs6000_legitimize_reload_address): Ditto.
286 (rs6000_legitimate_address_p): Ditto.
287 (abi_v4_pass_in_fpr): Ditto.
288 (setup_incoming_varargs): Ditto.
289 (rs6000_gimplify_va_arg): Ditto.
290 (rs6000_split_multireg_move): Ditto.
291 (rs6000_savres_strategy): Ditto.
292 (rs6000_emit_prologue_components): Ditto.
293 (rs6000_emit_epilogue_components): Ditto.
294 (rs6000_emit_prologue): Ditto.
295 (rs6000_emit_epilogue): Ditto.
296 (rs6000_elf_file_end): Ditto.
297 (rs6000_function_value): Ditto.
298 (rs6000_libcall_value): Ditto.
299 * config/rs6000/rs6000.h: Ditto.
300 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
301 (TARGET_MINMAX): ... this. New.
302 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
303 * config/rs6000/rs6000.md: Remove Xilinx FP support.
304 (*movsi_internal1_single): Delete.
305 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
306 mfpu=, mxilinx-fpu): Delete.
307 * config/rs6000/singlefp.h: Delete.
308 * config/rs6000/sysv4.h: Remove Xilinx FP support.
309 * config/rs6000/t-rs6000: Ditto.
310 * config/rs6000/t-xilinx: Delete.
311 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
312 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
313 (VStype_simple): Delete.
314 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
315 * config/rs6000/xfpu.h: Delete.
316 * config/rs6000/xfpu.md: Delete.
317 * config/rs6000/xilinx.h: Delete.
318 * config/rs6000/xilinx.opt: Delete.
319 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
320 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
322 2018-05-04 Tom de Vries <tom@codesourcery.com>
325 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
328 2018-05-04 Richard Biener <rguenther@suse.de>
331 * tree-complex.c (update_complex_assignment): We are always in SSA form.
332 (expand_complex_div_wide): Likewise.
333 (expand_complex_operations_1): Likewise.
334 (expand_complex_libcall): Preserve EH info of the original stmt.
335 (tree_lower_complex): Handle removed blocks.
336 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
337 on complex multiplication and division libcall builtins.
339 2018-05-04 Richard Biener <rguenther@suse.de>
342 * fold-const.c (negate_expr_p): Restrict negation of operand
343 zero of a division to when we know that can happen without
345 (fold_negate_expr_1): Likewise.
347 2018-05-04 Jakub Jelinek <jakub@redhat.com>
350 * real.h (real_nextafter): Declare.
351 * real.c (real_nextafter): New function.
352 * fold-const-call.c (fold_const_nextafter): New function.
353 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
355 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
356 even when arg1_mode is different from arg0_mode.
358 2018-05-03 Nathan Sidwell <nathan@acm.org>
360 * doc/extend.texi (Deprecated Features): Remove
362 (Backwards Compatibility): Likewise.
363 * doc/invoke.texi (C++ Language Options): Likewise.
364 (C++ Dialect Options): Likewise.
366 2018-05-03 Jakub Jelinek <jakub@redhat.com>
369 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
370 _mm512_mask_mullox_epi64): New intrinsics.
372 2018-05-03 Tom de Vries <tom@codesourcery.com>
375 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
376 dump files): Add offload-tree.
378 2018-05-03 Richard Biener <rguenther@suse.de>
380 PR tree-optimization/85615
381 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
382 to loops not nested in BBs loop father to avoid creating multi-entry
385 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
387 PR tree-optimization/70291
388 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
389 arguments. Change return type to tree. Emit libcall as a new
390 statement rather than replacing existing one when inplace_p is true.
391 (expand_complex_multiplication_components): New function.
392 (expand_complex_multiplication): Expand floating-point complex
393 multiplication using the above.
394 (expand_complex_division): Rename inner_type parameter to type.
395 Update expand_complex_libcall call-site.
396 (expand_complex_operations_1): Update expand_complex_multiplication
397 and expand_complex_division call-sites.
399 2018-05-02 Jakub Jelinek <jakub@redhat.com>
402 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
403 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
404 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
405 the highest significant bit of the shift count mask is clear. In
406 check whether and[sq]i3 is needed verify that all significant bits
407 of the shift count other than the highest are set.
409 2018-05-02 Tom de Vries <tom@codesourcery.com>
412 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
413 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
414 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
415 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
416 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
417 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
418 __builtin_goacc_parlevel_size.
420 2018-05-02 Richard Biener <rguenther@suse.de>
422 PR tree-optimization/85597
423 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
424 do not use split vect_get_vec_defs call but call vect_get_slp_defs
427 2018-05-02 Tom de Vries <tom@codesourcery.com>
430 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
431 dump files): Add ltrans-tree.
433 2018-05-02 Tom de Vries <tom@codesourcery.com>
436 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
437 dump files): Add wpa-ipa.
439 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
441 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
442 powerpc*-*-linux*paired* target.
443 * config/rs6000/750cl.h: Delete.
444 * config/rs6000/paired.h: Delete.
445 * config/rs6000/paired.md: Delete.
446 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
448 * config/rs6000/rs6000-builtin.def: Remove paired float support.
449 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
450 comment. Remove paired float support.
451 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
452 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
454 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
455 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
457 * config/rs6000/rs6000.c: Remove paired float support.
458 (paired_expand_vector_init, paired_expand_vector_move,
459 paired_emit_vector_compare, paired_emit_vector_cond_expr,
460 (paired_expand_lv_builtin, paired_expand_stv_builtin,
461 paired_expand_builtin, paired_expand_predicate_builtin,
462 paired_init_builtins): Delete.
463 * config/rs6000/rs6000.h: Remove paired float support.
464 * config/rs6000/rs6000.md: Remove paired float support.
465 (move_from_CR_ov_bit): Delete.
466 * config/rs6000/rs6000.opt (mpaired): Delete.
467 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
468 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
470 2018-05-02 Richard Biener <rguenther@suse.de>
473 * gimplify.c (gimplify_save_expr): When in SSA form allow
474 SAVE_EXPRs to compute to SSA vars.
476 2018-05-02 Jakub Jelinek <jakub@redhat.com>
479 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
480 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
481 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
482 clobber operands[2], instead use a new pseudo. Formatting fixes.
484 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
486 PR tree-optimization/85586
487 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
488 exit early for statements in the same group if the accesses are
491 2018-05-02 Tom de Vries <tom@codesourcery.com>
494 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
497 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
499 PR tree-optimization/85143
500 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
502 2018-05-01 Tom de Vries <tom@codesourcery.com>
505 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
506 not found" error message.
508 2018-05-01 Tom de Vries <tom@codesourcery.com>
511 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
512 * vec.c (test_ordered_remove_if): New function.
513 (vec_c_tests): Call test_ordered_remove_if.
514 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
515 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
516 * tree-vect-patterns.c (vect_pattern_recog_1): Use
517 VEC_ORDERED_REMOVE_IF.
519 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
521 PR tree-optimization/82665
522 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
523 pointer subtraction where arguments come from a memchr call.
525 2018-05-01 Jakub Jelinek <jakub@redhat.com>
527 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
528 --push-state --as-needed and --pop-state instead of --as-needed and
529 --no-as-needed if ld supports it.
530 * configure: Regenerated.
533 * doc/install.texi2html: Replace _002d with - and _002a with * in
534 generated html files using sed.
536 2018-04-30 David Malcolm <dmalcolm@redhat.com>
539 * gcc-rich-location.c (blank_line_before_p): New function.
540 (use_new_line): New function.
541 (gcc_rich_location::add_fixit_insert_formatted): New function.
542 * gcc-rich-location.h
543 (gcc_rich_location::add_fixit_insert_formatted): New function.
545 2018-04-30 David Malcolm <dmalcolm@redhat.com>
547 * selftest.c (assert_streq): Rename "expected" and "actual" to
548 "val1" and "val2". Extend NULL-handling to cover both inputs
549 symmetrically, while still requiring both to be non-NULL for a pass.
550 * selftest.h (assert_streq): Rename "expected" and "actual" to
552 (ASSERT_EQ): Likewise.
553 (ASSERT_EQ_AT): Likewise.
554 (ASSERT_KNOWN_EQ): Likewise.
555 (ASSERT_KNOWN_EQ_AT): Likewise.
556 (ASSERT_NE): Likewise.
557 (ASSERT_MAYBE_NE): Likewise.
558 (ASSERT_MAYBE_NE_AT): Likewise.
559 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
560 the assertion to pass.
561 (ASSERT_STREQ_AT): Likewise.
563 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
565 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
566 interaction with -pie.
568 2018-04-30 David Malcolm <dmalcolm@redhat.com>
570 * selftest.h: Fix alphabetization of per-source-file selftest
573 2018-04-30 Jason Merrill <jason@redhat.com>
575 PR c++/61982 - dead stores to destroyed objects.
576 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
579 2018-04-30 Jason Merrill <jason@redhat.com>
581 * tree.c (build_clobber): New.
582 * tree.h: Declare it.
583 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
585 2018-04-30 David Malcolm <dmalcolm@redhat.com>
587 * diagnostic-show-locus.c (layout::layout): Update for
588 location_get_source_line returning a char_span.
589 (struct char_span): Move to input.h.
590 (struct correction): Update for fields in char_span becoming
592 (struct source_line): Update for location_get_source_line
593 returning a char_span.
594 (layout::print_line): Likewise.
595 * edit-context.c (edited_file::print_content): Likewise.
596 (edited_file::print_diff_hunk): Likewise.
597 (edited_file::print_run_of_changed_lines): Likewise.
598 (edited_file::get_num_lines): Likewise.
599 (edited_line::edited_line): Likewise.
600 * final.c (asm_show_source): Likewise.
601 * input.c (location_get_source_line): Convert return type
602 from const char * to char_span, losing the final "line_len"
604 (dump_location_info): Update for the above.
605 (get_substring_ranges_for_loc): Likewise. Use a char_span
606 when handling the literal within the line.
607 (test_reading_source_line): Update for location_get_source_line
608 returning a char_span.
609 * input.h (class char_span): Move here from
610 diagnostic-show-locus.c, converting from a struct to a class.
611 Make data members private.
612 (char_span::operator bool): New.
613 (char_span::length): New.
614 (char_span::get_buffer): New.
615 (char_span::operator[]): New.
616 (char_span::subspan): Make const.
617 (char_span::xstrdup): New.
618 (location_get_source_line): Convert return type from const char *
619 to char_span, losing the final "line_size" param.
621 2018-04-30 Jan Hubicka <jh@suse.cz>
623 * lto-wrapper.c (ltrans_priorities): New static var.
625 (run_gcc): Read priorities and if doing parallel build order
626 the Makefile by them.
628 2018-04-30 David Malcolm <dmalcolm@redhat.com>
630 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
632 2018-04-30 Richard Biener <rguenther@suse.de>
634 * tree-cfg.c (verify_address): Remove base argument, add
635 flag whether to check TREE_ADDRESSABLE and do that.
636 (verify_expr): Remove.
637 (verify_types_in_gimple_reference): Add pieces from verify_expr.
638 (verify_gimple_assign_single): Likewise.
639 (verify_gimple_switch): Likewise.
640 (verify_expr_location_1): Dereference tp once. Add (disabled)
641 piece from verify_expr.
642 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
644 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
646 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
648 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
650 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
651 (small_data_pattern): Likewise.
652 (arc_rewrite_small_data): Likewise.
653 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
654 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
655 (get_symbol_alignment): New function.
656 (legitimate_small_data_address_p): Likewise.
657 (legitimate_scaled_address): Update, call
658 legitimate_small_data_address_p.
659 (output_sdata): New static variable.
660 (arc_print_operand): Update how we handle small data operands.
661 (arc_print_operand_address): Likewise.
662 (arc_legitimate_address_p): Update, use
663 legitimate_small_data_address_p.
664 (arc_rewrite_small_data_p): Remove.
665 (arc_rewrite_small_data_1): Likewise.
666 (arc_rewrite_small_data): Likewise.
667 (small_data_pattern): Likewise.
668 (compact_sda_memory_operand): Update to use
669 legitimate_small_data_address_p and get_symbol_alignment.
670 (prepare_move_operands): Don't rewite sdata pattern.
671 (prepare_extend_operands): Remove.
672 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
674 (zero_extendqisi2): Likewise.
675 (zero_extendhisi2): Likewise.
676 (extendqihi2): Likewise.
677 (extendqisi2): Likewise.
678 (extendhisi2): Likewise.
682 * config/arc/constraints.md (Usd): Change it to memory constraint.
684 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
686 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
687 as source of std instructions.
688 * config/arc/arc.md (movsi_insn): Update pattern predicate to
689 allow 6-bit constants as source for store instructions.
690 (movdi_insn): Update instruction pattern to allow 6-bit constants
691 as source for store instructions.
693 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
695 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
697 2018-04-30 Nathan Sidwell <nathan@acm.org>
698 Sandra Loosemore <sandra@codesourcery.com>
700 * dumpfile.c (dump_open): Allow '-' for stdout.
701 * doc/invoke.texi (Developer Options): Document dump filename
702 determination early. Document stdin/stdout selection.
704 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
706 Microblaze Target: PIC data text relative
708 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
709 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
711 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
712 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
713 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
714 New addressing mode for data-text relative position indepenedent code.
715 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
716 'ADDRESS_SYMBOLIC_TXT_REL'.
717 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
718 (microblaze_legitimate_pic_operand): Exclude function calls from
719 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
720 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
722 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
723 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
724 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
725 for 'address + offset'.
726 (microblaze_expand_prologue): Add new function prologue call for
728 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
729 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
730 table in case of TARGET_PIC_DATA_TEXT_REL.
731 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
732 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
733 Add new macros 'UNSPEC_TEXT',
734 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
735 + exclude function calls from 'UNSPEC_PLT' in case of data text
737 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
738 new target hook for generating address diff vector tables in case of
740 * doc/tm.texi : Regenerate.
741 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
742 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
743 of addr diff vector generation.
744 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
745 target hook definition.
746 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
747 Add default function for generate_pic_addr_diff_vec -> flag_pic.
748 * doc/invoke.texi (Add new pic option): Add new microblaze pic
749 option for data text relative.
751 2018-04-30 Richard Biener <rguenther@suse.de>
753 * tree-chrec.h (evolution_function_is_constant_p): Remove
755 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
757 2018-04-30 Richard Biener <rguenther@suse.de>
760 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
762 2018-04-30 Richard Biener <rguenther@suse.de>
764 PR tree-optimization/28364
765 PR tree-optimization/85275
766 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
767 copying first exit test.
769 2018-04-28 Mark Wielaard <mark@klomp.org>
771 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
773 (dwarf_AT): Handle DW_AT_addr_base.
774 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
776 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
779 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
780 (*ashl<dwi>3_doubleword_mask_1): Ditto.
781 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
782 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
784 2018-04-28 Richard Biener <rguenther@suse.de>
786 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
787 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
788 to reflect use. Only add interesting stmts.
790 2018-04-27 Martin Jambor <mjambor@suse.cz>
793 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
794 the jump function allows for passing through aggregate values.
796 2018-04-27 David Malcolm <dmalcolm@redhat.com>
798 * input.h (in_system_header_at): Convert from macro to inline
800 (from_macro_expansion_at): Likewise.
801 (from_macro_definition_at): Likewise.
803 2018-04-27 Jeff Law <law@redhat.com>
805 * config.gcc: Mark tile* targets as deprecated/obsolete.
807 2018-04-27 Richard Biener <rguenther@suse.de>
809 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
812 2018-04-27 Richard Biener <rguenther@suse.de>
814 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
816 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
818 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
819 with Yd constraint. Set "preferred_for_speed" attribute from
820 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
822 (*movdi_internal): Ditto.
823 (movti_interunit splitters): Remove
824 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
825 (movdi_interunit splitters): Ditto.
826 * config/i386/constraints.md (Ye): Remove.
827 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
829 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
832 * config/aarch64/constraints.md (Usg): Limit to 31.
835 2018-04-27 Jakub Jelinek <jakub@redhat.com>
837 PR tree-optimization/85529
838 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
839 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
840 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
841 zero extension or masking of the MSB bit.
842 (optimize_range_tests): Add FIRST_BB argument, pass it through
843 to optimize_range_tests_var_bound.
844 (maybe_optimize_range_tests, reassociate_bb): Adjust
845 optimize_range_tests callers.
847 2018-04-26 Richard Biener <rguenther@suse.de>
848 Jakub Jelinek <jakub@redhat.com>
850 * cgraph.h (symbol_table): Just declare debug method here.
851 * symtab.c (symbol_table::debug): Define.
853 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
855 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
857 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
859 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
860 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
861 (*movdi_internal): Substitute Yi and Yj constraint with x
862 and Ym and Yn constraint with y constraint. Update "isa"
863 attribute and set "preferred_for_speed" attribute from
864 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
865 (*movsi_internal): Ditto.
866 (*movdf_internal): Ditto.
867 (*movsf_internal): Ditto.
868 (*zero_extendsidi2): Ditto.
869 * config/i386/sse.md (vec_set<mode>_0): Ditto.
870 (sse2_loadld): Ditto.
871 (*vec_extract<ssevecmodelower>_0): Ditto.
872 (*vec_extractv4si_0_zext_sse4): Ditto.
873 (vec_concatv2di): Ditto.
874 (*vec_dup<mode>): Ditto.
875 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
876 * config/i386/constraints.md (Yi): Remove.
881 2018-04-26 Nathan Sidwell <nathan@acm.org>
883 * dumpfile.c (dump_open): New.
884 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
885 (dump_finish): Detect stdio/stderr by value not name.
887 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
889 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
891 2018-04-26 Tom de Vries <tom@codesourcery.com>
894 * config/nvptx/nvptx.c (verify_neutering_jumps)
895 (verify_neutering_labels): New function
896 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
898 2018-04-26 Tom de Vries <tom@codesourcery.com>
901 * config/nvptx/nvptx.c (needs_neutering_p): New function.
902 (nvptx_single): Use needs_neutering_p to skip over insns that do not
905 2018-04-26 Richard Biener <rguenther@suse.de>
906 Tom de Vries <tom@codesourcery.com>
909 * lto-streamer-out.c (output_function): Fixup loops if required to match
910 discovery done in the reader.
912 2018-04-26 Richard Biener <rguenther@suse.de>
914 PR tree-optimization/85116
915 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
916 have a loop exit from the single latch predecessor. Remove
917 case of header with just condition.
918 (ch_base::copy_headers): Exclude infinite loops from any
920 (pass_ch::execute): Record exits.
922 2018-04-26 Richard Biener <rguenther@suse.de>
924 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
925 prologue cost vector and pass it to vect_get_load_cost.
926 (vect_get_peeling_costs_all_drs): Likewise.
927 (vect_peeling_hash_get_lowest_cost): Likewise.
928 (vect_enhance_data_refs_alignment): Likewise.
930 2018-04-26 Richard Biener <rguenther@suse.de>
933 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
934 checking of integer<->pointer conversions.
935 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
936 sign-/zero-extending pointer types.
937 (expand_omp_for_static_chunk): Likewise.
939 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
940 Jean Lee <xiaoyur347@gmail.com>
942 * config/mips/mips.c (mips_asan_shadow_offset): New function.
943 (TARGET_ASAN_SHADOW_OFFSET): Define.
944 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
945 true for -fsanitize=address.
947 2018-04-25 Mark Wielaard <mark@klomp.org>
949 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
952 2018-04-25 Jakub Jelinek <jakub@redhat.com>
954 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
955 than "alu", remove explicit "memory" and "imm_disp" attributes.
956 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
959 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
960 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
963 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
966 * config/i386/i386.c (ix86_expand_builtin): Change memory
967 operand to XI, extend p0 to Pmode.
968 * config/i386/i386.md: Change unspec volatile and operand
969 1 mode to XI, change operand 0 mode to P.
971 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
973 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
974 GET_MODE_MASK before any checking.
975 (nds32_can_use_bset_p): Likewise.
976 (nds32_can_use_btgl_p): Likewise.
978 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
980 * config/nds32/nds32-doubleword.md: New define_split pattern for
981 illegal register number.
983 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
985 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
987 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
989 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
991 2018-04-25 Richard Biener <rguenther@suse.de>
993 * lto-streamer.h (LTO_major_version): Bump to 8.
995 2018-04-25 Jakub Jelinek <jakub@redhat.com>
997 * BASE-VER: Set to 9.0.0.
999 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1001 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1002 in __abskf2 and __powikf2.
1004 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1007 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1008 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1009 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1010 Use the above on operand 2. Reindent.
1011 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1013 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1016 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1018 * config/i386/i386.opt (mcet): Removed.
1019 * doc/install.texi: Remove -mcet documentation.
1020 * doc/invoke.texi: Likewise.
1022 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1025 * doc/install.texi: Remove -mcet from bootstrap-cet.
1027 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1030 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1031 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1035 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1036 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1037 containing a CONST_VECTOR.
1039 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1041 * doc/install.texi: Update newlib dependency for nvptx.
1043 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1046 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1047 instead of INTVAL when shifting x left.
1049 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1051 PR tree-optimization/85478
1052 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1053 vect_grouped_store_supported for single element vectors.
1055 2018-04-24 Richard Biener <rguenther@suse.de>
1058 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1059 load cost increase to the case of non-constant step.
1061 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1064 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1065 destination if any_malformed_asm.
1067 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1070 * expr.c (store_field): In the bitfield case, if the value comes from
1071 a function call and is returned in registers by means of a PARALLEL,
1072 do not change the mode of the temporary unless BLKmode and VOIDmode.
1074 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1076 PR rtl-optimization/85423
1077 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1078 dependencies to debug insns when the previous insn is non-debug.
1080 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1082 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1083 enums into a single definition.
1084 (fls): Fix predicates and printing.
1087 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1089 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1090 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1091 and short u6 immediate.
1092 (check_if_valid_sleep_operand): Remove.
1093 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1095 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1097 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1098 flag_always_save_lp condition.
1099 * config/nds32/nds32.opt (malways-save-lp): New option.
1101 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1103 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1104 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1105 * config/nds32/nds32.h
1106 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1107 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1109 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1111 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1112 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1114 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1115 Chung-Ju Wu <jasonwucj@gmail.com>
1117 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1119 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1120 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1122 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1124 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1126 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1128 * config/nds32/nds32-protos.h (nds32_data_alignment,
1129 nds32_local_alignment): Declare.
1130 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1131 nds32_local_alignment): New functions.
1132 (TARGET_CONSTANT_ALIGNMENT): Define.
1133 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1135 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1137 * config/nds32/nds32.c
1138 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1139 (TARGET_MODES_TIEABLE_P): Likewise.
1141 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1143 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1146 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1147 Chung-Ju Wu <jasonwucj@gmail.com>
1149 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1150 for unaligned access.
1151 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1153 * config/nds32/nds32-intrinsic.md: Likewise.
1154 * config/nds32/nds32_intrinsic.h: Likewise.
1155 * config/nds32/nds32.h (nds32_builtins): Likewise.
1156 * config/nds32/nds32.opt (munaligned-access): New option.
1157 * config/nds32/nds32.c (nds32_asm_file_start): Display
1158 flag_unaligned_access status.
1160 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1162 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1163 -mno-relax is present.
1164 * config/riscv/linux.h (LINK_SPEC): Ditto.
1166 2018-04-20 Martin Sebor <msebor@redhat.com>
1169 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1171 (gimple_fold_builtin_stxcpy_chk): Same.
1172 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1174 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1177 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1178 __powikf2 when long double is IEEE 128-bit.
1180 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1182 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1183 step to make sure stack always aligned.
1185 2018-04-20 Carl Love <cel@us.ibm.com>
1188 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1189 size check for arg0.
1191 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1192 Tom de Vries <tom@codesourcery.com>
1195 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1196 Emit insns for calls too.
1197 (nvptx_find_par): Always look for worker-level predecessor insn.
1198 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1200 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1201 (nvptx_process_pars): Propagate frames for calls.
1203 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1206 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1208 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1209 (ix86_handle_option): Don't handle OPT_mibt.
1210 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1212 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1214 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1215 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1216 (ix86_target_macros): Define __CET__ with flag_cf_protection
1217 for -fcf-protection.
1218 * config/i386/i386.c (isa2_opts): Remove -mibt.
1219 * config/i386/i386.h (TARGET_IBT): Removed.
1220 (TARGET_IBT_P): Likewise.
1221 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1222 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1223 * config/i386/i386.opt (mcet): Update help message.
1226 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1227 -mcet as an alias for -mshstk.
1229 2018-04-20 Richard Biener <rguenther@suse.de>
1232 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1233 complexity by forcing a single use of the multiply operand.
1235 2018-04-20 Martin Jambor <mjambor@suse.cz>
1238 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1239 recursion dependency to only apply to non-clones.
1241 2018-04-20 Martin Jambor <mjambor@suse.cz>
1244 * ipa-cp.c (create_specialized_node): Check that clones of
1245 self-recursive edges exist during IPA-CP.
1247 2018-04-19 Toon Moene <toon@moene.org>
1249 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1252 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1254 PR tree-optimization/85467
1255 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1256 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1257 VECTOR_CST element to type.
1259 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1262 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1263 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1264 (builtin_longjmp): Likewise.
1265 (save_stack_nonlocal): New pattern.
1266 (restore_stack_nonlocal): Likewise.
1268 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1271 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1272 Replace ASM_OUTPUT_LABEL with fprintf.
1274 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1277 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1278 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1279 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1280 define __IBT__ and __SHSTK__ for -fcf-protection.
1281 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1283 (ix86_trampoline_init): Likewise.
1284 (x86_output_mi_thunk): Likewise.
1285 (ix86_notrack_prefixed_insn_p): Likewise.
1286 (ix86_option_override_internal): Don't disallow -fcf-protection.
1287 * config/i386/i386.md (rdssp<mode>): Also enable for
1289 (incssp<mode>): Likewise.
1290 (nop_endbr): Likewise.
1291 * config/i386/i386.opt (mcet): Change help message to built-in
1295 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1296 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1297 enable CET built-in functions.
1299 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1301 * common/config/i386/i386-common.c
1302 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1303 OPTION_MASK_ISA_MOVDIRI_UNSET,
1304 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1305 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1306 * config.gcc (movdirintrin.h): New header.
1307 * config/i386/cpuid.h (bit_MOVDIRI,
1308 bit_MOVDIR64B): New bits.
1309 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1311 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1312 (VOID, PVOID, PCVOID)): New function types.
1313 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1314 __builtin_ia32_directstoreu_u64,
1315 __builtin_ia32_movdir64b): New builtins.
1316 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1317 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1319 (ix86_valid_target_attribute_inner_p): Ditto.
1320 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1321 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1322 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1323 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1324 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1325 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1326 (movdiri<mode>, movdir64b_<mode>): New.
1327 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1328 * config/i386/immintrin.h: Include movdirintrin.h.
1329 * config/i386/movdirintrin.h: New file.
1330 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1332 2018-04-19 Richard Biener <rguenther@suse.de>
1335 * cfg.c (clear_bb_flags): When loop state says we have
1336 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1338 2018-04-19 Richard Biener <rguenther@suse.de>
1340 PR tree-optimization/84737
1341 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1342 copying restrict info.
1343 (vect_setup_realignment): Use it.
1344 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1345 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1346 the first DR to all generated stores.
1347 (vectorizable_load): Likewise for loads.
1349 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1351 PR tree-optimization/85446
1352 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1353 the integral and pointer types to have the same precision.
1355 * doc/install.texi: Document --disable-cet being the default and
1358 2018-04-18 Martin Liska <mliska@suse.cz>
1360 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1363 2018-04-18 Martin Liska <mliska@suse.cz>
1366 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1369 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1370 arguments if they are comparable.
1372 2018-04-18 Martin Liska <mliska@suse.cz>
1375 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1378 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1381 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1384 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1385 ENDBR after calling __morestack.
1387 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1390 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1391 by using gcc_base_ver to generate a gcc_driver_version, and use
1392 it when generating GCC_DRIVER_NAME.
1393 * configure: Regenerate.
1395 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1398 * config.gcc: Obsolete powerpc*-*-*spe*.
1400 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1403 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1404 (stabstr_D): Change type of unum from unsigned int to
1405 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1408 2018-04-17 Jim Wilson <jimw@sifive.com>
1411 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1412 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1413 Set arg_pointer_offset after using pretend_args_size.
1415 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1417 PR rtl-optimization/85431
1418 * dse.c (record_store): Ignore zero width stores.
1421 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1422 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1423 __builtin_stack_restore rather than after it.
1424 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1425 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1426 argument instead of virtual_dynamic_stack_rtx.
1428 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1430 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1432 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1433 Add note to error message to explain internal mapping of overloaded
1434 built-in function name to non-overloaded built-in function name.
1435 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1438 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1441 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1442 where the inputs overlap with the output.
1444 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1446 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1447 (=v, v) alternative and explicit "memory" attribute.
1448 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1449 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1451 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1452 "sselog1" type instead of "sselog".
1453 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1454 "sselog". Remove explicit "memory" attribute.
1455 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1456 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1458 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1459 "sselog1" type instead of "sselog". Remove explicit "memory"
1461 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1462 use "sselog1" type instead of "sselog". Remove explicit "memory"
1464 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1465 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1467 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1468 "sselog1" type instead of "sselog". Remove explicit "memory"
1470 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1471 use "sselog1" type instead of "sselog". Remove explicit "memory"
1475 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1478 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1481 2018-04-17 Martin Jambor <mjambor@suse.cz>
1484 * ipa-cp.c (create_specialized_node): Call
1485 expand_all_artificial_thunks if necessary.
1487 2018-04-17 Martin Liska <mliska@suse.cz>
1490 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1491 in message, remote space in between '_G' and '('.
1493 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1496 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1497 avx512f_vmcmp<mode>3<round_saeonly_name>,
1498 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1499 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1500 avx512f_rndscale<mode><round_saeonly_name>,
1501 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1502 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1503 Use %<iptr>2 instead of %2 for -masm=intel.
1504 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1505 avx512f_vcvttss2usi<round_saeonly_name>,
1506 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1508 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1509 avx512f_vcvttsd2usi<round_saeonly_name>,
1510 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1511 Use %q1 instead of %1 for -masm=intel.
1512 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1513 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1514 of %3 for -masm=intel.
1515 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1517 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1519 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1521 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1522 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1524 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1526 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1527 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1528 %g1 and one with %0 and %1.
1529 (avx512er_vmrcp28<mode><round_saeonly_name>,
1530 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1532 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1533 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1534 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1535 of %0 and %{%4%} for -masm=intel.
1536 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1537 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1538 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1539 order of %0 and %{%5%}%{z%} for -masm=intel.
1541 2018-04-17 Jan Hubicka <jh@suse.cz>
1544 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1546 2018-04-17 Martin Liska <mliska@suse.cz>
1549 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1550 for target_clone error message. Make default implementation
1551 clone to be a local declaration.
1552 (separate_attrs): Add new argument and check for an empty
1554 (expand_target_clones): Handle it.
1555 (ipa_target_clone): Make redirection just for target_clones
1558 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1559 Tom de Vries <tom@codesourcery.com>
1562 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1563 tiled basic blocks without omp continue statements.
1565 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1568 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1569 vec_extract expression as having side effects to make sure it gets
1572 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1575 * config/i386/i386.c (get_builtin_code_for_version): Check
1578 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1581 * gcc/config.gcc: Support "skylake".
1582 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1584 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1585 (processor_target_table): Add "skylake".
1586 (ix86_option_override_internal): Add "skylake".
1587 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1588 PROCESSOR_CANNONLAKE.
1589 (get_builtin_code_for_version): Fix priority for
1590 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1591 PROCESSOR_SKYLAKE-AVX512.
1592 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1593 (processor_type): Add PROCESSOR_SKYLAKE.
1595 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1596 Jason Merrill <jason@redhat.com>
1599 * convert.c (convert_to_integer_1): Use direct recursion for
1600 enumeral types and types with a precision less than the number
1601 of bits in their mode.
1603 2018-04-16 Julia Koval <julia.koval@intel.com>
1606 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1607 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1609 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1612 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1613 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1614 and -mno-direct-move.
1616 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1619 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1620 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1621 Ensure negative shifts result in {0}.
1623 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1625 PR rtl-optimization/79916
1626 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1627 regs (if any) to define how to gnerate SD moves when LRA is in
1630 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1632 PR rtl-optimization/85393
1633 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1634 * except.c (expand_dw2_landing_pad_for_region): Make static.
1635 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1636 a label and unconditional jump to old_bb, rather than
1637 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1640 PR rtl-optimization/85376
1641 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1642 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1643 instead of a specific value.
1645 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1646 Bin Cheng <bin.cheng@arm.com>
1648 PR tree-optimization/82965
1649 PR tree-optimization/83991
1650 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1651 by_profile_only parameter.
1652 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1653 information if the loop was predicted to iterate too many times.
1654 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1656 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1659 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1662 2018-04-13 Martin Liska <mliska@suse.cz>
1663 Jakub Jelinek <jakub@redhat.com>
1666 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1667 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1668 * builtins.c (expand_builtin_memory_copy_args): Use
1669 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1670 handle dest_addr == pc_rtx.
1672 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1675 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1676 asked to not generate direct moves.
1677 (fix_trunc<mode>si2_stfiwx): Similar.
1678 (fix_trunc<mode>si2_internal): Similar.
1680 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1683 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1684 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1685 lookup if dest in some wider mode is known to be const0_rtx and
1686 if so, record permanent equivalence for it to be ZERO_EXTEND of
1687 the narrower mode destination.
1689 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1691 * lto-streamer-out.c (output_function): Revert 259346.
1692 * omp-expand.c (expand_oacc_for): Likewise.
1694 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1696 PR rtl-optimization/85354
1697 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1698 * sel-sched.c (sel_global_init): ... here.
1700 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1703 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1704 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1705 mode for PE-COFF targets.
1706 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1707 (i386_pe_asm_lto_end): Likewise.
1708 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1709 (TARGET_ASM_LTO_END): Likewise.
1710 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1711 (i386_pe_asm_lto_start): New function.
1712 (i386_pe_asm_lto_end): Likewise.
1714 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1715 Richard Biener <rguenther@suse.de>
1718 * lto-streamer-out.c (output_function): Fix CFG loop state before
1720 * omp-expand.c (expand_oacc_for): Handle calls to internal
1721 functions like regular functions.
1723 2018-04-12 Richard Biener <rguenther@suse.de>
1726 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1727 for the early LTO debug to properly generate references to it
1728 during DIE emission. Do not re-use that for the skeleton for
1730 (dwarf2out_early_finish): Likewise.
1732 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1735 * config/i386/sse.md
1736 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1737 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1738 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1739 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1740 and output is a reg, avoid creating invalid lowpart subreg, but
1741 instead split into a 512-bit move. Don't split if not AVX512VL,
1742 input is xmm16+ reg and output is a mem.
1743 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1744 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1745 xmm16+ reg and output is a mem.
1747 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1749 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1750 also for flag_dwarf2_cfi_asm.
1752 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1754 PR rtl-optimization/85342
1755 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1756 a bool scalar var inside of the loop instead. Don't try to update
1757 recog_data.operand after failed apply_change_group.
1759 2018-04-12 Tom de Vries <tom@codesourcery.com>
1762 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1763 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1764 array with flexible array member as array without given dimension.
1765 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1766 argument for undefined param to true.
1768 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1771 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1772 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1773 from PowerPC section.
1774 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1775 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1776 help text that is too long.
1777 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1778 help text that is too long.
1779 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1780 help text that is too long.
1782 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1784 * config/alpha/alpha.md (stack_probe_internal): Rename
1785 from "probe_stack". Update all callers.
1787 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1789 PR rtl-optimization/84566
1790 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1791 sched_macro_fuse_insns.
1793 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1796 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1797 (compute_block_dependences): ... from here.
1799 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1801 PR tree-optimization/85331
1802 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1803 from int to HOST_WIDE_INT.
1805 2018-04-11 Martin Jambor <mjambor@suse.cz>
1808 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1809 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1810 not the same as the source val.
1811 (cgraph_edge_brings_value_p): New parameter.
1812 (gather_edges_for_value): Pass destination value to
1813 cgraph_edge_brings_value_p.
1814 (perhaps_add_new_callers): Likewise.
1815 (get_info_about_necessary_edges): Likewise and exclude values brought
1816 only by self-recursive edges.
1817 (create_specialized_node): Redirect only clones of self-calling edges.
1818 (+self_recursive_pass_through_p): New function.
1819 (find_more_scalar_values_for_callers_subset): Use it.
1820 (find_aggregate_values_for_callers_subset): Likewise.
1821 (known_aggs_to_agg_replacement_list): Removed.
1822 (decide_whether_version_node): Re-calculate known constants for all
1823 remaining context clones.
1825 2018-04-11 Richard Biener <rguenther@suse.de>
1828 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1829 from early DWARF output.
1830 (dwarf2out_early_finish): Output line info unconditionally into
1831 early DWARF and add reference to it.
1833 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1836 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1837 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1838 other than V2DFmode using iptr mode attribute.
1839 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1841 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1843 PR rtl-optimization/84659
1844 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1846 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1849 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1851 (output_loc_list): Pass address of a dummy size variable even in the
1852 locview handling loop.
1853 (index_location_lists): Add comment on why skip_loc_list_entry can't
1856 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1859 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1862 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1865 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1866 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1867 and -mstring-compare-inline-limit.
1869 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1872 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1873 for stack clash protection in a register whenever we need it to be in
1876 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1878 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1879 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1881 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1884 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1886 (mlong-double-): Ditto.
1887 * config/rs6000/sysv4.opt (msdata=): Ditto.
1888 (mtls-size=): Ditto.
1890 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1892 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1893 erroneous entries for
1894 "vector int vec_ldl (int, long int *)", and
1895 "vector unsigned int vec_ldl (int, unsigned long int *)".
1896 Add comments and entries for
1897 "vector bool char vec_ldl (int, bool char *)",
1898 "vector bool short vec_ldl (int, bool short *)",
1899 "vector bool int vec_ldl (int, bool int *)",
1900 "vector bool long long vec_ldl (int, bool long long *)",
1901 "vector pixel vec_ldl (int, pixel *)",
1902 "vector long long vec_ldl (int, long long *)",
1903 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1904 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1905 type tree bool_long_long_type_node and correct definition of
1906 bool_V2DI_type_node to make reference to this new type tree.
1907 (rs6000_mangle_type): Replace erroneous reference to
1908 bool_long_type_node with bool_long_long_type_node.
1909 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1910 comments to emphasize sign distinctions for char and int types and
1911 replace RS6000_BTI_bool_long constant with
1912 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1913 use of RS6000_BTI_pixel.
1914 (bool_long_type_node): Remove this macro definition.
1915 (bool_long_long_type_node): New macro definition
1917 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1919 PR rtl-optimization/85300
1920 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1921 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1922 simplify_unary_operation fails.
1924 2018-04-10 Martin Liska <mliska@suse.cz>
1926 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1927 cgraph_edge and ipa_ref.
1929 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1933 * config/i386/sse.md
1934 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1935 computation of the VEC_MERGE selector from mask.
1936 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1937 Fix decoding of the VEC_MERGE selector into mask.
1939 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1941 PR tree-optimization/85286
1942 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1944 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1946 * final.c (final_1): Set insn_last_address as well as
1947 insn_current_address.
1949 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1952 * explow.c (emit_stack_probe): Call validize_mem on memory location
1953 before passing it to gen_probe_stack. Create address operand and
1954 legitimize it for the probe_stack_address case.
1956 2018-04-09 Jan Hubicka <jh@suse.cz>
1959 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1960 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1961 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1963 (free_lang_data): Rebuild type inheritance graph.
