1 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4 (*movdf_internal): Ditto.
6 (*rsqrtsf2_sse): Ditto.
7 (*sqrt<mode>2_sse): Ditto.
9 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
13 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18 * config/aarch64/arm_neon.h: Likewise.
19 * config/aarch64/iterators.md (VQ_I): New.
21 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
23 * config.gcc: Add arc/t-multilib-linux to tmake_file for
25 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
28 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
30 * config/nds32/constraints.md (S): New constraint.
31 * config/nds32/nds32.md (call_internal): Use constraint S.
32 (call_value_internal): Likewise.
33 (sibcall_internal): Likewise.
34 (sibcall_value_internal): Likewise.
36 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
37 Chung-Ju Wu <jasonwucj@gmail.com>
39 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
42 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
43 Chung-Ju Wu <jasonwucj@gmail.com>
45 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
46 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
47 (nds32_rtx_costs_impl): Simplify.
48 (nds32_address_cost_impl): Simplify.
49 (nds32_init_rtx_costs): New function.
50 (nds32_rtx_costs_speed_prefer): Likewise.
51 (nds32_rtx_costs_size_prefer): Likewise.
52 (nds32_address_cost_speed_prefer): Likewise.
53 (nds32_address_cost_speed_fwprop): Likewise.
54 (nds32_address_cost_size_prefer): Likewise.
55 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
56 * config/nds32/nds32.c (nds32_option_override): Use
57 nds32_init_rtx_costs function.
59 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
61 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
62 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
63 (TARGET_PIPELINE_N8): Likewise.
64 (TARGET_PIPELINE_N10): Likewise.
65 (TARGET_PIPELINE_N13): Likewise.
66 (TARGET_PIPELINE_GRAYWOLF): Likewise.
68 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
70 * config/nds32/nds32-fpu.md: Update copyright year.
72 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
74 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
76 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
78 * config/nds32/nds32.c
79 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
80 * config/nds32/nds32.opt (minline-asm-r15): New option.
82 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
84 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
86 * config/nds32/nds32.md (abssi2): New pattern.
88 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
90 * config/i386/i386.md (rex64namesuffix): New mode attribute.
91 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
92 Merge insn pattern from sse_cvtsi2ss<round_name> and
93 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
94 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
95 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
96 using SWI48 mode iterator.
97 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
98 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
99 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
100 pattern from sse_cvttss2si<round_saeonly_name>
101 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
102 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
103 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
104 using SWI48 mode iterator.
105 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
106 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
107 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
108 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
109 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
110 using SWI48 mode iterator.
111 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
112 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
113 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
114 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
115 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
117 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
118 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
119 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
120 pattern from sse_cvttsd2si<round_saeonly_name>
121 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
123 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
125 * config/nds32/nds32-md-auxiliary.c
126 (nds32_valid_smw_lwm_base_p): Refine.
127 (nds32_output_smw_single_word): Refine.
128 (nds32_output_smw_double_word): New.
129 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
131 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
133 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
134 (nds32_output_stack_pop): Refine.
135 (nds32_expand_unaligned_load): Refine.
136 (nds32_expand_unaligned_store): Refine.
138 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
139 Chung-Ju Wu <jasonwucj@gmail.com>
141 * config/nds32/constants.md: Add TP_REGNUM constant.
142 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
143 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
145 * config/nds32/nds32-doubleword.md: Consider flag_pic.
146 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
147 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
148 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
149 and PIC code generation.
150 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
152 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
154 * config/nds32/nds32.md: Support TLS and PIC.
155 * config/nds32/nds32.c: Support TLS and PIC.
156 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
157 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
160 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
162 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
165 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
166 Chung-Ju Wu <jasonwucj@gmail.com>
168 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
169 * config/nds32/nds32-md-auxiliary.c
170 (symbolic_reference_mentioned_p): New.
171 (nds32_legitimize_ict_address): New.
172 (nds32_expand_ict_move): New.
173 (nds32_indirect_call_referenced_p): New.
174 (nds32_symbol_binds_local_p): Delete.
175 (nds32_long_call_p): Modify.
176 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
177 * config/nds32/nds32-protos.h
178 (symbolic_reference_mentioned_p): Declare.
179 (nds32_legitimize_ict_address): Declare.
180 (nds32_expand_ict_move): Declare.
181 (nds32_indirect_call_referenced_p): Declare.
182 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
183 (nds32_relax_group): Use nds32_ict_const_p as condition.
184 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
185 (nds32_asm_file_start): Output ict_model directive in asm code.
186 (nds32_legitimate_address_p): Consider indirect call.
187 (nds32_print_operand): Consider indirect call.
188 (nds32_print_operand_address): Consider indirect call.
189 (nds32_insert_attributes): Handle "indirect_call" attribute.
190 (TARGET_LEGITIMATE_ADDRESS_P): Define.
191 (TARGET_LEGITIMATE_CONSTANT_P): Define.
192 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
193 (TARGET_DELEGITIMIZE_ADDRESS): Define.
194 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
195 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
196 (TARGET_ICT_MODEL_SMALL): Define.
197 (TARGET_ICT_MODEL_LARGE): Define.
198 * config/nds32/nds32.md (movsi): Consider ict model.
199 (call, call_value): Consider ict model.
200 (sibcall, sibcall_value): Consider ict model.
201 * config/nds32/nds32.opt (mict-model): New option.
202 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
205 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
206 Monk Chiang <sh.chiang04@gmail.com>
207 Jim Wilson <jimw@sifive.com>
209 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
210 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
211 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
212 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
213 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
214 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
215 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
216 compute save_libcall_adjustment properly.
217 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
218 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
219 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
220 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
221 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
222 (ABI_SPEC): Handle mabi=ilp32e.
223 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
225 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
226 <-march>: Add rv32e as an example.
228 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
231 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
232 (intra_create_variable_infos): Handle C++ constructors.
234 2018-05-18 Martin Liska <mliska@suse.cz>
236 * passes.def: Remove a redundant pass.
238 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
241 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
243 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
245 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
247 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
249 (armv2, armv2a, armv3, armv3m): Delete architectures.
250 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
251 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
252 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
254 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
255 (*mulsidi3adddi): Likewise.
256 (mulsidi3): Likewise.
257 (*mulsidi3_nov6): Likewise.
258 (umulsidi3): Likewise.
259 (umulsidi3_nov6): Likewise.
260 (umaddsidi4): Likewise.
261 (*umulsidi3adddi): Likewise.
262 (smulsi3_highpart): Likewise.
263 (*smulsi3_highpart_nov6): Likewise.
264 (umulsi3_highpart): Likewise.
265 (*umulsi3_highpart_nov6): Likewise.
266 * config/arm/arm.h (arm_arch3m): Delete.
267 * config/arm/arm.c (arm_arch3m): Delete.
268 (arm_option_override_internal): Update armv3-related comment.
269 (arm_configure_build_target): Delete use of isa_bit_mode32.
270 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
271 (arm_rtx_costs_internal): Delete check of arm_arch3m.
272 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
275 * config/arm/arm-protos.h (arm_arch3m): Delete.
276 * config/arm/arm-tables.opt: Regenerate.
277 * config/arm/arm-tune.md: Likewise.
278 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
279 deleted architectures.
281 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
283 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
284 (armv5t, armv5te): New features.
285 (ARMv5, ARMv5e): Delete fgroups.
286 (ARMv5t, ARMv5te): Adjust for above changes.
288 (armv5, armv5e): Delete arches.
289 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
291 (*call_reg_arm): Likewise.
292 (*call_value_reg_armv5): Likewise.
293 (*call_value_reg_arm): Likewise.
294 (*call_symbol): Likewise.
295 (*call_value_symbol): Likewise.
296 (*sibcall_insn): Likewise.
297 (*sibcall_value_insn): Likewise.
299 (prefetch): Likewise.
300 (define_split and define_peephole2 dependent on arm_arch5):
302 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
304 (TARGET_ARM_QBIT): Likewise.
305 (TARGET_DSP_MULTIPLY): Likewise.
306 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
307 (arm_arch5, arm_arch5e): Delete.
308 (arm_arch5t, arm_arch5te): Declare.
309 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
310 (arm_arch5t): Declare.
311 (arm_option_reconfigure_globals): Update for the above.
312 (arm_options_perform_arch_sanity_checks): Update comment, replace
313 use of arm_arch5 with arm_arch5t.
314 (use_return_insn): Likewise.
315 (arm_emit_call_insn): Likewise.
316 (output_return_instruction): Likewise.
317 (arm_final_prescan_insn): Likewise.
318 (arm_coproc_builtin_available): Likewise.