1965 2018-04-09 Martin Sebor <msebor@redhat.com>
1967 * invoke.texi (-finline-small-functions): Mention other optimization
1969 (-findirect-inlining, -fpartial-inlining): Same.
1970 (-finline-functions-called-once): Same.
1971 (-freorder-blocks-and-partition): Same.
1973 2018-04-09 Jan Hubicka <jh@suse.cz>
1976 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1977 jumps; choose last target that matches the criteria (i.e.
1978 no partition changes for non-crossing jumps).
1979 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1980 support for redirecting crossing jumps to non-crossing.
1982 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1984 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1985 also for naked functions.
1987 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1989 * config/arc/arc.md (add_shift): New pattern.
1990 (add_shift2): Likewise.
1991 (sub_shift): Likewise.
1992 (sub_shift_cmp0_noout): Likewise.
1993 (compare_si_ashiftsi): Likewise.
1994 (xbfu_cmp0_noout): New combine pattern.
1995 (xbfu_cmp0"): Likewise.
1996 (movsi_set_cc_insn): Place the predicable variant first.
1997 (commutative_binary_cmp0_noout): Remove clobber.
1998 (commutative_binary_cmp0): New pattern.
1999 (noncommutative_binary_cmp0): Likewise.
2000 (noncommutative_binary_cmp0_noout): Likewise.
2001 (noncommutative_binary_comparison_result_used): Removed.
2002 (rsub_cmp0): New pattern.
2003 (rsub_cmp0_noout): Likewise.
2004 (extzvsi): Changed, keep only meaningful variants.
2005 (SQH, SEZ): New iterators.
2006 (SQH_postfix): New mode attribute.
2007 (SEZ_prefix): New code attribute.
2008 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2009 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2010 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2012 (noncommutative_operator): Check the availability of barrel
2015 2018-04-09 Richard Biener <rguenther@suse.de>
2017 PR tree-optimization/85284
2018 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2019 Only use the niter constraining form of simple_iv when the exit
2022 2018-04-09 Tom de Vries <tom@codesourcery.com>
2025 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2026 (define_expand "*memory_barrier"): New define_expand.
2027 (define_insn "memory_barrier"): New insn.
2029 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2031 PR rtl-optimization/80463
2032 PR rtl-optimization/83972
2033 PR rtl-optimization/83480
2035 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2036 correct producer for the insn.
2037 (tidy_control_flow): Fixup seqnos in case of debug insns.
2039 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2041 PR rtl-optimization/83913
2043 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2044 different sched-times when merging exprs.
2046 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2048 PR rtl-optimization/83962
2050 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2051 tidy_fallthru_edge and tidy_control_flow.
2053 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2055 PR rtl-optimization/83530
2057 * sel-sched.c (force_next_insn): New global variable.
2058 (remove_insn_for_debug): When force_next_insn is true, also leave only
2059 next insn in the ready list.
2060 (sel_sched_region): When the region wasn't scheduled, make another pass
2061 over it with force_next_insn set to 1.
2063 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2065 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2067 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2068 for interrupt control.
2069 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2070 functions for interrupt control.
2071 * config/nds32/nds32-intrinsic.md: Likewise.
2072 * config/nds32/nds32_intrinsic.h: Likewise.
2073 * config/nds32/nds32.h (nds32_builtins): Likewise.
2075 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2077 * config/nds32/nds32.c (nds32_init_machine_status,
2078 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2079 strict_aligned_p field.
2080 (nds32_expand_to_rtl_hook): New function.
2081 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2082 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2084 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2085 Chung-Ju Wu <jasonwucj@gmail.com>
2087 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2088 * config/nds32/nds32-n7.md: New file.
2089 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2090 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2092 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2093 * config/nds32/nds32.md (pipeline_model): Add n7.
2094 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2095 * config/nds32/pipelines.md: Include n7 settings.
2097 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2098 Chung-Ju Wu <jasonwucj@gmail.com>
2100 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2101 * config/nds32/nds32-e8.md: New file.
2102 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2103 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2105 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2106 * config/nds32/nds32.md (pipeline_model): Add e8.
2107 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2108 * config/nds32/pipelines.md: Include e8 settings.
2110 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2111 Chung-Ju Wu <jasonwucj@gmail.com>
2113 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2114 * config/nds32/nds32-n8.md: New file.
2115 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2116 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2118 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2119 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2120 * config/nds32/nds32.md (pipeline_model): Add n8.
2121 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2122 * config/nds32/pipelines.md: Include n8 settings.
2124 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2125 Chung-Ju Wu <jasonwucj@gmail.com>
2127 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2128 * config/nds32/nds32-n9-2r1w.md: New file.
2129 * config/nds32/nds32-n9-3r2w.md: New file.
2130 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2131 nds32_register_ports): New or modify for cpu n9.
2132 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2134 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2135 * config/nds32/nds32-utils.c: New file.
2136 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2137 TARGET_MUL_SLOW): Define.
2138 * config/nds32/nds32.md (pipeline_model): New attribute.
2139 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2140 New options that support cpu n9.
2141 * config/nds32/pipelines.md: Include n9 settings.
2142 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2144 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2146 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2147 information if necessary.
2148 (output_cond_branch_compare_zero): Likewise.
2149 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2150 (nds32_target_alignment): Refine for alignment.
2151 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2152 (FUNCTION_BOUNDARY): Modify.
2153 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2155 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2157 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2159 * config/nds32/constants.md (unspec_volatile_element): Add values for
2160 TLB operation and data prefetch.
2161 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2162 functions for TLB operation and data prefetch.
2163 * config/nds32/nds32-intrinsic.md: Likewise.
2164 * config/nds32/nds32_intrinsic.h: Likewise.
2165 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2166 (nds32_print_operand): Likewise.
2167 * config/nds32/nds32.h (nds32_builtins): Likewise.
2169 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2170 Andrew Pinski <pinsika@gcc.gnu.org>
2173 * match.pd: Use constant_boolean_node of correct type instead of
2174 boolean_true_node or boolean_false_node for simplifying
2175 pointer comparisons to zero.
2177 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2179 PR tree-optimization/80021
2180 * tree.c (verify_type_variant): Make error call in verify_variant_match
2181 translatable and remove final full stop.
2183 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2185 * config/nds32/constants.md (unspec_volatile_element): Add
2186 UNSPEC_VOLATILE_EH_RETURN.
2187 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2188 nds32_output_stack_pop): Support dwarf exception handling process.
2189 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2190 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2191 exception handling process.
2192 (nds32_compute_stack_frame): Likewise.
2193 (nds32_return_addr_rtx): Likewise.
2194 (nds32_initial_elimination_offset): Likewise.
2195 (nds32_expand_prologue): Likewise.
2196 (nds32_expand_epilogue): Likewise.
2197 (nds32_dynamic_chain_address): New function.
2198 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2200 (DYNAMIC_CHAIN_ADDRESS): Define.
2201 (EH_RETURN_DATA_REGNO): Define.
2202 (EH_RETURN_STACKADJ_RTX): Define.
2203 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2204 patterns for dwarf exception handling.
2206 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2208 * config/nds32/nds32.h: Clean up obsolete macros.
2210 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2212 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2213 Add enum values for particular instructions.
2214 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2215 particular intrinsic functions.
2216 * config/nds32/nds32-intrinsic.md: Likewise.
2217 * config/nds32/nds32_intrinsic.h: Likewise.
2218 * config/nds32/nds32.h (nds32_builtins): Likewise.
2219 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2220 (btst, ave): New patterns for particular instructions.
2222 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2224 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2225 Add enum values for atomic load/store and memory sync.
2226 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2228 * config/nds32/nds32-intrinsic.md: Likewise.
2229 * config/nds32/nds32_intrinsic.h: Likewise.
2230 * config/nds32/nds32.h (nds32_builtins): Likewise.
2232 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2234 PR tree-optimization/85257
2235 * fold-const.c (native_encode_vector): If not all elts could fit
2236 and off is -1, return 0 rather than offset.
2237 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2238 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2239 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2240 adjust buffer in native_interpret_expr call.
2242 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2244 * config/nds32/constants.md (unspec_volatile_element): Add cache
2245 control enum values.
2246 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2247 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2248 * config/nds32/nds32.c (nds32_cctl_names): New.
2249 (nds32_print_operand): Handle cache control register names.
2250 * config/nds32/nds32.h (nds32_builtins): New enum values.
2251 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2253 * config/nds32/nds32.md (type): Add mmu.
2254 * config/nds32/pipelines.md (simple_insn): Add mmu.
2256 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2258 * config/nds32/nds32.md (type): Remove call.
2259 * config/nds32/pipelines.md (simple_insn): Likewise.
2261 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2263 * config/nds32/constants.md (unspec_volatile_element): Add
2264 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2265 UNSPEC_VOLATILE_FMFCFG.
2266 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2267 description for fmfcfg and fmfcsr.
2268 (bdesc_1arg): Add fmtcsr.
2269 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2270 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2271 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2272 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2273 unspec_fmfcfg): New patterns.
2274 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2275 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2276 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2277 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2278 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2279 __nds32__fmfcfg): Define.
2281 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2283 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2284 intrinsic register names.
2285 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2286 intrinsic register enum values and macros.
2288 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2290 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2291 for load/store addressing form.
2292 (nds32_print_operand_address): Likewise.
2294 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2297 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2298 based on LABEL_REF. Remove useless assertion.
2299 (pic_address_needs_scratch): Fix formatting.
2300 (sparc_legitimize_pic_address): Minor tweaks.
2301 (sparc_delegitimize_address): Adjust assertion accordingly.
2302 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2303 into symbolic_operand.
2304 (movsi_high_pic_label_ref): Likewise.
2305 (movsi_lo_sum_pic_label_ref): Likewise.
2306 (movdi_pic_label_ref): Likewise.
2307 (movdi_high_pic_label_ref): Likewise.
2308 (movdi_lo_sum_pic_label_ref): Likewise.
2310 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2312 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2313 custom LIB_SPEC setup.
2315 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2316 Kito Cheng <kito.cheng@gmail.com>
2318 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2319 * config/riscv/freebsd.h: New.
2321 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2323 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2324 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2327 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2328 Kito Cheng <kito.cheng@gmail.com>
2330 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2331 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2332 * config/nds32/nds32-protos.h (nds32_output_call,
2333 nds32_output_return): Declare.
2334 * config/nds32/nds32.md: Refine all the call and return patterns.
2336 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2339 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2340 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2342 PR rtl-optimization/84872
2343 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2344 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2347 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2349 * expr.c (copy_blkmode_to_reg): Revert 254862.
2350 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2352 2018-04-06 Richard Biener <rguenther@suse.de>
2355 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2356 after seeing a component reference with an adjacent field. Treat
2357 refs to arrays at struct end of external decls similar to
2358 refs to unconstrained commons.
2360 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2363 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2364 look through SAVE_EXPRs with non-side-effects argument. Adjust
2366 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2369 2018-04-06 Richard Biener <rguenther@suse.de>
2372 * alias.c (find_base_term): New wrapper around find_base_term
2373 unwinding CSELIB_VAL_PTR changes.
2374 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2377 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2379 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2381 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2382 constant definitions.
2383 ("nop"): lr 0,0 -> nopr r0
2384 ("nop_lr0", "nop_lr1"): New insn definitions.
2386 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2388 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2389 NDS32_V3PUSH_AVAILABLE_P macro.
2391 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2392 Chung-Ju Wu <jasonwucj@gmail.com>
2394 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2395 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2396 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2397 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2398 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2399 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2400 * config/nds32/constraints.md: New constraints and checking for hard
2401 float configuration.
2402 * config/nds32/iterators.md: New mode iterator and attribute for hard
2403 float configuration.
2404 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2406 * config/nds32/nds32-fpu.md: New file.
2407 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2408 deal with hard float code generation.
2409 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2411 (abi_type, float_reg_number): New enum type.
2412 * config/nds32/nds32-predicates.c: New predicates for hard float.
2413 * config/nds32/nds32-protos.h: Declare functions for hard float.
2414 * config/nds32/nds32.c: Implementation for hard float configuration.
2415 * config/nds32/nds32.h: Definitions for hard float configuration.
2416 * config/nds32/nds32.md: Include hard float machine description and
2417 modify patterns for hard float configuration.
2418 * config/nds32/nds32.opt: New options for hard float configuration.
2419 * config/nds32/predicates.md: New predicates for hard float
2422 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2424 * common/config/nds32/nds32-common.c
2425 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2427 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2430 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2431 CONSTRUCTOR_ELT (ctor, ...)->value.
2433 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2436 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2438 2018-04-05 Tom de Vries <tom@codesourcery.com>
2441 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2444 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2445 Kito Cheng <kito.cheng@gmail.com>
2447 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2448 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2449 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2450 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2452 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2453 Kito Cheng <kito.cheng@gmail.com>
2455 * config/nds32/constraints.md (Ufe): New memory constraint.
2456 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2457 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2458 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2460 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2461 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2463 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2465 * config/nds32/nds32.md: Use optimize_size in the condition for
2466 alu-shift instructions.
2468 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2470 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2472 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2474 * config/nds32/nds32.md (negsi2): Refine pattern.
2476 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2477 Chung-Ju Wu <jasonwucj@gmail.com>
2479 * config/nds32/iterators.md (shift_rotate): New code iterator.
2480 (shift): New code attribute.
2481 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2482 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2483 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2484 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2485 bit-wise operations.
2486 (andsi3, *andsi3): Ditto.
2487 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2488 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2489 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2490 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2491 nds32_ior_operand, nds32_xor_operand): New predicates.
2493 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2495 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2496 (addsi3, subsi3): ... this.
2498 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2500 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2502 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2504 * config/nds32/nds32.md: Adjust indention.
2506 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2508 * config/nds32/nds32.md (feature): New attribute.
2510 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2512 * config/nds32/nds32.md (subtype): New attribute.
2514 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2517 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2518 expansion to perform a bitwise AND of the argument followed by a
2519 boolean negation of the result.
2521 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2523 PR rtl-optimization/84878
2524 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2525 the basic block. Assert the use reference is not artificial and that
2526 it has an associated insn.
2528 2018-04-04 Michael Matz <matz@suse.de>
2530 * builtins.c (compute_objsize): Pass correct operand
2531 to array_at_struct_end_p.
2533 2018-04-04 Richard Biener <rguenther@suse.de>
2536 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2537 from contexts for DINFO_LEVEL_TERSE and below.
2539 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2541 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2542 resiter_operand condition.
2543 * config/nds32/nds32.md (*move<mode>): Ditto.
2545 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2546 Monk Chiang <sh.chiang04@gmail.com>
2548 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2550 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2552 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2554 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2555 Kito Cheng <kito.cheng@gmail.com>
2557 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2558 nds32_cond_code_str, output_cond_branch,
2559 output_cond_branch_compare_zero, nds32_expand_cbranch,
2560 nds32_expand_cstore, nds32_expand_movcc,
2561 nds32_output_cbranchsi4_equality_zero,
2562 nds32_output_cbranchsi4_equality_reg,
2563 nds32_output_cbranchsi4_equality_reg_or_const_int,
2564 nds32_output_cbranchsi4_greater_less_zero: New functions.
2565 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2566 nds32_expand_cstore, nds32_expand_movcc,
2567 nds32_output_cbranchsi4_equality_zero,
2568 nds32_output_cbranchsi4_equality_reg,
2569 nds32_output_cbranchsi4_equality_reg_or_const_int,
2570 nds32_output_cbranchsi4_greater_less_zero): Declare.
2571 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2572 nds32_rimm11s_operand): New predicates.
2573 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2574 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2577 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2579 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2580 * config/nds32/nds32.md: Ditto.
2581 * config/nds32/pipelines.md: Ditto.
2583 2018-04-04 Richard Biener <rguenther@suse.de>
2585 PR tree-optimization/85168
2586 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2587 propagating abnormals.
2589 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2591 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2593 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2594 Kito Cheng <kito.cheng@gmail.com>
2596 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2597 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2598 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2599 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2600 * config/nds32/nds32.md (sibcall_internal): New.
2601 (sibcall_register): Remove.
2602 (sibcall_immediate): Remove.
2603 (sibcall_value_internal): New.
2604 (sibcall_value_register): Remove.
2605 (sibcall_value_immediate): Remove.
2606 * config/nds32/predicates.md (nds32_general_register_operand): New.
2607 (nds32_call_address_operand): New.
2609 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2611 PR rtl-optimization/85167
2612 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2613 bb_defs if *split_p, instead preinitialize it to NULL.
2615 PR tree-optimization/85156
2616 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2617 evaluating the argument multiple times.
2619 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2621 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2623 (_mm_cvtpd_ps): Likewise.
2624 (_mm_cvttpd_epi32): Likewise.
2625 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2626 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2627 vector, pixel, and bool following altivec.h include.
2629 2018-04-03 Martin Sebor <msebor@redhat.com>
2631 * doc/extend.texi (Common Function Attributes): Clarify.
2632 (const attribute): Likewise.
2633 (pure attribute): Likewise.
2635 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2638 * config/i386/i386.c (ix86_expand_vector_set): Use
2639 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2641 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2643 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2644 instructions when changing rounding bits to preserve precision bits
2645 in the x87 control word.
2647 2018-04-03 Martin Liska <mliska@suse.cz>
2649 PR tree-optimization/82491
2650 * rtl.h (strip_offset_and_add): Replace += suboffset with
2651 poly_uint64 () + suboffset.
2653 2018-03-29 Martin Liska <mliska@suse.cz>
2654 Martin Jambor <mjambor@suse.cz>
2657 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2658 param_type is not an integral or pointer type.
2660 2018-04-03 Richard Biener <rguenther@suse.de>
2662 * sese.h (recompute_all_dominators): Remove.
2664 2018-04-02 Martin Sebor <msebor@redhat.com>
2666 * doc/invoke.texi (-Wrestrict): Fix typos.
2668 2018-04-02 Jim Wilson <jimw@sifive.com>
2670 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2671 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2672 (<optab>di3, <optab>si3_extend): Likewise.
2673 (<optab>si3_mask, <optab>si3_mask_1): New.
2674 (<optab>di3_mask, <optab>di3_mask_1): New.
2675 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2676 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2677 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2679 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2681 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2684 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2686 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2687 (nds32_canonicalize_comparison): New function.
2689 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2690 Kito Cheng <kito.cheng@gmail.com>
2691 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2693 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2694 * config/nds32/constants.md (unspec_volatile_element): Add
2695 UNSPEC_VOLATILE_RELAX_GROUP.
2696 * config/nds32/nds32-relax-opt.c: New file.
2697 * config/nds32/nds32-predicates.c
2698 (nds32_symbol_load_store_p): New function.
2699 * config/nds32/nds32-protos.h
2700 (nds32_symbol_load_store_p): Declare function.
2701 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2702 * config/nds32/nds32.c
2703 (nds32_register_pass): New function to register pass.
2704 (nds32_register_passes): New function to register passes.
2705 * config/nds32/nds32.md (relax_group): New pattern.
2706 * config/nds32/nds32.opt (mrelax-hint): New option.
2707 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2709 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2711 * config/nds32/t-nds32: Modify files dependency.
2713 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2715 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2716 (PROFILE_HOOK): Define its implementation.
2718 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2720 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2721 type and 32-bit size.
2723 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2726 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2727 (V_128_256): New mode iterator.
2728 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2729 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2730 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2732 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2733 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2735 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2738 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2739 NaN inputs correctly.
2741 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2744 * config/rs6000/vsx.md (??r): New mode attribute.
2745 (*vsx_mov<mode>_64bit): Use it.
2746 (*vsx_mov<mode>_32bit): Likewise.
2748 2018-03-30 Martin Sebor <msebor@redhat.com>
2750 PR tree-optimization/84818
2751 * builtins.c (check_access): Use warning_n.
2753 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2756 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2758 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2761 2018-03-30 Julia Koval <julia.koval@intel.com>
2764 * x86-tune.def (movx, partial_reg_dependency): Enable for
2767 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2770 * lra-constraints.c (process_alt_operands): Move setting
2771 this_alternative_matches below.
2773 2018-03-29 Martin Liska <mliska@suse.cz>
2776 * doc/invoke.texi: Document how LTO works with debug info.
2777 Describe auto-load support of binutils. Mention 'x86-64'
2778 as valid option value of -march option.
2780 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2782 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2785 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2786 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2787 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2790 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2793 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2794 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2795 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2796 for RS6000_BTM_POWERPC64.
2797 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2798 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2799 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2804 2018-03-28 Carl Love <cel@us.ibm.com>
2807 2017-09-27 Carl Love <cel@us.ibm.com>
2809 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2810 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2811 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2814 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2816 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2817 instead of __vector bool.
2818 (_mm_max_pu8): Likewise.
2819 (_mm_min_pi16): Likewise.
2821 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2824 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2825 (DIVWEUO): Likewise.
2827 (DIVDEUO): Likewise.
2828 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2829 DIVWEUO and DIVDEUO.
2830 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2831 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2832 (div_extend): Likewise.
2833 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2835 (__builtin_divweuo): Likewise.
2836 (__builtin_divdeo): Likewise.
2837 (__builtin_divdeuo): Likewise.
2839 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2842 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2843 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2845 PR tree-optimization/82004
2846 * gimple-match-head.c (optimize_pow_to_exp): New function.
2847 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2848 Don't fold to exp if optimize_pow_to_exp is false.
2850 2018-03-28 Martin Liska <mliska@suse.cz>
2853 * calls.c (initialize_argument_information): Fix trailing space.
2854 * common.opt: Fix typo and provide better explanation for
2855 -fsanitize-coverage option.
2856 * config/i386/i386.opt: Fix typo.
2858 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2859 Martin Liska <mliska@suse.cz>
2862 * gimplify.c (asan_poison_variable): Don't do the check for
2863 gimplify_omp_ctxp here.
2864 (gimplify_decl_expr): Do it here.
2865 (gimplify_target_expr): Likewise.
2867 2018-03-28 Martin Liska <mliska@suse.cz>
2870 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2871 chkp_type_bounds_count if MPX is not enabled.
2873 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2875 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2877 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2880 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2881 function to create the function decl for complex long double
2882 multiply and divide for -mabi=ieeelongdouble.
2883 (init_float128_ieee): Call it.
2885 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2888 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2889 -fcf-protection=branch -mibt.
2890 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2892 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2895 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2897 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2900 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2901 extern array declarations.
2903 2018-03-27 Richard Biener <rguenther@suse.de>
2906 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2907 explicit single_use checks.
2909 2018-03-27 Richard Biener <rguenther@suse.de>
2911 PR tree-optimization/85082
2912 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2915 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2917 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2918 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2919 Turn on fasynchronous-unwind-tables and funwind-tables.
2921 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2924 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2925 (*bmi_blsr_<mode>_ccz): Ditto.
2927 2018-03-26 Tom de Vries <tom@codesourcery.com>
2929 PR tree-optimization/85063
2930 * omp-general.c (offloading_function_p): New function. Factor out
2932 * omp-offload.c (pass_omp_target_link::gate): ... here.
2933 * omp-general.h (offloading_function_p): Declare.
2934 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2935 with attribute omp declare target for offloading functions.
2937 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2939 PR tree-optimization/84005
2940 * tree-data-ref.h (get_base_for_alignment): Declare.
2941 * tree-data-ref.c (get_base_for_alignment_1): New function.
2942 (get_base_for_alignment): Likewise.
2943 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2944 get_base_for_alignment to find a suitable base object, instead
2945 of always using drb->base_address.
2947 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2950 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2951 known size by default.
2953 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2956 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2957 and non BLKmode operands.
2959 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2962 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2963 Clean up attributes.
2965 2018-03-23 Richard Biener <rguenther@suse.de>
2968 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2969 we are going to emit early debug for LTO.
2971 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2974 * function.c (match_asm_constraints_1): Don't optimize if input
2975 doesn't satisfy general_operand predicate for output's mode.
2978 * alias.c (write_dependence_p): Don't require for x_canonicalized
2979 non-VOIDmode if x has VOIDmode.
2982 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2983 just don't try to optimize it rather than assert it never happens.
2985 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2987 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2988 macro expansions for definition of ST_INTERNAL_<mode> and
2989 LD_INTERNAL_<mode> builtins.
2990 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2992 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2994 (altivec_expand_st_builtin): Likewise.
2995 (altivec_expand_builtin): Remove calls to deleted functions.
2996 (rs6000_address_for_altivec): Delete this function.
2997 * config/rs6000/vector.md: Remove expands for
2998 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3000 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3003 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3004 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3005 re-computing once computed.
3006 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3007 (arm_init_machine_status): Initialize
3008 machine->static_chain_stack_bytes.
3010 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3013 * doc/extend.texi: Add four new prototypes for vec_ld.
3014 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3015 definitions for more logical presentation.
3016 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3017 entries for V1TI variants of __builtin_altivec_ld builtin.
3018 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3019 handling of V1TI variant of LVX icode pattern.
3020 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3021 (rs6000_gimple_fold_builtin): Likewise.
3022 (altivec_init_builtins): Add code to define
3023 __builtin_altivec_lvx_v1ti function.
3025 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3028 * function.c (match_asm_constraints_1): Don't do the optimization
3029 if input isn't a REG, SUBREG, MEM or constant.
3031 2018-03-22 Tom de Vries <tom@codesourcery.com>
3033 PR tree-optimization/84956
3034 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3035 bb_has_abnormal_pred.
3037 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3040 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3041 DECL_INITIAL (decl) to decl at the end.
3042 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3045 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3047 * doc/extend.texi (__builtin_tgmath): Document when complex
3048 integer types are treated as _Complex _Float64.
3050 2018-03-21 Tom de Vries <tom@codesourcery.com>
3052 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3054 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3056 PR tree-optimization/84960
3057 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3058 if it is ENTRY block, move them into single succ of ENTRY in that case.
3060 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3062 PR tree-optimization/84811
3063 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3064 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3065 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3066 (known_subrange_p): Update accordingly. Cast each value involved
3067 in the size comparison, rather than casting the result of the
3070 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3072 PR tree-optimization/84982
3073 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3074 by flipping the least significant bit rather than all bits from
3075 bitpos to bitpos + bitsize - 1.
3077 2018-03-21 Nathan Sidwell <nathan@acm.org>
3079 * doc/extend.texi (Deprecated Features): Remove mention of
3080 long-deleted deprecations.
3082 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3085 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3086 * configure: Regenerate.
3088 2018-03-21 Tom de Vries <tom@codesourcery.com>
3090 PR tree-optimization/83126
3091 * tree-parloops.c (num_phis): New function.
3092 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3094 2018-03-21 Nathan Sidwell <nathan@acm.org>
3096 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3097 mention anon-struct/union members and trailing attributes.
3099 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3101 PR tree-optimization/84969
3102 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3103 builtin memset partitions if they set different rhs values.
3105 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3107 PR rtl-optimization/84989
3108 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3109 VEC_DUPLICATE with scalar result mode.
3111 2018-03-21 Martin Liska <mliska@suse.cz>
3114 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3115 not intended return statement.
3117 2018-03-21 Martin Liska <mliska@suse.cz>
3120 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3121 (chkp_find_bound_slots_1): Limit number of iterations.
3123 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3126 * Minor grammar fixes for x86 options.
3128 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3131 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3132 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3134 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3137 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3138 (altivec_lvx_<mode>_1op): Likewise.
3139 (altivec_stvx_<mode>_2op): Likewise.
3140 (altivec_stvx_<mode>_1op): Likewise.
3141 (altivec_lvx_<VM2:mode>): New define_expand.
3142 (altivec_stvx_<VM2:mode>): Likewise.
3143 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3144 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3145 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3146 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3147 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3148 (rs6000_gen_lvx): Likewise.
3149 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3150 (altivec_expand_stv_builtin): Likewise.
3151 (altivec_expand_builtin): Likewise.
3152 * config/rs6000/vector.md: Likewise.
3154 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3157 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3160 2018-03-20 Richard Biener <rguenther@suse.de>
3163 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3164 sign-conversions as zero, fall back to standard scalar_stmt
3167 2018-03-20 Martin Liska <mliska@suse.cz>
3170 * predict.c (rebuild_frequencies): Handle case when we have
3171 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3173 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3176 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3177 flag_section_anchors.
3178 * varasm.c (use_blocks_for_decl_p): Remove hack for
3179 dw2_force_const_mem.
3182 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3184 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3185 be created, use lowpart_subreg of operands[0] rather than operands[0]
3187 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3188 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3189 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3190 and n constraint instead of aarch64_shift_imm_di and Usd.
3191 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3192 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3194 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3197 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3198 to favor GPR over NEON registers.
3199 (<shift>di3_neon): Likewise.
3201 2018-03-20 Tom de Vries <tom@codesourcery.com>
3204 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3205 (nvptx_process_pars): Emit bar.sync asap and alap.
3207 2018-03-20 Tom de Vries <tom@codesourcery.com>
3210 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3211 seen_label if seen_label is already set.
3213 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3216 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3217 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3218 Use 1U instead of 1. Formatting fixes.
3221 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3222 instead of TREE_TYPE (s1) for the return value.
3224 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3226 PR tree-optimization/84946
3227 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3228 bitsize + bitsize in poly_uint64 rather than poly_int64.
3231 * dwarf2asm.c: Include fold-const.c.
3232 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3233 of decl rather than decl itself.
3235 PR rtl-optimization/84643
3236 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3238 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3241 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3242 calling assemble_variable.
3244 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3247 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3248 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3250 2018-03-19 Jim Wilson <jimw@sifive.com>
3253 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3254 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3255 (riscv_first_stack_step): Likewise.
3256 (riscv_option_override): Use STACK_BOUNDARY instead of
3258 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3260 (BIGGEST_ALIGNMENT): Set to 128.
3261 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3262 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3265 2018-03-19 Richard Biener <rguenther@suse.de>
3267 PR tree-optimization/84933
3268 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3269 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3271 2018-03-19 Richard Biener <rguenther@suse.de>
3273 PR tree-optimization/84859
3274 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3275 (cond_if_else_store_replacement): Perform sinking operation on
3276 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3277 Generalize what a BB with a single eligible store is.
3279 2018-03-19 Richard Biener <rguenther@suse.de>
3281 PR tree-optimization/84929
3282 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3283 chrec_is_positive against non-chrec arg.
3285 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3288 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3290 2018-03-18 Martin Liska <mliska@suse.cz>
3292 PR rtl-optimization/84635
3293 * regrename.c (build_def_use): Use matches_mode only when
3296 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3298 PR tree-optimization/84913
3299 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3300 vectorize chains of COND_EXPRs.
3302 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3304 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3306 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3308 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3310 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3312 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3314 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3315 Kito Cheng <kito.cheng@gmail.com>
3317 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3318 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3319 (nds32_adjust_reg_alloc_order): New function.
3320 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3322 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3324 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3325 nds32_print_operand, nds32_print_operand_address): Use
3326 HOST_WIDE_INT_PRINT_DEC instead.
3328 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3330 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3332 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3335 * config/i386/i386.c (initial_ix86_tune_features,
3336 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3338 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3339 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3340 rather than 1u << ix86_tune. Formatting fix.
3341 (ix86_option_override_internal): Change ix86_arch_mask from
3342 unsigned int to unsigned HOST_WIDE_INT, initialize to
3343 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3344 (ix86_function_specific_restore): Likewise.
3346 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3349 * postreload.c (reload_combine_recognize_pattern): Perform
3350 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3351 truncate_int_for_mode the result for the destination's mode.
3354 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3355 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3357 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3360 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3363 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3364 regno_allocno_class_array and sorted_pseudos.
3365 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3366 insns where regno is used.
3368 2018-03-16 Martin Liska <mliska@suse.cz>
3371 * multiple_target.c (create_dispatcher_calls): Redirect
3372 reference in the symbol table.
3374 2018-03-16 Martin Liska <mliska@suse.cz>
3377 * multiple_target.c (create_dispatcher_calls): Redirect also
3380 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3384 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3385 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3386 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3388 2018-03-16 Julia Koval <julia.koval@intel.com>
3390 * doc/invoke.texi (Skylake Server): Add CLWB.
3391 Cannonlake): Remove CLWB.
3393 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3395 PR tree-optimization/84841
3396 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3398 (FLOAT_ONE_CONST_TYPE): Define.
3399 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3400 (sort_by_operand_rank): Put entries with higher constant_type last
3401 rather than first to match comments.
3403 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3405 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3408 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3411 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3412 check and use address of target always.
3414 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3417 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3418 (indirect_thunk_bnd_needed): Likewise.
3419 (indirect_thunks_used): Likewise.
3420 (indirect_thunks_bnd_used): Likewise.
3421 (indirect_return_needed): New.
3422 (indirect_return_bnd_needed): Likewise.
3423 (output_indirect_thunk_function): Add a bool argument for
3425 (output_indirect_thunk_function): Don't generate alias for
3426 function return thunk.
3427 (ix86_code_end): Call output_indirect_thunk_function to generate
3428 function return thunks.
3429 (ix86_output_function_return): Set indirect_return_bnd_needed
3430 and indirect_return_needed instead of indirect_thunk_bnd_needed
3431 and indirect_thunk_needed.
3433 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3435 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3436 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3437 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3439 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3440 Paul Hua <paul.hua.gm@gmail.com>
3443 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3445 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3447 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3448 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3451 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3454 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3455 instead of GET_MODE_SIZE when comparing Units.
3457 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3460 * varasm.c (hash_section): Return an unchangeble hash value
3461 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3462 Return !aarch64_can_use_per_function_literal_pools_p ().
3464 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3467 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3468 rather than address of cmode as last argument to prepare_cmp_insn.
3470 2018-03-15 Julia Koval <julia.koval@intel.com>
3472 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3473 F_AVX512VNNI, F_AVX512BITALG): New.
3475 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3478 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3479 insn for floating-point loads and stores.
3481 2018-03-14 Carl Love <cel@us.ibm.com>
3483 * config/rs6000/rs6000-c.c: Add macro definitions for
3484 ALTIVEC_BUILTIN_VEC_PERMXOR.
3485 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3486 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3487 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3488 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3490 * config/doc/extend.texi: Add prototypes for vec_permxor.
3492 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3495 * diagnostic-show-locus.c (class layout_point): Convert m_line
3496 from int to linenum_type.
3497 (line_span::comparator): Use linenum "compare" function when
3498 comparing line numbers.
3499 (test_line_span): New function.
3500 (layout_range::contains_point): Convert param "row" from int to
3502 (layout_range::intersects_line_p): Likewise.
3503 (layout::will_show_line_p): Likewise.
3504 (layout::print_source_line): Likewise.
3505 (layout::should_print_annotation_line_p): Likewise.
3506 (layout::print_annotation_line): Likewise.
3507 (layout::print_leading_fixits): Likewise.
3508 (layout::annotation_line_showed_range_p): Likewise.
3509 (struct line_corrections): Likewise for field m_row.
3510 (line_corrections::line_corrections): Likewise for param "row".
3511 (layout::print_trailing_fixits): Likewise.
3512 (layout::get_state_at_point): Likewise.
3513 (layout::get_x_bound_for_row): Likewise.
3514 (layout::print_line): Likewise.
3515 (diagnostic_show_locus): Likewise for locals "last_line" and
3517 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3518 * input.c (selftest::test_linenum_comparisons): New function.
3519 (selftest::input_c_tests): Call it.
3520 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3521 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3522 * selftest.h (ASSERT_GT): New macro.
3523 (ASSERT_GT_AT): New macro.
3524 (ASSERT_LT): New macro.
3525 (ASSERT_LT_AT): New macro.
3527 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3529 PR rtl-optimization/84780
3530 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3532 2018-03-14 Martin Liska <mliska@suse.cz>
3534 * tree.c (record_node_allocation_statistics): Use
3535 get_stats_node_kind.
3536 (get_stats_node_kind): New function extracted from
3537 record_node_allocation_statistics.
3538 (free_node): Use get_stats_node_kind.
3540 2018-03-14 Richard Biener <rguenther@suse.de>
3542 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3543 that the value-set of ANTIC_IN doesn't grow.
3546 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3548 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3549 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3551 2018-03-14 Julia Koval <julia.koval@intel.com>
3553 * config.gcc (icelake-client, icelake-server): New.
3555 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3556 (initial_ix86_arch_features): Ditto.
3557 (PTA_SKYLAKE): Add SGX.
3558 (PTA_ICELAKE): Remove.
3559 (PTA_ICELAKE_CLIENT): New.
3560 (PTA_ICELAKE_SERVER): New.
3561 (ix86_option_override_internal): Split up icelake on icelake client and
3563 (get_builtin_code_for_version): Ditto.
3564 (fold_builtin_cpu): Ditto.
3565 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3566 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3567 * config/i386/i386.h (processor_type): Ditto.
3568 * doc/invoke.texi: Ditto.
3570 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3573 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3574 INTEGER_CST offset, add it together with bitpos / 8 and
3575 sign extend based on POINTER_SIZE.
3579 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3582 * config/i386/constraints.md (Yc): New register constraint.
3583 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3584 Use Yc constraint for alternative 2 of operand 0. Remove
3585 preferred_for_speed attribute.
3587 2018-03-14 Richard Biener <rguenther@suse.de>
3589 PR tree-optimization/84830
3590 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3591 with the old one to avoid oscillations.
3593 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3596 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3598 (assign_by_spills): Return a flag of reload assignment failure.
3599 Do not process the reload assignment failures. Do not spill other
3600 reload pseudos if they has the same reg class. Update n if
3602 (lra_assign): Add a return arg. Set up from the result of
3603 assign_by_spills call.
3604 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3605 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3606 usage_insns if it is not NULL.
3607 (spill_hard_reg_in_range): New function.
3608 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3609 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3610 function prototypes.
3611 (lra_assign): Change prototype.
3612 * lra.c (lra): Add code to deal with fails by splitting hard reg
3615 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3617 * config/riscv/riscv.opt (mrelax): New option.
3618 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3619 "norelax" when riscv_mrelax is disabled.
3620 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3622 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3625 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3626 reassociation for int modes.
3628 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3630 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3631 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3633 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3634 * config/aarch64/aarch64-sve.md
3635 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3636 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3637 (*extend<mode><Vwide>2): Rename to...
3638 (aarch64_sve_extend<mode><Vwide>2): ...this.
3639 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3640 renaming the old pattern to...
3641 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3643 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3644 define_expand, renaming the old pattern to...
3645 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3646 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3647 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3648 account when deciding which SVE instruction the optab should use.
3649 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3651 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3653 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3654 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3655 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3656 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3657 (tlsdesc_small_<mode>): Turn a define_expand and use
3658 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3659 (tlsdesc_small_advsimd_<mode>): ...this.
3660 (tlsdesc_small_sve_<mode>): New pattern.
3662 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3664 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3665 (UNSPEC_UMUL_HIGHPART): New constants.
3666 (MUL_HIGHPART): New int iteraor.
3667 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3668 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3670 (*<su>mul<mode>3_highpart): New define_insn.
3672 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3675 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3678 2018-03-13 Martin Liska <mliska@suse.cz>
3681 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3683 (sem_item_optimizer::~sem_item_optimizer): Release it.
3684 (sem_item_optimizer::merge_classes): Register variable aliases.
3685 (sem_item_optimizer::fixup_pt_set): New function.
3686 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3687 * ipa-icf.h: Declare new variables and functions.
3689 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3692 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3693 integer_pow2p@2 and test integer_pow2p in condition.
3694 (A < 0 ? C : 0): Similarly for @1.
3697 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3698 characters starting at p contain '\0' character, don't look beyond
3702 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3703 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3706 * reg-stack.c (change_stack): Change update_end var from int to
3707 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3708 also call set_block_for_insn on the newly added insns and rescan.
3711 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3712 on the last operand.