319 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
320 arm_arch5e with arm_arch5t and arm_arch5te.
321 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
322 (arm_arch5t, arm_arch5te): Declare.
323 * config/arm/arm-tables.opt: Regenerate.
324 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
325 * config/arm/t-multilib: Likewise.
326 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
327 instead of arm_arch5.
328 (*call_reg_thumb1): Likewise.
329 (*call_value_reg_thumb1_v5): Likewise.
330 (*call_value_reg_thumb1): Likewise.
331 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
333 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
335 2018-05-18 Martin Liska <mliska@suse.cz>
337 PR gcov-profile/84846
338 * doc/gcov.texi: Document -t option of gcov tool.
340 2018-05-18 Martin Liska <mliska@suse.cz>
342 PR gcov-profile/84846
343 * gcov.c (print_usage): Add new -t option.
344 (process_args): Handle the option.
345 (generate_results): Use stdout as output when requested by
348 2018-05-18 Martin Liska <mliska@suse.cz>
350 PR gcov-profile/84846
351 * coverage.c (coverage_init): Write PWD to .gcno file.
352 * doc/gcov.texi: Document how working directory is printed.
353 * gcov-dump.c (dump_gcov_file): Print PWD.
354 * gcov.c (output_intermediate_file): Likewise.
355 (read_graph_file): Read PWD string.
356 (output_lines): Print PWD.
358 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
361 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
362 for retval and return false if all args to phi are zero.
364 2018-05-18 Richard Biener <rguenther@suse.de>
366 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
368 (evrp_dom_walker::before_dom_children): Call it.
370 2018-05-18 Richard Biener <rguenther@suse.de>
372 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
373 results when processing array refs with variable index.
375 2018-05-18 Toon Moene <toon@moene.org>
377 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
378 directly after that of -floop-interchange. Indicate that both
379 options are enabled by default when specifying -O3.
381 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
383 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
384 iterator. Delete separate integer-mode vec_set<mode> expander.
385 (aarch64_simd_vec_setv2di): Delete.
386 (vec_setv2di): Delete.
387 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
388 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
389 the "w, r" alternative.
391 2018-05-18 Martin Liska <mliska@suse.cz>
393 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
394 * tree-pass.h (make_pass_lower_switch_O0): New function.
395 * tree-switch-conversion.c (node_has_low_bound): Remove.
396 (node_has_high_bound): Likewise.
397 (node_is_bounded): Likewise.
398 (class pass_lower_switch): Make it a template type and create
400 (pass_lower_switch::execute): Add template argument.
401 (make_pass_lower_switch): New function.
402 (make_pass_lower_switch_O0): New function.
403 (do_jump_if_equal): Remove.
404 (emit_case_nodes): Simplify to just handle all 3 cases and leave
405 all the hard work to tree optimization passes.
407 2018-05-18 Martin Liska <mliska@suse.cz>
409 * dbgcnt.c (limit_low): Renamed from limit.
410 (limit_high): New variable.
411 (dbg_cnt_is_enabled): Check for upper limit.
412 (dbg_cnt): Adjust dumping.
413 (dbg_cnt_set_limit_by_index): Add new argument for high
415 (dbg_cnt_set_limit_by_name): Likewise.
416 (dbg_cnt_process_single_pair): Parse new format.
417 (dbg_cnt_process_opt): Use strtok.
418 (dbg_cnt_list_all_counters): Remove 'value' and add
420 * doc/invoke.texi: Document changes.
422 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
424 * doc/sourcebuild.texi (scalar_all_fma): Document.
425 * tree.def (FMA_EXPR): Delete.
426 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
427 * internal-fn.c (ternary_direct): New macro.
428 (expand_ternary_optab_fn): Likewise.
429 (direct_ternary_optab_supported_p): Likewise.
430 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
431 * builtins.c (fold_builtin_fma): Delete.
432 (fold_builtin_3): Don't call it.
433 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
434 * expr.c (expand_expr_real_2): Likewise.
435 * fold-const.c (operand_equal_p): Likewise.
436 (fold_ternary_loc): Likewise.
437 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
438 * gimple.c (DEFTREECODE): Likewise.
439 * gimplify.c (gimplify_expr): Likewise.
440 * optabs-tree.c (optab_for_tree_code): Likewise.
441 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
442 * tree-eh.c (operation_could_trap_p): Likewise.
443 (stmt_could_throw_1_p): Likewise.
444 * tree-inline.c (estimate_operator_cost): Likewise.
445 * tree-pretty-print.c (dump_generic_node): Likewise.
446 (op_code_prio): Likewise.
447 * tree-ssa-loop-im.c (stmt_cost): Likewise.
448 * tree-ssa-operands.c (get_expr_operands): Likewise.
449 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
450 * fold-const-call.h (fold_fma): Delete.
451 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
452 CFN_FNMA and CFN_FNMS.
454 * genmatch.c (combined_fn): New enum.
455 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
456 (commutative_op): New function.
457 (commutate): Use it. Handle more than 2 operands.
458 (dt_operand::gen_gimple_expr): Use commutative_op.
459 (parser::parse_expr): Allow :c to be used with non-binary
460 operators if the commutative operand is known.
461 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
462 CFN_FMS, CFN_FNMA and CFN_FNMS.
463 (backprop::process_assign_use): Remove FMA_EXPR handling.
464 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
465 (gen_hsa_fma): New function.
466 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
467 IFN_FNMA and IFN_FNMS.
468 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
469 * gimple-fold.h (follow_all_ssa_edges): Declare.
470 * gimple-fold.c (follow_all_ssa_edges): New function.
471 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
472 gimple_build interface and use follow_all_ssa_edges to fold the result.
473 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
474 instead of checking for optabs directly.
475 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
476 rather than FMA_EXPRs.
477 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
478 call to IFN_FMA instead of an FMA_EXPR.
480 2018-05-17 Jim Wilson <jimw@sifive.com>
482 * expr.c (do_tablejump): When converting index to Pmode, if we have a
483 sign extended promoted subreg, and the range does not have the sign bit
484 set, then do a sign extend.
486 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
487 test, check for sign extended subreg and/or constant operands, and
488 do a sign extend in that case.
490 2018-05-17 Steve Ellcey <sellcey@cavium.com>
492 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
493 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
495 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
496 Change logics_shift_reg to logics_shift_imm.
497 (thunderx2t99_fp_loadpair_basic): Delete.
498 (thunderx2t99_fp_storepair_basic): Delete.
499 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
500 (thunderx2t99_asimd_polynomial): Delete.
501 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
502 and neon_fp_mul_d_scalar_q.
503 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
504 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
505 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
506 (thunderx2t99_asimd_lut): Add missing tbl types.
507 (thunderx2t99_asimd_ext): Delete.
508 (thunderx2t99_asimd_load1_1_mult): Delete.
509 (thunderx2t99_asimd_load1_2_mult): Delete.
510 (thunderx2t99_asimd_load1_ldp): New.
511 (thunderx2t99_asimd_load1): New.
512 (thunderx2t99_asimd_load2): Add missing *load2* types.
513 (thunderx2t99_asimd_load3): New.
514 (thunderx2t99_asimd_load4): New.
515 (thunderx2t99_asimd_store1_1_mult): Delete.
516 (thunderx2t99_asimd_store1_2_mult): Delete.
517 (thunderx2t99_asimd_store2_mult): Delete.
518 (thunderx2t99_asimd_store2_onelane): Delete.
519 (thunderx2t99_asimd_store_stp): New.
520 (thunderx2t99_asimd_store1): New.
521 (thunderx2t99_asimd_store2): New.
522 (thunderx2t99_asimd_store3): New.
523 (thunderx2t99_asimd_store4): New.
525 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
527 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
528 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
530 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
531 Segher Boessenkool <segher@kernel.crashing.org>
534 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
537 2018-05-17 Richard Biener <rguenther@suse.de>
539 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
540 for pruning loop and prune defs feeding only already visited PHIs.
542 2018-05-17 Richard Biener <rguenther@suse.de>
544 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
546 2018-05-17 Bin Cheng <bin.cheng@arm.com>
547 Richard Biener <rguenther@suse.de>
549 PR tree-optimization/85793
550 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
551 for VMAT_ELEMENTWISE.
553 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
555 * internal-fn.h (lookup_internal_fn): Declare
556 * internal-fn.c (lookup_internal_fn): New function.
557 * gimple.c (gimple_build_call_from_tree): Handle calls to
559 * gimple-pretty-print.c (dump_gimple_call): Print "." before
560 internal function names.
561 * tree-pretty-print.c (dump_generic_node): Likewise.
562 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
564 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
566 * gimple-fold.h (gimple_build): Make the function forms take
567 combined_fn rather than built_in_function.