3715 * tree.c (stabilize_reference_1): Return save_expr (e) for
3716 STATEMENT_LIST even if it doesn't have side-effects.
3718 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3720 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3722 2018-03-12 Renlin Li <renlin.li@arm.com>
3724 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3725 aarch64_output_scalar_simd_mov_immediate.
3727 2018-03-12 Martin Sebor <msebor@redhat.com>
3729 PR tree-optimization/83456
3730 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3731 for perfectly overlapping calls to memcpy.
3732 (gimple_fold_builtin_memory_chk): Same.
3733 (gimple_fold_builtin_strcpy): Handle no-warning.
3734 (gimple_fold_builtin_stxcpy_chk): Same.
3735 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3737 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3739 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3740 parameter. Use it for SFmode.
3741 (rs6000_function_arg_advance_1): Adjust.
3742 (rs6000_function_arg): Adjust.
3743 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3745 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3747 PR rtl-optimization/84169
3748 PR rtl-optimization/84780
3749 * combine.c (can_combine_p): Check for a 2-insn combination whether
3750 the destination register is used between the two insns, too.
3752 2018-03-12 Richard Biener <rguenther@suse.de>
3754 PR tree-optimization/84803
3755 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3756 for refs DR analysis didn't process.
3758 2018-03-12 Richard Biener <rguenther@suse.de>
3760 PR tree-optimization/84777
3761 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3762 force-vectorize loops ignore whether we are optimizing for size.
3764 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3766 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3767 (TARGET_MD_ASM_ADJUST): Define.
3769 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3770 Kito Cheng <kito.cheng@gmail.com>
3771 Chung-Ju Wu <jasonwucj@gmail.com>
3773 * config/nds32/nds32.c (nds32_compute_stack_frame,
3774 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3775 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3776 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3777 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3778 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3779 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3780 * config/nds32/nds32.md (prologue, epilogue): Use macro
3781 NDS32_V3PUSH_AVAILABLE_P to do checking.
3783 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3786 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3787 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3788 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3789 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3790 addition of most attributes on !orig_type_die or the attribute not
3791 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3793 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3794 Chung-Ju Wu <jasonwucj@gmail.com>
3796 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3798 * config/nds32/nds32.opt (mvh): New option.
3800 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3801 Chung-Ju Wu <jasonwucj@gmail.com>
3803 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3805 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3806 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3809 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3810 Chung-Ju Wu <jasonwucj@gmail.com>
3812 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3814 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3815 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3817 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3818 Kito Cheng <kito.cheng@gmail.com>
3819 Chung-Ju Wu <jasonwucj@gmail.com>
3821 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3822 UNSPEC_FFMISM and UNSPEC_FLMISM.
3823 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3824 for ffb, ffmism and flmism.
3825 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3826 (unspec_ffmism): Ditto.
3827 (unspec_flmism): Ditto.
3828 (nds32_expand_builtin_impl): Check if string extension is available.
3829 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3830 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3832 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3835 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3838 * lra-assigns.c (assign_by_spills): Return a flag of reload
3839 assignment failure. Do not process the reload assignment
3840 failures. Do not spill other reload pseudos if they has the same
3842 (lra_assign): Add a return arg. Set up from the result of
3843 assign_by_spills call.
3844 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3845 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3846 usage_insns if it is not NULL.
3847 (spill_hard_reg_in_range): New function.
3848 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3849 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3850 function prototypes.
3851 (lra_assign): Change prototype.
3852 * lra.c (lra): Add code to deal with fails by splitting hard reg
3855 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3858 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3860 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3863 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3864 (dw_val_node): Add val_symbolic_view.
3865 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3866 (symview_upper_bound): New.
3867 (new_line_info_table): Initialize symviews_since_reset.
3868 (dwarf2out_source_line): Count symviews_since_reset and set
3869 symview_upper_bound.
3870 (dw_val_equal_p): Handle symview.
3871 (add_AT_symview): New.
3872 (print_dw_val): Handle symview.
3873 (attr_checksum, attr_checksum_ordered): Likewise.
3874 (same_dw_val_p, size_of_die): Likewise.
3875 (value_format, output_die): Likewise.
3876 (add_high_low_attributes): Use add_AT_symview for entry_view.
3877 (dwarf2out_finish): Reset symview_upper_bound, clear
3880 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3883 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3884 Add strict argument and use it.
3885 (rs6000_split_multireg_move): Update for new strict argument.
3886 (mem_operand_gpr): Disallow all non-offsettable addresses.
3887 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3889 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3892 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3893 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3894 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3897 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3898 decl, use remap_type if we want to use the type.
3900 2018-03-09 Martin Sebor <msebor@redhat.com>
3902 PR tree-optimization/84526
3903 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3905 (builtin_access::generic_overlap): Be prepared to handle non-array
3908 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3910 PR rtl-optimization/84682
3911 * lra-constraints.c (process_address_1): Check is_address flag
3912 for address constraints.
3913 (process_alt_operands): Likewise.
3914 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3915 preprocess_constraints.
3916 * recog.h (preprocess_constraints): Add oploc parameter.
3918 * recog.c (preprocess_constraints): Test address_operand for
3919 CT_ADDRESS constraints.
3921 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3924 * lra-assigns.c (assign_by_spills): Return a flag of reload
3925 assignment failure. Do not process the reload assignment
3926 failures. Do not spill other reload pseudos if they has the same
3928 (lra_assign): Add a return arg. Set up from the result of
3929 assign_by_spills call.
3930 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3931 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3932 usage_insns if it is not NULL.
3933 (spill_hard_reg_in_range): New function.
3934 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3935 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3936 function prototypes.
3937 (lra_assign): Change prototype.
3938 * lra.c (lra): Add code to deal with fails by splitting hard reg
3941 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3944 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3945 Accept complain bool parameter. Only emit errors if it is true.
3946 (arm_parse_cpu_option_name): Likewise.
3947 (arm_target_thumb_only): Adjust callers of the above.
3948 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3949 prototype to take a default true bool parameter.
3950 (arm_parse_arch_option_name): Likewise.
3952 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3953 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3957 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3958 * configure: Regenerate.
3959 * configure.ac ("linker --version-script option"): New.
3960 ("linker soname option"): New.
3962 2018-03-09 Richard Biener <rguenther@suse.de>
3964 PR tree-optimization/84775
3965 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3966 immediate uses of predicate stmts and mark them modified.
3969 PR tree-optimization/84178
3970 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3972 (version_loop_for_if_conversion): Delay update_ssa call.
3973 (tree_if_conversion): Delay update_ssa until after predicate
3976 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3979 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3980 when the function accesses prior frames.
3982 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3985 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3986 gen_llsym, otherwise call maybe_gen_llsym.
3989 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3990 has ',' character inside of it.
3992 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3995 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3996 as clobbering CC_REGNUM.
3998 2018-03-08 Richard Biener <rguenther@suse.de>
4001 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4002 (follow_copies_to_constant): Do not follow SSA names registered
4005 2018-03-08 Richard Biener <rguenther@suse.de>
4007 PR tree-optimization/84178
4008 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4010 (version_loop_for_if_conversion): Delay update_ssa call.
4011 (tree_if_conversion): Delay update_ssa until after predicate
4014 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4016 PR tree-optimization/84178
4017 * tree-if-conv.c (release_bb_predicate): Remove the
4018 the assertion that the stmts have NULL use_ops.
4019 Discard the statements, asserting that they haven't
4020 yet been added to a BB.
4022 2018-03-08 Richard Biener <rguenther@suse.de>
4024 PR tree-optimization/84746
4025 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4026 (phi_translate): Pass in destination ANTIC_OUT set.
4027 (phi_translate_1): Likewise. For a simplified result lookup
4028 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4029 (phi_translate_set): Adjust.
4030 (do_pre_regular_insertion): Likewise.
4031 (do_pre_partial_partial_insertion): Likewise.
4033 2018-03-08 Martin Liska <mliska@suse.cz>
4035 PR gcov-profile/84735
4036 * doc/gcov.texi: Document usage of profile files.
4037 * gcov-io.h: Document changes in the format.
4039 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4043 * dwarf2out.c (struct dw_line_info_table): Update comments for
4045 (FORCE_RESET_NEXT_VIEW): New.
4046 (FORCE_RESETTING_VIEW_P): New.
4047 (RESETTING_VIEW_P): Check for -1 too.
4048 (ZERO_VIEW_P): Likewise.
4049 (new_line_info_table): Force-reset next view.
4050 (dwarf2out_begin_function): Likewise.
4051 (dwarf2out_source_line): Simplify zero_view_p initialization.
4052 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4053 view directly. Omit view when omitting .loc at line 0.
4055 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4057 PR tree-optimization/84740
4058 * tree-switch-conversion.c (process_switch): Call build_constructors
4059 only if info.phi_count is non-zero.
4061 PR tree-optimization/84739
4062 * tree-tailcall.c (find_tail_calls): Check call arguments against
4063 DECL_ARGUMENTS (current_function_decl) rather than
4064 DECL_ARGUMENTS (func) when checking for tail recursion.
4066 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4068 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4069 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4070 Volker Reichelt's entry and add entries for people that perform
4071 GCC fuzzy testing and report numerous bugs.
4073 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4076 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4077 readonly data in sdata, if that is disabled.
4078 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4079 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4080 -mreadonly-in-sdata option.
4082 2018-03-07 Martin Sebor <msebor@redhat.com>
4084 PR tree-optimization/84468
4085 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4086 basic block when looking for nul assignment.
4088 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4091 * except.h (output_function_exception_table): Adjust prototype.
4092 * except.c (output_function_exception_table): Remove FNNAME parameter
4093 and add SECTION parameter. Ouput one part of the table at a time.
4094 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4095 the first part of the exception table and emit unwind directives.
4096 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4097 (i386_pe_seh_cold_init): Likewise.
4098 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4099 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4100 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4101 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4102 * config/i386/winnt.c: Include except.h.
4103 (struct seh_frame_state): Add reg_offset, after_prologue and
4104 in_cold_section fields.
4105 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4106 (i386_pe_seh_cold_init): New function.
4107 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4108 to seh->in_cold_section.
4109 (seh_emit_push): Record the offset of the push.
4110 (seh_emit_save): Record the offet of the save.
4111 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4112 Test seh->after_prologue to disregard the epilogue.
4113 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4114 (i386_pe_end_cold_function): New function.
4116 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4119 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4120 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4123 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4125 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4126 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4129 * multiple_target.c: Include tree-inline.h and intl.h.
4130 (expand_target_clones): Diagnose and fail if node->definition and
4131 !tree_versionable_function_p (node->decl).
4133 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4135 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4137 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4138 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4139 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4141 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4144 * combine.c (try_combine): Use reg_or_subregno instead of handling
4145 just paradoxical SUBREGs and REGs.
4147 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4149 * config/arc/arc.c (arc_finalize_pic): Remove function.
4150 (arc_must_save_register): We use single base PIC register, remove
4151 checks to save/restore the PIC register.
4152 (arc_expand_prologue): Likewise.
4153 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4155 (arc_verify_short): Likewise.
4156 (arc_attr_type): Likewise.
4157 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4158 (walk_stores): Likewise.
4159 (arc_address_cost): Make it static.
4160 (arc_verify_short): Likewise.
4161 (branch_dest): Likewise.
4162 (arc_attr_type): Likewise.
4163 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4164 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4165 (arc_final_prescan_insn): Remove inserting the nops due to
4166 hardware hazards. It is done in reorg step.
4167 (insn_length_variant_t): Remove.
4168 (insn_length_parameters_t): Likewise.
4169 (arc_insn_length_parameters): Likewise.
4170 (arc_get_insn_variants): Likewise.
4171 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4173 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4176 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4179 PR tree-optimization/84687
4180 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4182 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4184 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4186 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4187 Rename to ppc_speculation_barrier.
4188 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4189 __builtin_ppc_speculation_barrier.
4191 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4194 * combine.c (combine_simplify_rtx): Don't try to simplify if
4195 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4198 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4200 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4201 to 32 bytes when compiling for POWER9.
4203 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4206 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4207 regparm >= 3 with no arg reg available also for calls with
4208 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4211 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4213 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4215 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4218 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4220 2018-03-05 Richard Biener <rguenther@suse.de>
4222 PR tree-optimization/84486
4223 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4224 When inserting a __builtin_assume_aligned call set the LHS
4225 SSA name alignment info accordingly.
4227 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4229 PR tree-optimization/84114
4230 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4231 Avoid reassociation of FLOAT_MODE addition.
4233 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4235 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4236 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4237 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4238 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4239 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4240 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4241 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4243 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4244 __builtin_ia32_wbinvd): New builtins.
4245 (SPECIAL_ARGS2): New.
4246 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4247 (SPECIAL_ARGS2): New.
4248 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4249 (ix86_valid_target_attribute_inner_p): Ditto.
4250 (ix86_init_mmx_sse_builtins): Add special_args2.
4251 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4252 TARGET_WBNOINVD_P): New.
4253 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4254 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4255 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4256 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4257 * config/i386/pconfigintrin.h: New file.
4258 * config/i386/wbnoinvdintrin.h: Ditto.
4259 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4260 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4262 2018-03-05 Richard Biener <rguenther@suse.de>
4264 PR tree-optimization/84670
4265 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4267 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4268 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4269 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4270 doesn't grow if all successors (recursively) were visited at least
4273 2018-03-05 Richard Biener <rguenther@suse.de>
4275 PR tree-optimization/84650
4276 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4277 if executed in the loop pipeline.
4279 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4281 * doc/configfiles.texi (Configuration Files): Move info about
4282 conditionalizing $target-protos.h to...
4283 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4284 differs from $target-protos.h.
4286 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4287 Chung-Ju Wu <jasonwucj@gmail.com>
4289 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4290 * config/nds32/nds32-multiple.md (setmemsi): Define.
4291 * config/nds32/nds32-memory-manipulation.c
4292 (nds32_gen_dup_4_byte_to_word_value): New.
4293 (emit_setmem_word_loop): New.
4294 (emit_setmem_byte_loop): New.
4295 (nds32_expand_setmem_loop): New.
4296 (nds32_expand_setmem_loop_v3m): New.
4297 (nds32_expand_setmem_unroll): New.
4298 (nds32_expand_setmem): New.
4300 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4301 Chung-Ju Wu <jasonwucj@gmail.com>
4303 * config/nds32/nds32-memory-manipulation.c
4304 (nds32_emit_load_store): New.
4305 (nds32_emit_post_inc_load_store): New.
4306 (nds32_emit_mem_move): New.
4307 (nds32_emit_mem_move_block): New.
4308 (nds32_expand_movmemsi_loop_unknown_size): New.
4309 (nds32_expand_movmemsi_loop_known_size): New.
4310 (nds32_expand_movmemsi_loop): New.
4311 (nds32_expand_movmemsi_unroll): New.
4312 (nds32_expand_movmemqi): Rename ...
4313 (nds32_expand_movmemsi): ... to this.
4314 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4315 (movmemsi): ... to this.
4316 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4317 (nds32_expand_movmemsi): ... to this.
4319 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4320 Monk Chiang <sh.chiang04@gmail.com>
4321 Chung-Ju Wu <jasonwucj@gmail.com>
4323 * config/nds32/nds32-protos.h
4324 (nds32_expand_load_multiple): New arguments.
4325 (nds32_expand_store_multiple): Ditto.
4326 (nds32_valid_multiple_load_store): Rename ...
4327 (nds32_valid_multiple_load_store_p): ... to this.
4328 * config/nds32/nds32-memory-manipulation.c
4329 (nds32_expand_load_multiple): Refine implementation.
4330 (nds32_expand_store_multiple): Ditto.
4331 * config/nds32/nds32-multiple.md
4332 (load_multiple): Update nds32_expand_load_multiple interface.
4333 (store_multiple): Update nds32_expand_store_multiple interface.
4334 * config/nds32/nds32-predicates.c
4335 (nds32_valid_multiple_load_store): Rename ...
4336 (nds32_valid_multiple_load_store_p): ... to this and refine
4338 * config/nds32/predicates.md
4339 (nds32_load_multiple_and_update_address_operation): New predicate.
4340 (nds32_store_multiple_and_update_address_operation): New predicate.
4342 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4343 Chung-Ju Wu <jasonwucj@gmail.com>
4345 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4346 (combo): New attribute.
4347 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4349 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4351 * config/nds32/nds32.opt: Change -mcmodel= default value.
4353 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4354 Monk Chiang <sh.chiang04@gmail.com>
4355 Chung-Ju Wu <jasonwucj@gmail.com>
4357 * config/nds32/constants.md (unspec_element): New enum.
4358 * config/nds32/constraints.md (Umw): New constraint.
4359 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4360 * config/nds32/nds32-intrinsic.md: Likewise.
4361 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4362 (nds32_valid_smw_lwm_base_p): New.
4363 (nds32_output_smw_single_word): New.
4364 (nds32_output_lmw_single_word): New.
4365 (nds32_expand_unaligned_load): New.
4366 (nds32_expand_unaligned_store): New.
4367 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4368 (nds32_output_smw_single_word): Declare.
4369 (nds32_output_lmw_single_word): Declare.
4370 (nds32_expand_unaligned_load): Declare.
4371 (nds32_expand_unaligned_store): Declare.
4372 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4373 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4374 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4375 NDS32_BUILTIN_UASTORE_DW.
4376 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4379 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4380 Kito Cheng <kito.cheng@gmail.com>
4381 Chung-Ju Wu <jasonwucj@gmail.com>
4383 * config/nds32/nds32-intrinsic.c
4384 (nds32_expand_builtin_null_ftype_reg): Delete.
4385 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4386 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4387 (nds32_read_argument): New.
4388 (nds32_legitimize_target): Ditto.
4389 (nds32_legitimize_argument): Ditto.
4390 (nds32_check_constant_argument): Ditto.
4391 (nds32_expand_unop_builtin): Ditto.
4392 (nds32_expand_unopimm_builtin): Ditto.
4393 (nds32_expand_binop_builtin): Ditto.
4394 (nds32_builtin_decl_impl): Ditto.
4395 (builtin_description): Ditto.
4396 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4397 (nds32_init_builtins_impl): Ditto.
4398 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4399 (nds32_builtin_decl): New.
4400 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4401 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4403 2018-03-02 Jeff Law <law@redhat.com>
4405 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4406 (redundant_insn, fill_simple_delay_slots): Likewise.
4407 (fill_slots_from_thread): Likewise.
4408 * resource.c (mark_referenced_resources): Likewise.
4409 (mark_set_resources, find_dead_or_set_registers): Likewise.
4411 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4413 * substring-locations.h (format_warning_va): Formatting fix for
4415 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4417 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4418 * substring-locations.c: Include intl.h.
4419 (format_warning_va): Turned into small wrapper around
4420 format_warning_n_va, renamed to ...
4421 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4422 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4424 (format_warning_at_substring_n): New function.
4425 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4426 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4427 format_warning_at_substring with just a shorter name instead of
4428 const function pointer.
4429 (fmtwarn_n): New function.
4430 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4431 appropriate, get rid of all the fmtstr temporaries, move conditionals
4432 with G_() wrapped string literals directly into fmtwarn arguments,
4433 cast dir.len to (int), formatting fixes.
4435 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4437 * doc/invoke.texi: Remove "Cilk Plus" references.
4439 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4440 Richard Biener <rguenther@suse.de>
4443 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4444 for error or warning attributes if CALL_FROM_THUNK_P is set.
4447 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4450 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4451 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4454 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4455 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4457 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4458 instead of -1U in last predictors element's probability member.
4460 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4463 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4464 arguments if they are comparable.
4466 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4468 PR tree-optimization/84634
4469 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4470 masks and masked_loop_p with a single loop_masks, making sure it's
4471 null for bb vectorization.
4473 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4475 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4476 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4477 loop->force_vectorize to check whether there is no alias.
4479 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4482 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4484 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4486 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4487 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4488 instead of a loop around prev_real_insn.
4489 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4493 * config/i386/i386.c (ix86_print_operand): Use conditional
4494 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4497 2018-03-02 Richard Biener <rguenther@suse.de>
4499 PR tree-optimization/84427
4500 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4501 (bitmap_set_subtract_values): Rewrite to handle multiple
4504 (prune_clobbered_mems): Likewise.
4505 (phi_translate): Take edge instead of pred/phiblock.
4506 (phi_translate_1): Likewise.
4507 (phi_translate_set): Likewise. Insert all translated
4508 exprs for a value into the set, keeping possibly multiple
4509 expressions per value.
4510 (compute_antic_aux): Adjust for phi_translate changes.
4511 When intersecting union the expressions and prune those
4512 not in the final value set, keeping possibly multiple
4513 expressions per value. Do not use value-insertion
4514 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4515 all expressions. Add verification that the value-sets
4516 only shrink during iteration.
4517 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4518 (do_pre_regular_insertion): Likewise.
4519 (do_pre_partial_partial_insertion): Likewise.
4521 2018-03-02 Richard Biener <rguenther@suse.de>
4524 * config/darwin.c (saved_debug_info_level): New static global.
4525 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4526 (darwin_asm_lto_end): Restore debug info generation settings.
4528 2018-03-01 Martin Liska <mliska@suse.cz>
4531 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4534 2018-03-01 Richard Biener <rguenther@suse.de>
4537 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4538 type annotation with LTO when debug was disabled at compile-time.
4540 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4542 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4544 (mips_final_postscan_insn): Likewise.
4546 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4548 PR rtl-optimization/84528
4549 * alias.c (init_alias_target): Add commentary.
4550 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4551 a unique base value if the frame pointer is not eliminated
4552 to the stack pointer.
4554 2018-03-01 Tom de Vries <tom@codesourcery.com>
4556 PR rtl-optimization/83327
4557 * lra-int.h (hard_regs_spilled_into): Declare.
4558 * lra.c (hard_regs_spilled_into): Define.
4559 (init_reg_info): Init hard_regs_spilled_into.
4560 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4561 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4562 (process_bb_lives): Handle hard_regs_spilled_into.
4563 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4564 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4566 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4568 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4569 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4570 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4571 * config/rs6000/aix72.h: New file.
4573 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4575 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4576 instead of warning_at with conditional singular and plural messages
4580 * stor-layout.c (update_alignment_for_field): For
4581 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4582 && !DECL_PACKED (field), do the alignment update, just use
4583 only desired_align instead of MAX (type_align, desired_align)
4585 (place_field): Don't do known_align < desired_align handling
4586 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4587 is non-NULL, instead do it after rli->prev_field handling and
4588 only if not within a bitfield word. For DECL_PACKED (field)
4589 use type_align of BITS_PER_UNIT.
4591 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4593 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4594 superfluous parentheses and trailing spaces.
4596 2018-02-28 Richard Biener <rguenther@suse.de>
4598 PR tree-optimization/84584
4599 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4600 SCoPs with fake exit edge.
4602 2018-02-28 Martin Liska <mliska@suse.cz>
4605 * timevar.c (timer::print): Fix format to properly print 100%
4608 2018-02-28 Richard Biener <rguenther@suse.de>
4611 * genmatch.c (capture_info::walk_match): Do not mark
4612 captured expressions without operands as expr_p given
4613 they act more like predicates and should be subject to
4614 "lost tail" side-effect preserving.
4616 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4618 PR rtl-optimization/81611
4619 * auto-inc-dec.c (attempt_change): Move dead note from
4620 mem_insn if it's the next use of regno
4621 (find_address): Take address use of reg holding
4622 non-incremented value. Add parm to limit search to the named
4624 (merge_in_block): Attempt to use a mem insn that is the next
4625 use of the original regno.
4627 2018-02-27 Martin Sebor <msebor@redhat.com>
4630 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4631 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4633 2018-02-27 Martin Sebor <msebor@redhat.com>
4635 PR translation/84207
4636 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4637 n argument to unsigned HOST_WIDE_INT.
4638 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4639 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4640 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4641 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4643 2018-02-27 Richard Biener <rguenther@suse.de>
4645 PR tree-optimization/84512
4646 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4647 Do not use the estimate returned from record_stmt_cost for
4648 the scalar iteration cost but sum properly using add_stmt_cost.
4650 2018-02-27 Richard Biener <rguenther@suse.de>
4652 PR tree-optimization/84466
4653 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4654 Adjust last change to less strictly validate use operands.
4656 2018-02-27 Martin Liska <mliska@suse.cz>
4658 PR gcov-profile/84548
4659 * gcov.c (process_file): Allow partial overlap and consider it
4660 also as group functions.
4661 (output_lines): Properly calculate range of lines for a group.
4663 2018-02-27 Martin Liska <mliska@suse.cz>
4665 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4666 'ggc' suffixes. Change first column width.
4667 (timer::print): Fix formatting of the column.
4669 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4671 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4672 preserve inline entry blocks for the sake of debug inline
4673 entry point markers alone.
4674 (remove_unused_locals): Suggest in comments a better place to
4675 force the preservation of inline entry blocks that are
4676 otherwise unused, but do not preserve them.
4678 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4680 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4682 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4685 * config/i386/constraints.md (Bs): Replace
4686 ix86_indirect_branch_register with
4687 TARGET_INDIRECT_BRANCH_REGISTER.
4689 * config/i386/i386.md (indirect_jump): Likewise.
4690 (tablejump): Likewise.
4691 (*sibcall_memory): Likewise.
4692 (*sibcall_value_memory): Likewise.
4693 Peepholes of indirect call and jump via memory: Likewise.
4694 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4695 (*sibcall_value_GOT_32): Likewise.
4696 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4697 (GOT_memory_operand): Likewise.
4698 (call_insn_operand): Likewise.
4699 (sibcall_insn_operand): Likewise.
4700 (GOT32_symbol_operand): Likewise.
4701 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4703 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4705 PR rtl-optimization/83496
4706 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4707 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4708 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4709 redundant insn, if any.
4710 (relax_delay_slots): Likewise.
4711 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4713 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4715 PR tree-optimization/83965
4716 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4717 that grouped statements are part of a reduction chain. Return
4718 true if the statement is not marked as a reduction itself but
4720 (vect_recog_dot_prod_pattern): Don't check whether the statement
4721 is part of a group here.
4722 (vect_recog_sad_pattern): Likewise.
4723 (vect_recog_widen_sum_pattern): Likewise.
4725 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4728 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4730 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4733 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4735 (ix86_output_indirect_function_return): New prototype.
4736 (ix86_split_simple_return_pop_internal): Likewise.
4737 * config/i386/i386.c (indirect_return_via_cx): New.
4738 (indirect_return_via_cx_bnd): Likewise.
4739 (indirect_thunk_name): Handle return va CX_REG.
4740 (output_indirect_thunk_function): Create alias for
4741 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4742 (ix86_output_indirect_jmp): Remove the bool argument.
4743 (ix86_output_indirect_function_return): New function.
4744 (ix86_split_simple_return_pop_internal): Likewise.
4745 * config/i386/i386.md (*indirect_jump): Don't pass false
4746 to ix86_output_indirect_jmp.
4747 (*tablejump_1): Likewise.
4748 (simple_return_pop_internal): Change it to define_insn_and_split.
4749 Call ix86_split_simple_return_pop_internal to split it for
4751 (simple_return_indirect_internal): Call
4752 ix86_output_indirect_function_return instead of
4753 ix86_output_indirect_jmp.
4755 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4758 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4759 memset and value initialization afterwards.
4761 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4763 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4765 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4768 * common/config/aarch64/aarch64-common.c
4769 (aarch_option_optimization_table[]): Switch
4770 off fomit-frame-pointer
4772 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4773 Chung-Ju Wu <jasonwucj@gmail.com>
4775 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4777 (store_multiple): Ditto.
4779 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4781 * config.gcc: Add --with-cpu support for nds32 target.
4782 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4783 * config/nds32/nds32.opt: Add -mcpu= option.
4785 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4787 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4788 isel=yes): Warn for these deprecated options.
4790 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4792 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4793 ISA_2_5_MASKS_EMBEDDED.
4795 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4797 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4798 p->max as pointers rather than using iterative_hash_expr.
4800 2018-02-23 Carl Love <cel@us.ibm.com>
4802 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4803 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4805 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4806 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4807 P8V_BUILTIN_VEC_VUNSIGNED2.
4809 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4812 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4813 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4815 * lra-constraints.c (curr_insn_transform): Set up
4816 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4818 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4819 * lra-eliminations.c (spill_pseudos): Ditto.
4820 (process_insn_for_elimination): Ditto.
4821 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4822 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4823 LRA_NON_CLOBBERED_ALT.
4825 2018-02-22 Martin Sebor <msebor@redhat.com>
4827 PR tree-optimization/84480
4828 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4829 to maybe_diag_stxncpy_trunc. Call it.
4830 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4831 from gimple_fold_builtin_strcpy. Print inlining stack.
4832 (handle_builtin_stxncpy): Print inlining stack.
4833 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4835 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4838 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4839 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4840 and -fcheck-pointer-bounds are used together.
4841 (indirect_thunk_prefix): New enum.
4842 (indirect_thunk_need_prefix): New function.
4843 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4844 "_nt" instead of "_bnd" for NOTRACK prefix.
4845 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4846 (output_indirect_thunk_function): Likewise.
4848 (ix86_code_end): Update output_indirect_thunk_function calls.
4849 (ix86_output_indirect_branch_via_reg): Replace
4850 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4851 (ix86_output_indirect_branch_via_push): Likewise.
4852 (ix86_output_function_return): Likewise.
4853 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4854 incompatible with -fcf-protection=branch and
4855 -fcheck-pointer-bounds.
4857 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4860 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4861 Change gcc_assert call to output_operand_lossage.
4863 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4865 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4867 2018-02-22 DJ Delorie <dj@redhat.com>
4868 Sebastian Perta <sebastian.perta@renesas.com>
4869 Oleg Endo <olegendo@gcc.gnu.org>
4871 * config/rx/rx.c (rx_rtx_costs): New function.
4872 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4874 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4876 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4878 2018-02-22 Martin Liska <mliska@suse.cz>
4881 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4882 Add "native" as a possible value.
4884 2018-02-22 Martin Liska <mliska@suse.cz>
4887 * config/i386/i386.c (ix86_option_override_internal):
4888 Add "native" as a possible value for -march and -mtune.
4890 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4893 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4894 to all type variants.
4896 PR tree-optimization/84503
4897 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4898 width as info->bitpos + info->bitsize - start.
4899 (merged_store_group::merge_overlapping): Simplify width computation.
4900 (check_no_overlap): New function.
4901 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4902 start + width and last_order of the group, fail if check_no_overlap
4904 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4905 to group if check_no_overlap fails.
4907 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4909 * config/rs6000/altivec.md: Delete contraint arguments to
4910 define_expand, define_split, and define_peephole2, and in
4911 define_insn_and_split if always unused.
4912 * config/rs6000/darwin.md: Ditto.
4913 * config/rs6000/dfp.md: Ditto.
4914 * config/rs6000/rs6000.md: Ditto.
4915 * config/rs6000/sync.md: Ditto.
4916 * config/rs6000/vector.md: Ditto.
4917 * config/rs6000/vsx.md: Ditto.
4919 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4921 * config/rs6000/altivec.md: Write output control strings as braced
4922 blocks instead of double-quoted strings.
4923 * config/rs6000/darwin.md: Ditto.
4924 * config/rs6000/rs6000.md: Ditto.
4925 * config/rs6000/vector.md: Ditto.
4926 * config/rs6000/vsx.md: Ditto.
4928 2018-02-21 Jason Merrill <jason@redhat.com>
4930 PR c++/84314 - ICE with templates and fastcall attribute.
4931 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4933 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4935 * ipa-cp.c (determine_versionability): Fix comment typos.
4937 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4940 * ipa-cp.c (determine_versionability): Do not version functions caling
4943 2018-02-21 Martin Liska <mliska@suse.cz>
4946 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4947 Add "native" as a possible value.
4948 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4949 the macro when native cpu detection is available.
4951 2018-02-21 Martin Liska <mliska@suse.cz>
4954 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4955 Add "native" as a possible value.
4956 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4957 when native cpu detection is available.
4959 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4960 Martin Sebor <msebor@redhat.com>
4962 PR tree-optimization/84478
4963 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4965 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4966 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4967 support which is conservatively correct, for 2 only stay conservative
4968 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4969 argument to the 2 argument get_range_strlen, adjust 6 arg
4970 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4972 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4973 (gimple_fold_builtin_strlen): Pass true as last argument to
4976 2018-02-20 Martin Sebor <msebor@redhat.com>
4979 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4980 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4981 (builtin_memref::builtin_memref): Factor out parts into
4982 set_base_and_offset and call it.
4984 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4987 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4988 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4989 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4990 search at the associated MODE_INT.
4992 2018-02-20 Jeff Law <law@redhat.com>
4995 PR tree-optimization/81592
4997 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4998 for range data rather than using global data.
4999 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5000 range data rather than using global data.
5001 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5002 pass it to children as needed.
5003 (struct directive::fmtresult): Similarly.
5004 (struct directive::set_width): Similarly.
5005 (struct directive::set_precision): Similarly.
5006 (format_integer, format_directive, parse_directive): Similarly.
5007 (format_none): Accept unnamed vr_values parameter.
5008 (format_percent, format_floating, format_character): Similarly.
5009 (format_string, format_plain): Similarly.
5010 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5011 the EVRP range analyzer for range data rather than using global data.
5012 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5013 gimple-ssa-evrp-analyze.h
5014 (class sprintf_dom_walker): Add after_dom_children member function.
5015 Add evrp_range_analyzer member.
5016 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5017 range analyzer as needed.
5018 (sprintf_dom_walker::after_dom_children): New member function.
5019 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5021 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5022 (evrp_range_analyzer::pop_to_marker): Likewise.
5024 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5026 PR tree-optimization/84419
5027 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5028 with the required type if its current type is compatible but
5031 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5034 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5035 after vectorization.
5037 2018-02-20 Martin Liska <mliska@suse.cz>
5040 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5041 possible values if we don't have a hint.
5043 2018-02-20 Martin Liska <mliska@suse.cz>
5047 * final.c (shorten_branches): Build align_tab array with one
5049 * opts.c (finish_options): Add alignment option limit check.
5050 (MAX_CODE_ALIGN): Likewise.
5051 (MAX_CODE_ALIGN_VALUE): Likewise.
5052 * doc/invoke.texi: Document maximum allowed option value for
5053 all -falign-* options.
5055 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5058 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5059 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5060 * var-tracking.c (emit_note_insn_var_location): Remove all references
5061 to NOTE_INSN_CALL_ARG_LOCATION.
5062 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5063 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5064 Use copy_rtx_if_shared.
5065 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5066 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5067 (dwarf2out_var_location): Remove handling of
5068 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5070 * final.c (final_scan_insn): Remove all references to
5071 NOTE_INSN_CALL_ARG_LOCATION.
5072 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5073 before dumping final insns.
5074 * except.c (emit_note_eh_region_end): Remove all references to
5075 NOTE_INSN_CALL_ARG_LOCATION.
5076 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5077 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5078 * config/arc/arc.c (hwloop_optimize): Likewise.
5079 * config/arm/arm.c (create_fix_barrier): Likewise.
5080 * config/s390/s390.c (s390_chunkify_start): Likewise.
5081 * config/sh/sh.c (find_barrier): Likewise.
5082 * config/i386/i386.c (rest_of_insert_endbranch,
5083 ix86_seh_fixup_eh_fallthru): Likewise.
5084 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5085 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5086 * config/frv/frv.c (frv_function_prologue): Likewise.
5087 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5089 (note_outside_basic_block_p): Remove all references to
5090 NOTE_INSN_CALL_ARG_LOCATION.
5091 * gengtype.c (adjust_field_rtx_def): Likewise.
5092 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5094 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5095 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5098 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5101 PR tree-optimization/84452
5102 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5103 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5106 2018-02-19 Martin Liska <mliska@suse.cz>
5109 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5111 2018-02-19 Martin Liska <mliska@suse.cz>
5112 Richard Sandiford <richard.sandiford@linaro.org>
5114 PR tree-optimization/82491
5115 * gimple-fold.c (get_base_constructor): Make earlier bail out
5118 2018-02-19 Carl Love <cel@us.ibm.com>
5120 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5121 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5123 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5124 P8V_BUILTIN_VEC_NEG.
5126 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5128 * config/rl78/rl78.md (movdf): New define expand.
5130 2018-02-19 Martin Liska <mliska@suse.cz>
5133 * doc/invoke.texi: Fix typo.
5134 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5136 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5138 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5139 handle rs6000_single_float and rs6000_double_float specially for
5142 2018-02-16 Jeff Law <law@redhat.com>
5144 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5147 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5150 * ipa-inline.c (inline_small_functions): Fix a typo.
5152 2018-02-16 Nathan Sidwell <nathan@acm.org>
5154 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5156 2018-02-16 Carl Love <cel@us.ibm.com>
5158 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5159 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5160 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5161 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5162 expansion to P8V_BUILTIN_VEC_FLOAT2.
5164 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5166 PR rtl-optimization/70023
5167 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5168 src_regno into account.
5170 2018-02-16 Carl Love <cel@us.ibm.com>
5172 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5173 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5174 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5175 * config/rs6000/rs6000.c: Remove case statements for
5176 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5177 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5178 and P9V_BUILTIN_VEC_VINSERT4B.
5179 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5180 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5181 * config/rs6000/vsx.md:
5182 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5185 2018-02-16 Carl Love <cel@us.ibm.com>
5187 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5189 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5191 * config/rs6000/rs6000-c.c: Add the definitions for
5192 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5193 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5194 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5195 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5196 definition for insert4b and define insn *insert3b_internal.
5197 * doc/extend.texi: Add documentation for vec_extract4b.
5199 2018-02-16 Nathan Sidwell <nathan@acm.org>
5201 * doc/extend.texi (Backwards Compatibility): Mention friend
5202 injection. Note for-scope is deprecated.
5203 * doc/invoke.texi (-ffriend-injection): Deprecate.
5205 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5207 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5208 that moved to I2, also allow destinations that are a paradoxical
5209 subreg (instead of a normal reg).
5211 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5214 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5217 2018-02-16 Richard Biener <rguenther@suse.de>
5219 PR tree-optimization/84037
5220 PR tree-optimization/84016
5222 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5223 Adjust vec_construct for the fact we need additional higher latency
5224 128bit inserts for AVX256 and AVX512 vector builds.
5225 (ix86_add_stmt_cost): Scale vector construction cost for
5228 2018-02-16 Richard Biener <rguenther@suse.de>
5230 PR tree-optimization/84417
5231 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5232 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5233 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5235 2018-02-16 Martin Liska <mliska@suse.cz>
5238 * internal-fn.def (ASAN_CHECK): Set proper flags.
5239 (ASAN_MARK): Likewise.
5241 2018-02-16 Julia Koval <julia.koval@intel.com>
5243 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5244 from PTA_CANNONLAKE.
5246 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5249 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5250 Use ++iter rather than iter++ for std::list iterators.
5251 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5252 defer deleting them until all nodes in the forest are processed. Do
5253 free even leaf nodes. Change to_process into auto_vec.
5256 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5257 * vec.h (vec_default_construct): Use memset instead of placement new
5258 if BROKEN_VALUE_INITIALIZATION is defined.
5259 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5260 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5263 PR rtl-optimization/83723
5264 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5265 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5266 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5268 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5270 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5272 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5274 PR rtl-optimization/81443
5275 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5276 from inner REGs to paradoxical SUBREGs.
5278 2018-02-16 Richard Biener <rguenther@suse.de>
5280 PR tree-optimization/84399
5281 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5282 For operands we can analyze at their definition make sure we can
5283 analyze them at each use as well.
5285 2018-02-16 Richard Biener <rguenther@suse.de>
5287 PR tree-optimization/84190
5288 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5289 volatile accesses if the decl isn't volatile.
5291 2018-02-15 Jason Merrill <jason@redhat.com>
5293 PR c++/84314 - ICE with templates and fastcall attribute.
5294 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5295 TYPE_CANONICAL on an existing type.
5297 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5299 PR tree-optimization/84383
5300 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5301 dstoff nor call operand_equal_p if dstbase is NULL.
5303 PR tree-optimization/84334
5304 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5305 also a CONSTANT_CLASS_P, punt.
5307 2018-02-14 Jim Wilson <jimw@sifive.com>
5309 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5310 first SMALL_OPERAND check. New local min_second_step. Move assert
5311 to where locals are set. Add TARGET_RVC support.
5312 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5314 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5316 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5318 2018-02-14 Martin Sebor <msebor@redhat.com>
5320 PR tree-optimization/83698
5321 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5322 arrays constrain the offset range to their bounds.
5323 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5324 (builtin_access::overlap): Avoid setting the size of overlap if it's
5326 (maybe_diag_overlap): Also consider arrays when deciding what values
5327 of offsets to include in diagnostics.
5329 2018-02-14 Martin Sebor <msebor@redhat.com>
5332 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5333 that correspond to the kind of a declaration.
5335 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5338 * config/pa/pa.md: Load address of PIC label using the linkage table
5339 if the label is nonlocal.
5341 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5343 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5344 warning message if user requests -maltivec=be.
5345 * doc/invoke.texi: Document deprecation of -maltivec=be.
5347 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5350 * config/rs6000/rs6000-c.c: Update definitions for
5351 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5352 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5354 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5357 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5358 add _get_ssp intrinsics. Remove argument from
5359 __builtin_ia32_rdssp[d|q].
5360 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5361 * config/i386/i386-builtin.def: Remove argument from
5362 __builtin_ia32_rdssp[d|q].
5363 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5364 ix86_expand_special_args_builtin for _rdssp[d|q].
5365 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5366 Clear register before usage.
5367 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5368 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5370 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5372 PR tree-optimization/84357
5373 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5374 operand 1 of an ARRAY_REF too.
5376 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5379 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5380 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5382 (set_of_reg): New struct.
5383 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5384 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5385 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5387 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5388 Split into bitclr, bitset, bitinvert patterns if appropriate.
5389 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5390 use rx_fuse_in_memory_bitop.
5391 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5392 to named insn, correct maximum insn length.
5394 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5397 * machmode.def: Define a complex mode for PARTIAL_INT.
5398 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5400 * doc/rtl.texi: Document CSPImode.
5401 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5403 (msp430_hard_regno_nregs_with_padding): Likewise.
5405 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5408 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5410 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5412 PR rtl-optimization/84169
5413 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5414 we generated a parallel as new i3 and we split that to new i2 and i3
5415 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5416 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5417 those to i2, not i1. Partially rewrite this scan code.
5419 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5422 * stor-layout.c (place_field): For variable length fields, adjust
5423 offset_align afterwards not just based on the field's alignment,
5424 but also on the size.
5427 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5428 of exps and logs in the use_exp2 case.
5430 2018-02-13 Jeff Law <law@redhat.com>
5432 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5435 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5438 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5442 * common.opt (gas-loc-support, gas-locview-support): New.
5443 (ginline-points, ginternal-reset-location-views): New.
5444 * doc/invoke.texi: Document them. Use @itemx where intended.
5445 (gvariable-location-views): Adjust.
5446 * target.def (reset_location_view): New.
5447 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5448 (TARGET_RESET_LOCATION_VIEW): New.
5449 * doc/tm.texi: Rebuilt.
5450 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5451 (dwarf2out_default_as_locview_support): New.
5452 (output_asm_line_debug_info): Use option variables.
5453 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5454 (output_loc_list): Likewise.
5455 (add_high_low_attributes): Check option variables.
5456 Don't output entry view attribute in strict mode.
5457 (gen_inlined_subroutine_die): Check option variables.
5458 (dwarf2out_inline_entry): Likewise.
5459 (init_sections_and_labels): Likewise.
5460 (dwarf2out_early_finish): Likewise.
5461 (maybe_reset_location_view): New, from...
5462 (dwarf2out_var_location): ... here. Call it.
5463 * debug.h (dwarf2out_default_as_loc_support): Declare.
5464 (dwarf2out_default_as_locview_support): Declare.
5465 * hooks.c (hook_int_rtx_insn_0): New.
5466 * hooks.h (hook_int_rtx_insn_0): Declare.
5467 * toplev.c (process_options): Take -gas-loc-support and
5468 -gas-locview-support from dwarf2out. Enable
5469 -gvariable-location-views by default only with locview
5470 assembler support. Enable -ginternal-reset-location-views by
5471 default only if the target defines the corresponding hook.
5472 Enable -ginline-points by default if location views are
5473 enabled; force it disabled if statement frontiers are
5475 * tree-inline.c (expand_call_inline): Check option variables.
5476 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5478 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5480 PR tree-optimization/84321
5481 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5482 handling. Also check whether the anti-range contains any values
5483 that satisfy the mask; switch to a VR_RANGE if not.
5485 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5488 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5490 2018-02-13 Martin Jambor <mjambor@suse.cz>
5493 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5494 of call statements, also set location of a load to a temporary.
5496 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5498 * config/rl78/rl78.c (add_vector_labels): New function.
5499 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5500 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5501 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5502 which checks that no arguments are passed.
5503 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5504 * doc/extend.texi: Documentation for the new attribute.
5506 2018-02-13 Andreas Schwab <schwab@suse.de>
5508 * config/riscv/linux.h (CPP_SPEC): Define.
5510 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5513 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5514 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5515 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5516 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5517 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5518 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5519 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5520 temporarily for AES and PCLMUL builtins.
5522 PR tree-optimization/84339
5523 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5524 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5528 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5529 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5530 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5532 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5534 * omp-simd-clone.h: New file.
5535 * omp-simd-clone.c: Include omp-simd-clone.h.
5536 (expand_simd_clones): No longer static.
5537 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5538 cgraph.h and omp-simd-clone.h.
5539 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5540 (vect_recog_widen_shift_pattern): Formatting fix.
5541 (vect_pattern_recog_1): Don't check optab for calls.
5544 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5545 operands[2] into a REG before using gen_lowpart on it.
5547 2018-02-12 Jeff Law <law@redhat.com>
5550 * config/sh/sh.c (find_barrier): Consider a sibling call
5553 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5554 successfully back substituting a reg.
5556 2018-02-12 Richard Biener <rguenther@suse.de>
5558 PR tree-optimization/84037
5559 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5560 parameter, move visited init to caller.
5561 (vect_slp_analyze_operations): Separate cost from validity
5562 check, initialize visited once for all instances.
5563 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5565 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5567 (vect_model_promotion_demotion_cost): Likewise.
5568 (vectorizable_bswap): Guard cost modeling with !slp_node
5569 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5571 (vectorizable_call): Likewise.
5572 (vectorizable_conversion): Likewise.
5573 (vectorizable_assignment): Likewise.
5574 (vectorizable_shift): Likewise.
5575 (vectorizable_operation): Likewise.
5576 (vectorizable_store): Likewise.
5577 (vectorizable_load): Likewise.
5578 (vectorizable_condition): Likewise.
5579 (vectorizable_comparison): Likewise.
5581 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5584 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5585 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5586 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5588 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5591 * params.def (inline-min-speedup): Increase from 8 to 15.
5592 (max-inline-insns-auto): Decrease from 40 to 30.
5593 * ipa-split.c (consider_split): Add some buffer for function to
5594 be considered inlining candidate.
5595 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5598 2018-02-12 Richard Biener <rguenther@suse.de>
5600 PR tree-optimization/84037
5601 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5602 matched stmts if we cannot swap the non-matched ones.
5604 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5606 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5607 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5608 _mm_maskz_scalef_round_ss): New intrinsics.
5609 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5610 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5611 __builtin_ia32_scalefss_round): Remove.
5612 (__builtin_ia32_scalefsd_mask_round,
5613 __builtin_ia32_scalefss_mask_round): New intrinsics.
5614 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5615 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5616 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5617 "<round_constraint>")): Changed to ...
5618 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5619 "<round_scalar_constraint>")): ... this.
5620 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5621 %0, %1, %2<round_op3>}"): Changed to ...
5622 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5623 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5624 %2<round_scalar_mask_op3>}"): ... this.
5625 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5627 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5629 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5630 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5631 (_mm_maskz_sqrt_round_ss): New intrinsics.
5632 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5633 (__builtin_ia32_sqrtsd_mask_round)
5634 (__builtin_ia32_sqrtss_mask_round): New builtins.
5635 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5636 (__builtin_ia32_sqrtss_round): Remove.
5637 (__builtin_ia32_sqrtsd_mask_round)
5638 (__builtin_ia32_sqrtss_mask_round): New builtins.
5639 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5640 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5641 ((match_operand:VF_128 1 "vector_operand"
5642 "xBm,<round_constraint>")): Changed to ...
5643 ((match_operand:VF_128 1 "vector_operand"
5644 "xBm,<round_scalar_constraint>")): ... this.
5645 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5646 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5647 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5648 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5649 %<iptr>1<round_scalar_mask_op3>}): ... this.
5650 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5651 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5653 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5656 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5657 Cast vec_cmpeq result to correct type.
5658 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5659 Cast vec_cmpgt result to correct type.
5661 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5663 * final.c (final_scan_insn_1): Renamed from...
5664 (final_scan_insn): ... this. New wrapper, to recover
5665 seen from the outermost call in recursive ones.
5666 * config/sparc/sparc.c (output_return): Drop seen from call.
5667 (output_sibcall): Likewise.
5668 * config/visium/visium.c (output_branch): Likewise.
5670 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5672 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5675 2018-02-10 Alan Modra <amodra@gmail.com>
5678 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5679 Specify LR as an input.
5681 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5684 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5685 remove_member_access_dummy_vars): New functions.
5686 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5687 lower_omp_1, execute_lower_omp): Use them.
5689 PR rtl-optimization/84308
5690 * shrink-wrap.c (spread_components): Release todo vector.
5692 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5694 PR rtl-optimization/57193
5695 * ira-color.c (struct allocno_color_data): Add member
5696 conflict_allocno_hard_prefs.
5697 (update_conflict_allocno_hard_prefs): New.
5698 (bucket_allocno_compare_func): Add a preference based on
5699 conflict_allocno_hard_prefs.
5700 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5701 (color_allocnos): Remove a dead code. Initiate
5702 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5704 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5707 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5708 constraint from =wa to wa. Avoid a subreg on the output operand,
5709 instead use a pseudo and subreg it in a move.
5710 (p9_xxbrd_<mode>): Changed to ...
5711 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5712 (p9_xxbrd_v2df): New expander.
5713 (p9_xxbrw_<mode>): Changed to ...
5714 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5715 (p9_xxbrw_v4sf): New expander.
5717 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5719 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5721 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5724 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5725 multiply in 32-bit mode.
5726 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5727 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5730 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5732 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5733 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5734 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5735 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5737 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5740 * dwarf2out.c (is_trivial_indirect_ref): New function.
5741 (dwarf2out_late_global_decl): Do not generate a location
5742 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5743 and that are not defined in the current unit.
5745 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5747 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5748 instead of a libcall for UNORDERED.
5750 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5753 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5754 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5756 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5759 * config/s390/s390.c (s390_set_current_function): Invoke
5760 s390_indirect_branch_settings also if fndecl didn't change.
5762 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5764 * config/rs6000/rs6000.md (blockage): Set length to zero.
5766 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5768 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5770 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5773 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5774 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5778 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5779 PARALLEL incoming that failed vt_get_decl_and_offset check.
5782 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5783 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5785 (get_variable_section): For decls in named .bss* sections pass true as
5786 second argument to bss_initializer_p.
5788 2018-02-09 Marek Polacek <polacek@redhat.com>
5789 Jakub Jelinek <jakub@redhat.com>
5792 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5793 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5794 Sync some changes from cxx_fold_indirect_ref.
5796 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5798 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5800 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5801 (BLOCK_INLINE_ENTRY_LABEL): New.
5802 (dwarf2out_var_location): Disregard inline entry markers.
5803 (inline_entry_data): New struct.
5804 (inline_entry_data_hasher): New hashtable type.
5805 (inline_entry_data_hasher::hash): New.
5806 (inline_entry_data_hasher::equal): New.
5807 (inline_entry_data_table): New variable.
5808 (add_high_low_attributes): Add DW_AT_entry_pc and
5809 DW_AT_GNU_entry_view attributes if a pending entry is found
5810 in inline_entry_data_table. Add old entry_pc attribute only
5811 if debug nonbinding markers are disabled.
5812 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5813 markers are enabled.
5814 (block_within_block_p, dwarf2out_inline_entry): New.
5815 (dwarf2out_finish): Check that no entries remained in
5816 inline_entry_data_table.
5817 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5818 (final_scan_insn, notice_source_line): Likewise.
5819 (rest_of_clean_state): Skip inline entry markers.
5820 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5822 * gimple.c (gimple_build_debug_inline_entry): New.
5823 * gimple.h (enum gimple_debug_subcode): Add
5824 GIMPLE_DEBUG_INLINE_ENTRY.
5825 (gimple_build_debug_inline_entry): Declare.
5826 (gimple_debug_inline_entry_p): New.
5827 (gimple_debug_nonbind_marker_p): Adjust.
5828 * insn-notes.def (INLINE_ENTRY): New.
5829 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5830 inline entry marker notes.
5831 (print_insn): Likewise.
5832 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5833 (INSN_DEBUG_MARKER_KIND): Likewise.
5834 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5835 * tree-inline.c (expand_call_inline): Build and insert
5836 debug_inline_entry stmt.
5837 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5838 inline entry blocks early, if nonbind markers are enabled.
5839 (dump_scope_block): Dump fragment info.
5840 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5841 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5842 (gimple_build_debug_inline_entry): New.
5843 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5844 Enable/disable inline entry points too.
5845 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5846 (DEBUG_INSN): Describe inline entry markers.
5848 * common.opt (gvariable-location-views): New.
5849 (gvariable-location-views=incompat5): New.
5850 * config.in: Rebuilt.
5851 * configure: Rebuilt.
5852 * configure.ac: Test assembler for view support.
5853 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5854 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5855 * dwarf2out.c (var_loc_view): New typedef.
5856 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5857 (dwarf2out_locviews_in_attribute): New.
5858 (dwarf2out_locviews_in_loclist): New.
5859 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5860 (enum dw_line_info_opcode): Add LI_adv_address.
5861 (struct dw_line_info_table): Add view.
5862 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5863 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5864 (zero_view_p): New variable.
5865 (ZERO_VIEW_P): New macro.
5866 (output_asm_line_debug_info): New.
5867 (struct var_loc_node): Add view.
5868 (add_AT_view_list, AT_loc_list): New.
5869 (add_var_loc_to_decl): Add view param. Test it against last.
5870 (new_loc_list): Add view params. Record them.
5871 (AT_loc_list_ptr): Handle loc and view lists.
5872 (view_list_to_loc_list_val_node): New.
5873 (print_dw_val): Handle dw_val_class_view_list.
5874 (size_of_die): Likewise.
5875 (value_format): Likewise.
5876 (loc_list_has_views): New.
5877 (gen_llsym): Set vl_symbol too.
5878 (maybe_gen_llsym, skip_loc_list_entry): New.
5879 (dwarf2out_maybe_output_loclist_view_pair): New.
5880 (output_loc_list): Output view list or entries too.
5881 (output_view_list_offset): New.
5882 (output_die): Handle dw_val_class_view_list.
5883 (output_dwarf_version): New.
5884 (output_compilation_unit_header): Use it.
5885 (output_skeleton_debug_sections): Likewise.
5886 (output_rnglists, output_line_info): Likewise.
5887 (output_pubnames, output_aranges): Update version comments.
5888 (output_one_line_info_table): Output view numbers in asm comments.
5889 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5890 Call maybe_gen_llsym.
5891 (loc_list_from_tree_1): Adjust.
5892 (add_AT_location_description): Create view list attribute if
5893 needed, check it's absent otherwise.
5894 (convert_cfa_to_fb_loc_list): Adjust.
5895 (maybe_emit_file): Call output_asm_line_debug_info for test.
5896 (dwarf2out_var_location): Reset views as needed. Precompute
5897 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5898 attribute. Set view.
5899 (new_line_info_table): Reset next view.
5900 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5901 (dwarf2out_source_line): Likewise. Output view resets and labels to
5902 the assembler, or select appropriate line info opcodes.
5903 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5904 (optimize_string_length): Catch it. Adjust.
5905 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5906 dw_val_class_view_list, and remove it if no longer needed.
5907 (hash_loc_list): Hash view numbers.
5908 (loc_list_hasher::equal): Compare them.
5909 (optimize_location_lists): Check whether a view list symbol is
5910 needed, and whether the locview attribute is present, and
5911 whether they match. Remove the locview attribute if no longer
5913 (index_location_lists): Call skip_loc_list_entry for test.
5914 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5915 Use output_dwarf_version.
5916 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5917 (struct dw_val_node): Add val_view_list.
5918 * final.c (SEEN_NEXT_VIEW): New.
5919 (set_next_view_needed): New.
5920 (clear_next_view_needed): New.
5921 (maybe_output_next_view): New.
5922 (final_start_function): Rename to...
5923 (final_start_function_1): ... this. Take pointer to FIRST,
5924 add SEEN parameter. Emit param bindings in the initial view.
5925 (final_start_function): Reintroduce SEEN-less interface.
5926 (final): Rename to...
5927 (final_1): ... this. Take SEEN parameter. Output final pending
5928 next view at the end.
5929 (final): Reintroduce seen-less interface.
5930 (final_scan_insn): Output pending next view before switching
5931 sections or ending a block. Mark the next view as needed when
5932 outputting variable locations. Notify debug backend of section
5933 changes, and of location view changes.
5934 (rest_of_handle_final): Adjust.
5935 * toplev.c (process_options): Autodetect value for debug variable
5936 location views option. Warn on incompat5 without -gdwarf-5.
5937 * doc/invoke.texi (gvariable-location-views): New.
5938 (gvariable-location-views=incompat5): New.
5939 (gno-variable-location-views): New.
5941 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5943 PR tree-optimization/84136
5944 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5945 that the result of find_edge is non-NULL.
5947 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5950 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5951 storing integer register in SImode. Fix cost of 256 and 512
5952 byte aligned SSE register store.
5954 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5956 * config/i386/i386.c (ix86_multiplication_cost): Fix
5957 multiplication cost for TARGET_AVX512DQ.
5959 2018-02-08 Marek Polacek <polacek@redhat.com>
5961 PR tree-optimization/84238
5962 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5965 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5967 PR tree-optimization/84265
5968 * tree-vect-stmts.c (vectorizable_store): Don't treat
5969 VMAT_CONTIGUOUS accesses as grouped.
5970 (vectorizable_load): Likewise.
5972 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5974 PR tree-optimization/81635
5975 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5976 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5977 (test_round_for_mask): New functions.
5978 (wide_int_cc_tests): Call test_round_for_mask.
5979 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5980 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5981 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5982 range returned by get_range_info.
5984 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5987 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5988 * symtab.c: Include builtins.h
5989 (symtab_node::output_to_lto_symbol_table_p): Move here
5990 from lto-streamer-out.c:output_symbol_p.
5991 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5992 (output_symbol_p): Move all logic to symtab.c
5993 (produce_symtab): Update.
5995 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5997 * config/s390/s390-opts.h (enum indirect_branch): Define.
5998 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5999 (s390_indirect_branch_via_thunk)
6000 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6001 (enum s390_indirect_branch_type): Define.
6002 * config/s390/s390.c (struct s390_frame_layout, struct
6003 machine_function): Remove.
6004 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6005 (indirect_branch_table_label_no, indirect_branch_table_name):
6007 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6008 (enum s390_indirect_branch_option): Define.
6009 (s390_return_addr_from_memory): New function.
6010 (s390_handle_string_attribute): New function.
6011 (s390_attribute_table): Add new attribute handler.
6012 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6013 (s390_indirect_branch_via_thunk): New function.
6014 (s390_indirect_branch_via_inline_thunk): New function.
6015 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6016 sibling call optimization for non z10 compiles.
6017 (s390_emit_call): Force indirect branch target to be a single
6018 register. Add r1 clobber for non-z10 compiles.
6019 (s390_emit_epilogue): Emit return jump via return_use expander.
6020 (s390_reorg): Handle JUMP_INSNs as execute targets.
6021 (s390_option_override_internal): Perform validity checks for the
6022 new command line options.
6023 (s390_indirect_branch_attrvalue): New function.
6024 (s390_indirect_branch_settings): New function.
6025 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6026 (s390_output_indirect_thunk_function): New function.
6027 (s390_code_end): Implement target hook.
6028 (s390_case_values_threshold): Implement target hook.
6029 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6031 * config/s390/s390.h (struct s390_frame_layout)
6032 (struct machine_function): Move here from s390.c.
6033 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6034 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6035 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6036 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6037 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6038 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6039 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6040 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6041 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6042 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6043 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6044 (mnemonic attribute): Add values which aren't recognized
6046 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6047 pattern for branch conversion. Fix mnemonic attribute.
6048 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6049 indirect branch via thunk if requested.
6050 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6051 ("*indirect_jump"): Disable for branch conversion using out of
6053 ("indirect_jump_via_thunk<mode>_z10")
6054 ("indirect_jump_via_thunk<mode>")
6055 ("indirect_jump_via_inlinethunk<mode>_z10")
6056 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6057 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6058 ("casesi_jump_via_inlinethunk<mode>_z10")
6059 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6060 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6061 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6062 ("*indirect2_jump"): Disable for branch conversion.
6063 ("casesi_jump"): Turn into expander and expand patterns for branch
6065 ("return_use"): New expander.
6066 ("*return"): Emit return via thunk and rename it to ...
6067 ("*return<mode>"): ... this one.
6068 * config/s390/s390.opt: Add new options and and enum for the
6071 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6073 * lra-constraints.c (match_reload): Unconditionally use
6074 gen_lowpart_SUBREG, rather than selecting between that
6075 and equivalent gen_rtx_SUBREG code.
6077 2018-02-08 Richard Biener <rguenther@suse.de>
6079 PR tree-optimization/84233
6080 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6081 changed flag instead of boguously re-using phi_inserted.
6083 2018-02-08 Martin Jambor <mjambor@suse.cz>
6085 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6086 static local variables.
6088 2018-02-08 Richard Biener <rguenther@suse.de>
6090 PR tree-optimization/84278
6091 * tree-vect-stmts.c (vectorizable_store): When looking for
6092 smaller vector types to perform grouped strided loads/stores
6093 make sure the mode is supported by the target.
6094 (vectorizable_load): Likewise.
6096 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6098 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6099 Increase LDP/STP opportunities by adding adjacent callee-saves.
6101 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6103 PR rtl-optimization/84068
6104 PR rtl-optimization/83459
6105 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6107 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6109 PR tree-optimization/84224
6110 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6111 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6114 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6117 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6118 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6119 count, remove one USE.
6121 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6123 * doc/install.texi (Configuration): Document the
6124 --with-long-double-format={ibm,ieee} PowerPC configuration
6128 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6129 Convert from define_expand to be define_insn_and_split. Rework
6130 float/double/_Float128 conversions to QI/HI/SImode to work with
6131 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6132 conversions to QI/HImode types did a store and then a load to
6133 truncate the value. For conversions to VSX registers, don't split
6134 the insn, instead emit the code directly. Use the code iterator
6135 any_fix to combine signed and unsigned conversions.
6136 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6137 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6138 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6139 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6140 (fix_<mode>di2_hw): Likewise.
6141 (fixuns_<mode>di2_hw): Likewise.
6142 (fix_<mode>si2_hw): Likewise.
6143 (fixuns_<mode>si2_hw): Likewise.
6144 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6145 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6146 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6147 fix<uns>_trunc<SFDF:mode>si2_p8.
6148 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6150 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6151 (fix<uns>_<mode>_mem): Likewise.
6152 (fctiw<u>z_<mode>_mem): Likewise.
6153 (fix<uns>_<mode>_mem): Likewise.
6154 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6155 the register allocator from doing a direct move to the GPRs to do
6156 a store, and instead use the ISA 3.0 store byte/half-word from
6157 vector register instruction. For IEEE 128-bit floating point,
6158 also optimize stores of 32-bit ints.
6159 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6161 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6163 * genextract.c (push_pathstr_operand): New function to support
6165 (walk_rtx): Call push_pathstr_operand.
6166 (print_path): Support [a-zA-Z].
6168 2018-02-07 Richard Biener <rguenther@suse.de>
6170 PR tree-optimization/84037
6171 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6172 (cse_and_gimplify_to_preheader): Declare.
6173 (vect_get_place_in_interleaving_chain): Likewise.
6174 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6176 (_loop_vec_info::~_loop_vec_info): Delete it.
6177 (cse_and_gimplify_to_preheader): New function.
6178 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6179 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6180 (vectorizable_load): Likewise. For grouped stores always base
6181 the IV on the first element.
6182 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6183 condition before gimplifying.
6185 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6187 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6188 *DIV_EXPR and *MOD_EXPR.
6190 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6193 * config/i386/i386.c (ix86_option_override_internal): Mask out
6194 the CF_SET bit when checking -fcf-protection.
6196 2018-02-07 Tom de Vries <tom@codesourcery.com>
6199 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6202 2018-02-07 Richard Biener <rguenther@suse.de>
6204 PR tree-optimization/84204
6205 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6208 PR tree-optimization/84205
6209 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6210 special-case isl_ast_op_zdiv_r.
6212 PR tree-optimization/84223
6213 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6214 Only add conditions from within the region.
6215 (gather_bbs::after_dom_children): Adjust.
6217 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6220 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6221 * config/avr/avr.md: Only post-reload split REG-REG moves if
6222 either register is GENERAL_REG_P.
6224 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6226 PR tree-optimization/84235
6227 * tree-ssa-scopedtables.c
6228 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6229 if the subtraction is performed in floating point type where NaNs are
6230 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6231 build 1. Formatting fix.
6233 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6236 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6237 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6238 and skip it regardless of bb boundaries. Use CALL_P macro,
6239 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6240 unnecessarily, formatting fix.
6242 2018-02-06 Michael Collison <michael.collison@arm.com>
6244 * config/arm/thumb2.md:
6245 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6246 (*thumb_mov_notscc): Ditto.
6248 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6251 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6252 unsigned_fix, not "s".
6254 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6256 * configure.ac (gcc_fn_eh_frame_ro): New function.
6257 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6258 correct .eh_frame permissions.
6259 * configure: Regenerate.
6261 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6263 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6266 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6268 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6269 Display warning message for -mno-speculate-indirect-jumps.
6271 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6273 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6275 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6277 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6279 PR tree-optimization/84225
6280 * tree-eh.c (find_trapping_overflow): Only call
6281 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6283 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6286 * config/i386/i386.c: Reimplement the check of possible options
6287 -mibt/-mshstk conbination. Change error messages.
6288 * doc/invoke.texi: Fix a typo: remove extra '='.
6290 2018-02-06 Marek Polacek <polacek@redhat.com>
6292 PR tree-optimization/84228
6293 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6295 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6298 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6299 emitted arch directives.
6300 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6301 __ARM_FEATURE_COPROC before changing architectures.
6303 2018-02-06 Richard Biener <rguenther@suse.de>
6305 * config/i386/i386.c (print_reg): Fix typo.
6306 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6308 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6310 * configure: Regenerate.
6312 2018-02-05 Martin Sebor <msebor@redhat.com>
6314 PR tree-optimization/83369
6315 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6318 2018-02-05 Martin Liska <mliska@suse.cz>
6320 * doc/invoke.texi: Cherry-pick upstream r323995.
6322 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6324 * ira.c (ira_init_register_move_cost): Adjust comment.
6326 2018-02-05 Martin Liska <mliska@suse.cz>
6328 PR gcov-profile/84137
6329 * doc/gcov.texi: Fix typo in documentation.
6331 2018-02-05 Martin Liska <mliska@suse.cz>
6333 PR gcov-profile/83879
6334 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6335 using dlopen functionality.
6337 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6339 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6340 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6341 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6342 _mm_maskz_range_round_ss): New intrinsics.
6343 (__builtin_ia32_rangesd128_round)
6344 (__builtin_ia32_rangess128_round): Remove.
6345 (__builtin_ia32_rangesd128_mask_round,
6346 __builtin_ia32_rangess128_mask_round): New builtins.
6347 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6348 __builtin_ia32_rangess128_round): Remove.
6349 (__builtin_ia32_rangesd128_mask_round,
6350 __builtin_ia32_rangess128_mask_round): New builtins.
6351 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6352 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6353 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6354 "<round_saeonly_constraint>")): Changed to ...
6355 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6356 "<round_saeonly_scalar_constraint>")): ... this.
6357 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6358 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6359 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6360 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6361 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6363 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6365 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6367 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6368 Remove all values except native, 8540 and 8548.
6370 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6372 * config/i386/i386.c (ix86_output_function_return): Pass
6373 INVALID_REGNUM, instead of -1, as invalid register number to
6374 indirect_thunk_name and output_indirect_thunk.
6376 2018-02-02 Julia Koval <julia.koval@intel.com>
6378 * config.gcc: Add -march=icelake.
6379 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6380 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6381 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6382 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6383 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6384 (processor_target_table): Add icelake.
6385 (ix86_option_override_internal): Handle new PTAs.
6386 (get_builtin_code_for_version): Handle icelake.
6387 (M_INTEL_COREI7_ICELAKE): New.
6388 (fold_builtin_cpu): Handle icelake.
6389 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6390 * doc/invoke.texi: Add -march=icelake.
6392 2018-02-02 Julia Koval <julia.koval@intel.com>
6394 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6395 to wide_int_bitmask.
6396 * wide-int-bitmask.h: New.
6398 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6401 * config/i386/i386.md: Replace Pmode with word_mode in
6402 builtin_setjmp_setup and builtin_longjmp to support x32.
6404 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6408 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6410 (rs6000_supported_cpu_names): New static variable.
6411 (linux_cpu_translation_table): Likewise.
6412 (elf_platform) <cpu>: Define new static variable and use it.
6413 Translate kernel AT_PLATFORM name to canonical name if needed.
6414 Error if platform name is unknown.
6416 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6419 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6421 2018-02-01 Jeff Law <law@redhat.com>
6424 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6425 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6426 the scratch if RELEASE_VIA_POP is false.
6427 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6428 If we have to save a temporary register, decrement SIZE appropriately.
6429 Pass new arguments to release_scratch_register_on_entry.
6430 (ix86_adjust_stack_and_probe): Likewise.
6431 (ix86_emit_probe_stack_range): Pass new arguments to
6432 release_scratch_register_on_entry.
6434 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6436 PR rtl-optimization/84157
6437 * combine.c (change_zero_ext): Use REG_P predicate in
6438 front of HARD_REGISTER_P predicate.
6440 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6442 * config/avr/avr.c (avr_option_override): Move disabling of
6443 -fdelete-null-pointer-checks to...
6444 * common/config/avr/avr-common.c (avr_option_optimization_table):
6447 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6449 PR tree-optimization/81635
6450 * tree-data-ref.c (split_constant_offset_1): For types that
6451 wrap on overflow, try to use range info to prove that wrapping
6454 2018-02-01 Renlin Li <renlin.li@arm.com>
6457 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6459 (aarch64_register_move_cost): Likewise.
6460 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6462 (REG_CLASS_NAMES): Likewise.
6463 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6464 TAILCALL_ADDR_REGS. Remove IP registers.
6465 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6467 2018-02-01 Richard Biener <rguenther@suse.de>
6469 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6470 for specifying RPO order and allow NULL for that.
6471 * domwalk.c (dom_walker::dom_walker): Likewise.
6472 (dom_walker::walk): Handle NULL RPO order.
6473 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6475 (rewrite_update_dom_walker): Likewise.
6476 (mark_def_dom_walker): Likewise.
6478 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6480 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6481 (aarch64_maybe_expand_sve_subreg_move): Declare.
6482 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6483 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6485 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6486 that are semantically a reverse operation.
6487 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6488 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6489 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6491 (aarch64_can_change_mode_class): For big-endian, forbid changes
6492 between two SVE modes if they have different element sizes.
6494 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6496 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6497 the TImode handling for big-endian targets.
6499 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6501 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6502 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6504 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6505 Remove BSWAP handing for big-endian targets and use the form of
6506 LD1RQ appropariate for the mode.
6508 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6510 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6511 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6514 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6517 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6518 check for operands that need to go through aarch64_sve_reload_be.
6520 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6522 PR tree-optimization/81661
6523 PR tree-optimization/84117
6524 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6525 * tree-eh.c: Include gimplify.h.
6526 (find_trapping_overflow, replace_trapping_overflow,
6527 rewrite_to_non_trapping_overflow): New functions.
6528 * tree-vect-loop.c: Include tree-eh.h.
6529 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6530 * tree-data-ref.c: Include tree-eh.h.
6531 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6533 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6535 PR rtl-optimization/84123
6536 * combine.c (change_zero_ext): Check if hard register satisfies
6537 can_change_dest_mode before calling gen_lowpart_SUBREG.
6539 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6542 * ira.c (ira_init_register_move_cost): Remove assert.
6544 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6546 PR rtl-optimization/84071
6547 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6548 * doc/tm.texi: Regenerate.
6550 2018-01-31 Richard Biener <rguenther@suse.de>
6552 PR tree-optimization/84132
6553 * tree-data-ref.c (analyze_miv_subscript): Properly
6554 check whether evolution_function_is_affine_multivariate_p
6555 before calling gcd_of_steps_may_divide_p.
6557 2018-01-31 Julia Koval <julia.koval@intel.com>
6560 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6561 * config/i386/i386.md (rdpid_rex64) New.
6562 (rdpid): Make 32bit only.
6564 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6567 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6568 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6570 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6573 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6575 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6577 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6579 PR rtl-optimization/84071
6580 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6581 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6583 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6585 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6586 (arc_attribute_table): Add 'aux' attribute.
6587 (arc_in_small_data_p): Consider aux like variables.
6588 (arc_is_aux_reg_p): New function.
6589 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6590 (arc_get_aux_arg): New function.
6591 (prepare_move_operands): Handle aux-register access.
6592 (arc_handle_aux_attribute): New function.
6593 * doc/extend.texi (ARC Variable attributes): Add subsection.
6595 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6597 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6598 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6599 (arc_attribute_table): Add 'uncached' attribute.
6600 (arc_print_operand): Print '.di' flag for uncached memory
6602 (arc_in_small_data_p): Do not consider for small data the uncached
6604 (arc_is_uncached_mem_p): New function.
6605 * config/arc/predicates.md (compact_store_memory_operand): Check
6606 for uncached memory accesses.
6607 (nonvol_nonimm_operand): Likewise.
6608 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6610 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6613 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6614 falign-loops=): Add Optimization flag.
6616 2018-01-30 Jeff Law <law@redhat.com>
6619 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6620 INT_REGISTERS_SAVED. Check it prior to calling
6621 get_scratch_register_on_entry.
6622 (ix86_adjust_stack_and_probe): Similarly.
6623 (ix86_emit_probe_stack_range): Similarly.
6624 (ix86_expand_prologue): Corresponding changes.
6626 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6629 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6630 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6632 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6635 * lra-constraints.c (curr_insn_transform): Process AND in the
6638 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6640 PR rtl-optimization/83986
6641 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6642 dependence against last_pending_memory_flush in addition to
6645 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6647 PR tree-optimization/81611
6648 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6651 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6654 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6657 2018-01-30 Richard Biener <rguenther@suse.de>
6658 Jakub Jelinek <jakub@redhat.com>
6660 PR tree-optimization/84111
6661 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6662 inner loops added during recursion, as they don't have up-to-date
6665 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6668 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6669 (can_inline_edge_by_limits_p): ... here.
6670 (can_early_inline_edge_p, check_callers,
6671 update_caller_keys, update_callee_keys, recursive_inlining,
6672 add_new_edges_to_heap, speculation_useful_p,
6673 inline_small_functions,
6674 inline_small_functions, flatten_function,
6675 inline_to_all_callers_1): Update.
6677 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6679 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6682 2018-01-30 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/83008
6685 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6686 invariant and constant vector uses in stmts when they need
6689 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6692 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6693 * configure: Regenerate.
6695 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6697 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6699 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6700 Use gen_rtx_REG rather than gen_lowpart.
6702 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6704 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6705 rather than 0 when creating partial subregs.
6707 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6709 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6712 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6715 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6716 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6717 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6718 flags. This restores the settings used before the 2017-07-24.
6719 Turning off pre increment/decrement/modify allows IVOPTS to
6720 optimize DF/SF loops where the index is an int.
6722 2018-01-29 Richard Biener <rguenther@suse.de>
6723 Kelvin Nilsen <kelvin@gcc.gnu.org>
6726 * tree-vect-stmts.c (vectorizable_call): Don't call
6727 targetm.vectorize_builtin_md_vectorized_function if callee is
6730 2018-01-22 Carl Love <cel@us.ibm.com>
6732 * doc/extend.tex: Fix typo in second arg in
6733 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6735 2018-01-29 Richard Biener <rguenther@suse.de>
6737 PR tree-optimization/84086
6738 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6739 (flush_ssaname_freelist): When SSA names were released reset
6740 the SCEV hash table.
6742 2018-01-29 Richard Biener <rguenther@suse.de>
6744 PR tree-optimization/84057
6745 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6746 removed paths when removing edges.
6748 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6750 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6751 -mfunction-return=@var{choice}.
6753 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6756 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6758 (layout::print_source_line): Likewise.
6759 (test_get_line_width_without_trailing_whitespace): Add test cases.
6761 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6764 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6767 2018-01-26 Jim Wilson <jimw@sifive.com>
6769 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6771 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6774 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6776 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6777 and CMP + SUB-immediate -> SUBS.
6779 2018-01-26 Martin Sebor <msebor@redhat.com>
6781 PR tree-optimization/83896
6782 * tree-ssa-strlen.c (get_string_len): Rename...
6783 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6784 Avoid assuming length is constant.
6785 (handle_char_store): Use HOST_WIDE_INT for string length.
6787 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6790 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6791 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6793 2018-01-26 Richard Biener <rguenther@suse.de>
6795 PR rtl-optimization/84003
6796 * dse.c (record_store): Only record redundant stores when
6797 the earlier store aliases at least all accesses the later one does.
6799 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6801 PR rtl-optimization/83985
6802 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6803 REG_CFA_RESTORE insns.
6804 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6805 REG_CFA_RESTORE insns here.
6808 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6809 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6811 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6813 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6814 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6815 (arc_init): Likewise.
6816 (arc_override_options): Likewise.
6817 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6819 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6821 * config/arc/arc.h (TARGET_DBNZ): Define.
6822 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6823 properly set the tune attribute.
6824 (dbnz): Use TARGET_DBNZ guard.
6825 * config/arc/arc.opt (mtune): Add core3 option.
6827 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6829 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6830 recognize new pic like addresses.
6831 (arc_delegitimize_address): Clean up.
6833 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6835 * config/arc/arc-arches.def: Option mrf16 valid for all
6837 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6838 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6839 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6840 * config/arc/arc-tables.opt: Regenerate.
6841 * config/arc/arc.c (arc_conditional_register_usage): Handle
6842 reduced register file case.
6843 (arc_file_start): Set must have build attributes.
6844 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6846 * config/arc/arc.opt (mrf16): Add new option.
6847 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6848 * config/arc/genmultilib.awk: Handle new mrf16 option.