568 (gimple_simplify): Likewise.
569 * gimple-match-head.c (gimple_simplify): Likewise.
570 * gimple-fold.c (gimple_build): Likewise.
571 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
572 rather than gimple_build_call_internal.
573 (get_initial_defs_for_reduction): Likewise.
574 (vect_create_epilog_for_reduction): Likewise.
575 (vectorizable_live_operation): Likewise.
577 2018-05-17 Martin Liska <mliska@suse.cz>
579 * gimple-ssa-sprintf.c (format_directive): Do not use
580 space in between 'G_' and '('.
582 2018-05-17 Jakub Jelinek <jakub@redhat.com>
585 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
586 even if the mask is not all ones.
589 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
591 (ix86_gimple_fold_builtin): Likewise.
594 * config/i386/i386.c: Include tree-vector-builder.h.
595 (ix86_vector_shift_count): New function.
596 (ix86_fold_builtin): Fold shift builtins by scalar count.
597 (ix86_gimple_fold_builtin): Likewise.
599 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
600 _mm512_setzero): New intrinsics.
602 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
603 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
605 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
606 code generation for cases where splatting a value is not useful.
607 * simplify-rtx.c (simplify_ternary_operation): Simplify
608 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
609 mode to a vec_concat.
611 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
613 * config.gcc: Support "goldmont-plus".
614 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
616 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
617 PROCESSOR_GOLDMONT_PLUS.
618 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
619 (processor_target_table): Add "goldmont-plus".
620 (PTA_GOLDMONT_PLUS): Define.
621 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
622 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
623 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
624 (fold_builtin_cpu): Add "goldmont-plus".
625 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
626 (ix86_option_override_internal): Add "goldmont-plus".
627 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
628 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
629 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
630 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
632 2018-05-17 Richard Biener <rguenther@suse.de>
634 PR tree-optimization/85757
635 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
636 remove defs that only feed that PHI from further processing.
638 2018-05-16 Jim Wilson <jimw@sifive.com>
640 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
642 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
644 2018-05-16 Mark Wielaard <mark@klomp.org>
646 * dwarf2out.c (count_index_strings): New function.
647 (output_indirect_strings): Call count_index_strings and generate
648 header for dwarf_version >= 5.
650 2018-05-16 Mark Wielaard <mark@klomp.org>
652 * dwarf2out.c (dwarf_FORM): New function.
653 (set_indirect_string): Use dwarf_FORM.
654 (reset_indirect_string): Likewise.
655 (size_of_die): Likewise.
656 (value_format): Likewise.
657 (output_die): Likewise.
658 (add_skeleton_AT_string): Likewise.
659 (output_macinfo_op): Likewise.
660 (index_string): Likewise.
661 (output_index_string_offset): Likewise.
662 (output_index_string): Likewise.
663 (count_index_strings): Likewise.
665 2018-05-16 Carl Love <cel@us.ibm.com>
667 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
668 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
670 2018-05-16 Martin Jambor <mjambor@suse.cz>
672 * ipa-prop.c (ipa_free_all_edge_args): Remove.
673 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
675 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
677 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
678 (fnma<mode>4): Likewise.
679 (fms<mode>4): Likewise.
680 (fnms<mode>4): Likewise.
681 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
682 (aarch64_fnma<mode>4): Likewise.
683 (aarch64_fms<mode>4): Likewise.
684 (aarch64_fnms<mode>4): Likewise.
685 (aarch64_fnmadd<mode>4): Likewise.
687 2018-05-16 Jason Merrill <jason@redhat.com>
689 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
691 2018-05-16 Richard Biener <rguenther@suse.de>
693 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
694 (dump_stmt_cost): Declare.
695 (add_stmt_cost): Dump cost we add.
696 (add_stmt_costs): New function.
697 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
699 (vect_analyze_stmt): Adjust prototype.
700 (vectorizable_condition): Likewise.
701 (vectorizable_live_operation): Likewise.
702 (vectorizable_reduction): Likewise.
703 (vectorizable_induction): Likewise.
704 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
705 cost vector to pass to vectorizable_ and record afterwards.
706 (vect_model_reduction_cost): Take cost vector argument and adjust.
707 (vect_model_induction_cost): Likewise.
708 (vectorizable_reduction): Likewise.
709 (vectorizable_induction): Likewise.
710 (vectorizable_live_operation): Likewise.
711 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
712 SLP_TREE_NUMBER_OF_VEC_STMTS.
713 (vect_analyze_slp_cost_1): Remove.
714 (vect_analyze_slp_cost): Likewise.
715 (vect_slp_analyze_node_operations): Take visited args and
716 a target cost vector. Avoid processing already visited stmt sets.
717 (vect_slp_analyze_operations): Use a local cost vector to gather
718 costs and register those of non-discarded instances.
719 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
720 (vect_schedule_slp_instance): Remove copying of
721 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
723 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
724 adding cost. Record cost entry location.
725 (vect_prologue_cost_for_slp_op): Function to compute cost of
726 a constant or invariant generated for SLP vect in the prologue,
727 split out from vect_analyze_slp_cost_1.
728 (vect_model_simple_cost): Make static. Adjust for SLP costing.
729 (vect_model_promotion_demotion_cost): Likewise.
730 (vect_model_store_cost): Likewise, make static.
731 (vect_model_load_cost): Likewise.
732 (vectorizable_bswap): Add cost vector arg and adjust.
733 (vectorizable_call): Likewise.
734 (vectorizable_simd_clone_call): Likewise.
735 (vectorizable_conversion): Likewise.
736 (vectorizable_assignment): Likewise.
737 (vectorizable_shift): Likewise.
738 (vectorizable_operation): Likewise.
739 (vectorizable_store): Likewise.
740 (vectorizable_load): Likewise.
741 (vectorizable_condition): Likewise.
742 (vectorizable_comparison): Likewise.
743 (can_vectorize_live_stmts): Likewise.
744 (vect_analyze_stmt): Likewise.
745 (vect_transform_stmt): Adjust calls to vectorizable_*.
746 * tree-vectorizer.c: Include gimple-pretty-print.h.
747 (dump_stmt_cost): New function.
749 2018-05-16 Richard Biener <rguenther@suse.de>
751 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
752 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
753 * tree-ssa-dse.c: Include tree-ssa-loop.h.
754 (check_name): New callback.
755 (dse_classify_store): Track cycles via a visited bitmap of PHI
756 defs and simplify handling of in-loop and across loop dead stores
757 and properly fail for loop-variant refs. Handle byte-tracking with
758 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
761 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
763 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
764 (vect_get_mask_type_for_stmt): Likewise.
765 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
767 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
768 to determine the statement's vector type and the vector type that
769 should be used for calculating nunits. Deal with cases in which
770 the type has to be deferred.
771 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
772 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
773 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
774 (vect_determine_vf_for_stmt): New functions, split out from...
775 (vect_determine_vectorization_factor): ...here.
776 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
777 (vect_get_mask_type_for_stmt): New functions, split out from
778 vect_determine_vectorization_factor.
780 2018-05-16 Richard Biener <rguenther@suse.de>
782 * tree-cfg.c (verify_gimple_assign_ternary): Properly
783 verify the [VEC_]COND_EXPR embedded comparison.
785 2018-05-15 Martin Sebor <msebor@redhat.com>
787 PR tree-optimization/85753
788 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
789 RECORD_TYPE in addition to ARRAY_TYPE.
791 2018-05-15 Martin Sebor <msebor@redhat.com>
794 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
796 2018-05-15 Richard Biener <rguenther@suse.de>
798 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
799 add by_clobber_p one. Change algorithm to collect all defs
800 representing uses we need to walk and try reducing them to
801 a single one before failing.
802 (dse_dom_walker::dse_optimize_stmt): Adjust.
804 2018-05-13 Mark Wielaard <mark@klomp.org>
806 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
807 (size_of_loc_descr): Likewise.
808 (output_loc_operands): Likewise.
809 (output_loc_operands_raw): Likewise.
810 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
811 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
812 (hash_loc_operands): Likewise.
813 (compare_loc_operands): Likewise.
815 2018-05-14 Mark Wielaard <mark@klomp.org>
817 * dwarf2out.c (count_index_addrs): New function.
818 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
820 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
822 PR tree-optimization/83648
823 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
824 return value as malloc candidate.
826 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
829 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
830 as true in call to suggest_attribute.
832 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
834 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
837 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
839 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
841 (aarch64_crypto_aesd_fused): Likewise.
843 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
845 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
846 (movsi_aarch64): Likewise.
847 (load_pairsi): Likewise.
848 (load_pairdi): Likewise.