6849 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6850 * config/arc/t-multilib: Regenerate.
6851 * doc/invoke.texi (ARC Options): Document mrf16 option.
6853 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6855 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6856 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6857 (arc_attribute_table): Add 'secure_call' attribute.
6858 (arc_print_operand): Print secure call operand.
6859 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6861 (arc_is_secure_call_p): New function. * config/arc/arc.md
6862 (call_i): Add support for sjli instruction.
6863 (call_value_i): Likewise.
6864 * config/arc/constraints.md (Csc): New constraint.
6866 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6867 John Eric Martin <John.Martin@emmicro-us.com>
6869 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6870 * config/arc/arc.c (_arc_jli_section): New struct.
6871 (arc_jli_section): New type.
6872 (rc_jli_sections): New static variable.
6873 (arc_handle_jli_attribute): New function.
6874 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6875 (arc_file_end): New function.
6876 (TARGET_ASM_FILE_END): Define.
6877 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6878 (arc_add_jli_section): New function.
6879 (jli_call_scan): Likewise.
6880 (arc_reorg): Call jli_call_scan.
6881 (arc_output_addsi): Remove 'S' from printing asm operand.
6882 (arc_is_jli_call_p): New function.
6883 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6885 (movhi_insn): Likewise.
6886 (movsi_insn): Likewise.
6887 (movsi_set_cc_insn): Likewise.
6888 (loadqi_update): Likewise.
6889 (load_zeroextendqisi_update): Likewise.
6890 (load_signextendqisi_update): Likewise.
6891 (loadhi_update): Likewise.
6892 (load_zeroextendhisi_update): Likewise.
6893 (load_signextendhisi_update): Likewise.
6894 (loadsi_update): Likewise.
6895 (loadsf_update): Likewise.
6896 (movsicc_insn): Likewise.
6897 (bset_insn): Likewise.
6898 (bxor_insn): Likewise.
6899 (bclr_insn): Likewise.
6900 (bmsk_insn): Likewise.
6901 (bicsi3_insn): Likewise.
6902 (cmpsi_cc_c_insn): Likewise.
6903 (movsi_ne): Likewise.
6904 (movsi_cond_exec): Likewise.
6905 (clrsbsi2): Likewise.
6916 (call_i): Remove 'S' asm letter, add jli instruction.
6917 (call_value_i): Likewise.
6918 * config/arc/arc.op (mjli-always): New option.
6919 * config/arc/constraints.md (Cji): New constraint.
6920 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6922 (subsf3_fpx): Likewise.
6923 (mulsf3_fpx): Likewise.
6924 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6926 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6928 * doc/invoke.texi (ARC): Document mjli-always option.
6930 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6932 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6933 avoid addition with 0 and use incw and decw where possible.
6935 2018-01-26 Richard Biener <rguenther@suse.de>
6937 PR tree-optimization/81082
6938 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6939 association if it requires casting to unsigned.
6940 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6941 from fold_plusminus_mult_expr to catch important cases late when
6942 range info is available.
6944 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6946 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6947 * configure.ac (hidden_linkonce): New test.
6948 * configure: Regenerate.
6949 * config.in: Regenerate.
6951 2018-01-26 Julia Koval <julia.koval@intel.com>
6953 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6954 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6955 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6956 _mm_mask_bitshuffle_epi64_mask): Fix type.
6957 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6958 USI_FTYPE_V4DI_V4DI_USI): Remove.
6959 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6960 __builtin_ia32_vpshufbitqmb256_mask,
6961 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6962 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6963 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6965 2018-01-26 Alan Modra <amodra@gmail.com>
6968 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6969 UNSPEC_VBPERMQ. Sort other unspecs.
6971 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6973 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6975 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6978 * predict.c (drop_profile): Do not push/pop cfun; update also
6980 (handle_missing_profiles): Fix logic looking for zero profiles.
6982 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6985 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6986 on functions with #pragma omp declare simd or functions with simd
6988 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6989 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6990 Remove trailing \n from warning_at calls.
6992 2018-01-25 Tom de Vries <tom@codesourcery.com>
6995 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6996 for neutered workers.
6998 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7001 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7002 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7004 2018-01-24 Jeff Law <law@redhat.com>
7007 * i386.c (get_probe_interval): Move to earlier point.
7008 (ix86_compute_frame_layout): If -fstack-clash-protection and
7009 the frame is larger than the probe interval, then use pushes
7010 to save registers rather than reg->mem moves.
7011 (ix86_expand_prologue): Remove conditional for int_registers_saved
7014 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7017 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7018 min/max for never referenced object.
7020 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7023 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7025 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7026 attributes from DECL_ATTRIBUTES (decl) without affecting
7027 DECL_ATTRIBUTES (current_function_decl).
7028 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7029 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7031 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7033 PR tree-optimization/83979
7034 * fold-const.c (fold_comparison): Use constant_boolean_node
7035 instead of boolean_{true,false}_node.
7037 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7039 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7042 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7044 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7045 Simplify the clause that sets the length attribute.
7046 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7047 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7048 clause that sets the length attribute.
7049 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7051 2018-01-24 Tom de Vries <tom@codesourcery.com>
7054 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7055 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7056 Add strict parameter.
7057 (prevent_branch_around_nothing): Insert dummy insn between branch to
7058 label and label with no ptx insn inbetween.
7059 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7061 2018-01-24 Tom de Vries <tom@codesourcery.com>
7064 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7065 for neutered threads in warp.
7066 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7068 2018-01-24 Richard Biener <rguenther@suse.de>
7070 PR tree-optimization/83176
7071 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7074 2018-01-24 Richard Biener <rguenther@suse.de>
7076 PR tree-optimization/82819
7077 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7078 code generating pluses that are no-ops in the target precision.
7080 2018-01-24 Richard Biener <rguenther@suse.de>
7083 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7085 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7087 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7088 to merge probabilities.
7089 * predict.c (probably_never_executed): Also mark as cold functions
7090 with global 0 profile and guessed local profile.
7091 * profile-count.c (profile_probability::combine_with_count): New
7093 * profile-count.h (profile_probability::operator*,
7094 profile_probability::operator*=, profile_probability::operator/,
7095 profile_probability::operator/=): Reduce precision to adjusted
7096 and set value to guessed on contradictory divisions.
7097 (profile_probability::combine_with_freq): Remove.
7098 (profile_probability::combine_wiht_count): Declare.
7099 (profile_count::force_nonzero):: Set to adjusted.
7100 (profile_count::probability_in):: Set quality to adjusted.
7101 * tree-ssa-tail-merge.c (replace_block_by): Use
7104 2018-01-23 Andrew Waterman <andrew@sifive.com>
7105 Jim Wilson <jimw@sifive.com>
7107 * config/riscv/riscv.c (riscv_stack_boundary): New.
7108 (riscv_option_override): Set riscv_stack_boundary. Handle
7109 riscv_preferred_stack_boundary_arg.
7110 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7111 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7112 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7113 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7114 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7115 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7117 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7120 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7121 of struct ix86_frame.
7122 (ix86_expand_epilogue): Likewise. Add a local variable for
7123 the reg_save_offset field in struct ix86_frame.
7125 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7127 PR tree-optimization/82604
7128 * tree-loop-distribution.c (enum partition_kind): New enum item
7129 PKIND_PARTIAL_MEMSET.
7130 (partition_builtin_p): Support above new enum item.
7131 (generate_code_for_partition): Ditto.
7132 (compute_access_range): Differentiate cases that equality can be
7133 proven at all loops, the innermost loops or no loops.
7134 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7135 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7136 (finalize_partitions, distribute_loop): Don't fuse partition of
7137 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7138 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7141 2018-01-23 Martin Liska <mliska@suse.cz>
7143 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7144 order to ignore the predictor.
7145 (PRED_POLYMORPHIC_CALL): Likewise.
7146 (PRED_RECURSIVE_CALL): Likewise.
7148 2018-01-23 Martin Liska <mliska@suse.cz>
7150 * tree-profile.c (tree_profiling): Print function header to
7151 aware reader which function we are working on.
7152 * value-prof.c (gimple_find_values_to_profile): Do not print
7153 not interesting value histograms.
7155 2018-01-23 Martin Liska <mliska@suse.cz>
7157 * profile-count.h (enum profile_quality): Add
7158 profile_uninitialized as the first value. Do not number values
7159 as they are zero based.
7160 (profile_count::verify): Update sanity check.
7161 (profile_probability::verify): Likewise.
7163 2018-01-23 Nathan Sidwell <nathan@acm.org>
7165 * doc/invoke.texi (ffor-scope): Deprecate.
7167 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7169 PR tree-optimization/83510
7170 * domwalk.c (set_all_edges_as_executable): New function.
7171 (dom_walker::dom_walker): Convert bool param
7172 "skip_unreachable_blocks" to enum reachability. Move setup of
7173 edge flags to set_all_edges_as_executable and only do it when
7174 reachability is REACHABLE_BLOCKS.
7175 * domwalk.h (enum dom_walker::reachability): New enum.
7176 (dom_walker::dom_walker): Convert bool param
7177 "skip_unreachable_blocks" to enum reachability.
7178 (set_all_edges_as_executable): New decl.
7179 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7180 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7182 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7183 but converting true to REACHABLE_BLOCKS.
7184 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7186 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7187 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7188 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7190 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7191 if check_all_array_refs will be called.
7193 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7195 * tree.c (selftest::test_location_wrappers): Add more test
7198 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7200 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7201 (selftest::test_bit_in_range): Likewise.
7203 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7206 * doc/sourcebuild.texi (vect_float): Say that the selector
7207 only describes the situation when -funsafe-math-optimizations is on.
7208 (vect_float_strict): Document.
7210 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7212 PR tree-optimization/83965
7213 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7214 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7215 instead of checking only for a reduction.
7216 (vect_recog_widen_sum_pattern): Likewise.
7218 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7220 * predict.c (probably_never_executed): Only use precise profile info.
7221 (compute_function_frequency): Skip after inlining hack since we now
7222 have quality checking.
7224 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7226 * profile-count.h (profile_probability::very_unlikely,
7227 profile_probability::unlikely, profile_probability::even): Set
7228 precision to guessed.
7230 2018-01-23 Richard Biener <rguenther@suse.de>
7232 PR tree-optimization/83963
7233 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7234 Properly terminate dominator walk when crossing the exit edge not
7235 when visiting its source block.
7237 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7240 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7241 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7243 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7245 PR tree-optimization/83957
7246 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7247 semicolon after for body surrounded by braces.
7249 PR tree-optimization/83081
7250 * profile-count.h (profile_probability::split): New method.
7251 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7252 Use profile_probability::split.
7253 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7254 when splitting a single conditional jump into 2.
7256 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7258 PR tree-optimization/69452
7259 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7262 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7264 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7265 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7266 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7268 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7270 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7271 * config/rl78/rl78.md (movdi): New define_expand.
7272 * config/rl78/rl78.c (rl78_split_movdi): New function.
7274 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7277 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7279 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7280 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7281 128-bit to produce an UNSPEC move to get the double word with the
7282 signbit and then a shift directly to do signbit.
7283 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7284 implementation with a new version that just does either a direct
7285 move or a regular move. Move memory interface to separate insns.
7286 Move insns so they are next to the expander.
7287 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7288 with signbit move. Split big and little endian case.
7289 (signbit<mode>2_dm_mem_le): Likewise.
7290 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7291 (signbit<mode>2_dm2): Likewise.
7293 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7295 * config/rl78/rl78.md (anddi3): New define_expand.
7297 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7299 * config/rl78/rl78.md (umindi3): New define_expand.
7301 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7303 * config/rl78/rl78.md (smindi3): New define_expand.
7305 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7307 * config/rl78/rl78.md (smaxdi3): New define_expand.
7309 2018-01-22 Carl Love <cel@us.ibm.com>
7311 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7312 LVX_V1TI): Add macro expansion.
7313 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7314 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7315 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7316 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7317 Change check to determine if the instruction is a byte reversing
7318 entry. Fix typo in comment.
7319 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7320 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7321 Add def_builtin calls for new builtins.
7322 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7323 Add define_insn expansion.
7325 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7327 * config/rl78/rl78.md (umaxdi3): New define_expand.
7329 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7331 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7332 for non-QImode registers.
7334 2018-01-22 Richard Biener <rguenther@suse.de>
7336 PR tree-optimization/83963
7337 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7338 including the loop exit block.
7339 (scop_detection::merge_sese): Likewise.
7340 (scop_detection::add_scop): Do it here instead.
7342 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7344 * doc/sourcebuild.texi (arm_softfloat): Document.
7346 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7349 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7350 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7351 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7353 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7354 David Edelsohn <dje.gcc@gmail.com>
7357 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7358 Change "crset eq" to "crset 2".
7359 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7360 (*call_indirect_aix<mode>_nospec): Likewise.
7361 (*call_value_indirect_aix<mode>_nospec): Likewise.
7362 (*call_indirect_elfv2<mode>_nospec): Likewise.
7363 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7364 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7365 change assembly output from . to $.
7366 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7367 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7368 (*tablejump<mode>_internal1_nospec): Likewise.
7370 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7373 * config/sh/sh_optimize_sett_clrt.cc:
7374 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7376 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7378 PR tree-optimization/83940
7379 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7380 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7381 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7382 use it to pass back the definition type.
7383 (vect_check_store_rhs): Likewise rhs_dt_out.
7384 (vect_build_gather_load_calls): Add a mask_dt argument and use
7385 it instead of a call to vect_is_simple_use.
7386 (vectorizable_store): Update calls to vect_check_load_store_mask
7387 and vect_check_store_rhs. Use the dt returned by the latter instead
7388 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7389 instead of calls to vect_is_simple_use. Pass the scalar rather
7390 than the vector operand to vect_is_simple_use when handling
7391 second and subsequent copies of an rhs value.
7392 (vectorizable_load): Update calls to vect_check_load_store_mask
7393 and vect_build_gather_load_calls. Use the cached mask_dt and
7394 gs_info.offset_dt instead of calls to vect_is_simple_use.
7396 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7399 * tree-emutls.c: Include gimplify.h.
7400 (lower_emutls_2): New function.
7401 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7402 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7403 it before further processing.
7406 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7407 UINTVAL (trueop1) instead of INTVAL (op1).
7409 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7413 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7414 INCOMING_FRAME_SP_OFFSET if not defined.
7415 (scan_trace): Add ENTRY argument. If true and
7416 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7417 emit a note to adjust the CFA offset.
7418 (create_cfi_notes): Adjust scan_trace callers.
7419 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7420 INCOMING_FRAME_SP_OFFSET in the CIE.
7421 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7422 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7424 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7425 * doc/tm.texi: Regenerated.
7427 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7429 PR rtl-optimization/83147
7430 * lra-constraints.c (remove_inheritance_pseudos): Use
7431 lra_substitute_pseudo_within_insn.
7433 2018-01-19 Tom de Vries <tom@codesourcery.com>
7434 Cesar Philippidis <cesar@codesourcery.com>
7437 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7439 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7442 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7443 spaces for function labels.
7445 2018-01-19 Martin Liska <mliska@suse.cz>
7447 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7448 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7449 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7450 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7451 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7452 (PRED_CONST_RETURN): Change from 69 to 65.
7453 (PRED_NULL_RETURN): Change from 91 to 71.
7454 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7455 (PRED_LOOP_GUARD): Change from 66 to 73.
7457 2018-01-19 Martin Liska <mliska@suse.cz>
7459 * predict.c (predict_insn_def): Add new assert.
7460 (struct branch_predictor): Change type to signed integer.
7461 (test_prediction_value_range): Amend test to cover
7463 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7464 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7465 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7466 (PRED_LOOP_IV_COMPARE): Likewise.
7467 * predict.h (PROB_UNINITIALIZED): Define new constant.
7469 2018-01-19 Martin Liska <mliska@suse.cz>
7471 * predict.c (dump_prediction): Add new format for
7472 analyze_brprob.py script which is enabled with -details
7474 * profile-count.h (precise_p): New function.
7476 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7478 PR tree-optimization/83922
7479 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7480 there are no statements that need masking.
7481 (vect_active_double_reduction_p): New function.
7482 (vect_analyze_loop_operations): Use it when handling phis that
7483 are not in the loop header.
7485 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7487 PR tree-optimization/83914
7488 * tree-vect-loop.c (vectorizable_induction): Don't convert
7489 init_expr or apply the peeling adjustment for inductions
7490 that are nested within the vectorized loop.
7492 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7494 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7497 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7501 * function.h (gimplify_parameters): Add gimple_seq * argument.
7502 * function.c: Include gimple.h and options.h.
7503 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7504 for the added local temporaries if needed.
7505 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7506 if there are any parameter cleanups, wrap whole body into a
7507 try/finally with the cleanups.
7509 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7512 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7513 Use GET_MODE_CLASS for scalar floating point.
7515 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7519 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7520 Fix call of call_cgraph_insertion_hooks.
7522 2018-01-18 Martin Sebor <msebor@redhat.com>
7524 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7526 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7529 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7532 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7535 * common.opt: (-ffile-prefix-map): New option.
7536 * opts.c (common_handle_option): Defer it.
7537 * opts-global.c (handle_common_deferred_options): Handle it.
7538 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7539 * file-prefix-map.h: New file.
7540 (remap_debug_filename, add_debug_prefix_map): ...here.
7541 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7542 * final.c (debug_prefix_map, add_debug_prefix_map
7543 remap_debug_filename): Move to...
7544 * file-prefix-map.c: New file.
7545 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7546 generalize, get rid of alloca(), use strrchr() instead of strchr().
7547 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7548 Implement in terms of add_prefix_map().
7549 (remap_macro_filename, remap_debug_filename): Implement in term of
7551 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7552 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7553 * dbxout.c: Include file-prefix-map.h.
7554 * varasm.c: Likewise.
7555 * vmsdbgout.c: Likewise.
7556 * xcoffout.c: Likewise.
7557 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7558 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7559 * doc/invoke.texi (-ffile-prefix-map): Document.
7560 (-fdebug-prefix-map): Update description.
7562 2018-01-18 Martin Liska <mliska@suse.cz>
7564 * config/i386/i386.c (indirect_thunk_name): Document that also
7566 (output_indirect_thunk): Document why both instructions
7567 (pause and lfence) are generated.
7569 2018-01-18 Richard Biener <rguenther@suse.de>
7571 PR tree-optimization/83887
7572 * graphite-scop-detection.c
7573 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7574 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7575 (scop_detection::merge_sese): Re-implement with a flood-fill
7576 algorithm that properly finds a SESE region if it exists.
7578 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7581 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7582 pointer_diff optimizations use view_convert instead of convert.
7584 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7586 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7587 Generate different code for -mno-speculate-indirect-jumps.
7588 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7589 (*call_indirect_aix<mode>): Disable for
7590 -mno-speculate-indirect-jumps.
7591 (*call_indirect_aix<mode>_nospec): New define_insn.
7592 (*call_value_indirect_aix<mode>): Disable for
7593 -mno-speculate-indirect-jumps.
7594 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7595 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7596 -mno-speculate-indirect-jumps.
7597 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7599 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7601 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7602 long double type, set the flags for noting the default long double
7603 type, even if we don't pass or return a long double type.
7605 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7608 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7611 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7613 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7614 support for merge[hl].
7615 (fold_mergehl_helper): New helper function.
7616 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7617 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7618 (altivec_vmrglw_direct): Add xxmrglw insn.
7620 2018-01-17 Andrew Waterman <andrew@sifive.com>
7622 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7623 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7625 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7628 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7629 call the lto_location_cache before reading the
7630 DECL_SOURCE_LOCATION of the types.
7632 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7633 Richard Sandiford <richard.sandiford@linaro.org>
7635 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7636 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7637 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7638 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7639 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7641 * config/aarch64/constraints.md (aarch64_movti_operand):
7643 * config/aarch64/predicates.md (Uti): Add new constraint.
7645 2018-01-17 Carl Love <cel@us.ibm.com>
7647 * config/rs6000/vsx.md (define_expand xl_len_r,
7648 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7649 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7651 (define_expand, define_insn): Move the shift left from the
7652 define_insn to the define_expand for lxvl and stxvl instructions.
7653 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7654 and XL_LEN_R definitions to PURE.
7656 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7658 * config/i386/i386.c (indirect_thunk_name): Declare regno
7659 as unsigned int. Compare regno with INVALID_REGNUM.
7660 (output_indirect_thunk): Ditto.
7661 (output_indirect_thunk_function): Ditto.
7662 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7663 in the call to output_indirect_thunk_function.
7665 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7668 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7669 rather than the size of inner_type to determine the stack slot size
7670 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7672 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7675 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7678 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7680 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7681 endian Linux systems to optionally enable multilibs for selecting
7682 the long double type if the user configured an explicit type.
7683 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7684 have no long double multilibs if not defined.
7685 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7686 warn if the user used -mabi={ieee,ibm}longdouble and we built
7687 multilibs for long double.
7688 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7689 appropriate multilib option.
7690 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7692 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7693 for building long double multilibs.
7694 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7696 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7698 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7701 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7703 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7706 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7709 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7712 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7714 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7715 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7717 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7719 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7720 different rtl trees depending on TARGET_64BIT.
7721 (rs6000_gen_lvx): Likewise.
7723 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7725 * config/visium/visium.md (nop): Tweak comment.
7726 (hazard_nop): Likewise.
7728 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7730 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7731 -mspeculate-indirect-jumps.
7732 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7733 for -mno-speculate-indirect-jumps.
7734 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7735 (*call_value_indirect_elfv2<mode>): Disable for
7736 -mno-speculate-indirect-jumps.
7737 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7738 (indirect_jump): Emit different RTL for
7739 -mno-speculate-indirect-jumps.
7740 (*indirect_jump<mode>): Disable for
7741 -mno-speculate-indirect-jumps.
7742 (*indirect_jump<mode>_nospec): New define_insn.
7743 (tablejump): Emit different RTL for
7744 -mno-speculate-indirect-jumps.
7745 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7746 (tablejumpsi_nospec): New define_expand.
7747 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7748 (tablejumpdi_nospec): New define_expand.
7749 (*tablejump<mode>_internal1): Disable for
7750 -mno-speculate-indirect-jumps.
7751 (*tablejump<mode>_internal1_nospec): New define_insn.
7752 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7755 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7757 * caller-save.c (insert_save): Drop unnecessary parameter. All
7760 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7761 Richard Biener <rguenth@suse.de>
7764 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7765 return early, inline manually is_gimple_sizepos. Make sure if we
7766 call gimplify_expr we don't end up with a gimple constant.
7767 * tree.c (variably_modified_type_p): Don't return true for
7768 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7769 * gimplify.h (is_gimple_sizepos): Remove.
7771 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7773 PR tree-optimization/83857
7774 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7775 vectorizable_live_operation for pure SLP statements.
7776 (vectorizable_live_operation): Handle PHIs.
7778 2018-01-16 Richard Biener <rguenther@suse.de>
7780 PR tree-optimization/83867
7781 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7782 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7784 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7787 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7788 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7789 If off is not INTEGER_CST, issue a may not be aligned warning
7790 rather than isn't aligned. Use isn%'t rather than isn't.
7791 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7793 <case MULT_EXPR>: Improve the case when bottom and one of the
7794 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7795 operand, in that case check if the other operand is multiple of
7796 bottom divided by the INTEGER_CST operand.
7798 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7801 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7802 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7803 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7804 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7805 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7806 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7807 (pa_function_arg_size): New function.
7809 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7811 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7812 in a separate statement.
7814 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7816 PR tree-optimization/83847
7817 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7818 group gathers and scatters.
7820 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7822 PR rtl-optimization/86620
7823 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7825 PR rtl-optimization/83213
7826 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7827 to last if both are JUMP_INSNs.
7829 PR tree-optimization/83843
7830 * gimple-ssa-store-merging.c
7831 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7832 store_immediate_info for bswap/nop orig_stores.
7834 2018-01-15 Andrew Waterman <andrew@sifive.com>
7836 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7838 <UDIV>: Increase cost if !TARGET_DIV.
7840 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7842 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7843 (define_attr "cr_logical_3op"): New.
7844 (cceq_ior_compare): Adjust.
7845 (cceq_ior_compare_complement): Adjust.
7846 (*cceq_rev_compare): Adjust.
7847 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7848 (is_cracked_insn): Adjust.
7849 (insn_must_be_first_in_group): Adjust.
7850 * config/rs6000/40x.md: Adjust.
7851 * config/rs6000/440.md: Adjust.
7852 * config/rs6000/476.md: Adjust.
7853 * config/rs6000/601.md: Adjust.
7854 * config/rs6000/603.md: Adjust.
7855 * config/rs6000/6xx.md: Adjust.
7856 * config/rs6000/7450.md: Adjust.
7857 * config/rs6000/7xx.md: Adjust.
7858 * config/rs6000/8540.md: Adjust.
7859 * config/rs6000/cell.md: Adjust.
7860 * config/rs6000/e300c2c3.md: Adjust.
7861 * config/rs6000/e500mc.md: Adjust.
7862 * config/rs6000/e500mc64.md: Adjust.
7863 * config/rs6000/e5500.md: Adjust.
7864 * config/rs6000/e6500.md: Adjust.
7865 * config/rs6000/mpc.md: Adjust.
7866 * config/rs6000/power4.md: Adjust.
7867 * config/rs6000/power5.md: Adjust.
7868 * config/rs6000/power6.md: Adjust.
7869 * config/rs6000/power7.md: Adjust.
7870 * config/rs6000/power8.md: Adjust.
7871 * config/rs6000/power9.md: Adjust.
7872 * config/rs6000/rs64.md: Adjust.
7873 * config/rs6000/titan.md: Adjust.
7875 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7877 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7878 ix86_indirect_branch_register logic.
7880 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7882 * config/i386/constraints.md (Bs): Update
7883 ix86_indirect_branch_register check. Don't check
7884 ix86_indirect_branch_register with GOT_memory_operand.
7886 * config/i386/predicates.md (GOT_memory_operand): Don't check
7887 ix86_indirect_branch_register here.
7888 (GOT32_symbol_operand): Likewise.
7890 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7892 * config/i386/predicates.md (constant_call_address_operand):
7893 Rewrite ix86_indirect_branch_register logic.
7894 (sibcall_insn_operand): Likewise.
7896 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7898 * config/i386/constraints.md (Bs): Replace
7899 ix86_indirect_branch_thunk_register with
7900 ix86_indirect_branch_register.
7902 * config/i386/i386.md (indirect_jump): Likewise.
7903 (tablejump): Likewise.
7904 (*sibcall_memory): Likewise.
7905 (*sibcall_value_memory): Likewise.
7906 Peepholes of indirect call and jump via memory: Likewise.
7907 * config/i386/i386.opt: Likewise.
7908 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7909 (GOT_memory_operand): Likewise.
7910 (call_insn_operand): Likewise.
7911 (sibcall_insn_operand): Likewise.
7912 (GOT32_symbol_operand): Likewise.
7914 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7917 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7918 type rather than type addr's type points to.
7919 (expand_omp_atomic_mutex): Likewise.
7920 (expand_omp_atomic): Likewise.
7922 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7925 * config/i386/i386.c (output_indirect_thunk_function): Use
7926 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7927 for __x86_return_thunk.
7929 2018-01-15 Richard Biener <rguenther@suse.de>
7932 * expmed.c (extract_bit_field_1): Fix typo.
7934 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7937 * config/arm/iterators.md (VF): New mode iterator.
7938 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7939 Remove integer-related logic from pattern.
7940 (neon_vabd<mode>_3): Likewise.
7942 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7945 * common.opt (fstrict-overflow): No longer an alias.
7946 (fwrapv-pointer): New option.
7947 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7948 also for pointer types based on flag_wrapv_pointer.
7949 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7950 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7951 opts->x_flag_wrapv got set.
7952 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7953 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7954 POINTER_TYPE_OVERFLOW_UNDEFINED.
7955 * match.pd: Likewise in address comparison pattern.
7956 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7958 2018-01-15 Richard Biener <rguenther@suse.de>
7961 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7962 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7963 Reset type names to their identifier if their TYPE_DECL doesn't
7964 have linkage (and thus is used for ODR and devirt).
7965 (save_debug_info_for_decl): Remove.
7966 (save_debug_info_for_type): Likewise.
7967 (add_tree_to_fld_list): Adjust.
7968 * tree-pretty-print.c (dump_generic_node): Make dumping of
7969 type names more robust.
7971 2018-01-15 Richard Biener <rguenther@suse.de>
7973 * BASE-VER: Bump to 8.0.1.
7975 2018-01-14 Martin Sebor <msebor@redhat.com>
7978 * builtins.c (check_access): Avoid warning when the no-warning bit
7981 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7983 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7984 * ira-color (allocno_hard_regs_compare): Likewise.
7986 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7989 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7990 Use .pushsection/.popsection.
7992 2018-01-14 Martin Sebor <msebor@redhat.com>
7995 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7997 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7999 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8000 entry from extra_headers.
8001 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8002 extra_headers, make the list bitwise identical to the i?86-*-* one.
8004 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8006 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8007 -mcmodel=large with -mindirect-branch=thunk,
8008 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8009 -mfunction-return=thunk-extern.
8010 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8011 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8012 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8014 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8016 * config/i386/i386.c (print_reg): Print the name of the full
8017 integer register without '%'.
8018 (ix86_print_operand): Handle 'V'.
8019 * doc/extend.texi: Document 'V' modifier.
8021 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8023 * config/i386/constraints.md (Bs): Disallow memory operand for
8024 -mindirect-branch-register.
8026 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8027 (GOT_memory_operand): Likewise.
8028 (call_insn_operand): Likewise.
8029 (sibcall_insn_operand): Likewise.
8030 (GOT32_symbol_operand): Likewise.
8031 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8032 for -mindirect-branch-register.
8033 (tablejump): Likewise.
8034 (*sibcall_memory): Likewise.
8035 (*sibcall_value_memory): Likewise.
8036 Disallow peepholes of indirect call and jump via memory for
8037 -mindirect-branch-register.
8038 (*call_pop): Replace m with Bw.
8039 (*call_value_pop): Likewise.
8040 (*sibcall_pop_memory): Replace m with Bs.
8041 * config/i386/i386.opt (mindirect-branch-register): New option.
8042 * doc/invoke.texi: Document -mindirect-branch-register option.
8044 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8046 * config/i386/i386-protos.h (ix86_output_function_return): New.
8047 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8048 set function_return_type.
8049 (indirect_thunk_name): Add ret_p to indicate thunk for function
8051 (output_indirect_thunk_function): Pass false to
8052 indirect_thunk_name.
8053 (ix86_output_indirect_branch_via_reg): Likewise.
8054 (ix86_output_indirect_branch_via_push): Likewise.
8055 (output_indirect_thunk_function): Create alias for function
8056 return thunk if regno < 0.
8057 (ix86_output_function_return): New function.
8058 (ix86_handle_fndecl_attribute): Handle function_return.
8059 (ix86_attribute_table): Add function_return.
8060 * config/i386/i386.h (machine_function): Add
8061 function_return_type.
8062 * config/i386/i386.md (simple_return_internal): Use
8063 ix86_output_function_return.
8064 (simple_return_internal_long): Likewise.
8065 * config/i386/i386.opt (mfunction-return=): New option.
8066 (indirect_branch): Mention -mfunction-return=.
8067 * doc/extend.texi: Document function_return function attribute.
8068 * doc/invoke.texi: Document -mfunction-return= option.
8070 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8072 * config/i386/i386-opts.h (indirect_branch): New.
8073 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8074 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8075 with local indirect jump when converting indirect call and jump.
8076 (ix86_set_indirect_branch_type): New.
8077 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8078 (indirectlabelno): New.
8079 (indirect_thunk_needed): Likewise.
8080 (indirect_thunk_bnd_needed): Likewise.
8081 (indirect_thunks_used): Likewise.
8082 (indirect_thunks_bnd_used): Likewise.
8083 (INDIRECT_LABEL): Likewise.
8084 (indirect_thunk_name): Likewise.
8085 (output_indirect_thunk): Likewise.
8086 (output_indirect_thunk_function): Likewise.
8087 (ix86_output_indirect_branch_via_reg): Likewise.
8088 (ix86_output_indirect_branch_via_push): Likewise.
8089 (ix86_output_indirect_branch): Likewise.
8090 (ix86_output_indirect_jmp): Likewise.
8091 (ix86_code_end): Call output_indirect_thunk_function if needed.
8092 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8094 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8095 (ix86_attribute_table): Add indirect_branch.
8096 * config/i386/i386.h (machine_function): Add indirect_branch_type
8097 and has_local_indirect_jump.
8098 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8100 (tablejump): Likewise.
8101 (*indirect_jump): Use ix86_output_indirect_jmp.
8102 (*tablejump_1): Likewise.
8103 (simple_return_indirect_internal): Likewise.
8104 * config/i386/i386.opt (mindirect-branch=): New option.
8105 (indirect_branch): New.
8108 (thunk-inline): Likewise.
8109 (thunk-extern): Likewise.
8110 * doc/extend.texi: Document indirect_branch function attribute.
8111 * doc/invoke.texi: Document -mindirect-branch= option.
8113 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8116 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8118 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8120 * ipa-inline.c (want_inline_small_function_p): Return false if
8121 inlining has already failed with CIF_FINAL_ERROR.
8122 (update_caller_keys): Call want_inline_small_function_p before
8124 (update_callee_keys): Likewise.
8126 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8128 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8130 (rs6000_quadword_masked_address_p): Likewise.
8131 (quad_aligned_load_p): Likewise.
8132 (quad_aligned_store_p): Likewise.
8133 (const_load_sequence_p): Add comment to describe the outer-most loop.
8134 (mimic_memory_attributes_and_flags): New function.
8135 (rs6000_gen_stvx): Likewise.
8136 (replace_swapped_aligned_store): Likewise.
8137 (rs6000_gen_lvx): Likewise.
8138 (replace_swapped_aligned_load): Likewise.
8139 (replace_swapped_load_constant): Capitalize argument name in
8140 comment describing this function.
8141 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8142 and stores that access quad-word aligned addresses and replace
8143 with stvx or lvx instructions when appropriate.
8144 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8145 New function prototype.
8146 (rs6000_quadword_masked_address_p): Likewise.
8147 (rs6000_gen_lvx): Likewise.
8148 (rs6000_gen_stvx): Likewise.
8149 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8150 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8151 when memory address is aligned.
8152 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8153 this split to select lvx instruction when memory address is aligned.
8154 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8155 instruction when memory address is aligned.
8156 (*vsx_le_perm_load_v16qi): Likewise.
8157 (four unnamed splitters): Modify to select the stvx instruction
8158 when memory is aligned.
8160 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8162 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8163 which appears in the queue multiple times.
8165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8166 Alan Hayward <alan.hayward@arm.com>
8167 David Sherwood <david.sherwood@arm.com>
8169 * tree-vectorizer.h (vec_lower_bound): New structure.
8170 (_loop_vec_info): Add check_nonzero and lower_bounds.
8171 (LOOP_VINFO_CHECK_NONZERO): New macro.
8172 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8173 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8174 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8175 fields. Make seg_len the distance travelled, not including the
8177 (dr_direction_indicator): Declare.
8178 (dr_zero_step_indicator): Likewise.
8179 (dr_known_forward_stride_p): Likewise.
8180 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8182 (runtime_alias_check_p): Allow runtime alias checks with
8184 (operator ==): Compare access_size and align.
8185 (prune_runtime_alias_test_list): Rework for new distinction between
8186 the access_size and seg_len.
8187 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8189 (get_segment_min_max): New function.
8190 (create_intersect_range_checks): Use it.
8191 (dr_step_indicator): New function.
8192 (dr_direction_indicator): Likewise.
8193 (dr_zero_step_indicator): Likewise.
8194 (dr_known_forward_stride_p): Likewise.
8195 * tree-loop-distribution.c (data_ref_segment_size): Return
8196 DR_STEP * (niters - 1).
8197 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8199 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8200 (vect_preserves_scalar_order_p): New function, split out from...
8201 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8202 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8203 (vect_vfa_access_size): New function.
8204 (vect_vfa_align): Likewise.
8205 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8206 (dump_lower_bound): New function.
8207 (vect_check_lower_bound): Likewise.
8208 (vect_small_gap_p): Likewise.
8209 (vectorizable_with_step_bound_p): Likewise.
8210 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8211 depencies if the vectorization factor is 1. Convert the checks
8212 for nonzero steps into checks on the bounds of DR_STEP. Try using
8213 a bunds check for variable steps if the minimum required step is
8214 relatively small. Update calls to the dr_with_seg_len
8215 constructor and to vect_compile_time_alias.
8216 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8218 (vect_loop_versioning): Call it.
8219 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8221 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8224 Alan Hayward <alan.hayward@arm.com>
8225 David Sherwood <david.sherwood@arm.com>
8227 * doc/sourcebuild.texi (vect_scatter_store): Document.
8228 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8230 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8232 * genopinit.c (main): Add supports_vec_scatter_store and
8233 supports_vec_scatter_store_cached to target_optabs.
8234 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8235 IFN_MASK_SCATTER_STORE.
8236 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8238 * internal-fn.h (internal_store_fn_p): Declare.
8239 (internal_fn_stored_value_index): Likewise.
8240 * internal-fn.c (scatter_store_direct): New macro.
8241 (expand_scatter_store_optab_fn): New function.
8242 (direct_scatter_store_optab_supported_p): New macro.
8243 (internal_store_fn_p): New function.
8244 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8245 IFN_MASK_SCATTER_STORE.
8246 (internal_fn_mask_index): Likewise.
8247 (internal_fn_stored_value_index): New function.
8248 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8250 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8251 * optabs-query.c (supports_vec_scatter_store_p): New function.
8252 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8253 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8254 true for scatter stores.
8255 (vect_gather_scatter_fn_p): Handle scatter stores too.
8256 (vect_check_gather_scatter): Consider using scatter stores if
8257 supports_vec_scatter_store_p.
8258 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8260 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8261 internal_fn_stored_value_index.
8262 (check_load_store_masking): Handle scatter stores too.
8263 (vect_get_store_rhs): Make public.
8264 (vectorizable_call): Use internal_store_fn_p.
8265 (vectorizable_store): Handle scatter store internal functions.
8266 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8267 when deciding whether the end of the group has been reached.
8268 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8269 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8270 (mask_scatter_store<mode>): New insns.
8272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8273 Alan Hayward <alan.hayward@arm.com>
8274 David Sherwood <david.sherwood@arm.com>
8276 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8277 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8278 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8280 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8281 Use vect_truncate_gather_scatter_offset if we can't treat the
8282 operation as a normal gather load or scatter store.
8283 (get_group_load_store_type): Take the gather_scatter_info
8284 as argument. Try using a gather load or scatter store for
8285 single-element groups.
8286 (get_load_store_type): Update calls to get_group_load_store_type
8287 and vect_use_strided_gather_scatters_p.
8289 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8290 Alan Hayward <alan.hayward@arm.com>
8291 David Sherwood <david.sherwood@arm.com>
8293 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8294 optional tree argument.
8295 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8297 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8298 but continue to use the current value as a fallback.
8299 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8300 to compare the updates.
8301 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8302 (get_load_store_type): Use it when handling a strided access.
8303 (vect_get_strided_load_store_ops): New function.
8304 (vect_get_data_ptr_increment): Likewise.
8305 (vectorizable_load): Handle strided gather loads. Always pass
8306 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8309 Alan Hayward <alan.hayward@arm.com>
8310 David Sherwood <david.sherwood@arm.com>
8312 * doc/md.texi (gather_load@var{m}): Document.
8313 (mask_gather_load@var{m}): Likewise.
8314 * genopinit.c (main): Add supports_vec_gather_load and
8315 supports_vec_gather_load_cached to target_optabs.