849 (store_pairsi): Likewise.
850 (store_pairdi): Likewise.
851 (load_pairsf): Likewise.
852 (load_pairdf): Likewise.
853 (store_pairsf): Likewise.
854 (store_pairdf): Likewise.
855 (zero_extend): Likewise.
856 (trunc): Swap alternatives.
857 (fcvt_target): Add '?' to prefer w over r.
859 2018-05-14 Jakub Jelinek <jakub@redhat.com>
862 * config/i386/i386.md: Disallow non-commutative arithmetics in
863 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
864 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
865 in the peephole2 before it.
867 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
869 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
870 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
871 (ix86_handle_option): Handle -mcldemote.
872 * config.gcc: New header.
873 * config/i386/cldemoteintrin.h: New file.
874 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
875 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
877 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
878 OPTION_MASK_ISA_CLDEMOTE.
879 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
880 (ix86_valid_target_attribute_inner_p): Ditto.
881 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
882 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
883 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
884 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
885 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
887 * config/i386/i386.opt: Add -mcldemote.
888 * config/i386/x86intrin.h: New header.
889 * doc/invoke.texi: Add -mcldemote.
891 2018-05-14 Richard Biener <rguenther@suse.de>
893 * doc/match-and-simplify.texi: Adjust :s documentation.
895 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
897 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
898 intended memcpy size.
899 (REORDER_45): Likewise.
901 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
904 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
905 * vec.c (qsort_chk): Use gcc_qsort.
906 * Makefile.in (OBJS-libcommon): Add sort.o.
907 (build/sort.o): New target. Use it...
908 (BUILD_RTL): ... here, and...
909 (build/gencfn-macros): ... here, and...
910 (build/genmatch): ... here.
912 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
913 Chung-Ju Wu <jasonwucj@gmail.com>
915 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
916 * config/nds32/nds32-graywolf.md: New file.
917 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
918 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
920 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
921 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
922 * config/nds32/nds32.md (pipeline_model): Add graywolf.
923 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
924 * config/nds32/pipelines.md: Include n15 settings.
926 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
927 Chung-Ju Wu <jasonwucj@gmail.com>
929 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
930 * config/nds32/nds32-n13.md: New file.
931 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
932 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
934 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
935 * config/nds32/nds32.md (pipeline_model): Add n13.
936 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
937 * config/nds32/pipelines.md: Include n13 settings.
939 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
940 Chung-Ju Wu <jasonwucj@gmail.com>
942 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
943 * config/nds32/nds32-n10.md: New file.
944 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
945 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
947 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
948 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
949 * config/nds32/nds32.md (pipeline_model): Add n10.
950 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
951 * config/nds32/pipelines.md: Include n10 settings.
953 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
954 Kito Cheng <kito.cheng@gmail.com>
955 Chung-Ju Wu <jasonwucj@gmail.com>
957 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
958 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
959 Add enum values for DSP extension instructions.
960 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
962 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
963 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
965 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
966 * config/nds32/nds32-dspext.md: New file for DSP implementation.
967 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
968 * config/nds32/nds32-intrinsic.md: Likewise.
969 * config/nds32/nds32_intrinsic.h: Likewise.
970 * config/nds32/nds32-md-auxiliary.c: Likewise.
971 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
972 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
973 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
974 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
975 * config/nds32/nds32-protos.h: New declarations for DSP extension.
976 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
977 TYPE_DMAC in switch statement.
978 * config/nds32/nds32.c: New checking and implementation for DSP
979 extension instructions.
980 * config/nds32/nds32.h: Likewise.
981 * config/nds32/nds32.md: Likewise.
982 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
983 * config/nds32/predicates.md: Implement new predicates for DSP
986 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
988 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
989 Reformat alternatives and attributes so it is easier to identify
990 which constraints/attributes go with which instruction.
991 (mov<mode>_hardfloat32, FMOVE64): Likewise.
992 (mov<mode>_softfloat32, FMOVE64): Likewise.
993 (mov<mode>_hardfloat64, FMOVE64): Likewise.
994 (mov<mode>_softfloat64, FMOVE64): Likewise.
996 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
998 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1000 (Basic PowerPC Built-in Functions): The new name of the
1001 subsection previously known as "PowerPC Built-in Functions".
1002 (Basic PowerPC Built-in Functions Available on all Configurations):
1004 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1005 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1006 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1007 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1009 2018-05-11 Martin Jambor <mjambor@suse.cz>
1012 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1015 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1018 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1020 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1022 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1023 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1024 (ix86_handle_option): Handle -mwaitpkg.
1025 * config.gcc: New header.
1026 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1027 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1028 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1030 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1031 OPTION_MASK_ISA_WAITPKG.
1032 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1033 (ix86_option_override_internal): Add PTA_WAITPKG.
1034 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1035 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1036 IX86_BUILTIN_TPAUSE.
1037 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1038 __builtin_ia32_umwait and __builtin_ia32_tpause.
1039 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1040 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1041 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1042 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1043 UNSPECV_TPAUSE): New.
1044 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1045 * config/i386/i386.opt: Add -mwaitpkg.
1046 * config/i386/waitpkgintrin.h: New file.
1047 * config/i386/x86intrin.h: New header.
1048 * doc/invoke.texi: Add -mwaitpkg.
1050 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1053 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1055 (cortex-m0): Use armv6s-m isa.
1056 (cortex-m0plus): Likewise.
1057 (cortex-m1): Likewise.
1058 (cortex-m0.small-multiply): Likewise.
1059 (cortex-m0plus.small-multiply): Likewise.
1060 (cortex-m1.small-multiply): Likewise.
1062 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1063 Jakub Jelinek <jakub@redhat.com>
1065 PR tree-optimization/85692
1066 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1067 source permute as well.
1069 2018-05-11 Martin Liska <mliska@suse.cz>
1072 * doc/extend.texi: Document LLVM style format for no_sanitize
1075 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1077 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1078 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1079 (mode_supports_vsx_dform_quad): Likewise.
1080 (mode_supports_vmx_dform): Move these functions to be next to the
1081 other mode_supports functions.
1082 (mode_supports_dq_form): Likewise.
1083 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1084 mode_supports_dq_form.
1085 (reg_offset_addressing_ok_p): Likewise.
1086 (offsettable_ok_by_alignment): Likewise.
1087 (rs6000_legitimate_offset_address_p): Likewise.
1088 (legitimate_lo_sum_address_p): Likewise.
1089 (rs6000_legitimize_address): Likewise.
1090 (rs6000_legitimize_reload_address): Likewise.
1091 (rs6000_secondary_reload_inner): Likewise.
1092 (rs6000_preferred_reload_class): Likewise.
1093 (rs6000_output_move_128bit): Likewise.
1095 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1097 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1098 Generate SImode target register for null target.
1099 <case IX86_BUILTIN_XGETBV>: Ditto.
1100 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1101 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1103 2018-05-10 Carl Love <cel@us.ibm.com>
1105 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1106 dcbtt and dcbtstt if operands[2] is 0.
1108 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1111 * config/i386/sse.md (usadv64qi): New expander.
1113 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1115 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1116 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1117 -maltivec=be support.
1118 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1119 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1120 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1121 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1122 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1123 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1124 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1125 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1126 altivec_vsumsws): Adjust.
1127 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1128 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1129 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1131 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1132 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1133 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1134 (altivec_lve<VI_char>x): Delete expand.
1135 (*altivec_lve<VI_char>x_internal): Rename to...
1136 (altivec_lve<VI_char>x): ... this.
1137 (altivec_lvxl_<mode>): Delete expand.
1138 (*altivec_lvxl_<mode>_internal): Rename to ...
1139 (altivec_lvxl_<mode>): ... this.
1140 (altivec_stvxl_<mode>): Delete expand.
1141 (*altivec_stvxl_<mode>_internal): Rename to ...
1142 (altivec_stvxl_<mode>): ... this.
1143 (altivec_stve<VI_char>x): Delete expand.
1144 (*altivec_stve<VI_char>x_internal): Rename to ...
1145 (altivec_stve<VI_char>x): ... this.
1146 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1147 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1148 reduc_plus_scal_<mode>): Adjust.
1149 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1151 (rs6000_cpu_cpp_builtins): Adjust.
1152 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1153 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1154 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1155 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1156 -maltivec=be support.
1157 (rs6000_split_vec_extract_var): Adjust.
1158 (rs6000_split_v4si_init): Adjust.
1159 (swap_selector_for_mode): Delete.
1160 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1161 altivec_expand_stvex_be): Delete.
1162 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1163 -maltivec=be support.
1164 (rs6000_gimple_fold_builtin): Ditto.