8316 * optabs-tree.c (init_tree_optimization_optabs): Use
8317 ggc_cleared_alloc to allocate target_optabs.
8318 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8319 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8321 * internal-fn.h (internal_load_fn_p): Declare.
8322 (internal_gather_scatter_fn_p): Likewise.
8323 (internal_fn_mask_index): Likewise.
8324 (internal_gather_scatter_fn_supported_p): Likewise.
8325 * internal-fn.c (gather_load_direct): New macro.
8326 (expand_gather_load_optab_fn): New function.
8327 (direct_gather_load_optab_supported_p): New macro.
8328 (direct_internal_fn_optab): New function.
8329 (internal_load_fn_p): Likewise.
8330 (internal_gather_scatter_fn_p): Likewise.
8331 (internal_fn_mask_index): Likewise.
8332 (internal_gather_scatter_fn_supported_p): Likewise.
8333 * optabs-query.c (supports_at_least_one_mode_p): New function.
8334 (supports_vec_gather_load_p): Likewise.
8335 * optabs-query.h (supports_vec_gather_load_p): Declare.
8336 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8337 and memory_type field.
8338 (NUM_PATTERNS): Bump to 15.
8339 * tree-vect-data-refs.c: Include internal-fn.h.
8340 (vect_gather_scatter_fn_p): New function.
8341 (vect_describe_gather_scatter_call): Likewise.
8342 (vect_check_gather_scatter): Try using internal functions for
8343 gather loads. Recognize existing calls to a gather load function.
8344 (vect_analyze_data_refs): Consider using gather loads if
8345 supports_vec_gather_load_p.
8346 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8347 (vect_get_gather_scatter_offset_type): Likewise.
8348 (vect_convert_mask_for_vectype): Likewise.
8349 (vect_add_conversion_to_patterm): Likewise.
8350 (vect_try_gather_scatter_pattern): Likewise.
8351 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8352 (vect_vect_recog_func_ptrs): Add it.
8353 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8354 internal_fn_mask_index and internal_gather_scatter_fn_p.
8355 (check_load_store_masking): Take the gather_scatter_info as an
8356 argument and handle gather loads.
8357 (vect_get_gather_scatter_ops): New function.
8358 (vectorizable_call): Check internal_load_fn_p.
8359 (vectorizable_load): Likewise. Handle gather load internal
8361 (vectorizable_store): Update call to check_load_store_masking.
8362 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8363 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8364 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8365 (aarch64_gather_scale_operand_d): New predicates.
8366 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8367 (mask_gather_load<mode>): New insns.
8369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8370 Alan Hayward <alan.hayward@arm.com>
8371 David Sherwood <david.sherwood@arm.com>
8373 * optabs.def (fold_left_plus_optab): New optab.
8374 * doc/md.texi (fold_left_plus_@var{m}): Document.
8375 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8376 * internal-fn.c (fold_left_direct): Define.
8377 (expand_fold_left_optab_fn): Likewise.
8378 (direct_fold_left_optab_supported_p): Likewise.
8379 * fold-const-call.c (fold_const_fold_left): New function.
8380 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8381 * tree-parloops.c (valid_reduction_p): New function.
8382 (gather_scalar_reductions): Use it.
8383 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8384 (vect_finish_replace_stmt): Declare.
8385 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8386 (needs_fold_left_reduction_p): New function, split out from...
8387 (vect_is_simple_reduction): ...here. Accept reductions that
8388 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8389 (vect_force_simple_reduction): Also store the reduction type in
8390 the assignment's STMT_VINFO_REDUC_TYPE.
8391 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8392 (merge_with_identity): New function.
8393 (vect_expand_fold_left): Likewise.
8394 (vectorize_fold_left_reduction): Likewise.
8395 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8396 scalar phi in place for it. Check for target support and reject
8397 cases that would reassociate the operation. Defer the transform
8398 phase to vectorize_fold_left_reduction.
8399 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8400 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8401 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8405 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8406 call to ifc_temp_var.
8408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8409 Alan Hayward <alan.hayward@arm.com>
8410 David Sherwood <david.sherwood@arm.com>
8412 * target.def (legitimize_address_displacement): Take the original
8413 offset as a poly_int.
8414 * targhooks.h (default_legitimize_address_displacement): Update
8416 * targhooks.c (default_legitimize_address_displacement): Likewise.
8417 * doc/tm.texi: Regenerate.
8418 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8419 as an argument, moving assert of ad->disp == ad->disp_term to...
8420 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8421 Try calling targetm.legitimize_address_displacement before expanding
8422 the address rather than afterwards, and adjust for the new interface.
8423 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8424 Match the new hook interface. Handle SVE addresses.
8425 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8430 * Makefile.in (OBJS): Add early-remat.o.
8431 * target.def (select_early_remat_modes): New hook.
8432 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8433 * doc/tm.texi: Regenerate.
8434 * targhooks.h (default_select_early_remat_modes): Declare.
8435 * targhooks.c (default_select_early_remat_modes): New function.
8436 * timevar.def (TV_EARLY_REMAT): New timevar.
8437 * passes.def (pass_early_remat): New pass.
8438 * tree-pass.h (make_pass_early_remat): Declare.
8439 * early-remat.c: New file.
8440 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8442 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8445 Alan Hayward <alan.hayward@arm.com>
8446 David Sherwood <david.sherwood@arm.com>
8448 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8449 vfm1 with a bound_epilog parameter.
8450 (vect_do_peeling): Update calls accordingly, and move the prologue
8451 call earlier in the function. Treat the base bound_epilog as 0 for
8452 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8453 this base when peeling for gaps.
8454 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8455 with fully-masked loops.
8456 (vect_estimate_min_profitable_iters): Handle the single peeled
8457 iteration in that case.
8459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8460 Alan Hayward <alan.hayward@arm.com>
8461 David Sherwood <david.sherwood@arm.com>
8463 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8464 single-element interleaving even if the size is not a power of 2.
8465 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8466 accesses for single-element interleaving if the group size is
8469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8470 Alan Hayward <alan.hayward@arm.com>
8471 David Sherwood <david.sherwood@arm.com>
8473 * doc/md.texi (fold_extract_last_@var{m}): Document.
8474 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8475 * optabs.def (fold_extract_last_optab): New optab.
8476 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8477 * internal-fn.c (fold_extract_direct): New macro.
8478 (expand_fold_extract_optab_fn): Likewise.
8479 (direct_fold_extract_optab_supported_p): Likewise.
8480 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8481 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8482 EXTRACT_LAST_REDUCTION.
8483 (get_initial_def_for_reduction): Do not create an initial vector
8484 for EXTRACT_LAST_REDUCTION reductions.
8485 (vectorizable_reduction): Leave the scalar phi in place for
8486 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8487 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8488 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8489 transform phase to vectorizable_condition.
8490 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8492 (vect_finish_stmt_generation): ...here.
8493 (vect_finish_replace_stmt): New function.
8494 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8495 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8497 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8500 Alan Hayward <alan.hayward@arm.com>
8501 David Sherwood <david.sherwood@arm.com>
8503 * doc/md.texi (extract_last_@var{m}): Document.
8504 * optabs.def (extract_last_optab): New optab.
8505 * internal-fn.def (EXTRACT_LAST): New internal function.
8506 * internal-fn.c (cond_unary_direct): New macro.
8507 (expand_cond_unary_optab_fn): Likewise.
8508 (direct_cond_unary_optab_supported_p): Likewise.
8509 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8510 loops using EXTRACT_LAST.
8511 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8512 (extract_last_<mode>): ...this optab.
8513 (vec_extract<mode><Vel>): Update accordingly.
8515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8516 Alan Hayward <alan.hayward@arm.com>
8517 David Sherwood <david.sherwood@arm.com>
8519 * target.def (empty_mask_is_expensive): New hook.
8520 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8521 * doc/tm.texi: Regenerate.
8522 * targhooks.h (default_empty_mask_is_expensive): Declare.
8523 * targhooks.c (default_empty_mask_is_expensive): New function.
8524 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8525 if the target says that empty masks are expensive.
8526 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8528 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8530 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8531 Alan Hayward <alan.hayward@arm.com>
8532 David Sherwood <david.sherwood@arm.com>
8534 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8535 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8536 (vect_use_loop_mask_for_alignment_p): New function.
8537 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8538 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8539 niters_skip argument. Make sure that the first niters_skip elements
8540 of the first iteration are inactive.
8541 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8542 Update call to vect_set_loop_masks_directly.
8543 (get_misalign_in_elems): New function, split out from...
8544 (vect_gen_prolog_loop_niters): ...here.
8545 (vect_update_init_of_dr): Take a code argument that specifies whether
8546 the adjustment should be added or subtracted.
8547 (vect_update_init_of_drs): Likewise.
8548 (vect_prepare_for_masked_peels): New function.
8549 (vect_do_peeling): Skip prologue peeling if we're using a mask
8550 instead. Update call to vect_update_inits_of_drs.
8551 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8553 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8554 alignment. Do not include the number of peeled iterations in
8555 the minimum threshold in that case.
8556 (vectorizable_induction): Adjust the start value down by
8557 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8558 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8559 Take the number of skipped iterations into account when calculating
8561 * tree-vect-stmts.c (vect_gen_while_not): New function.
8563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8564 Alan Hayward <alan.hayward@arm.com>
8565 David Sherwood <david.sherwood@arm.com>
8567 * doc/sourcebuild.texi (vect_fully_masked): Document.
8568 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8570 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8572 (vect_analyze_loop_2): ...here. Don't check the vectorization
8573 factor against the number of loop iterations if the loop is
8576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8577 Alan Hayward <alan.hayward@arm.com>
8578 David Sherwood <david.sherwood@arm.com>
8580 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8581 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8582 (dump_groups): Update accordingly.
8583 (iv_use::mem_type): New member variable.
8584 (address_p): New function.
8585 (record_use): Add a mem_type argument and initialize the new
8587 (record_group_use): Add a mem_type argument. Use address_p.
8588 Remove obsolete null checks of base_object. Update call to record_use.
8589 (find_interesting_uses_op): Update call to record_group_use.
8590 (find_interesting_uses_cond): Likewise.
8591 (find_interesting_uses_address): Likewise.
8592 (get_mem_type_for_internal_fn): New function.
8593 (find_address_like_use): Likewise.
8594 (find_interesting_uses_stmt): Try find_address_like_use before
8595 calling find_interesting_uses_op.
8596 (addr_offset_valid_p): Use the iv mem_type field as the type
8597 of the addressed memory.
8598 (add_autoinc_candidates): Likewise.
8599 (get_address_cost): Likewise.
8600 (split_small_address_groups_p): Use address_p.
8601 (split_address_groups): Likewise.
8602 (add_iv_candidate_for_use): Likewise.
8603 (autoinc_possible_for_pair): Likewise.
8604 (rewrite_groups): Likewise.
8605 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8606 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8607 (get_alias_ptr_type_for_ptr_address): New function.
8608 (rewrite_use_address): Rewrite address uses in calls that were
8609 identified by find_address_like_use.
8611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8612 Alan Hayward <alan.hayward@arm.com>
8613 David Sherwood <david.sherwood@arm.com>
8615 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8617 * gimple-expr.h (is_gimple_addressable: Likewise.
8618 * gimple-expr.c (is_gimple_address): Likewise.
8619 * internal-fn.c (expand_call_mem_ref): New function.
8620 (expand_mask_load_optab_fn): Use it.
8621 (expand_mask_store_optab_fn): Likewise.
8623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8624 Alan Hayward <alan.hayward@arm.com>
8625 David Sherwood <david.sherwood@arm.com>
8627 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8628 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8629 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8630 (cond_umax@var{mode}): Document.
8631 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8632 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8633 (cond_umin_optab, cond_umax_optab): New optabs.
8634 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8635 (COND_IOR, COND_XOR): New internal functions.
8636 * internal-fn.h (get_conditional_internal_fn): Declare.
8637 * internal-fn.c (cond_binary_direct): New macro.
8638 (expand_cond_binary_optab_fn): Likewise.
8639 (direct_cond_binary_optab_supported_p): Likewise.
8640 (get_conditional_internal_fn): New function.
8641 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8642 Cope with reduction statements that are vectorized as calls rather
8644 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8645 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8646 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8647 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8648 (UNSPEC_COND_EOR): New unspecs.
8649 (optab): Add mappings for them.
8650 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8651 (sve_int_op, sve_fp_op): New int attributes.
8653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8654 Alan Hayward <alan.hayward@arm.com>
8655 David Sherwood <david.sherwood@arm.com>
8657 * optabs.def (while_ult_optab): New optab.
8658 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8659 * internal-fn.def (WHILE_ULT): New internal function.
8660 * internal-fn.h (direct_internal_fn_supported_p): New override
8661 that takes two types as argument.
8662 * internal-fn.c (while_direct): New macro.
8663 (expand_while_optab_fn): New function.
8664 (convert_optab_supported_p): Likewise.
8665 (direct_while_optab_supported_p): New macro.
8666 * wide-int.h (wi::udiv_ceil): New function.
8667 * tree-vectorizer.h (rgroup_masks): New structure.
8668 (vec_loop_masks): New typedef.
8669 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8671 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8672 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8673 (vect_max_vf): New function.
8674 (slpeel_make_loop_iterate_ntimes): Delete.
8675 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8676 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8677 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8678 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8679 internal-fn.h, stor-layout.h and optabs-query.h.
8680 (vect_set_loop_mask): New function.
8681 (add_preheader_seq): Likewise.
8682 (add_header_seq): Likewise.
8683 (interleave_supported_p): Likewise.
8684 (vect_maybe_permute_loop_masks): Likewise.
8685 (vect_set_loop_masks_directly): Likewise.
8686 (vect_set_loop_condition_masked): Likewise.
8687 (vect_set_loop_condition_unmasked): New function, split out from
8688 slpeel_make_loop_iterate_ntimes.
8689 (slpeel_make_loop_iterate_ntimes): Rename to..
8690 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8691 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8692 (vect_do_peeling): Update call accordingly.
8693 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8695 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8696 mask_compare_type, can_fully_mask_p and fully_masked_p.
8697 (release_vec_loop_masks): New function.
8698 (_loop_vec_info): Use it to free the loop masks.
8699 (can_produce_all_loop_masks_p): New function.
8700 (vect_get_max_nscalars_per_iter): Likewise.
8701 (vect_verify_full_masking): Likewise.
8702 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8703 retries, and free the mask rgroups before retrying. Check loop-wide
8704 reasons for disallowing fully-masked loops. Make the final decision
8705 about whether use a fully-masked loop or not.
8706 (vect_estimate_min_profitable_iters): Do not assume that peeling
8707 for the number of iterations will be needed for fully-masked loops.
8708 (vectorizable_reduction): Disable fully-masked loops.
8709 (vectorizable_live_operation): Likewise.
8710 (vect_halve_mask_nunits): New function.
8711 (vect_double_mask_nunits): Likewise.
8712 (vect_record_loop_mask): Likewise.
8713 (vect_get_loop_mask): Likewise.
8714 (vect_transform_loop): Handle the case in which the final loop
8715 iteration might handle a partial vector. Call vect_set_loop_condition
8716 instead of slpeel_make_loop_iterate_ntimes.
8717 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8718 (check_load_store_masking): New function.
8719 (prepare_load_store_mask): Likewise.
8720 (vectorizable_store): Handle fully-masked loops.
8721 (vectorizable_load): Likewise.
8722 (supportable_widening_operation): Use vect_halve_mask_nunits for
8724 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8725 (vect_gen_while): New function.
8726 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8727 (aarch64_uqdec<mode>): New insn.
8729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8730 Alan Hayward <alan.hayward@arm.com>
8731 David Sherwood <david.sherwood@arm.com>
8733 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8734 (reduc_xor_scal_optab): New optabs.
8735 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8736 (reduc_xor_scal_@var{m}): Document.
8737 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8738 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8740 * fold-const-call.c (fold_const_call): Handle them.
8741 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8742 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8743 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8744 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8745 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8746 (UNSPEC_XORV): New unspecs.
8747 (optab): Add entries for them.
8748 (BITWISEV): New int iterator.
8749 (bit_reduc_op): New int attributes.
8751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8752 Alan Hayward <alan.hayward@arm.com>
8753 David Sherwood <david.sherwood@arm.com>
8755 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8756 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8757 * optabs.def (vec_shl_insert_optab): New optab.
8758 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8759 (duplicate_and_interleave): Likewise.
8760 * tree-vect-loop.c: Include internal-fn.h.
8761 (neutral_op_for_slp_reduction): New function, split out from
8762 get_initial_defs_for_reduction.
8763 (get_initial_def_for_reduction): Handle option 2 for variable-length
8764 vectors by loading the neutral value into a vector and then shifting
8765 the initial value into element 0.
8766 (get_initial_defs_for_reduction): Replace the code argument with
8767 the neutral value calculated by neutral_op_for_slp_reduction.
8768 Use gimple_build_vector for constant-length vectors.
8769 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8770 but the first group_size elements have a neutral value.
8771 Use duplicate_and_interleave otherwise.
8772 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8773 Update call to get_initial_defs_for_reduction. Handle SLP
8774 reductions for variable-length vectors by creating one vector
8775 result for each scalar result, with the elements associated
8776 with other scalar results stubbed out with the neutral value.
8777 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8778 Require IFN_VEC_SHL_INSERT for double reductions on
8779 variable-length vectors, or SLP reductions that have
8780 a neutral value. Require can_duplicate_and_interleave_p
8781 support for variable-length unchained SLP reductions if there
8782 is no neutral value, such as for MIN/MAX reductions. Also require
8783 the number of vector elements to be a multiple of the number of
8784 SLP statements when doing variable-length unchained SLP reductions.
8785 Update call to vect_create_epilog_for_reduction.
8786 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8787 and remove initial values.
8788 (duplicate_and_interleave): Make public.
8789 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8790 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8793 Alan Hayward <alan.hayward@arm.com>
8794 David Sherwood <david.sherwood@arm.com>
8796 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8797 (can_duplicate_and_interleave_p): New function.
8798 (vect_get_and_check_slp_defs): Take the vector of statements
8799 rather than just the current one. Remove excess parentheses.
8800 Restriction rejectinon of vect_constant_def and vect_external_def
8801 for variable-length vectors to boolean types, or types for which
8802 can_duplicate_and_interleave_p is false.
8803 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8804 (duplicate_and_interleave): New function.
8805 (vect_get_constant_vectors): Use gimple_build_vector for
8806 constant-length vectors and suitable variable-length constant
8807 vectors. Use duplicate_and_interleave for other variable-length
8808 vectors. Don't defer the update when inserting new statements.
8810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8811 Alan Hayward <alan.hayward@arm.com>
8812 David Sherwood <david.sherwood@arm.com>
8814 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8815 min_profitable_iters doesn't go negative.
8817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8818 Alan Hayward <alan.hayward@arm.com>
8819 David Sherwood <david.sherwood@arm.com>
8821 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8822 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8823 * optabs.def (vec_mask_load_lanes_optab): New optab.
8824 (vec_mask_store_lanes_optab): Likewise.
8825 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8826 (MASK_STORE_LANES): Likewise.
8827 * internal-fn.c (mask_load_lanes_direct): New macro.
8828 (mask_store_lanes_direct): Likewise.
8829 (expand_mask_load_optab_fn): Handle masked operations.
8830 (expand_mask_load_lanes_optab_fn): New macro.
8831 (expand_mask_store_optab_fn): Handle masked operations.
8832 (expand_mask_store_lanes_optab_fn): New macro.
8833 (direct_mask_load_lanes_optab_supported_p): Likewise.
8834 (direct_mask_store_lanes_optab_supported_p): Likewise.
8835 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8837 (vect_load_lanes_supported): Likewise.
8838 * tree-vect-data-refs.c (strip_conversion): New function.
8839 (can_group_stmts_p): Likewise.
8840 (vect_analyze_data_ref_accesses): Use it instead of checking
8841 for a pair of assignments.
8842 (vect_store_lanes_supported): Take a masked_p parameter.
8843 (vect_load_lanes_supported): Likewise.
8844 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8845 vect_store_lanes_supported and vect_load_lanes_supported.
8846 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8847 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8848 parameter. Don't allow gaps for masked accesses.
8849 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8850 and vect_load_lanes_supported.
8851 (get_load_store_type): Take a masked_p parameter and update
8852 call to get_group_load_store_type.
8853 (vectorizable_store): Update call to get_load_store_type.
8854 Handle IFN_MASK_STORE_LANES.
8855 (vectorizable_load): Update call to get_load_store_type.
8856 Handle IFN_MASK_LOAD_LANES.
8858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8859 Alan Hayward <alan.hayward@arm.com>
8860 David Sherwood <david.sherwood@arm.com>
8862 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8864 * config/aarch64/aarch64-protos.h
8865 (aarch64_sve_struct_memory_operand_p): Declare.
8866 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8867 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8868 (VPRED, vpred): Handle SVE structure modes.
8869 * config/aarch64/constraints.md (Utx): New constraint.
8870 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8871 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8872 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8873 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8874 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8875 structure modes. Split into pieces after RA.
8876 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8877 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8879 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8880 SVE structure modes.
8881 (aarch64_classify_address): Likewise.
8882 (sizetochar): Move earlier in file.
8883 (aarch64_print_operand): Handle SVE register lists.
8884 (aarch64_array_mode): New function.
8885 (aarch64_sve_struct_memory_operand_p): Likewise.
8886 (TARGET_ARRAY_MODE): Redefine.
8888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8889 Alan Hayward <alan.hayward@arm.com>
8890 David Sherwood <david.sherwood@arm.com>
8892 * target.def (array_mode): New target hook.
8893 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8894 * doc/tm.texi: Regenerate.
8895 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8896 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8897 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8899 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8903 Alan Hayward <alan.hayward@arm.com>
8904 David Sherwood <david.sherwood@arm.com>
8906 * fold-const.c (fold_binary_loc): Check the argument types
8907 rather than the result type when testing for a vector operation.
8909 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8911 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8912 * doc/tm.texi: Regenerate.
8914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8915 Alan Hayward <alan.hayward@arm.com>
8916 David Sherwood <david.sherwood@arm.com>
8918 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8919 (sve): Document new AArch64 extension.
8920 * doc/md.texi (w): Extend the description of the AArch64
8921 constraint to include SVE vectors.
8922 (Upl, Upa): Document new AArch64 predicate constraints.
8923 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8925 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8926 (msve-vector-bits=): New option.
8927 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8928 SVE when these are disabled.
8929 (sve): New extension.
8930 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8931 modes. Adjust their number of units based on aarch64_sve_vg.
8932 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8933 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8934 aarch64_addr_query_type.
8935 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8936 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8937 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8938 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8939 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8940 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8941 (aarch64_simd_imm_zero_p): Delete.
8942 (aarch64_check_zero_based_sve_index_immediate): Declare.
8943 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8944 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8945 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8946 (aarch64_sve_float_mul_immediate_p): Likewise.
8947 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8949 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8950 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8951 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8952 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8953 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8954 (aarch64_regmode_natural_size): Likewise.
8955 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8956 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8958 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8959 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8960 for VG and the SVE predicate registers.
8961 (V_ALIASES): Add a "z"-prefixed alias.
8962 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8963 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8964 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8965 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8966 (REG_CLASS_NAMES): Add entries for them.
8967 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8968 and the predicate registers.
8969 (aarch64_sve_vg): Declare.
8970 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8971 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8972 (REGMODE_NATURAL_SIZE): Define.
8973 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8975 * config/aarch64/aarch64.c: Include cfgrtl.h.
8976 (simd_immediate_info): Add a constructor for series vectors,
8977 and an associated step field.
8978 (aarch64_sve_vg): New variable.
8979 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8980 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8981 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8982 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8983 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8984 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8985 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8986 (aarch64_get_mask_mode): New functions.
8987 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8988 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8989 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8990 predicate modes and predicate registers. Explicitly restrict
8991 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8992 to store a vector mode if it is recognized by
8993 aarch64_classify_vector_mode.
8994 (aarch64_regmode_natural_size): New function.
8995 (aarch64_hard_regno_caller_save_mode): Return the original mode
8997 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8998 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8999 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9000 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9002 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9003 does not overlap dest if the function is frame-related. Handle
9005 (aarch64_split_add_offset): New function.
9006 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9007 them aarch64_add_offset.
9008 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9009 and update call to aarch64_sub_sp.
9010 (aarch64_add_cfa_expression): New function.
9011 (aarch64_expand_prologue): Pass extra temporary registers to the
9012 functions above. Handle the case in which we need to emit new
9013 DW_CFA_expressions for registers that were originally saved
9014 relative to the stack pointer, but now have to be expressed
9015 relative to the frame pointer.
9016 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9018 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9019 IP0 and IP1 values for SVE frames.
9020 (aarch64_expand_vec_series): New function.
9021 (aarch64_expand_sve_widened_duplicate): Likewise.
9022 (aarch64_expand_sve_const_vector): Likewise.
9023 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9024 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9025 into the register, rather than emitting a SET directly.
9026 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9027 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9028 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9029 (offset_9bit_signed_scaled_p): New functions.
9030 (aarch64_replicate_bitmask_imm): New function.
9031 (aarch64_bitmask_imm): Use it.
9032 (aarch64_cannot_force_const_mem): Reject expressions involving
9033 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9034 (aarch64_classify_index): Handle SVE indices, by requiring
9035 a plain register index with a scale that matches the element size.
9036 (aarch64_classify_address): Handle SVE addresses. Assert that
9037 the mode of the address is VOIDmode or an integer mode.
9038 Update call to aarch64_classify_symbol.
9039 (aarch64_classify_symbolic_expression): Update call to
9040 aarch64_classify_symbol.
9041 (aarch64_const_vec_all_in_range_p): New function.
9042 (aarch64_print_vector_float_operand): Likewise.
9043 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9044 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9045 and the FP immediates 1.0 and 0.5.
9046 (aarch64_print_address_internal): Handle SVE addresses.
9047 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9048 (aarch64_regno_regclass): Handle predicate registers.
9049 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9051 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9052 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9053 (aarch64_convert_sve_vector_bits): New function.
9054 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9055 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9057 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9058 Handle SVE vector and predicate modes. Accept VL-based constants
9059 that need only one temporary register, and VL offsets that require
9060 no temporary registers.
9061 (aarch64_conditional_register_usage): Mark the predicate registers
9062 as fixed if SVE isn't available.
9063 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9064 Return true for SVE vector and predicate modes.
9065 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9066 rather than an unsigned int. Handle SVE modes.
9067 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9069 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9071 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9072 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9073 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9074 (aarch64_sve_float_mul_immediate_p): New functions.
9075 (aarch64_sve_valid_immediate): New function.
9076 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9077 Explicitly reject structure modes. Check for INDEX constants.
9078 Handle PTRUE and PFALSE constants.
9079 (aarch64_check_zero_based_sve_index_immediate): New function.
9080 (aarch64_simd_imm_zero_p): Delete.
9081 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9082 vector modes. Accept constants in the range of CNT[BHWD].
9083 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9084 ask for an Advanced SIMD mode.
9085 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9086 (aarch64_simd_vector_alignment): Handle SVE predicates.
9087 (aarch64_vectorize_preferred_vector_alignment): New function.
9088 (aarch64_simd_vector_alignment_reachable): Use it instead of
9090 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9091 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9093 (MAX_VECT_LEN): Delete.
9094 (expand_vec_perm_d): Add a vec_flags field.
9095 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9096 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9097 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9099 (aarch64_evpc_rev): Rename to...
9100 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9101 (aarch64_evpc_rev_global): New function.
9102 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9103 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9105 (aarch64_evpc_sve_tbl): New function.
9106 (aarch64_expand_vec_perm_const_1): Update after rename of
9107 aarch64_evpc_rev. Handle SVE permutes too, trying
9108 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9109 than aarch64_evpc_tbl.
9110 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9111 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9112 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9113 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9114 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9115 (aarch64_expand_sve_vcond): New functions.
9116 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9117 of aarch64_vector_mode_p.
9118 (aarch64_dwarf_poly_indeterminate_value): New function.
9119 (aarch64_compute_pressure_classes): Likewise.
9120 (aarch64_can_change_mode_class): Likewise.
9121 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9122 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9123 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9124 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9125 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9126 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9127 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9128 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9130 (Dn, Dl, Dr): Accept const as well as const_vector.
9131 (Dz): Likewise. Compare against CONST0_RTX.
9132 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9133 of "vector" where appropriate.
9134 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9135 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9136 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9137 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9138 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9139 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9140 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9141 (v_int_equiv): Extend to SVE modes.
9142 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9144 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9145 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9146 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9147 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9148 (SVE_COND_FP_CMP): New int iterators.
9149 (perm_hilo): Handle the new unpack unspecs.
9150 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9152 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9153 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9154 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9155 (aarch64_equality_operator, aarch64_constant_vector_operand)
9156 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9157 (aarch64_sve_nonimmediate_operand): Likewise.
9158 (aarch64_sve_general_operand): Likewise.
9159 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9160 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9161 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9162 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9163 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9164 (aarch64_sve_float_arith_immediate): Likewise.
9165 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9166 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9167 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9168 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9169 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9170 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9171 (aarch64_sve_float_arith_operand): Likewise.
9172 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9173 (aarch64_sve_float_mul_operand): Likewise.
9174 (aarch64_sve_vec_perm_operand): Likewise.
9175 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9176 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9177 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9178 as well as const_vector.
9179 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9180 in file. Use CONST0_RTX and CONSTM1_RTX.
9181 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9182 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9183 Use aarch64_simd_imm_zero.
9184 * config/aarch64/aarch64-sve.md: New file.
9185 * config/aarch64/aarch64.md: Include it.
9186 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9187 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9188 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9189 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9190 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9191 (sve): New attribute.
9192 (enabled): Disable instructions with the sve attribute unless
9194 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9195 aarch64_expand_mov_immediate.
9196 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9197 CNT[BHSD] immediates.
9198 (movti): Split CONST_POLY_INT moves into two halves.
9199 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9200 Split additions that need a temporary here if the destination
9201 is the stack pointer.
9202 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9203 (*add<mode>3_poly_1): New instruction.
9204 (set_clobber_cc): New expander.
9206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9208 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9209 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9210 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9211 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9212 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9213 Change innermode from fixed_mode_size to machine_mode.
9214 (simplify_subreg): Update call accordingly. Handle a constant-sized
9215 subreg of a variable-length CONST_VECTOR.
9217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9218 Alan Hayward <alan.hayward@arm.com>
9219 David Sherwood <david.sherwood@arm.com>
9221 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9222 (add_offset_to_base): New function, split out from...
9223 (create_mem_ref): ...here. When handling a scale other than 1,
9224 check first whether the address is valid without the offset.
9225 Add it into the base if so, leaving the index and scale as-is.
9227 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9230 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9231 fold_for_warn before checking if arg2 is INTEGER_CST.
9233 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9235 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9236 (store_multiple_operation): Delete.
9237 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9238 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9239 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9240 guarded by TARGET_STRING.
9241 (rs6000_output_load_multiple): Delete.
9242 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9243 OPTION_MASK_STRING / TARGET_STRING handling.
9244 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9245 (const rs6000_opt_masks) <"string">: Change mask to 0.
9246 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9247 (MASK_STRING): Delete.
9248 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9250 (load_multiple): Delete.
9257 (store_multiple): Delete.
9264 (movmemsi_8reg): Delete.
9265 (corresponding unnamed define_insn): Delete.
9266 (movmemsi_6reg): Delete.
9267 (corresponding unnamed define_insn): Delete.
9268 (movmemsi_4reg): Delete.
9269 (corresponding unnamed define_insn): Delete.
9270 (movmemsi_2reg): Delete.
9271 (corresponding unnamed define_insn): Delete.
9272 (movmemsi_1reg): Delete.
9273 (corresponding unnamed define_insn): Delete.
9274 * config/rs6000/rs6000.opt (mno-string): New.
9275 (mstring): Replace by deprecation warning stub.
9276 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9278 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9280 * regrename.c (regrename_do_replace): If replacing the same
9281 reg multiple times, try to reuse last created gen_raw_REG.
9284 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9285 main to workaround a bug in GDB.
9287 2018-01-12 Tom de Vries <tom@codesourcery.com>
9290 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9292 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9294 PR rtl-optimization/80481
9295 * ira-color.c (get_cap_member): New function.
9296 (allocnos_conflict_by_live_ranges_p): Use it.
9297 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9298 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9300 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9303 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9304 (*saddl_se_1): Ditto.
9306 (*ssubl_se_1): Ditto.
9308 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9310 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9311 rather than wi::to_widest for DR_INITs.
9312 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9313 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9314 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9316 (vect_analyze_group_access_1): Note that here.
9318 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9320 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9321 polynomial type sizes.
9323 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9325 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9326 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9327 (gimple_add_tmp_var): Likewise.
9329 2018-01-12 Martin Liska <mliska@suse.cz>
9331 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9332 (gimple_alloc_sizes): Likewise.
9333 (dump_gimple_statistics): Use PRIu64 in printf format.
9334 * gimple.h: Change uint64_t to int.
9336 2018-01-12 Martin Liska <mliska@suse.cz>
9338 * tree-core.h: Use uint64_t instead of int.
9339 * tree.c (tree_node_counts): Likewise.
9340 (tree_node_sizes): Likewise.
9341 (dump_tree_statistics): Use PRIu64 in printf format.
9343 2018-01-12 Martin Liska <mliska@suse.cz>
9345 * Makefile.in: As qsort_chk is implemented in vec.c, add
9346 vec.o to linkage of gencfn-macros.
9347 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9348 passing the info to record_node_allocation_statistics.
9349 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9351 * ggc-common.c (struct ggc_usage): Add operator== and use
9352 it in operator< and compare function.
9353 * mem-stats.h (struct mem_usage): Likewise.
9354 * vec.c (struct vec_usage): Remove operator< and compare
9355 function. Can be simply inherited.
9357 2018-01-12 Martin Jambor <mjambor@suse.cz>
9360 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9361 * tree-ssa-math-opts.c: Include domwalk.h.
9362 (convert_mult_to_fma_1): New function.
9363 (fma_transformation_info): New type.
9364 (fma_deferring_state): Likewise.
9365 (cancel_fma_deferring): New function.
9366 (result_of_phi): Likewise.
9367 (last_fma_candidate_feeds_initial_phi): Likewise.
9368 (convert_mult_to_fma): Added deferring logic, split actual
9369 transformation to convert_mult_to_fma_1.
9370 (math_opts_dom_walker): New type.
9371 (math_opts_dom_walker::after_dom_children): New method, body moved
9372 here from pass_optimize_widening_mul::execute, added deferring logic
9374 (pass_optimize_widening_mul::execute): Moved most of code to
9375 math_opts_dom_walker::after_dom_children.
9376 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9377 * config/i386/i386.c (ix86_option_override_internal): Added
9378 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9380 2018-01-12 Richard Biener <rguenther@suse.de>
9383 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9384 inline instance vars.
9386 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9389 * config/rx/rx.c (rx_is_restricted_memory_address):
9392 2018-01-12 Richard Biener <rguenther@suse.de>
9394 PR tree-optimization/80846
9395 * target.def (split_reduction): New target hook.
9396 * targhooks.c (default_split_reduction): New function.
9397 * targhooks.h (default_split_reduction): Declare.
9398 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9399 target requests first reduce vectors by combining low and high
9401 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9402 (get_vectype_for_scalar_type_and_size): Export.
9403 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9404 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9405 * doc/tm.texi: Regenerate.
9406 * config/i386/i386.c (ix86_split_reduction): Implement
9407 TARGET_VECTORIZE_SPLIT_REDUCTION.
9409 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9412 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9413 in PIC mode except for TARGET_VXWORKS_RTP.
9414 * config/sparc/sparc.c: Include cfgrtl.h.
9415 (TARGET_INIT_PIC_REG): Define.
9416 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9417 (sparc_pic_register_p): New predicate.
9418 (sparc_legitimate_address_p): Use it.
9419 (sparc_legitimize_pic_address): Likewise.
9420 (sparc_delegitimize_address): Likewise.
9421 (sparc_mode_dependent_address_p): Likewise.
9422 (gen_load_pcrel_sym): Remove 4th parameter.
9423 (load_got_register): Adjust call to above. Remove obsolete stuff.
9424 (sparc_expand_prologue): Do not call load_got_register here.
9425 (sparc_flat_expand_prologue): Likewise.
9426 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9427 (sparc_use_pseudo_pic_reg): New function.
9428 (sparc_init_pic_reg): Likewise.
9429 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9430 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9432 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9434 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9435 Add item for branch_cost.
9437 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9439 PR rtl-optimization/83565
9440 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9441 not extend the result to a larger mode for rotate operations.
9442 (num_sign_bit_copies1): Likewise.
9444 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9447 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9449 Use values-Xc.o for -pedantic.
9450 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9452 2018-01-12 Martin Liska <mliska@suse.cz>
9455 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9457 (possible_polymorphic_call_targets): Use it.
9458 (ipa_devirt): Likewise.
9460 2018-01-12 Martin Liska <mliska@suse.cz>
9462 * profile-count.h (enum profile_quality): Use 0 as invalid
9463 enum value of profile_quality.
9465 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9467 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9468 -mext-string options.
9470 2018-01-12 Richard Biener <rguenther@suse.de>
9472 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9473 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9474 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9476 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9478 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9480 * configure.ac (--with-long-double-format): Add support for the
9481 configuration option to change the default long double format on
9483 * config.gcc (powerpc*-linux*-*): Likewise.
9484 * configure: Regenerate.
9485 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9486 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9487 used without modification.
9489 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9491 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9492 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9493 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9494 MISC_BUILTIN_SPEC_BARRIER.
9495 (rs6000_init_builtins): Likewise.
9496 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9498 (speculation_barrier): New define_insn.
9499 * doc/extend.texi: Document __builtin_speculation_barrier.
9501 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9504 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9505 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9506 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9508 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9509 integral modes instead of "ss" and "sd".
9510 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9511 vectors with 32-bit and 64-bit elements.
9512 (vecdupssescalarmodesuffix): New mode attribute.
9513 (vec_dup<mode>): Use it.
9515 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9518 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9519 frame if argument is passed on stack.
9521 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9524 * ree.c (combine_reaching_defs): Optimize also
9525 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9526 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9528 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9531 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9533 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9536 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9537 after they are computed.
9539 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9541 PR tree-optimization/83695
9542 * gimple-loop-linterchange.cc
9543 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9544 reset cached scev information after interchange.
9545 (pass_linterchange::execute): Remove call to scev_reset_htab.
9547 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9549 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9550 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9551 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9552 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9553 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9554 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9555 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9556 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9557 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9558 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9559 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9560 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9561 (V_lane_reg): Likewise.
9562 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9564 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9565 (vfmal_lane_low<mode>_intrinsic,
9566 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9567 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9568 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9569 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9570 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9571 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9573 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9575 * config/arm/arm-cpus.in (fp16fml): New feature.
9576 (ALL_SIMD): Add fp16fml.
9577 (armv8.2-a): Add fp16fml as an option.
9578 (armv8.3-a): Likewise.
9579 (armv8.4-a): Add fp16fml as part of fp16.
9580 * config/arm/arm.h (TARGET_FP16FML): Define.
9581 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9583 * config/arm/arm-modes.def (V2HF): Define.
9584 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9585 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9586 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9587 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9588 vfmsl_low, vfmsl_high): New set of builtins.
9589 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9590 (vfml_op): New code attribute.
9591 (VFMLHALVES): New int iterator.
9592 (VFML, VFMLSEL): New mode attributes.
9593 (V_reg): Define mapping for V2HF.