1165 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1167 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1168 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1169 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1170 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1171 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1172 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1173 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1174 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1175 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1176 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1177 anonymous split): Adjust.
1178 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1179 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1181 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1183 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1184 when --with-gxx-include-dir is also specified.
1185 * configure: Regenerate.
1187 2018-05-09 Jim Wilson <jimw@sifive.com>
1190 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1191 * config/riscv/t-withmultilib: New.
1192 * config/riscv/withmultilib.h: New.
1193 * doc/install.texi: Document RISC-V --with-multilib-list support.
1195 2018-05-09 Richard Biener <rguenther@suse.de>
1197 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1199 (vect_bb_vectorization_profitable_p): Adjust. Compute
1200 actual scalar cost using the cost vector and the add_stmt_cost
1203 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1205 PR rtl-optimization/85645
1206 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1207 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1209 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1211 PR rtl-optimization/85645
1212 * shrink-wrap.c (spread_components): Return a boolean saying if
1213 anything was changed.
1214 (try_shrink_wrapping_separate): Iterate spread_components until
1215 nothing changes anymore.
1217 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1219 PR rtl-optimization/85645
1220 * regrename.c (build_def_use): Also kill the chains that include the
1221 destination of a REG_CFA_REGISTER note.
1223 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1225 PR rtl-optimization/85645
1226 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1227 insn that has a REG_CFA_REGISTER note.
1229 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1231 * cfgexpand.c (expand_clobber): New function.
1232 (expand_gimple_stmt_1): Use it.
1233 * tree-vect-stmts.c (vect_clobber_variable): New function,
1235 (vectorizable_simd_clone_call): ...here.
1236 (vectorizable_store): Emit a clobber either side of an
1237 IFN_STORE_LANES sequence.
1238 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1240 2018-05-09 Tom de Vries <tom@codesourcery.com>
1243 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1244 (define_insn "trap_if_false"): Add exit after trap.
1246 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1248 PR rtl-optimization/85638
1249 * bb-reorder.c: Include common/common-target.h.
1250 (create_forwarder_block): New function extracted from...
1251 (fix_up_crossing_landing_pad): ...here. Rename into...
1252 (dw2_fix_up_crossing_landing_pad): ...this.
1253 (sjlj_fix_up_crossing_landing_pad): New function.
1254 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1255 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1256 from both partitions and exit the loop after one iteration.
1258 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1261 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1263 (Basic PowerPC Built-in Functions): The new name of the
1264 subsection previously known as "PowerPC Built-in Functions".
1265 (Basic PowerPC Built-in Functions Available on all Configurations):
1267 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1269 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1271 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1273 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1276 2018-05-08 Jim Wilson <jimw@sifive.com>
1278 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1279 (LD_EMUL_SUFFIX): New.
1280 (LINK_SPEC): Use it.
1282 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1284 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1286 (Basic PowerPC Built-in Functions): The new name of the
1287 subsection previously known as "PowerPC Built-in Functions".
1288 (Basic PowerPC Built-in Functions Available on all Configurations):
1290 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1292 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1294 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1296 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1299 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1302 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1303 after cmpelim optimization.
1305 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1307 * config.gcc: Support "goldmont".
1308 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1309 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1311 * config/i386/i386.c (m_GOLDMONT): Define.
1312 (processor_target_table): Add "goldmont".
1313 (PTA_GOLDMONT): Define.
1314 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1315 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1316 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1317 (fold_builtin_cpu): Add "goldmont".
1318 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1319 (ix86_option_override_internal): Add "goldmont".
1320 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1321 (processor_type): Add PROCESSOR_GOLDMONT.
1322 * config/i386/i386.md: Add CPU "glm".
1323 * config/i386/glm.md: New file.
1324 * config/i386/x86-tune.def: Add m_GOLDMONT.
1325 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1327 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1330 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1332 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1333 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1334 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1337 * config/i386/i386.c (ix86_fold_builtin): Handle
1338 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1341 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1342 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1344 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1347 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1348 (check_arch): Likewise.
1349 (check_fpu): Return the result rather than printing it.
1350 (end arch): Fix operator precedence.
1351 (end cpu): Likewise.
1352 (END): Print the result from check_fpu.
1354 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1355 Alan Hayward <alan.hayward@arm.com>
1356 David Sherwood <david.sherwood@arm.com>
1358 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1359 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1360 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1361 (*fcmuo<mode>_and): New patterns.
1363 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1365 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1366 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1367 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1368 (cmp_op, sve_imm_con): New code attributes.
1369 (SVE_COND_INT_CMP, imm_con): Delete.
1370 (cmp_op): Remove above unspecs from int attribute.
1371 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1373 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1374 comparison-specific unspecs.
1375 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1376 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1377 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1378 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1379 (*vec_fcm<cmp_op><mode>): Rename to...
1380 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1381 (*vec_fcmuo<mode>): Rename to...
1382 (*fcmuo<mode>): ...this and adjust likewise.
1383 (*pred_fcm<cmp_op><mode>): New pattern.
1384 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1385 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1387 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1389 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1390 (aarch64_emit_sve_predicated_cond): New function.
1391 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1392 (aarch64_emit_unspec_cond_or): Replace with...
1393 (aarch64_emit_sve_or_conds): ...this new function. Use
1394 aarch64_emit_sve_ptrue_op for the individual comparisons and
1395 aarch64_emit_binop to OR them together.
1396 (aarch64_emit_inverted_unspec_cond): Replace with...
1397 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1398 aarch64_emit_sve_ptrue_op for the comparison and
1399 aarch64_emit_unop to invert the result.
1400 (aarch64_expand_sve_vec_cmp_float): Update after the above
1401 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1403 2018-05-07 Nathan Sidwell <nathan@acm.org>
1405 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1406 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1407 (Backwards Compatibility): Likewise.
1409 2018-05-07 Luis Machado <luis.machado@linaro.org>
1413 2018-05-07 Luis Machado <luis.machado@linaro.org>
1415 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1416 <prefetch_dynamic_strides>: New const bool field.
1417 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1418 prefetch_dynamic_strides.
1419 (exynosm1_prefetch_tune): Likewise.
1420 (thunderxt88_prefetch_tune): Likewise.
1421 (thunderx_prefetch_tune): Likewise.
1422 (thunderx2t99_prefetch_tune): Likewise.
1423 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1424 (aarch64_override_options_internal): Update to set
1425 PARAM_PREFETCH_DYNAMIC_STRIDES.
1426 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1427 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1428 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1429 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1430 prefetch-dynamic-strides setting.
1432 2018-05-07 Luis Machado <luis.machado@linaro.org>
1434 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1435 <minimum_stride>: New const int field.
1436 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1437 minimum_stride field.
1438 (exynosm1_prefetch_tune): Likewise.
1439 (thunderxt88_prefetch_tune): Likewise.
1440 (thunderx_prefetch_tune): Likewise.
1441 (thunderx2t99_prefetch_tune): Likewise.
1442 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1443 (aarch64_override_options_internal): Update to set
1444 PARAM_PREFETCH_MINIMUM_STRIDE.
1445 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1446 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1447 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1448 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1449 stride is constant and is below the minimum stride threshold.
1451 2018-05-07 Luis Machado <luis.machado@linaro.org>
1453 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1456 2018-05-07 Luis Machado <luis.machado@linaro.org>
1458 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1459 <prefetch_dynamic_strides>: New const bool field.
1460 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1461 prefetch_dynamic_strides.
1462 (exynosm1_prefetch_tune): Likewise.
1463 (thunderxt88_prefetch_tune): Likewise.
1464 (thunderx_prefetch_tune): Likewise.
1465 (thunderx2t99_prefetch_tune): Likewise.
1466 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1467 (aarch64_override_options_internal): Update to set
1468 PARAM_PREFETCH_DYNAMIC_STRIDES.
1469 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1470 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1471 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1472 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1473 prefetch-dynamic-strides setting.
1475 2018-05-07 Luis Machado <luis.machado@linaro.org>
1477 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1478 <minimum_stride>: New const int field.
1479 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1480 minimum_stride field.
1481 (exynosm1_prefetch_tune): Likewise.
1482 (thunderxt88_prefetch_tune): Likewise.
1483 (thunderx_prefetch_tune): Likewise.
1484 (thunderx2t99_prefetch_tune): Likewise.
1485 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1486 (aarch64_override_options_internal): Update to set
1487 PARAM_PREFETCH_MINIMUM_STRIDE.
1488 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1489 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1490 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1491 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1492 stride is constant and is below the minimum stride threshold.