9594 (V_hi, V_lo): New mode attributes.
9595 (VF_constraint): Likewise.
9596 (vfml_half, vfml_half_selector): New int attributes.
9597 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9599 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9600 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9602 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9603 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9604 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9605 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9607 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9608 Document new effective target and option set.
9610 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9612 * config/arm/arm-cpus.in (armv8_4): New feature.
9613 (ARMv8_4a): New fgroup.
9614 (armv8.4-a): New arch.
9615 * config/arm/arm-tables.opt: Regenerate.
9616 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9617 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9618 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9619 Add matching rules for -march=armv8.4-a and extensions.
9620 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9622 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9625 * config/rx/rx.md (BW): New mode attribute.
9626 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9628 2018-01-11 Richard Biener <rguenther@suse.de>
9630 PR tree-optimization/83435
9631 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9632 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9633 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9635 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9636 Alan Hayward <alan.hayward@arm.com>
9637 David Sherwood <david.sherwood@arm.com>
9639 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9641 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9642 (aarch64_print_address_internal): Use it to check for a zero offset.
9644 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9645 Alan Hayward <alan.hayward@arm.com>
9646 David Sherwood <david.sherwood@arm.com>
9648 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9649 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9650 Return a poly_int64 rather than a HOST_WIDE_INT.
9651 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9652 rather than a HOST_WIDE_INT.
9653 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9654 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9655 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9656 final_offset from HOST_WIDE_INT to poly_int64.
9657 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9658 to_constant when getting the number of units in an Advanced SIMD
9660 (aarch64_builtin_vectorized_function): Check for a constant number
9662 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9664 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9665 attribute instead of GET_MODE_NUNITS.
9666 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9667 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9668 GET_MODE_SIZE for fixed-size registers.
9669 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9670 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9671 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9672 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9673 (aarch64_print_operand, aarch64_print_address_internal)
9674 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9675 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9676 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9677 Handle polynomial GET_MODE_SIZE.
9678 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9679 wider than SImode without modification.
9680 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9681 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9682 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9683 passing and returning SVE modes.
9684 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9685 rather than GEN_INT.
9686 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9687 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9688 (aarch64_allocate_and_probe_stack_space): Likewise.
9689 (aarch64_layout_frame): Cope with polynomial offsets.
9690 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9691 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9693 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9694 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9695 poly_int64 rather than a HOST_WIDE_INT.
9696 (aarch64_get_separate_components, aarch64_process_components)
9697 (aarch64_expand_prologue, aarch64_expand_epilogue)
9698 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9699 (aarch64_anchor_offset): New function, split out from...
9700 (aarch64_legitimize_address): ...here.
9701 (aarch64_builtin_vectorization_cost): Handle polynomial
9702 TYPE_VECTOR_SUBPARTS.
9703 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9705 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9706 number of elements from the PARALLEL rather than the mode.
9707 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9708 rather than GET_MODE_BITSIZE.
9709 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9710 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9711 (aarch64_expand_vec_perm_const_1): Handle polynomial
9712 d->perm.length () and d->perm elements.
9713 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9714 Apply to_constant to d->perm elements.
9715 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9716 polynomial CONST_VECTOR_NUNITS.
9717 (aarch64_move_pointer): Take amount as a poly_int64 rather
9719 (aarch64_progress_pointer): Avoid temporary variable.
9720 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9721 the mode attribute instead of GET_MODE.
9723 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9724 Alan Hayward <alan.hayward@arm.com>
9725 David Sherwood <david.sherwood@arm.com>
9727 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9728 x exists before using it.
9729 (aarch64_add_constant_internal): Rename to...
9730 (aarch64_add_offset_1): ...this. Replace regnum with separate
9731 src and dest rtxes. Handle the case in which they're different,
9732 including when the offset is zero. Replace scratchreg with an rtx.
9733 Use 2 additions if there is no spare register into which we can
9734 move a 16-bit constant.
9735 (aarch64_add_constant): Delete.
9736 (aarch64_add_offset): Replace reg with separate src and dest
9737 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9738 Use aarch64_add_offset_1.
9739 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9740 an rtx rather than an int. Take the delta as a poly_int64
9741 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9742 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9743 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9744 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9745 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9747 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9748 aarch64_add_constant.
9750 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9752 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9753 Use scalar_float_mode.
9755 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9757 * config/aarch64/aarch64-simd.md
9758 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9759 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9760 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9761 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9762 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9763 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9764 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9765 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9766 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9767 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9769 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9772 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9773 targ_options->x_arm_arch_string is non NULL.
9775 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9777 * config/aarch64/aarch64.h
9778 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9780 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9783 * expmed.c (emit_store_flag_force): Swap if const op0
9784 and change VOIDmode to mode of op0.
9786 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9788 PR rtl-optimization/83761
9789 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9790 than bytes to mode_for_size.
9792 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9795 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9796 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9799 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9802 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9803 when in layout mode.
9804 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9805 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9808 2018-01-10 Michael Collison <michael.collison@arm.com>
9810 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9811 * config/aarch64/aarch64-option-extension.def: Add
9812 AARCH64_OPT_EXTENSION of 'fp16fml'.
9813 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9814 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9815 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9816 * config/aarch64/constraints.md (Ui7): New constraint.
9817 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9818 (VFMLA_SEL_W): Ditto.
9821 (VFMLA16_LOW): New int iterator.
9822 (VFMLA16_HIGH): Ditto.
9823 (UNSPEC_FMLAL): New unspec.
9824 (UNSPEC_FMLSL): Ditto.
9825 (UNSPEC_FMLAL2): Ditto.
9826 (UNSPEC_FMLSL2): Ditto.
9827 (f16mac): New code attribute.
9828 * config/aarch64/aarch64-simd-builtins.def
9829 (aarch64_fmlal_lowv2sf): Ditto.
9830 (aarch64_fmlsl_lowv2sf): Ditto.
9831 (aarch64_fmlalq_lowv4sf): Ditto.
9832 (aarch64_fmlslq_lowv4sf): Ditto.
9833 (aarch64_fmlal_highv2sf): Ditto.
9834 (aarch64_fmlsl_highv2sf): Ditto.
9835 (aarch64_fmlalq_highv4sf): Ditto.
9836 (aarch64_fmlslq_highv4sf): Ditto.
9837 (aarch64_fmlal_lane_lowv2sf): Ditto.
9838 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9839 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9840 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9841 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9842 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9843 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9844 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9845 (aarch64_fmlal_lane_highv2sf): Ditto.
9846 (aarch64_fmlsl_lane_highv2sf): Ditto.
9847 (aarch64_fmlal_laneq_highv2sf): Ditto.
9848 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9849 (aarch64_fmlalq_lane_highv4sf): Ditto.
9850 (aarch64_fmlsl_lane_highv4sf): Ditto.
9851 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9852 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9853 * config/aarch64/aarch64-simd.md:
9854 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9855 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9856 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9857 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9858 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9859 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9860 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9861 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9862 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9863 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9864 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9865 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9866 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9867 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9868 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9869 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9870 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9871 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9872 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9873 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9874 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9875 (vfmlsl_low_u32): Ditto.
9876 (vfmlalq_low_u32): Ditto.
9877 (vfmlslq_low_u32): Ditto.
9878 (vfmlal_high_u32): Ditto.
9879 (vfmlsl_high_u32): Ditto.
9880 (vfmlalq_high_u32): Ditto.
9881 (vfmlslq_high_u32): Ditto.
9882 (vfmlal_lane_low_u32): Ditto.
9883 (vfmlsl_lane_low_u32): Ditto.
9884 (vfmlal_laneq_low_u32): Ditto.
9885 (vfmlsl_laneq_low_u32): Ditto.
9886 (vfmlalq_lane_low_u32): Ditto.
9887 (vfmlslq_lane_low_u32): Ditto.
9888 (vfmlalq_laneq_low_u32): Ditto.
9889 (vfmlslq_laneq_low_u32): Ditto.
9890 (vfmlal_lane_high_u32): Ditto.
9891 (vfmlsl_lane_high_u32): Ditto.
9892 (vfmlal_laneq_high_u32): Ditto.
9893 (vfmlsl_laneq_high_u32): Ditto.
9894 (vfmlalq_lane_high_u32): Ditto.
9895 (vfmlslq_lane_high_u32): Ditto.
9896 (vfmlalq_laneq_high_u32): Ditto.
9897 (vfmlslq_laneq_high_u32): Ditto.
9898 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9899 (AARCH64_FL_FOR_ARCH8_4): New.
9900 (AARCH64_ISA_F16FML): New ISA flag.
9901 (TARGET_F16FML): New feature flag for fp16fml.
9902 (doc/invoke.texi): Document new fp16fml option.
9904 2018-01-10 Michael Collison <michael.collison@arm.com>
9906 * config/aarch64/aarch64-builtins.c:
9907 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9908 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9909 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9910 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9911 (AARCH64_ISA_SHA3): New ISA flag.
9912 (TARGET_SHA3): New feature flag for sha3.
9913 * config/aarch64/iterators.md (sha512_op): New int attribute.
9914 (CRYPTO_SHA512): New int iterator.
9915 (UNSPEC_SHA512H): New unspec.
9916 (UNSPEC_SHA512H2): Ditto.
9917 (UNSPEC_SHA512SU0): Ditto.
9918 (UNSPEC_SHA512SU1): Ditto.
9919 * config/aarch64/aarch64-simd-builtins.def
9920 (aarch64_crypto_sha512hqv2di): New builtin.
9921 (aarch64_crypto_sha512h2qv2di): Ditto.
9922 (aarch64_crypto_sha512su0qv2di): Ditto.
9923 (aarch64_crypto_sha512su1qv2di): Ditto.
9924 (aarch64_eor3qv8hi): Ditto.
9925 (aarch64_rax1qv2di): Ditto.
9926 (aarch64_xarqv2di): Ditto.
9927 (aarch64_bcaxqv8hi): Ditto.
9928 * config/aarch64/aarch64-simd.md:
9929 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9930 (aarch64_crypto_sha512su0qv2di): Ditto.
9931 (aarch64_crypto_sha512su1qv2di): Ditto.
9932 (aarch64_eor3qv8hi): Ditto.
9933 (aarch64_rax1qv2di): Ditto.
9934 (aarch64_xarqv2di): Ditto.
9935 (aarch64_bcaxqv8hi): Ditto.
9936 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9937 (vsha512h2q_u64): Ditto.
9938 (vsha512su0q_u64): Ditto.
9939 (vsha512su1q_u64): Ditto.
9940 (veor3q_u16): Ditto.
9941 (vrax1q_u64): Ditto.
9943 (vbcaxq_u16): Ditto.
9944 * config/arm/types.md (crypto_sha512): New type attribute.
9945 (crypto_sha3): Ditto.
9946 (doc/invoke.texi): Document new sha3 option.
9948 2018-01-10 Michael Collison <michael.collison@arm.com>
9950 * config/aarch64/aarch64-builtins.c:
9951 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9952 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9953 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9954 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9955 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9956 (AARCH64_ISA_SM4): New ISA flag.
9957 (TARGET_SM4): New feature flag for sm4.
9958 * config/aarch64/aarch64-simd-builtins.def
9959 (aarch64_sm3ss1qv4si): Ditto.
9960 (aarch64_sm3tt1aq4si): Ditto.
9961 (aarch64_sm3tt1bq4si): Ditto.
9962 (aarch64_sm3tt2aq4si): Ditto.
9963 (aarch64_sm3tt2bq4si): Ditto.
9964 (aarch64_sm3partw1qv4si): Ditto.
9965 (aarch64_sm3partw2qv4si): Ditto.
9966 (aarch64_sm4eqv4si): Ditto.
9967 (aarch64_sm4ekeyqv4si): Ditto.
9968 * config/aarch64/aarch64-simd.md:
9969 (aarch64_sm3ss1qv4si): Ditto.
9970 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9971 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9972 (aarch64_sm4eqv4si): Ditto.
9973 (aarch64_sm4ekeyqv4si): Ditto.
9974 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9975 (sm3part_op): Ditto.
9976 (CRYPTO_SM3TT): Ditto.
9977 (CRYPTO_SM3PART): Ditto.
9978 (UNSPEC_SM3SS1): New unspec.
9979 (UNSPEC_SM3TT1A): Ditto.
9980 (UNSPEC_SM3TT1B): Ditto.
9981 (UNSPEC_SM3TT2A): Ditto.
9982 (UNSPEC_SM3TT2B): Ditto.
9983 (UNSPEC_SM3PARTW1): Ditto.
9984 (UNSPEC_SM3PARTW2): Ditto.
9985 (UNSPEC_SM4E): Ditto.
9986 (UNSPEC_SM4EKEY): Ditto.
9987 * config/aarch64/constraints.md (Ui2): New constraint.
9988 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9989 * config/arm/types.md (crypto_sm3): New type attribute.
9990 (crypto_sm4): Ditto.
9991 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9992 (vsm3tt1aq_u32): Ditto.
9993 (vsm3tt1bq_u32): Ditto.
9994 (vsm3tt2aq_u32): Ditto.
9995 (vsm3tt2bq_u32): Ditto.
9996 (vsm3partw1q_u32): Ditto.
9997 (vsm3partw2q_u32): Ditto.
9998 (vsm4eq_u32): Ditto.
9999 (vsm4ekeyq_u32): Ditto.
10000 (doc/invoke.texi): Document new sm4 option.
10002 2018-01-10 Michael Collison <michael.collison@arm.com>
10004 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10005 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10006 (AARCH64_FL_FOR_ARCH8_4): New.
10007 (AARCH64_FL_V8_4): New flag.
10008 (doc/invoke.texi): Document new armv8.4-a option.
10010 2018-01-10 Michael Collison <michael.collison@arm.com>
10012 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10013 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10014 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10015 * config/aarch64/aarch64-option-extension.def: Add
10016 AARCH64_OPT_EXTENSION of 'sha2'.
10017 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10018 (crypto): Disable sha2 and aes if crypto disabled.
10019 (crypto): Enable aes and sha2 if enabled.
10020 (simd): Disable sha2 and aes if simd disabled.
10021 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10023 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10024 (TARGET_SHA2): New feature flag for sha2.
10025 (TARGET_AES): New feature flag for aes.
10026 * config/aarch64/aarch64-simd.md:
10027 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10028 conditional on TARGET_AES.
10029 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10030 (aarch64_crypto_sha1hsi): Make pattern conditional
10032 (aarch64_crypto_sha1hv4si): Ditto.
10033 (aarch64_be_crypto_sha1hv4si): Ditto.
10034 (aarch64_crypto_sha1su1v4si): Ditto.
10035 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10036 (aarch64_crypto_sha1su0v4si): Ditto.
10037 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10038 (aarch64_crypto_sha256su0v4si): Ditto.
10039 (aarch64_crypto_sha256su1v4si): Ditto.
10040 (doc/invoke.texi): Document new aes and sha2 options.
10042 2018-01-10 Martin Sebor <msebor@redhat.com>
10044 PR tree-optimization/83781
10045 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10048 2018-01-11 Martin Sebor <msebor@gmail.com>
10049 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10051 PR tree-optimization/83501
10052 PR tree-optimization/81703
10054 * tree-ssa-strlen.c (get_string_cst): Rename...
10055 (get_string_len): ...to this. Handle global constants.
10056 (handle_char_store): Adjust.
10058 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10059 Jim Wilson <jimw@sifive.com>
10061 * config/riscv/riscv-protos.h (riscv_output_return): New.
10062 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10063 (riscv_attribute_table, riscv_output_return),
10064 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10065 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10066 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10068 (riscv_expand_prologue): Add early return for naked function.
10069 (riscv_expand_epilogue): Likewise.
10070 (riscv_function_ok_for_sibcall): Return false for naked function.
10071 (riscv_set_current_function): New.
10072 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10073 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10074 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10075 * doc/extend.texi (RISC-V Function Attributes): New.
10077 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10079 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10080 check for 128-bit long double before checking TCmode.
10081 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10082 128-bit long doubles before checking TFmode or TCmode.
10083 (FLOAT128_IBM_P): Likewise.
10085 2018-01-10 Martin Sebor <msebor@redhat.com>
10087 PR tree-optimization/83671
10088 * builtins.c (c_strlen): Unconditionally return zero for the empty
10090 Use -Warray-bounds for warnings.
10091 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10092 for non-constant array indices with COMPONENT_REF, arrays of
10093 arrays, and pointers to arrays.
10094 (gimple_fold_builtin_strlen): Determine and set length range for
10095 non-constant character arrays.
10097 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10099 PR middle-end/81897
10100 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10103 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10105 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10107 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10110 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10111 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10112 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10113 indexed_or_indirect_operand predicate.
10114 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10115 (*vsx_le_perm_load_v8hi): Likewise.
10116 (*vsx_le_perm_load_v16qi): Likewise.
10117 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10118 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10119 (*vsx_le_perm_store_v8hi): Likewise.
10120 (*vsx_le_perm_store_v16qi): Likewise.
10121 (eight unnamed splitters): Likewise.
10123 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10125 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10126 * config/rs6000/emmintrin.h: Likewise.
10127 * config/rs6000/mmintrin.h: Likewise.
10128 * config/rs6000/xmmintrin.h: Likewise.
10130 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10133 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10135 * tree.c (tree_nop_conversion): Return true for location wrapper
10137 (maybe_wrap_with_location): New function.
10138 (selftest::check_strip_nops): New function.
10139 (selftest::test_location_wrappers): New function.
10140 (selftest::tree_c_tests): Call it.
10141 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10142 (maybe_wrap_with_location): New decl.
10143 (EXPR_LOCATION_WRAPPER_P): New macro.
10144 (location_wrapper_p): New inline function.
10145 (tree_strip_any_location_wrapper): New inline function.
10147 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10150 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10151 stack_realign_offset for the largest alignment of stack slot
10153 (ix86_find_max_used_stack_alignment): New function.
10154 (ix86_finalize_stack_frame_flags): Use it. Set
10155 max_used_stack_alignment if we don't realign stack.
10156 * config/i386/i386.h (machine_function): Add
10157 max_used_stack_alignment.
10159 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10161 * config/arm/arm.opt (-mbranch-cost): New option.
10162 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10165 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10168 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10169 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10171 2018-01-10 Richard Biener <rguenther@suse.de>
10174 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10175 early out so it also covers the case where we have a non-NULL
10178 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10180 PR tree-optimization/83753
10181 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10182 for non-strided grouped accesses if the number of elements is 1.
10184 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10187 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10188 * i386.h (TARGET_USE_GATHER): Define.
10189 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10191 2018-01-10 Martin Liska <mliska@suse.cz>
10194 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10195 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10197 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10198 CLEANUP_NO_PARTITIONING is not set.
10200 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10202 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10203 for vectors, as a partial revert of r254296.
10204 * rtl.h (const_vec_p): Delete.
10205 (const_vec_duplicate_p): Don't test for vector CONSTs.
10206 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10207 * expmed.c (make_tree): Likewise.
10210 * common.md (E, F): Use CONSTANT_P instead of checking for
10212 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10213 checking for CONST_VECTOR.
10215 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10217 PR middle-end/83575
10218 * predict.c (force_edge_cold): Handle in more sane way edges
10219 with no prediction.
10221 2018-01-09 Carl Love <cel@us.ibm.com>
10223 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10225 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10226 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10227 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10228 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10229 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10230 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10231 * config/rs6000/rs6000-protos.h: Add extern defition for
10232 rs6000_generate_float2_double_code.
10233 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10235 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10236 (float2_v2df): Add define_expand.
10238 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10241 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10242 op_mode in the force_to_mode call.
10244 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10246 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10247 instead of checking each element individually.
10248 (aarch64_evpc_uzp): Likewise.
10249 (aarch64_evpc_zip): Likewise.
10250 (aarch64_evpc_ext): Likewise.
10251 (aarch64_evpc_rev): Likewise.
10252 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10253 instead of checking each element individually. Return true without
10255 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10256 whether all selected elements come from the same input, instead of
10257 checking each element individually. Remove calls to gen_rtx_REG,
10258 start_sequence and end_sequence and instead assert that no rtl is
10261 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10263 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10264 order of HIGH and CONST checks.
10266 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10268 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10269 if the destination isn't an SSA_NAME.
10271 2018-01-09 Richard Biener <rguenther@suse.de>
10273 PR tree-optimization/83668
10274 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10276 (canonicalize_loop_form): ... here, renamed from ...
10277 (canonicalize_loop_closed_ssa_form): ... this and amended to
10278 swap successor edges for loop exit blocks to make us use
10279 the RPO order we need for initial schedule generation.
10281 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10283 PR tree-optimization/64811
10284 * match.pd: When optimizing comparisons with Inf, avoid
10285 introducing or losing exceptions from comparisons with NaN.
10287 2018-01-09 Martin Liska <mliska@suse.cz>
10290 * asan.c (shadow_mem_size): Add gcc_assert.
10292 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10294 Don't save registers in main().
10297 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10298 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10299 * config/avr/avr.c (avr_set_current_function): Don't error if
10300 naked, OS_task or OS_main are specified at the same time.
10301 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10303 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10305 * common/config/avr/avr-common.c (avr_option_optimization_table):
10306 Switch on -mmain-is-OS_task for optimizing compilations.
10308 2018-01-09 Richard Biener <rguenther@suse.de>
10310 PR tree-optimization/83572
10311 * graphite.c: Include cfganal.h.
10312 (graphite_transform_loops): Connect infinite loops to exit
10313 and remove fake edges at the end.
10315 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10317 * ipa-inline.c (edge_badness): Revert accidental checkin.
10319 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10322 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10323 symbols; not inline clones.
10325 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10328 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10329 hard registers. Formatting fixes.
10331 PR preprocessor/83722
10332 * gcc.c (try_generate_repro): Pass
10333 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10334 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10337 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10338 Kito Cheng <kito.cheng@gmail.com>
10340 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10341 (riscv_leaf_function_p): Delete.
10342 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10344 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10346 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10348 (do_ifelse): New function.
10349 (do_isel): New function.
10350 (do_sub3): New function.
10351 (do_add3): New function.
10352 (do_load_mask_compare): New function.
10353 (do_overlap_load_compare): New function.
10354 (expand_compare_loop): New function.
10355 (expand_block_compare): Call expand_compare_loop() when appropriate.
10356 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10357 option description.
10358 (-mblock-compare-inline-loop-limit): New option.
10360 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10363 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10364 Reverse order of second and third operands in first alternative.
10365 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10366 of first and second elements in UNSPEC_VPERMR vector.
10367 (altivec_expand_vec_perm_le): Likewise.
10369 2018-01-08 Jeff Law <law@redhat.com>
10371 PR rtl-optimizatin/81308
10372 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10373 (process_switch): If group_case_labels makes a change, then set
10375 (pass_convert_switch::execute): If a switch is converted, then
10376 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10378 PR rtl-optimization/81308
10379 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10382 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10384 PR target/83663 - Revert r255946
10385 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10386 generation for cases where splatting a value is not useful.
10387 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10388 across a vec_duplicate and a paradoxical subreg forming a vector
10389 mode to a vec_concat.
10391 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10393 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10394 -march=armv8.3-a variants.
10395 * config/arm/t-multilib: Likewise.
10396 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10398 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10400 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10402 (cceq_ior_compare_complement): Give it a name so I can use it, and
10403 change boolean_or_operator predicate to boolean_operator so it can
10404 be used to generate a crand.
10405 (eqne): New code iterator.
10406 (bd/bd_neg): New code_attrs.
10407 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10408 a single define_insn.
10409 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10410 decrement (bdnzt/bdnzf/bdzt/bdzf).
10411 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10412 with the new names of the branch decrement patterns, and added the
10413 names of the branch decrement conditional patterns.
10415 2018-01-08 Richard Biener <rguenther@suse.de>
10417 PR tree-optimization/83563
10418 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10421 2018-01-08 Richard Biener <rguenther@suse.de>
10423 PR middle-end/83713
10424 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10426 2018-01-08 Richard Biener <rguenther@suse.de>
10428 PR tree-optimization/83685
10429 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10430 references to abnormals.
10432 2018-01-08 Richard Biener <rguenther@suse.de>
10435 * dwarf2out.c (output_indirect_strings): Handle empty
10436 skeleton_debug_str_hash.
10437 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10439 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10441 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10442 (emit_store_direct): Likewise.
10443 (arc_trampoline_adjust_address): Likewise.
10444 (arc_asm_trampoline_template): New function.
10445 (arc_initialize_trampoline): Use asm_trampoline_template.
10446 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10447 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10448 * config/arc/arc.md (flush_icache): Delete pattern.
10450 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10452 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10453 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10456 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10459 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10460 by not USED_FOR_TARGET.
10461 (make_pass_resolve_sw_modes): Likewise.
10463 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10465 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10468 2018-01-08 Richard Biener <rguenther@suse.de>
10470 PR middle-end/83580
10471 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10473 2018-01-08 Richard Biener <rguenther@suse.de>
10475 PR middle-end/83517
10476 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10478 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10480 PR middle-end/81897
10481 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10482 basic blocks with a small number of successors.
10483 (convert_control_dep_chain_into_preds): Improve handling of
10485 (dump_predicates): Split apart into...
10486 (dump_pred_chain): ...here...
10487 (dump_pred_info): ...and here.
10488 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10489 (can_chain_union_be_invalidated_p): Improve check for invalidation
10491 (uninit_uses_cannot_happen): Avoid unnecessary if
10492 convert_control_dep_chain_into_preds yielded nothing.
10494 2018-01-06 Martin Sebor <msebor@redhat.com>
10496 PR tree-optimization/83640
10497 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10498 subtracting negative offset from size.
10499 (builtin_access::overlap): Adjust offset bounds of the access to fall
10500 within the size of the object if possible.
10502 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10504 PR rtl-optimization/83699
10505 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10506 extract_bit_field_as_subreg to cases in which the extracted
10507 value is also a vector.
10509 * lra-constraints.c (process_alt_operands): Test for the equivalence
10510 substitutions when detecting a possible reload cycle.
10512 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10515 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10516 by default if flag_selective_schedling{,2}. Formatting fixes.
10518 PR rtl-optimization/83682
10519 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10520 if it has non-VECTOR_MODE element mode.
10521 (vec_duplicate_p): Likewise.
10523 PR middle-end/83694
10524 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10525 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10527 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10530 * config/i386/i386-builtin.def
10531 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10532 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10533 Require also OPTION_MASK_ISA_AVX512F in addition to
10534 OPTION_MASK_ISA_GFNI.
10535 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10536 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10537 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10538 to OPTION_MASK_ISA_GFNI.
10539 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10540 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10541 OPTION_MASK_ISA_AVX512BW.
10542 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10543 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10544 addition to OPTION_MASK_ISA_GFNI.
10545 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10546 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10547 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10548 to OPTION_MASK_ISA_GFNI.
10549 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10550 a requirement for all ISAs rather than any of them with a few
10552 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10554 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10555 bitmasks to be enabled with 3 exceptions, instead of requiring any
10556 enabled ISA with lots of exceptions.
10557 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10558 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10559 Change avx512bw in isa attribute to avx512f.
10560 * config/i386/sgxintrin.h: Add license boilerplate.
10561 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10562 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10563 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10564 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10566 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10567 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10568 temporarily sse2 rather than sse if not enabled already.
10571 * config/i386/sse.md (VI248_VLBW): Rename to ...
10572 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10573 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10574 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10575 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10576 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10577 mode iterator instead of VI248_VLBW.
10579 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10581 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10582 (record_modified): Skip clobbers; add debug output.
10583 (param_change_prob): Use sreal frequencies.
10585 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10587 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10588 punt for user-aligned variables.
10590 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10592 * tree-chrec.c (chrec_contains_symbols): Return true for
10595 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10598 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10599 of (x|y) == x for BICS pattern.
10601 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10603 PR tree-optimization/83605
10604 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10605 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10608 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10610 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10611 * config/epiphany/rtems.h: New file.
10613 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10614 Uros Bizjak <ubizjak@gmail.com>
10617 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10618 QIreg_operand instead of register_operand predicate.
10619 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10620 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10621 comments instead of -fmitigate[-_]rop.
10623 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10626 * cgraphunit.c (symbol_table::compile): Switch to text_section
10627 before calling assembly_start debug hook.
10628 * run-rtl-passes.c (run_rtl_passes): Likewise.
10631 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10633 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10634 range_int_cst_p rather than !symbolic_range_p before calling
10635 extract_range_from_multiplicative_op_1.
10637 2018-01-04 Jeff Law <law@redhat.com>
10639 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10640 redundant test in assertion.
10642 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10644 * doc/rtl.texi: Document machine_mode wrapper classes.
10646 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10648 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10649 using tree_to_uhwi.
10651 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10653 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10654 the VEC_PERM_EXPR fold to fail.
10656 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10659 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10660 to switched_sections.
10662 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10665 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10666 test for d.testing.
10668 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10671 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10672 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10674 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10677 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10678 is BLKmode and bitpos not zero or mode change is needed.
10680 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10683 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10686 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10689 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10690 instead of MULT rtx. Update all corresponding splitters.
10691 (*saddl_se): Ditto.
10692 (*ssub<modesuffix>): Ditto.
10693 (*ssubl_se): Ditto.
10694 (*cmp_sadd_di): Update split patterns.
10695 (*cmp_sadd_si): Ditto.
10696 (*cmp_sadd_sidi): Ditto.
10697 (*cmp_ssub_di): Ditto.
10698 (*cmp_ssub_si): Ditto.
10699 (*cmp_ssub_sidi): Ditto.
10700 * config/alpha/predicates.md (const23_operand): New predicate.
10701 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10702 Look for ASHIFT, not MULT inner operand.
10703 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10705 2018-01-04 Martin Liska <mliska@suse.cz>
10707 PR gcov-profile/83669
10708 * gcov.c (output_intermediate_file): Add version to intermediate
10710 * doc/gcov.texi: Document new field 'version' in intermediate
10711 file format. Fix location of '-k' option of gcov command.
10713 2018-01-04 Martin Liska <mliska@suse.cz>
10716 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10718 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10720 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10722 2018-01-03 Martin Sebor <msebor@redhat.com>
10724 PR tree-optimization/83655
10725 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10726 checking calls with invalid arguments.
10728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10730 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10731 (vectorizable_mask_load_store): Delete.
10732 (vectorizable_call): Return false for masked loads and stores.
10733 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10734 instead of gimple_assign_rhs1.
10735 (vectorizable_load): Handle IFN_MASK_LOAD.
10736 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10740 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10742 (vectorizable_mask_load_store): ...here.
10743 (vectorizable_load): ...and here.
10745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10747 * tree-vect-stmts.c (vect_build_all_ones_mask)
10748 (vect_build_zero_merge_argument): New functions, split out from...
10749 (vectorizable_load): ...here.
10751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10753 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10755 (vectorizable_mask_load_store): ...here.
10756 (vectorizable_store): ...and here.
10758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10760 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10762 (vectorizable_mask_load_store): ...here.
10764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10766 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10767 (vect_model_store_cost): Take a vec_load_store_type instead of a
10769 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10770 (vect_model_store_cost): Take a vec_load_store_type instead of a
10772 (vectorizable_mask_load_store): Update accordingly.
10773 (vectorizable_store): Likewise.
10774 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10778 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10779 IFN_MASK_LOAD calls here rather than...
10780 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * expmed.c (extract_bit_field_1): For vector extracts,
10787 fall back to extract_bit_field_as_subreg if vec_extract
10790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10791 Alan Hayward <alan.hayward@arm.com>
10792 David Sherwood <david.sherwood@arm.com>
10794 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10795 they are variable or constant sized.
10796 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10797 slots for constant-sized data.
10799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10800 Alan Hayward <alan.hayward@arm.com>
10801 David Sherwood <david.sherwood@arm.com>
10803 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10804 handling COND_EXPRs with boolean comparisons, try to find a better
10805 basis for the mask type than the boolean itself.
10807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10809 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10810 is calculated and how it can be overridden.
10811 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10812 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10814 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10818 Alan Hayward <alan.hayward@arm.com>
10819 David Sherwood <david.sherwood@arm.com>
10821 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10822 Remove the mode argument.
10823 (aarch64_simd_valid_immediate): Remove the mode and inverse
10825 * config/aarch64/iterators.md (bitsize): New iterator.
10826 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10827 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10828 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10829 aarch64_simd_valid_immediate.
10830 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10831 (aarch64_reg_or_bic_imm): Likewise.
10832 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10833 with an insn_type enum and msl with a modifier_type enum.
10834 Replace element_width with a scalar_mode. Change the shift
10835 to unsigned int. Add constructors for scalar_float_mode and
10836 scalar_int_mode elements.
10837 (aarch64_vect_float_const_representable_p): Delete.
10838 (aarch64_can_const_movi_rtx_p)
10839 (aarch64_simd_scalar_immediate_valid_for_move)
10840 (aarch64_simd_make_constant): Update call to
10841 aarch64_simd_valid_immediate.
10842 (aarch64_advsimd_valid_immediate_hs): New function.
10843 (aarch64_advsimd_valid_immediate): Likewise.
10844 (aarch64_simd_valid_immediate): Remove mode and inverse
10845 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10846 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10847 and aarch64_float_const_representable_p on the result.
10848 (aarch64_output_simd_mov_immediate): Remove mode argument.
10849 Update call to aarch64_simd_valid_immediate and use of
10850 simd_immediate_info.
10851 (aarch64_output_scalar_simd_mov_immediate): Update call
10854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10855 Alan Hayward <alan.hayward@arm.com>
10856 David Sherwood <david.sherwood@arm.com>
10858 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10859 (mode_nunits): Likewise CONST_MODE_NUNITS.
10860 * machmode.def (ADJUST_NUNITS): Document.
10861 * genmodes.c (mode_data::need_nunits_adj): New field.
10862 (blank_mode): Update accordingly.
10863 (adj_nunits): New variable.
10864 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10866 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10867 listed in adj_nunits.
10868 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10869 listed in adj_nunits. Don't emit case statements for such modes.
10870 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10871 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10872 nothing if adj_nunits is nonnull.
10873 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10874 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10875 (emit_mode_fbit): Update use of print_maybe_const_decl.
10876 (emit_move_size): Likewise. Treat the array as non-const
10878 (emit_mode_adjustments): Handle adj_nunits.
10880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10882 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10883 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10884 (VECTOR_MODES): Use it.
10885 (make_vector_modes): Take the prefix as an argument.
10887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10888 Alan Hayward <alan.hayward@arm.com>
10889 David Sherwood <david.sherwood@arm.com>
10891 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10892 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10893 for MODE_VECTOR_BOOL.
10894 * machmode.def (VECTOR_BOOL_MODE): Document.
10895 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10896 (make_vector_bool_mode): New function.
10897 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10899 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10900 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10902 * stor-layout.c (int_mode_for_mode): Likewise.
10903 * tree.c (build_vector_type_for_mode): Likewise.
10904 * varasm.c (output_constant_pool_2): Likewise.
10905 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10906 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10907 for MODE_VECTOR_BOOL.
10908 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10909 of mode class checks.
10910 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10911 instead of a list of mode class checks.
10912 (expand_vector_scalar_condition): Likewise.
10913 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10916 Alan Hayward <alan.hayward@arm.com>
10917 David Sherwood <david.sherwood@arm.com>
10919 * machmode.h (mode_size): Change from unsigned short to
10921 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10922 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10923 or if measurement_type is not polynomial.
10924 (fixed_size_mode::includes_p): Check for constant-sized modes.
10925 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10926 return a poly_uint16 rather than an unsigned short.
10927 (emit_mode_size): Change the type of mode_size from unsigned short
10928 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10929 (emit_mode_adjustments): Cope with polynomial vector sizes.
10930 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10932 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10934 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10935 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10936 * caller-save.c (setup_save_areas): Likewise.
10937 (replace_reg_with_saved_mem): Likewise.
10938 * calls.c (emit_library_call_value_1): Likewise.
10939 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10940 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10941 (gen_lowpart_for_combine): Likewise.
10942 * convert.c (convert_to_integer_1): Likewise.
10943 * cse.c (equiv_constant, cse_insn): Likewise.
10944 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10945 (cselib_subst_to_values): Likewise.
10946 * dce.c (word_dce_process_block): Likewise.
10947 * df-problems.c (df_word_lr_mark_ref): Likewise.
10948 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10949 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10950 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10951 (rtl_for_decl_location): Likewise.
10952 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10953 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10954 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10955 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10956 (expand_expr_real_1): Likewise.
10957 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10958 (pad_below): Likewise.
10959 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10960 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10961 * ira.c (get_subreg_tracking_sizes): Likewise.
10962 * ira-build.c (ira_create_allocno_objects): Likewise.
10963 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10964 (ira_sort_regnos_for_alter_reg): Likewise.
10965 * ira-costs.c (record_operand_costs): Likewise.
10966 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10967 (resolve_simple_move): Likewise.
10968 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10969 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10970 (lra_constraints): Likewise.
10971 (CONST_POOL_OK_P): Reject variable-sized modes.
10972 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10973 (add_pseudo_to_slot, lra_spill): Likewise.
10974 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10975 * optabs-query.c (get_best_extraction_insn): Likewise.
10976 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10977 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10978 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10979 * recog.c (offsettable_address_addr_space_p): Likewise.
10980 * regcprop.c (maybe_mode_change): Likewise.
10981 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10982 * regrename.c (build_def_use): Likewise.
10983 * regstat.c (dump_reg_info): Likewise.
10984 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10985 (find_reloads, find_reloads_subreg_address): Likewise.
10986 * reload1.c (eliminate_regs_1): Likewise.
10987 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10988 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10989 (simplify_binary_operation_1, simplify_subreg): Likewise.
10990 * targhooks.c (default_function_arg_padding): Likewise.
10991 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10992 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10993 (verify_gimple_assign_ternary): Likewise.
10994 * tree-inline.c (estimate_move_cost): Likewise.
10995 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10996 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10997 (get_address_cost_ainc): Likewise.
10998 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10999 (vect_supportable_dr_alignment): Likewise.
11000 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11001 (vectorizable_reduction): Likewise.
11002 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11003 (vectorizable_operation, vectorizable_load): Likewise.
11004 * tree.c (build_same_sized_truth_vector_type): Likewise.
11005 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11006 * var-tracking.c (emit_note_insn_var_location): Likewise.
11007 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11008 (ADDR_VEC_ALIGN): Likewise.
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_to_bits): Return a poly_uint16 rather than an
11016 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11017 or if measurement_type is polynomial.
11018 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11019 * combine.c (make_extraction): Likewise.
11020 * dse.c (find_shift_sequence): Likewise.
11021 * dwarf2out.c (mem_loc_descriptor): Likewise.
11022 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11023 (extract_bit_field, extract_low_bits): Likewise.
11024 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11025 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11026 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11027 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11028 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11029 * reload.c (find_reloads): Likewise.
11030 * reload1.c (alter_reg): Likewise.
11031 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11032 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11033 * tree-if-conv.c (predicate_mem_writes): Likewise.
11034 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11035 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11036 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11037 * valtrack.c (dead_debug_insert_temp): Likewise.
11038 * varasm.c (mergeable_constant_section): Likewise.
11039 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * expr.c (expand_assignment): Cope with polynomial mode sizes
11046 when assigning to a CONCAT.
11048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11049 Alan Hayward <alan.hayward@arm.com>
11050 David Sherwood <david.sherwood@arm.com>
11052 * machmode.h (mode_precision): Change from unsigned short to
11054 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11056 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11057 or if measurement_type is not polynomial.
11058 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11059 in which the mode is already known to be a scalar_int_mode.
11060 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11061 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11063 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11064 for GET_MODE_PRECISION.
11065 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11066 for GET_MODE_PRECISION.