1494 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1497 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1498 the type is addressable. Don't force op into register if it has
1501 2018-05-05 Roland McGrath <mcgrathr@google.com>
1504 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1505 any section for which we don't know a specific type it should have,
1506 regardless of name. Previously this was done only for the exact
1507 names ".init_array", ".fini_array", and ".preinit_array".
1508 (default_elf_asm_named_section): Add comment about
1509 relationship with default_section_type_flags and SECTION_NOTYPE.
1510 (get_section): Don't consider it a type conflict if one side has
1511 SECTION_NOTYPE and the other doesn't, as long as neither has the
1512 SECTION_BSS et al used in the default_section_type_flags logic.
1514 2018-05-05 Tom de Vries <tom@codesourcery.com>
1517 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1518 (workaround_barsyncs): New function.
1519 (nvptx_reorg): Use workaround_barsyncs.
1520 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1521 (define_expand "nvptx_membar_cta"): New define_expand.
1522 (define_insn "*nvptx_membar_cta"): New insn.
1524 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1526 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1527 To improve optimization opportunities.
1528 * builtin-types.def: The new needed builtin types for the above.
1530 2018-05-04 Richard Biener <rguenther@suse.de>
1532 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1533 * gimple-ssa-store-merging.c
1534 (imm_store_chain_info::output_merged_store): Remove redundant create,
1535 release split_store vector contents on failure.
1536 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1537 scalar stmt vector on cache hit.
1539 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1541 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1543 * config.gcc (powerpc-xilinx-eabi*): Remove.
1544 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1546 (fusion_addis_mem_combo_load): Ditto.
1547 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1549 (rs6000_cpu_cpp_builtins): Ditto.
1550 * config/rs6000/rs6000-linux.c
1551 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1552 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1553 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1555 (rs6000_setup_reg_addr_masks): Ditto.
1556 (rs6000_init_hard_regno_mode_ok): Ditto.
1557 (rs6000_option_override_internal): Ditto.
1558 (legitimate_lo_sum_address_p): Ditto.
1559 (rs6000_legitimize_address): Ditto.
1560 (rs6000_legitimize_reload_address): Ditto.
1561 (rs6000_legitimate_address_p): Ditto.
1562 (abi_v4_pass_in_fpr): Ditto.
1563 (setup_incoming_varargs): Ditto.
1564 (rs6000_gimplify_va_arg): Ditto.
1565 (rs6000_split_multireg_move): Ditto.
1566 (rs6000_savres_strategy): Ditto.
1567 (rs6000_emit_prologue_components): Ditto.
1568 (rs6000_emit_epilogue_components): Ditto.
1569 (rs6000_emit_prologue): Ditto.
1570 (rs6000_emit_epilogue): Ditto.
1571 (rs6000_elf_file_end): Ditto.
1572 (rs6000_function_value): Ditto.
1573 (rs6000_libcall_value): Ditto.
1574 * config/rs6000/rs6000.h: Ditto.
1575 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1576 (TARGET_MINMAX): ... this. New.
1577 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1578 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1579 (*movsi_internal1_single): Delete.
1580 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1581 mfpu=, mxilinx-fpu): Delete.
1582 * config/rs6000/singlefp.h: Delete.
1583 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1584 * config/rs6000/t-rs6000: Ditto.
1585 * config/rs6000/t-xilinx: Delete.
1586 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1587 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1588 (VStype_simple): Delete.
1589 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1590 * config/rs6000/xfpu.h: Delete.
1591 * config/rs6000/xfpu.md: Delete.
1592 * config/rs6000/xilinx.h: Delete.
1593 * config/rs6000/xilinx.opt: Delete.
1594 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1595 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1597 2018-05-04 Tom de Vries <tom@codesourcery.com>
1600 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1603 2018-05-04 Richard Biener <rguenther@suse.de>
1606 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1607 (expand_complex_div_wide): Likewise.
1608 (expand_complex_operations_1): Likewise.
1609 (expand_complex_libcall): Preserve EH info of the original stmt.
1610 (tree_lower_complex): Handle removed blocks.
1611 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1612 on complex multiplication and division libcall builtins.
1614 2018-05-04 Richard Biener <rguenther@suse.de>
1617 * fold-const.c (negate_expr_p): Restrict negation of operand
1618 zero of a division to when we know that can happen without
1620 (fold_negate_expr_1): Likewise.
1622 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1625 * real.h (real_nextafter): Declare.
1626 * real.c (real_nextafter): New function.
1627 * fold-const-call.c (fold_const_nextafter): New function.
1628 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1629 CASE_CFN_NEXTTOWARD.
1630 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1631 even when arg1_mode is different from arg0_mode.
1633 2018-05-03 Nathan Sidwell <nathan@acm.org>
1635 * doc/extend.texi (Deprecated Features): Remove
1637 (Backwards Compatibility): Likewise.
1638 * doc/invoke.texi (C++ Language Options): Likewise.
1639 (C++ Dialect Options): Likewise.
1641 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1644 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1645 _mm512_mask_mullox_epi64): New intrinsics.
1647 2018-05-03 Tom de Vries <tom@codesourcery.com>
1650 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1651 dump files): Add offload-tree.
1653 2018-05-03 Richard Biener <rguenther@suse.de>
1655 PR tree-optimization/85615
1656 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1657 to loops not nested in BBs loop father to avoid creating multi-entry
1660 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1662 PR tree-optimization/70291
1663 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1664 arguments. Change return type to tree. Emit libcall as a new
1665 statement rather than replacing existing one when inplace_p is true.
1666 (expand_complex_multiplication_components): New function.
1667 (expand_complex_multiplication): Expand floating-point complex
1668 multiplication using the above.
1669 (expand_complex_division): Rename inner_type parameter to type.
1670 Update expand_complex_libcall call-site.
1671 (expand_complex_operations_1): Update expand_complex_multiplication
1672 and expand_complex_division call-sites.
1674 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1677 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1678 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1679 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1680 the highest significant bit of the shift count mask is clear. In
1681 check whether and[sq]i3 is needed verify that all significant bits
1682 of the shift count other than the highest are set.
1684 2018-05-02 Tom de Vries <tom@codesourcery.com>
1687 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1688 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1689 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1690 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1691 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1692 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1693 __builtin_goacc_parlevel_size.
1695 2018-05-02 Richard Biener <rguenther@suse.de>
1697 PR tree-optimization/85597
1698 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1699 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1702 2018-05-02 Tom de Vries <tom@codesourcery.com>
1705 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1706 dump files): Add ltrans-tree.
1708 2018-05-02 Tom de Vries <tom@codesourcery.com>
1711 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1712 dump files): Add wpa-ipa.
1714 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1716 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1717 powerpc*-*-linux*paired* target.
1718 * config/rs6000/750cl.h: Delete.
1719 * config/rs6000/paired.h: Delete.
1720 * config/rs6000/paired.md: Delete.
1721 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1723 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1724 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1725 comment. Remove paired float support.
1726 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1727 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1729 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1730 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1732 * config/rs6000/rs6000.c: Remove paired float support.
1733 (paired_expand_vector_init, paired_expand_vector_move,
1734 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1735 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1736 paired_expand_builtin, paired_expand_predicate_builtin,
1737 paired_init_builtins): Delete.
1738 * config/rs6000/rs6000.h: Remove paired float support.
1739 * config/rs6000/rs6000.md: Remove paired float support.
1740 (move_from_CR_ov_bit): Delete.
1741 * config/rs6000/rs6000.opt (mpaired): Delete.
1742 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1743 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1745 2018-05-02 Richard Biener <rguenther@suse.de>
1748 * gimplify.c (gimplify_save_expr): When in SSA form allow
1749 SAVE_EXPRs to compute to SSA vars.
1751 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1754 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1755 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1756 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1757 clobber operands[2], instead use a new pseudo. Formatting fixes.
1759 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1761 PR tree-optimization/85586
1762 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1763 exit early for statements in the same group if the accesses are
1766 2018-05-02 Tom de Vries <tom@codesourcery.com>
1769 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1772 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1774 PR tree-optimization/85143
1775 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1777 2018-05-01 Tom de Vries <tom@codesourcery.com>
1780 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1781 not found" error message.
1783 2018-05-01 Tom de Vries <tom@codesourcery.com>
1786 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1787 * vec.c (test_ordered_remove_if): New function.
1788 (vec_c_tests): Call test_ordered_remove_if.
1789 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1790 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1791 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1792 VEC_ORDERED_REMOVE_IF.
1794 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1796 PR tree-optimization/82665
1797 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1798 pointer subtraction where arguments come from a memchr call.
1800 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1802 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1803 --push-state --as-needed and --pop-state instead of --as-needed and
1804 --no-as-needed if ld supports it.
1805 * configure: Regenerated.
1808 * doc/install.texi2html: Replace _002d with - and _002a with * in
1809 generated html files using sed.
1811 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1814 * gcc-rich-location.c (blank_line_before_p): New function.
1815 (use_new_line): New function.
1816 (gcc_rich_location::add_fixit_insert_formatted): New function.
1817 * gcc-rich-location.h
1818 (gcc_rich_location::add_fixit_insert_formatted): New function.
1820 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1822 * selftest.c (assert_streq): Rename "expected" and "actual" to
1823 "val1" and "val2". Extend NULL-handling to cover both inputs
1824 symmetrically, while still requiring both to be non-NULL for a pass.
1825 * selftest.h (assert_streq): Rename "expected" and "actual" to
1827 (ASSERT_EQ): Likewise.
1828 (ASSERT_EQ_AT): Likewise.
1829 (ASSERT_KNOWN_EQ): Likewise.
1830 (ASSERT_KNOWN_EQ_AT): Likewise.
1831 (ASSERT_NE): Likewise.
1832 (ASSERT_MAYBE_NE): Likewise.
1833 (ASSERT_MAYBE_NE_AT): Likewise.
1834 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1835 the assertion to pass.
1836 (ASSERT_STREQ_AT): Likewise.
1838 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1840 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1841 interaction with -pie.
1843 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1845 * selftest.h: Fix alphabetization of per-source-file selftest
1848 2018-04-30 Jason Merrill <jason@redhat.com>
1850 PR c++/61982 - dead stores to destroyed objects.
1851 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1854 2018-04-30 Jason Merrill <jason@redhat.com>
1856 * tree.c (build_clobber): New.
1857 * tree.h: Declare it.
1858 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1860 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1862 * diagnostic-show-locus.c (layout::layout): Update for
1863 location_get_source_line returning a char_span.
1864 (struct char_span): Move to input.h.
1865 (struct correction): Update for fields in char_span becoming
1867 (struct source_line): Update for location_get_source_line
1868 returning a char_span.
1869 (layout::print_line): Likewise.
1870 * edit-context.c (edited_file::print_content): Likewise.
1871 (edited_file::print_diff_hunk): Likewise.
1872 (edited_file::print_run_of_changed_lines): Likewise.
1873 (edited_file::get_num_lines): Likewise.
1874 (edited_line::edited_line): Likewise.
1875 * final.c (asm_show_source): Likewise.
1876 * input.c (location_get_source_line): Convert return type
1877 from const char * to char_span, losing the final "line_len"
1879 (dump_location_info): Update for the above.
1880 (get_substring_ranges_for_loc): Likewise. Use a char_span
1881 when handling the literal within the line.
1882 (test_reading_source_line): Update for location_get_source_line
1883 returning a char_span.
1884 * input.h (class char_span): Move here from
1885 diagnostic-show-locus.c, converting from a struct to a class.
1886 Make data members private.
1887 (char_span::operator bool): New.
1888 (char_span::length): New.
1889 (char_span::get_buffer): New.
1890 (char_span::operator[]): New.
1891 (char_span::subspan): Make const.
1892 (char_span::xstrdup): New.
1893 (location_get_source_line): Convert return type from const char *
1894 to char_span, losing the final "line_size" param.
1896 2018-04-30 Jan Hubicka <jh@suse.cz>
1898 * lto-wrapper.c (ltrans_priorities): New static var.
1899 (cmp_priority): New.
1900 (run_gcc): Read priorities and if doing parallel build order
1901 the Makefile by them.
1903 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1905 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1907 2018-04-30 Richard Biener <rguenther@suse.de>
1909 * tree-cfg.c (verify_address): Remove base argument, add
1910 flag whether to check TREE_ADDRESSABLE and do that.
1911 (verify_expr): Remove.
1912 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1913 (verify_gimple_assign_single): Likewise.
1914 (verify_gimple_switch): Likewise.
1915 (verify_expr_location_1): Dereference tp once. Add (disabled)
1916 piece from verify_expr.
1917 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1919 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1921 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1923 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1925 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1926 (small_data_pattern): Likewise.
1927 (arc_rewrite_small_data): Likewise.
1928 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1929 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1930 (get_symbol_alignment): New function.
1931 (legitimate_small_data_address_p): Likewise.
1932 (legitimate_scaled_address): Update, call
1933 legitimate_small_data_address_p.
1934 (output_sdata): New static variable.
1935 (arc_print_operand): Update how we handle small data operands.
1936 (arc_print_operand_address): Likewise.
1937 (arc_legitimate_address_p): Update, use
1938 legitimate_small_data_address_p.
1939 (arc_rewrite_small_data_p): Remove.
1940 (arc_rewrite_small_data_1): Likewise.
1941 (arc_rewrite_small_data): Likewise.
1942 (small_data_pattern): Likewise.
1943 (compact_sda_memory_operand): Update to use
1944 legitimate_small_data_address_p and get_symbol_alignment.
1945 (prepare_move_operands): Don't rewite sdata pattern.
1946 (prepare_extend_operands): Remove.
1947 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1949 (zero_extendqisi2): Likewise.
1950 (zero_extendhisi2): Likewise.
1951 (extendqihi2): Likewise.
1952 (extendqisi2): Likewise.
1953 (extendhisi2): Likewise.
1957 * config/arc/constraints.md (Usd): Change it to memory constraint.
1959 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1961 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1962 as source of std instructions.
1963 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1964 allow 6-bit constants as source for store instructions.
1965 (movdi_insn): Update instruction pattern to allow 6-bit constants
1966 as source for store instructions.
1968 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1970 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1972 2018-04-30 Nathan Sidwell <nathan@acm.org>
1973 Sandra Loosemore <sandra@codesourcery.com>
1975 * dumpfile.c (dump_open): Allow '-' for stdout.
1976 * doc/invoke.texi (Developer Options): Document dump filename
1977 determination early. Document stdin/stdout selection.
1979 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1981 Microblaze Target: PIC data text relative
1983 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1984 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1986 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1987 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1988 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1989 New addressing mode for data-text relative position indepenedent code.
1990 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1991 'ADDRESS_SYMBOLIC_TXT_REL'.
1992 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1993 (microblaze_legitimate_pic_operand): Exclude function calls from
1994 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1995 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1997 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1998 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1999 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2000 for 'address + offset'.
2001 (microblaze_expand_prologue): Add new function prologue call for
2003 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2004 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2005 table in case of TARGET_PIC_DATA_TEXT_REL.
2006 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2007 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2008 Add new macros 'UNSPEC_TEXT',
2009 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2010 + exclude function calls from 'UNSPEC_PLT' in case of data text
2012 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2013 new target hook for generating address diff vector tables in case of
2015 * doc/tm.texi : Regenerate.
2016 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2017 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2018 of addr diff vector generation.
2019 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2020 target hook definition.
2021 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2022 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2023 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2024 option for data text relative.
2026 2018-04-30 Richard Biener <rguenther@suse.de>
2028 * tree-chrec.h (evolution_function_is_constant_p): Remove
2030 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2032 2018-04-30 Richard Biener <rguenther@suse.de>
2035 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2037 2018-04-30 Richard Biener <rguenther@suse.de>
2039 PR tree-optimization/28364
2040 PR tree-optimization/85275
2041 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2042 copying first exit test.
2044 2018-04-28 Mark Wielaard <mark@klomp.org>
2046 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2048 (dwarf_AT): Handle DW_AT_addr_base.
2049 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2051 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2054 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2055 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2056 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2057 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2059 2018-04-28 Richard Biener <rguenther@suse.de>
2061 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2062 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2063 to reflect use. Only add interesting stmts.
2065 2018-04-27 Martin Jambor <mjambor@suse.cz>
2068 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2069 the jump function allows for passing through aggregate values.
2071 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2073 * input.h (in_system_header_at): Convert from macro to inline
2075 (from_macro_expansion_at): Likewise.
2076 (from_macro_definition_at): Likewise.
2078 2018-04-27 Jeff Law <law@redhat.com>
2080 * config.gcc: Mark tile* targets as deprecated/obsolete.
2082 2018-04-27 Richard Biener <rguenther@suse.de>
2084 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2087 2018-04-27 Richard Biener <rguenther@suse.de>
2089 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2091 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2093 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2094 with Yd constraint. Set "preferred_for_speed" attribute from
2095 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2097 (*movdi_internal): Ditto.
2098 (movti_interunit splitters): Remove
2099 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2100 (movdi_interunit splitters): Ditto.
2101 * config/i386/constraints.md (Ye): Remove.
2102 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2104 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2107 * config/aarch64/constraints.md (Usg): Limit to 31.