11067 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11069 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11070 (expand_field_assignment, make_extraction): Likewise.
11071 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11072 (get_last_value): Likewise.
11073 * convert.c (convert_to_integer_1): Likewise.
11074 * cse.c (cse_insn): Likewise.
11075 * expr.c (expand_expr_real_1): Likewise.
11076 * lra-constraints.c (simplify_operand_subreg): Likewise.
11077 * optabs-query.c (can_atomic_load_p): Likewise.
11078 * optabs.c (expand_atomic_load): Likewise.
11079 (expand_atomic_store): Likewise.
11080 * ree.c (combine_reaching_defs): Likewise.
11081 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11082 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11083 * tree.h (type_has_mode_precision_p): Likewise.
11084 * ubsan.c (instrument_si_overflow): Likewise.
11086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11087 Alan Hayward <alan.hayward@arm.com>
11088 David Sherwood <david.sherwood@arm.com>
11090 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11091 polynomial numbers of units.
11092 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11093 (valid_vector_subparts_p): New function.
11094 (build_vector_type): Remove temporary shim and take the number
11095 of units as a poly_uint64 rather than an int.
11096 (build_opaque_vector_type): Take the number of units as a
11097 poly_uint64 rather than an int.
11098 * tree.c (build_vector_from_ctor): Handle polynomial
11099 TYPE_VECTOR_SUBPARTS.
11100 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11101 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11102 (build_vector_from_val): If the number of units is variable,
11103 use build_vec_duplicate_cst for constant operands and
11104 VEC_DUPLICATE_EXPR otherwise.
11105 (make_vector_type): Remove temporary is_constant ().
11106 (build_vector_type, build_opaque_vector_type): Take the number of
11107 units as a poly_uint64 rather than an int.
11108 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11110 * cfgexpand.c (expand_debug_expr): Likewise.
11111 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11112 (store_constructor, expand_expr_real_1): Likewise.
11113 (const_scalar_mask_from_tree): Likewise.
11114 * fold-const-call.c (fold_const_reduction): Likewise.
11115 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11116 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11117 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11118 (fold_relational_const): Likewise.
11119 (native_interpret_vector): Likewise. Change the size from an
11120 int to an unsigned int.
11121 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11122 TYPE_VECTOR_SUBPARTS.
11123 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11124 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11125 duplicating a non-constant operand into a variable-length vector.
11126 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11127 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11128 * ipa-icf.c (sem_variable::equals): Likewise.
11129 * match.pd: Likewise.
11130 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11131 * print-tree.c (print_node): Likewise.
11132 * stor-layout.c (layout_type): Likewise.
11133 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11134 * tree-cfg.c (verify_gimple_comparison): Likewise.
11135 (verify_gimple_assign_binary): Likewise.
11136 (verify_gimple_assign_ternary): Likewise.
11137 (verify_gimple_assign_single): Likewise.
11138 * tree-pretty-print.c (dump_generic_node): Likewise.
11139 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11140 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11141 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11142 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11143 (vect_shift_permute_load_chain): Likewise.
11144 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11145 (expand_vector_condition, optimize_vector_constructor): Likewise.
11146 (lower_vec_perm, get_compute_type): Likewise.
11147 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11148 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11149 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11150 (vect_recog_mask_conversion_pattern): Likewise.
11151 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11152 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11153 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11154 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11155 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11156 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11157 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11158 (supportable_widening_operation): Likewise.
11159 (supportable_narrowing_operation): Likewise.
11160 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11162 * varasm.c (output_constant): Likewise.
11164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11165 Alan Hayward <alan.hayward@arm.com>
11166 David Sherwood <david.sherwood@arm.com>
11168 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11169 so that both the length == 3 and length != 3 cases set up their
11170 own permute vectors. Add comments explaining why we know the
11171 number of elements is constant.
11172 (vect_permute_load_chain): Likewise.
11174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11175 Alan Hayward <alan.hayward@arm.com>
11176 David Sherwood <david.sherwood@arm.com>
11178 * machmode.h (mode_nunits): Change from unsigned char to
11180 (ONLY_FIXED_SIZE_MODES): New macro.
11181 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11182 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11183 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11185 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11186 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11187 or if measurement_type is not polynomial.
11188 * genmodes.c (ZERO_COEFFS): New macro.
11189 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11191 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11192 Use ZERO_COEFFS when emitting initializers.
11193 * data-streamer.h (bp_pack_poly_value): New function.
11194 (bp_unpack_poly_value): Likewise.
11195 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11196 for GET_MODE_NUNITS.
11197 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11198 for GET_MODE_NUNITS.
11199 * tree.c (make_vector_type): Remove temporary shim and make
11200 the real function take the number of units as a poly_uint64
11201 rather than an int.
11202 (build_vector_type_for_mode): Handle polynomial nunits.
11203 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11204 * emit-rtl.c (const_vec_series_p_1): Likewise.
11205 (gen_rtx_CONST_VECTOR): Likewise.
11206 * fold-const.c (test_vec_duplicate_folding): Likewise.
11207 * genrecog.c (validate_pattern): Likewise.
11208 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11209 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11210 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11211 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11212 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11213 * rtlanal.c (subreg_get_info): Likewise.
11214 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11215 (vect_grouped_load_supported): Likewise.
11216 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11217 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11218 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11219 (simplify_const_unary_operation, simplify_binary_operation_1)
11220 (simplify_const_binary_operation, simplify_ternary_operation)
11221 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11222 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11223 instead of CONST_VECTOR_NUNITS.
11224 * varasm.c (output_constant_pool_2): Likewise.
11225 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11226 explicit-encoded elements in the XVEC for variable-length vectors.
11228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11230 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11233 Alan Hayward <alan.hayward@arm.com>
11234 David Sherwood <david.sherwood@arm.com>
11236 * coretypes.h (fixed_size_mode): Declare.
11237 (fixed_size_mode_pod): New typedef.
11238 * builtins.h (target_builtins::x_apply_args_mode)
11239 (target_builtins::x_apply_result_mode): Change type to
11240 fixed_size_mode_pod.
11241 * builtins.c (apply_args_size, apply_result_size, result_vector)
11242 (expand_builtin_apply_args_1, expand_builtin_apply)
11243 (expand_builtin_return): Update accordingly.
11245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11247 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11248 * cselib.c (cselib_hash_rtx): Likewise.
11249 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11250 CONST_VECTOR encoding.
11252 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11253 Jeff Law <law@redhat.com>
11256 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11257 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11258 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11259 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11262 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11263 explicitly probe *sp in a noreturn function if there were any callee
11264 register saves or frame pointer is needed.
11266 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11269 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11270 BLKmode for ternary, binary or unary expressions.
11273 * var-tracking.c (delete_vta_debug_insn): New inline function.
11274 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11275 insns from get_insns () to NULL instead of each bb separately.
11276 Use delete_vta_debug_insn. No longer static.
11277 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11278 delete_vta_debug_insns callers.
11279 * rtl.h (delete_vta_debug_insns): Declare.
11280 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11281 instead of variable_tracking_main.
11283 2018-01-03 Martin Sebor <msebor@redhat.com>
11285 PR tree-optimization/83603
11286 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11287 arguments past the endof the argument list in functions declared
11288 without a prototype.
11289 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11290 Avoid checking when arguments are null.
11292 2018-01-03 Martin Sebor <msebor@redhat.com>
11295 * doc/extend.texi (attribute const): Fix a typo.
11296 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11297 issuing -Wsuggest-attribute for void functions.
11299 2018-01-03 Martin Sebor <msebor@redhat.com>
11301 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11302 offset_int::from instead of wide_int::to_shwi.
11303 (maybe_diag_overlap): Remove assertion.
11304 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11305 * gimple-ssa-sprintf.c (format_directive): Same.
11306 (parse_directive): Same.
11307 (sprintf_dom_walker::compute_format_length): Same.
11308 (try_substitute_return_value): Same.
11310 2018-01-03 Jeff Law <law@redhat.com>
11312 PR middle-end/83654
11313 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11314 non-constant residual for zero at runtime and avoid probing in
11315 that case. Reorganize code for trailing problem to mirror handling
11318 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11320 PR tree-optimization/83501
11321 * tree-ssa-strlen.c (get_string_cst): New.
11322 (handle_char_store): Call get_string_cst.
11324 2018-01-03 Martin Liska <mliska@suse.cz>
11326 PR tree-optimization/83593
11327 * tree-ssa-strlen.c: Include tree-cfg.h.
11328 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11329 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11330 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11332 (strlen_dom_walker::before_dom_children): Call
11333 gimple_purge_dead_eh_edges. Dump tranformation with details
11335 (strlen_dom_walker::before_dom_children): Update call by adding
11336 new argument cleanup_eh.
11337 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11339 2018-01-03 Martin Liska <mliska@suse.cz>
11342 * cif-code.def (VARIADIC_THUNK): New enum value.
11343 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11346 2018-01-03 Jan Beulich <jbeulich@suse.com>
11348 * sse.md (mov<mode>_internal): Tighten condition for when to use
11349 vmovdqu<ssescalarsize> for TI and OI modes.
11351 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11353 Update copyright years.
11355 2018-01-03 Martin Liska <mliska@suse.cz>
11358 * ipa-visibility.c (function_and_variable_visibility): Skip
11359 functions with noipa attribure.
11361 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11363 * gcc.c (process_command): Update copyright notice dates.
11364 * gcov-dump.c (print_version): Ditto.
11365 * gcov.c (print_version): Ditto.
11366 * gcov-tool.c (print_version): Ditto.
11367 * gengtype.c (create_file): Ditto.
11368 * doc/cpp.texi: Bump @copying's copyright year.
11369 * doc/cppinternals.texi: Ditto.
11370 * doc/gcc.texi: Ditto.
11371 * doc/gccint.texi: Ditto.
11372 * doc/gcov.texi: Ditto.
11373 * doc/install.texi: Ditto.
11374 * doc/invoke.texi: Ditto.
11376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11378 * vector-builder.h (vector_builder::m_full_nelts): Change from
11379 unsigned int to poly_uint64.
11380 (vector_builder::full_nelts): Update prototype accordingly.
11381 (vector_builder::new_vector): Likewise.
11382 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11383 (vector_builder::operator ==): Likewise.
11384 (vector_builder::finalize): Likewise.
11385 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11386 Take the number of elements as a poly_uint64 rather than an
11388 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11389 from unsigned int to poly_uint64.
11390 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11391 (vec_perm_indices::new_vector): Likewise.
11392 (vec_perm_indices::length): Likewise.
11393 (vec_perm_indices::nelts_per_input): Likewise.
11394 (vec_perm_indices::input_nelts): Likewise.
11395 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11396 number of elements per input as a poly_uint64 rather than an
11397 unsigned int. Use the original encoding for variable-length
11398 vectors, rather than clamping each individual element.
11399 For the second and subsequent elements in each pattern,
11400 clamp the step and base before clamping their sum.
11401 (vec_perm_indices::series_p): Handle polynomial element counts.
11402 (vec_perm_indices::all_in_range_p): Likewise.
11403 (vec_perm_indices_to_tree): Likewise.
11404 (vec_perm_indices_to_rtx): Likewise.
11405 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11406 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11407 (tree_vector_builder::new_binary_operation): Handle polynomial
11408 element counts. Return false if we need to know the number
11409 of elements at compile time.
11410 * fold-const.c (fold_vec_perm): Punt if the number of elements
11411 isn't known at compile time.
11413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 * vec-perm-indices.h (vec_perm_builder): Change element type
11416 from HOST_WIDE_INT to poly_int64.
11417 (vec_perm_indices::element_type): Update accordingly.
11418 (vec_perm_indices::clamp): Handle polynomial element_types.
11419 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11420 (vec_perm_indices::all_in_range_p): Likewise.
11421 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11423 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11424 polynomial vec_perm_indices element types.
11425 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11426 * fold-const.c (fold_vec_perm): Likewise.
11427 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11428 * tree-vect-generic.c (lower_vec_perm): Likewise.
11429 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11430 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11431 element type to HOST_WIDE_INT.
11433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11434 Alan Hayward <alan.hayward@arm.com>
11435 David Sherwood <david.sherwood@arm.com>
11437 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11438 rather than an int. Use plus_constant.
11439 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11440 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11443 Alan Hayward <alan.hayward@arm.com>
11444 David Sherwood <david.sherwood@arm.com>
11446 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11447 a HOST_WIDE_INT to a poly_int64.
11449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11450 Alan Hayward <alan.hayward@arm.com>
11451 David Sherwood <david.sherwood@arm.com>
11453 * calls.c (load_register_parameters): Cope with polynomial
11454 mode sizes. Require a constant size for BLKmode parameters
11455 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11456 forces a parameter to be padded at the lsb end in order to
11457 fill a complete number of words, require the parameter size
11458 to be ordered wrt UNITS_PER_WORD.
11460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11461 Alan Hayward <alan.hayward@arm.com>
11462 David Sherwood <david.sherwood@arm.com>
11464 * reload1.c (spill_stack_slot_width): Change element type
11465 from unsigned int to poly_uint64_pod.
11466 (alter_reg): Treat mode sizes as polynomial.
11468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11469 Alan Hayward <alan.hayward@arm.com>
11470 David Sherwood <david.sherwood@arm.com>
11472 * reload.c (complex_word_subreg_p): New function.
11473 (reload_inner_reg_of_subreg, push_reload): Use it.
11475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11476 Alan Hayward <alan.hayward@arm.com>
11477 David Sherwood <david.sherwood@arm.com>
11479 * lra-constraints.c (process_alt_operands): Reject matched
11480 operands whose sizes aren't ordered.
11481 (match_reload): Refer to this check here.
11483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11484 Alan Hayward <alan.hayward@arm.com>
11485 David Sherwood <david.sherwood@arm.com>
11487 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11488 that the mode size is in the set {1, 2, 4, 8, 16}.
11490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11491 Alan Hayward <alan.hayward@arm.com>
11492 David Sherwood <david.sherwood@arm.com>
11494 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11495 Use plus_constant instead of gen_rtx_PLUS.
11497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11498 Alan Hayward <alan.hayward@arm.com>
11499 David Sherwood <david.sherwood@arm.com>
11501 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11502 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11503 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11504 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11505 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11506 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11507 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11508 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11509 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11510 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11512 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11513 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11514 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11515 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11516 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11517 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11518 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11519 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11520 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11521 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11523 * expr.c (emit_move_resolve_push): Treat the input and result
11524 of PUSH_ROUNDING as a poly_int64.
11525 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11526 (emit_push_insn): Likewise.
11527 * lra-eliminations.c (mark_not_eliminable): Likewise.
11528 * recog.c (push_operand): Likewise.
11529 * reload1.c (elimination_effects): Likewise.
11530 * rtlanal.c (nonzero_bits1): Likewise.
11531 * calls.c (store_one_arg): Likewise. Require the padding to be
11532 known at compile time.
11534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11535 Alan Hayward <alan.hayward@arm.com>
11536 David Sherwood <david.sherwood@arm.com>
11538 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11539 Use plus_constant instead of gen_rtx_PLUS.
11541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11542 Alan Hayward <alan.hayward@arm.com>
11543 David Sherwood <david.sherwood@arm.com>
11545 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11546 rather than an int.
11548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11549 Alan Hayward <alan.hayward@arm.com>
11550 David Sherwood <david.sherwood@arm.com>
11552 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11553 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11554 via stack temporaries. Treat the mode size as polynomial too.
11556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11557 Alan Hayward <alan.hayward@arm.com>
11558 David Sherwood <david.sherwood@arm.com>
11560 * expr.c (expand_expr_real_2): When handling conversions involving
11561 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11562 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11563 as a poly_uint64 too.
11565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11566 Alan Hayward <alan.hayward@arm.com>
11567 David Sherwood <david.sherwood@arm.com>
11569 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11572 Alan Hayward <alan.hayward@arm.com>
11573 David Sherwood <david.sherwood@arm.com>
11575 * combine.c (can_change_dest_mode): Handle polynomial
11576 REGMODE_NATURAL_SIZE.
11577 * expmed.c (store_bit_field_1): Likewise.
11578 * expr.c (store_constructor): Likewise.
11579 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11580 and polynomial REGMODE_NATURAL_SIZE.
11581 (gen_lowpart_common): Likewise.
11582 * reginfo.c (record_subregs_of_mode): Likewise.
11583 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11590 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
11598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11599 Alan Hayward <alan.hayward@arm.com>
11600 David Sherwood <david.sherwood@arm.com>
11602 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11603 in a POINTER_PLUS_EXPR.
11605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11606 Alan Hayward <alan.hayward@arm.com>
11607 David Sherwood <david.sherwood@arm.com>
11609 * omp-simd-clone.c (simd_clone_subparts): New function.
11610 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11611 (ipa_simd_modify_function_body): Likewise.
11613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11614 Alan Hayward <alan.hayward@arm.com>
11615 David Sherwood <david.sherwood@arm.com>
11617 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11618 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11619 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11620 (expand_vector_condition, vector_element): Likewise.
11621 (subparts_gt): New function.
11622 (get_compute_type): Use subparts_gt.
11623 (count_type_subparts): Delete.
11624 (expand_vector_operations_1): Use subparts_gt instead of
11625 count_type_subparts.
11627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11628 Alan Hayward <alan.hayward@arm.com>
11629 David Sherwood <david.sherwood@arm.com>
11631 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11632 (vect_compile_time_alias): ...this new function. Do the calculation
11633 on poly_ints rather than trees.
11634 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11642 (vect_schedule_slp_instance): Likewise.
11644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11645 Alan Hayward <alan.hayward@arm.com>
11646 David Sherwood <david.sherwood@arm.com>
11648 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11649 constant and extern definitions for variable-length vectors.
11650 (vect_get_constant_vectors): Note that the number of units
11651 is known to be constant.
11653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11654 Alan Hayward <alan.hayward@arm.com>
11655 David Sherwood <david.sherwood@arm.com>
11657 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11658 of units as polynomial. Choose between WIDE and NARROW based
11661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11662 Alan Hayward <alan.hayward@arm.com>
11663 David Sherwood <david.sherwood@arm.com>
11665 * tree-vect-stmts.c (simd_clone_subparts): New function.
11666 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11669 Alan Hayward <alan.hayward@arm.com>
11670 David Sherwood <david.sherwood@arm.com>
11672 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11673 vectors as polynomial. Use build_index_vector for
11674 IFN_GOMP_SIMD_LANE.
11676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11677 Alan Hayward <alan.hayward@arm.com>
11678 David Sherwood <david.sherwood@arm.com>
11680 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11681 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11682 for variable-length vectors.
11683 (vectorizable_mask_load_store): Treat the number of units as
11684 polynomial, asserting that it is constant if the condition has
11685 already been enforced.
11686 (vectorizable_store, vectorizable_load): Likewise.
11688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11689 Alan Hayward <alan.hayward@arm.com>
11690 David Sherwood <david.sherwood@arm.com>
11692 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11693 of units as polynomial. Punt if we can't tell at compile time
11694 which vector contains the final result.
11696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11697 Alan Hayward <alan.hayward@arm.com>
11698 David Sherwood <david.sherwood@arm.com>
11700 * tree-vect-loop.c (vectorizable_induction): Treat the number
11701 of units as polynomial. Punt on SLP inductions. Use an integer
11702 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11703 cast of such a series for variable-length floating-point
11706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11707 Alan Hayward <alan.hayward@arm.com>
11708 David Sherwood <david.sherwood@arm.com>
11710 * tree.h (build_index_vector): Declare.
11711 * tree.c (build_index_vector): New function.
11712 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11713 of units as polynomial, forcibly converting it to a constant if
11714 vectorizable_reduction has already enforced the condition.
11715 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11716 to create a {1,2,3,...} vector.
11717 (vectorizable_reduction): Treat the number of units as polynomial.
11718 Choose vectype_in based on the largest scalar element size rather
11719 than the smallest number of units. Enforce the restrictions
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 (vector_alignment_reachable_p): Treat the
11727 number of units as polynomial.
11729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11730 Alan Hayward <alan.hayward@arm.com>
11731 David Sherwood <david.sherwood@arm.com>
11733 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11734 * target.def (autovectorize_vector_sizes): Return the vector sizes
11735 by pointer, using vector_sizes rather than a bitmask.
11736 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11737 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11738 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11740 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11741 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11742 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11743 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11744 * omp-general.c (omp_max_vf): Likewise.
11745 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11746 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11747 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11748 * tree-vect-slp.c (vect_slp_bb): Likewise.
11749 * doc/tm.texi: Regenerate.
11750 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11752 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11753 the vector size as a poly_uint64 rather than an unsigned int.
11754 (current_vector_size): Change from an unsigned int to a poly_uint64.
11755 (get_vectype_for_scalar_type): Update accordingly.
11756 * tree.h (build_truth_vector_type): Take the size and number of
11757 units as a poly_uint64 rather than an unsigned int.
11758 (build_vector_type): Add a temporary overload that takes
11759 the number of units as a poly_uint64 rather than an unsigned int.
11760 * tree.c (make_vector_type): Likewise.
11761 (build_truth_vector_type): Take the number of units as a poly_uint64
11762 rather than an unsigned int.
11764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11765 Alan Hayward <alan.hayward@arm.com>
11766 David Sherwood <david.sherwood@arm.com>
11768 * target.def (get_mask_mode): Take the number of units and length
11769 as poly_uint64s rather than unsigned ints.
11770 * targhooks.h (default_get_mask_mode): Update accordingly.
11771 * targhooks.c (default_get_mask_mode): Likewise.
11772 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11773 * doc/tm.texi: Regenerate.
11775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11776 Alan Hayward <alan.hayward@arm.com>
11777 David Sherwood <david.sherwood@arm.com>
11779 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11780 * omp-general.c (omp_max_vf): Likewise.
11781 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11782 (expand_omp_simd): Handle polynomial safelen.
11783 * omp-low.c (omplow_simd_context): Add a default constructor.
11784 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11785 (lower_rec_simd_input_clauses): Update accordingly.
11786 (lower_rec_input_clauses): Likewise.
11788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11789 Alan Hayward <alan.hayward@arm.com>
11790 David Sherwood <david.sherwood@arm.com>
11792 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11793 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11794 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11795 (vect_analyze_slp_cost): Likewise.
11796 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11797 (vect_model_load_cost): Likewise.
11799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11800 Alan Hayward <alan.hayward@arm.com>
11801 David Sherwood <david.sherwood@arm.com>
11803 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11804 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11805 from an unsigned int * to a poly_uint64_pod *.
11806 (calculate_unrolling_factor): New function.
11807 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11810 Alan Hayward <alan.hayward@arm.com>
11811 David Sherwood <david.sherwood@arm.com>
11813 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11814 from an unsigned int to a poly_uint64.
11815 (_loop_vec_info::slp_unrolling_factor): Likewise.
11816 (_loop_vec_info::vectorization_factor): Change from an int
11818 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11819 (vect_get_num_vectors): New function.
11820 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11821 (vect_get_num_copies): Use vect_get_num_vectors.
11822 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11823 to an unsigned int *.
11824 (vect_analyze_data_refs): Change min_vf from an int * to a
11826 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11827 than an unsigned HOST_WIDE_INT.
11828 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11829 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11830 to an unsigned int *.
11831 (vect_analyze_data_ref_dependences): Likewise.
11832 (vect_compute_data_ref_alignment): Handle polynomial vf.
11833 (vect_enhance_data_refs_alignment): Likewise.
11834 (vect_prune_runtime_alias_test_list): Likewise.
11835 (vect_shift_permute_load_chain): Likewise.
11836 (vect_supportable_dr_alignment): Likewise.
11837 (dependence_distance_ge_vf): Take the vectorization factor as a
11838 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11839 (vect_analyze_data_refs): Change min_vf from an int * to a
11841 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11842 vfm1 as a poly_uint64 rather than an int. Make the same change
11843 for the returned bound_scalar.
11844 (vect_gen_vector_loop_niters): Handle polynomial vf.
11845 (vect_do_peeling): Likewise. Update call to
11846 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11847 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11849 * tree-vect-loop.c (vect_determine_vectorization_factor)
11850 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11851 (vect_get_known_peeling_cost): Likewise.
11852 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11853 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11854 (vect_transform_loop): Likewise. Use the lowest possible VF when
11855 updating the upper bounds of the loop.
11856 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11857 rather than an int.
11858 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11859 polynomial unroll factors.
11860 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11861 (vect_make_slp_decision): Likewise.
11862 (vect_supported_load_permutation_p): Likewise, and polynomial
11864 (vect_analyze_slp_cost): Handle polynomial vf.
11865 (vect_slp_analyze_node_operations): Likewise.
11866 (vect_slp_analyze_bb_1): Likewise.
11867 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11868 than an unsigned HOST_WIDE_INT.
11869 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11870 (vectorizable_load): Handle polynomial vf.
11871 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11873 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11876 Alan Hayward <alan.hayward@arm.com>
11877 David Sherwood <david.sherwood@arm.com>
11879 * match.pd: Handle bit operations involving three constants
11880 and try to fold one pair.
11882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11884 * tree-vect-loop-manip.c: Include gimple-fold.h.
11885 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11886 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11887 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11888 Add a path that uses a step of VF instead of 1, but disable it
11890 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11891 and niters_no_overflow parameters. Update calls to
11892 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11893 Create a new SSA name if the latter choses to use a ste other
11894 than zero, and return it via niters_vector_mult_vf_var.
11895 * tree-vect-loop.c (vect_transform_loop): Update calls to
11896 vect_do_peeling, vect_gen_vector_loop_niters and
11897 slpeel_make_loop_iterate_ntimes.
11898 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11899 (vect_gen_vector_loop_niters): Update declarations after above changes.
11901 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11903 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11904 128-bit round to integer instructions.
11905 (ceil<mode>2): Likewise.
11906 (btrunc<mode>2): Likewise.
11907 (round<mode>2): Likewise.
11909 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11911 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11912 unaligned VSX load/store on P8/P9.
11913 (expand_block_clear): Allow the use of unaligned VSX
11914 load/store on P8/P9.
11916 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11918 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11920 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11921 swap associated with both a load and a store.
11923 2018-01-02 Andrew Waterman <andrew@sifive.com>
11925 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11926 * config/riscv/riscv.md (clear_cache): Use it.
11928 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11930 * web.c: Remove out-of-date comment.
11932 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11934 * expr.c (fixup_args_size_notes): Check that any existing
11935 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11936 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11937 (emit_single_push_insn): ...here.
11939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11941 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11942 (const_vector_encoded_nelts): New function.
11943 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11944 (const_vector_int_elt, const_vector_elt): Declare.
11945 * emit-rtl.c (const_vector_int_elt_1): New function.
11946 (const_vector_elt): Likewise.
11947 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11948 of CONST_VECTOR_ELT.
11950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11952 * expr.c: Include rtx-vector-builder.h.
11953 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11954 directly on the tree encoding.
11955 (const_vector_from_tree): Likewise.
11956 * optabs.c: Include rtx-vector-builder.h.
11957 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11958 sequence of "u" values.
11959 * vec-perm-indices.c: Include rtx-vector-builder.h.
11960 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11961 directly on the vec_perm_indices encoding.
11963 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11965 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11966 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11967 * rtx-vector-builder.h: New file.
11968 * rtx-vector-builder.c: Likewise.
11969 * rtl.h (rtx_def::u2): Add a const_vector field.
11970 (CONST_VECTOR_NPATTERNS): New macro.
11971 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11972 (CONST_VECTOR_DUPLICATE_P): Likewise.
11973 (CONST_VECTOR_STEPPED_P): Likewise.
11974 (CONST_VECTOR_ENCODED_ELT): Likewise.
11975 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11976 (unwrap_const_vec_duplicate): Likewise.
11977 (const_vec_series_p): Check for a non-duplicated vector encoding.
11978 Say that the function only returns true for integer vectors.
11979 * emit-rtl.c: Include rtx-vector-builder.h.
11980 (gen_const_vec_duplicate_1): Delete.
11981 (gen_const_vector): Call gen_const_vec_duplicate instead of
11982 gen_const_vec_duplicate_1.
11983 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11984 (gen_const_vec_duplicate): Use rtx_vector_builder.
11985 (gen_const_vec_series): Likewise.
11986 (gen_rtx_CONST_VECTOR): Likewise.
11987 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11988 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11989 Build a new vector rather than modifying a CONST_VECTOR in-place.
11990 (handle_special_swappables): Update call accordingly.
11991 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11992 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11993 Build a new vector rather than modifying a CONST_VECTOR in-place.
11994 (handle_special_swappables): Update call accordingly.
11996 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11998 * simplify-rtx.c (simplify_const_binary_operation): Use
11999 CONST_VECTOR_ELT instead of XVECEXP.
12001 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12003 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12004 the selector elements to be different from the data elements
12005 if the selector is a VECTOR_CST.
12006 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12007 ssizetype for the selector.
12009 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12011 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12012 before testing each element individually.
12013 * tree-vect-generic.c (lower_vec_perm): Likewise.
12015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12017 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12018 * selftest-run-tests.c (selftest::run_tests): Call it.
12019 * vector-builder.h (vector_builder::operator ==): New function.
12020 (vector_builder::operator !=): Likewise.
12021 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12022 (vec_perm_indices::all_from_input_p): New function.
12023 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12024 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12025 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12026 instead of reading the VECTOR_CST directly. Detect whether both
12027 vector inputs are the same before constructing the vec_perm_indices,
12028 and update the number of inputs argument accordingly. Use the
12029 utility functions added above. Only construct sel2 if we need to.
12031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12033 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12034 the broadcast of the low byte.
12035 (expand_mult_highpart): Use an explicit encoding for the permutes.
12036 * optabs-query.c (can_mult_highpart_p): Likewise.
12037 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12038 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12039 (vectorizable_bswap): Likewise.
12040 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12041 explicit encoding for the power-of-2 permutes.
12042 (vect_permute_store_chain): Likewise.
12043 (vect_grouped_load_supported): Likewise.
12044 (vect_permute_load_chain): Likewise.
12046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12048 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12049 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12050 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12051 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12052 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12053 (vect_gen_perm_mask_any): Likewise.
12055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12057 * int-vector-builder.h: New file.
12058 * vec-perm-indices.h: Include int-vector-builder.h.
12059 (vec_perm_indices): Redefine as an int_vector_builder.
12060 (auto_vec_perm_indices): Delete.
12061 (vec_perm_builder): Redefine as a stand-alone class.
12062 (vec_perm_indices::vec_perm_indices): New function.
12063 (vec_perm_indices::clamp): Likewise.
12064 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12065 (vec_perm_indices::new_vector): New function.
12066 (vec_perm_indices::new_expanded_vector): Update for new
12067 vec_perm_indices class.
12068 (vec_perm_indices::rotate_inputs): New function.
12069 (vec_perm_indices::all_in_range_p): Operate directly on the
12070 encoded form, without computing elided elements.
12071 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12072 encoding. Update for new vec_perm_indices class.
12073 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12074 the given vec_perm_builder.
12075 (expand_vec_perm_var): Update vec_perm_builder constructor.
12076 (expand_mult_highpart): Use vec_perm_builder instead of
12077 auto_vec_perm_indices.
12078 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12079 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12080 or double series encoding as appropriate.
12081 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12082 vec_perm_indices instead of auto_vec_perm_indices.
12083 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12084 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12085 (vect_permute_store_chain): Likewise.
12086 (vect_grouped_load_supported): Likewise.
12087 (vect_permute_load_chain): Likewise.
12088 (vect_shift_permute_load_chain): Likewise.
12089 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12090 (vect_transform_slp_perm_load): Likewise.
12091 (vect_schedule_slp_instance): Likewise.
12092 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12093 (vectorizable_mask_load_store): Likewise.
12094 (vectorizable_bswap): Likewise.
12095 (vectorizable_store): Likewise.
12096 (vectorizable_load): Likewise.
12097 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12098 vec_perm_indices instead of auto_vec_perm_indices. Use
12099 tree_to_vec_perm_builder to read the vector from a tree.
12100 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12101 vec_perm_builder instead of a vec_perm_indices.
12102 (have_whole_vector_shift): Use vec_perm_builder and
12103 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12104 truncation to calc_vec_perm_mask_for_shift.
12105 (vect_create_epilog_for_reduction): Likewise.
12106 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12107 from auto_vec_perm_indices to vec_perm_indices.
12108 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12109 instead of changing individual elements.
12110 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12111 the vector in d.perm.
12112 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12113 from auto_vec_perm_indices to vec_perm_indices.
12114 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12115 instead of changing individual elements.
12116 (arm_vectorize_vec_perm_const): Use new_vector to install
12117 the vector in d.perm.
12118 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12119 Update vec_perm_builder constructor.
12120 (rs6000_expand_interleave): Likewise.
12121 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12122 (rs6000_expand_interleave): Likewise.
12124 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12126 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12127 to qimode could truncate the indices.
12128 * optabs.c (expand_vec_perm_var): Likewise.
12130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12132 * Makefile.in (OBJS): Add vec-perm-indices.o.
12133 * vec-perm-indices.h: New file.
12134 * vec-perm-indices.c: Likewise.
12135 * target.h (vec_perm_indices): Replace with a forward class
12137 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12138 * optabs.h: Include vec-perm-indices.h.
12139 (expand_vec_perm): Delete.
12140 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12141 (expand_vec_perm_const): Declare.
12142 * target.def (vec_perm_const_ok): Replace with...
12143 (vec_perm_const): ...this new hook.
12144 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12145 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12146 * doc/tm.texi: Regenerate.
12147 * optabs.def (vec_perm_const): Delete.
12148 * doc/md.texi (vec_perm_const): Likewise.
12149 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12150 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12151 expand_vec_perm for constant permutation vectors. Assert that
12152 the mode of variable permutation vectors is the integer equivalent
12153 of the mode that is being permuted.
12154 * optabs-query.h (selector_fits_mode_p): Declare.
12155 * optabs-query.c: Include vec-perm-indices.h.
12156 (selector_fits_mode_p): New function.
12157 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12158 is defined, instead of checking whether the vec_perm_const_optab
12159 exists. Use targetm.vectorize.vec_perm_const instead of
12160 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12161 fit in the vector mode before using a variable permute.
12162 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12163 vec_perm_indices instead of an rtx.
12164 (expand_vec_perm): Replace with...
12165 (expand_vec_perm_const): ...this new function. Take the selector
12166 as a vec_perm_indices rather than an rtx. Also take the mode of
12167 the selector. Update call to shift_amt_for_vec_perm_mask.
12168 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12169 Use vec_perm_indices::new_expanded_vector to expand the original
12170 selector into bytes. Check whether the indices fit in the vector
12171 mode before using a variable permute.
12172 (expand_vec_perm_var): Make global.
12173 (expand_mult_highpart): Use expand_vec_perm_const.
12174 * fold-const.c: Includes vec-perm-indices.h.
12175 * tree-ssa-forwprop.c: Likewise.
12176 * tree-vect-data-refs.c: Likewise.
12177 * tree-vect-generic.c: Likewise.
12178 * tree-vect-loop.c: Likewise.
12179 * tree-vect-slp.c: Likewise.
12180 * tree-vect-stmts.c: Likewise.
12181 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12183 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12184 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12185 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12186 (aarch64_vectorize_vec_perm_const): ...this new function.
12187 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12188 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12189 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12190 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12191 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12192 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12193 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12195 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12196 check for NEON modes.
12197 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12198 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12199 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12200 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12202 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12203 the old VEC_PERM_CONST conditions.
12204 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12205 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12206 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12207 (ia64_vectorize_vec_perm_const_ok): Merge into...
12208 (ia64_vectorize_vec_perm_const): ...this new function.
12209 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12210 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12211 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12212 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12213 * config/mips/mips.c (mips_expand_vec_perm_const)
12214 (mips_vectorize_vec_perm_const_ok): Merge into...
12215 (mips_vectorize_vec_perm_const): ...this new function.
12216 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12217 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12218 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12219 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12220 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12221 (rs6000_expand_vec_perm_const): Delete.
12222 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12224 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12225 (altivec_expand_vec_perm_const_le): Take each operand individually.
12226 Operate on constant selectors rather than rtxes.
12227 (altivec_expand_vec_perm_const): Likewise. Update call to
12228 altivec_expand_vec_perm_const_le.
12229 (rs6000_expand_vec_perm_const): Delete.
12230 (rs6000_vectorize_vec_perm_const_ok): Delete.
12231 (rs6000_vectorize_vec_perm_const): New function.
12232 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12233 an element count and rtx array.
12234 (rs6000_expand_extract_even): Update call accordingly.
12235 (rs6000_expand_interleave): Likewise.
12236 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12237 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12238 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12239 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12240 (rs6000_expand_vec_perm_const): Delete.
12241 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12242 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12243 (altivec_expand_vec_perm_const_le): Take each operand individually.
12244 Operate on constant selectors rather than rtxes.
12245 (altivec_expand_vec_perm_const): Likewise. Update call to
12246 altivec_expand_vec_perm_const_le.
12247 (rs6000_expand_vec_perm_const): Delete.
12248 (rs6000_vectorize_vec_perm_const_ok): Delete.
12249 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12250 reference to the SPE evmerge intructions.
12251 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12252 an element count and rtx array.
12253 (rs6000_expand_extract_even): Update call accordingly.
12254 (rs6000_expand_interleave): Likewise.
12255 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12256 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12258 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12260 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12262 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12263 vector mode and that that mode matches the mode of the data
12265 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12266 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12267 directly using expand_vec_perm_1 when forcing selectors into
12269 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12273 * optabs-query.h (can_vec_perm_p): Delete.
12274 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12275 * optabs-query.c (can_vec_perm_p): Split into...
12276 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12277 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12278 particular selector is valid.
12279 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12280 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12281 (vect_grouped_load_supported): Likewise.
12282 (vect_shift_permute_load_chain): Likewise.
12283 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12284 (vect_transform_slp_perm_load): Likewise.
12285 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12286 (vectorizable_bswap): Likewise.
12287 (vect_gen_perm_mask_checked): Likewise.
12288 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12289 implementations of variable permutation vectors into account
12290 when deciding which selector to use.
12291 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12292 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12293 with a false third argument.
12294 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12295 to test whether the constant selector is valid and can_vec_perm_var_p
12296 to test whether a variable selector is valid.
12298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12300 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12301 * optabs-query.c (can_vec_perm_p): Likewise.
12302 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12303 instead of vec_perm_indices.
12304 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12305 (vect_gen_perm_mask_checked): Likewise,
12306 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12307 (vect_gen_perm_mask_checked): Likewise,
12309 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12311 * optabs-query.h (qimode_for_vec_perm): Declare.
12312 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12313 (qimode_for_vec_perm): ...this new function.
12314 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12316 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12318 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12319 does not have a conditional at the top.
12321 2018-01-02 Richard Biener <rguenther@suse.de>
12323 * ipa-inline.c (big_speedup_p): Fix expression.
12325 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12328 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12331 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12335 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12336 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12337 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12338 cond_taken_branch_cost 3->4.
12340 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12342 PR tree-optimization/83581
12343 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12344 TODO_cleanup_cfg if any changes have been made.
12346 PR middle-end/83608
12347 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12348 convert_modes if target mode has the right side, but different mode
12351 PR middle-end/83609
12352 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12353 last argument when extracting from CONCAT. If either from_real or
12354 from_imag is NULL, use expansion through memory. If result is not
12355 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12356 the parts directly to inner mode, if even that fails, use expansion
12359 PR middle-end/83623
12360 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12361 check for bswap in mode rather than HImode and use that in expand_unop
12364 Copyright (C) 2018 Free Software Foundation, Inc.
12366 Copying and distribution of this file, with or without modification,
12367 are permitted in any medium without royalty provided the copyright
12368 notice and this notice are preserved.