2110 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2112 PR tree-optimization/85529
2113 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2114 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2115 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2116 zero extension or masking of the MSB bit.
2117 (optimize_range_tests): Add FIRST_BB argument, pass it through
2118 to optimize_range_tests_var_bound.
2119 (maybe_optimize_range_tests, reassociate_bb): Adjust
2120 optimize_range_tests callers.
2122 2018-04-26 Richard Biener <rguenther@suse.de>
2123 Jakub Jelinek <jakub@redhat.com>
2125 * cgraph.h (symbol_table): Just declare debug method here.
2126 * symtab.c (symbol_table::debug): Define.
2128 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2130 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2132 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2134 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2135 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2136 (*movdi_internal): Substitute Yi and Yj constraint with x
2137 and Ym and Yn constraint with y constraint. Update "isa"
2138 attribute and set "preferred_for_speed" attribute from
2139 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2140 (*movsi_internal): Ditto.
2141 (*movdf_internal): Ditto.
2142 (*movsf_internal): Ditto.
2143 (*zero_extendsidi2): Ditto.
2144 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2145 (sse2_loadld): Ditto.
2146 (*vec_extract<ssevecmodelower>_0): Ditto.
2147 (*vec_extractv4si_0_zext_sse4): Ditto.
2148 (vec_concatv2di): Ditto.
2149 (*vec_dup<mode>): Ditto.
2150 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2151 * config/i386/constraints.md (Yi): Remove.
2156 2018-04-26 Nathan Sidwell <nathan@acm.org>
2158 * dumpfile.c (dump_open): New.
2159 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2160 (dump_finish): Detect stdio/stderr by value not name.
2162 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2164 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2166 2018-04-26 Tom de Vries <tom@codesourcery.com>
2169 * config/nvptx/nvptx.c (verify_neutering_jumps)
2170 (verify_neutering_labels): New function
2171 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2173 2018-04-26 Tom de Vries <tom@codesourcery.com>
2176 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2177 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2180 2018-04-26 Richard Biener <rguenther@suse.de>
2181 Tom de Vries <tom@codesourcery.com>
2184 * lto-streamer-out.c (output_function): Fixup loops if required to match
2185 discovery done in the reader.
2187 2018-04-26 Richard Biener <rguenther@suse.de>
2189 PR tree-optimization/85116
2190 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2191 have a loop exit from the single latch predecessor. Remove
2192 case of header with just condition.
2193 (ch_base::copy_headers): Exclude infinite loops from any
2195 (pass_ch::execute): Record exits.
2197 2018-04-26 Richard Biener <rguenther@suse.de>
2199 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2200 prologue cost vector and pass it to vect_get_load_cost.
2201 (vect_get_peeling_costs_all_drs): Likewise.
2202 (vect_peeling_hash_get_lowest_cost): Likewise.
2203 (vect_enhance_data_refs_alignment): Likewise.
2205 2018-04-26 Richard Biener <rguenther@suse.de>
2208 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2209 checking of integer<->pointer conversions.
2210 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2211 sign-/zero-extending pointer types.
2212 (expand_omp_for_static_chunk): Likewise.
2214 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2215 Jean Lee <xiaoyur347@gmail.com>
2217 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2218 (TARGET_ASAN_SHADOW_OFFSET): Define.
2219 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2220 true for -fsanitize=address.
2222 2018-04-25 Mark Wielaard <mark@klomp.org>
2224 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2227 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2229 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2230 than "alu", remove explicit "memory" and "imm_disp" attributes.
2231 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2234 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2235 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2236 gen_lowpart_no_emit.
2238 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2241 * config/i386/i386.c (ix86_expand_builtin): Change memory
2242 operand to XI, extend p0 to Pmode.
2243 * config/i386/i386.md: Change unspec volatile and operand
2244 1 mode to XI, change operand 0 mode to P.
2246 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2248 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2249 GET_MODE_MASK before any checking.
2250 (nds32_can_use_bset_p): Likewise.
2251 (nds32_can_use_btgl_p): Likewise.
2253 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2255 * config/nds32/nds32-doubleword.md: New define_split pattern for
2256 illegal register number.
2258 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2260 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2262 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2264 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2266 2018-04-25 Richard Biener <rguenther@suse.de>
2268 * lto-streamer.h (LTO_major_version): Bump to 8.
2270 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2272 * BASE-VER: Set to 9.0.0.
2274 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2276 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2277 in __abskf2 and __powikf2.
2279 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2282 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2283 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2284 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2285 Use the above on operand 2. Reindent.
2286 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2288 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2291 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2293 * config/i386/i386.opt (mcet): Removed.
2294 * doc/install.texi: Remove -mcet documentation.
2295 * doc/invoke.texi: Likewise.
2297 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2300 * doc/install.texi: Remove -mcet from bootstrap-cet.
2302 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2305 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2306 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2310 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2311 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2312 containing a CONST_VECTOR.
2314 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2316 * doc/install.texi: Update newlib dependency for nvptx.
2318 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2321 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2322 instead of INTVAL when shifting x left.
2324 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2326 PR tree-optimization/85478
2327 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2328 vect_grouped_store_supported for single element vectors.
2330 2018-04-24 Richard Biener <rguenther@suse.de>
2333 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2334 load cost increase to the case of non-constant step.
2336 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2339 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2340 destination if any_malformed_asm.
2342 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2345 * expr.c (store_field): In the bitfield case, if the value comes from
2346 a function call and is returned in registers by means of a PARALLEL,
2347 do not change the mode of the temporary unless BLKmode and VOIDmode.
2349 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2351 PR rtl-optimization/85423
2352 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2353 dependencies to debug insns when the previous insn is non-debug.
2355 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2357 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2358 enums into a single definition.
2359 (fls): Fix predicates and printing.
2362 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2364 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2365 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2366 and short u6 immediate.
2367 (check_if_valid_sleep_operand): Remove.
2368 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2370 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2372 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2373 flag_always_save_lp condition.
2374 * config/nds32/nds32.opt (malways-save-lp): New option.
2376 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2378 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2379 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2380 * config/nds32/nds32.h
2381 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2382 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2384 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2386 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2387 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2389 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2390 Chung-Ju Wu <jasonwucj@gmail.com>
2392 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2394 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2395 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2397 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2399 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2401 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2403 * config/nds32/nds32-protos.h (nds32_data_alignment,
2404 nds32_local_alignment): Declare.
2405 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2406 nds32_local_alignment): New functions.
2407 (TARGET_CONSTANT_ALIGNMENT): Define.
2408 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2410 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2412 * config/nds32/nds32.c
2413 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2414 (TARGET_MODES_TIEABLE_P): Likewise.
2416 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2418 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2421 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2422 Chung-Ju Wu <jasonwucj@gmail.com>
2424 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2425 for unaligned access.
2426 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2428 * config/nds32/nds32-intrinsic.md: Likewise.
2429 * config/nds32/nds32_intrinsic.h: Likewise.
2430 * config/nds32/nds32.h (nds32_builtins): Likewise.
2431 * config/nds32/nds32.opt (munaligned-access): New option.
2432 * config/nds32/nds32.c (nds32_asm_file_start): Display
2433 flag_unaligned_access status.
2435 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2437 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2438 -mno-relax is present.
2439 * config/riscv/linux.h (LINK_SPEC): Ditto.
2441 2018-04-20 Martin Sebor <msebor@redhat.com>
2444 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2446 (gimple_fold_builtin_stxcpy_chk): Same.
2447 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2449 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2452 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2453 __powikf2 when long double is IEEE 128-bit.
2455 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2457 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2458 step to make sure stack always aligned.
2460 2018-04-20 Carl Love <cel@us.ibm.com>
2463 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2464 size check for arg0.
2466 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2467 Tom de Vries <tom@codesourcery.com>
2470 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2471 Emit insns for calls too.
2472 (nvptx_find_par): Always look for worker-level predecessor insn.
2473 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2475 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2476 (nvptx_process_pars): Propagate frames for calls.
2478 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2481 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2483 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2484 (ix86_handle_option): Don't handle OPT_mibt.
2485 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2487 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2489 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2490 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2491 (ix86_target_macros): Define __CET__ with flag_cf_protection
2492 for -fcf-protection.
2493 * config/i386/i386.c (isa2_opts): Remove -mibt.
2494 * config/i386/i386.h (TARGET_IBT): Removed.
2495 (TARGET_IBT_P): Likewise.
2496 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2497 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2498 * config/i386/i386.opt (mcet): Update help message.
2501 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2502 -mcet as an alias for -mshstk.
2504 2018-04-20 Richard Biener <rguenther@suse.de>
2507 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2508 complexity by forcing a single use of the multiply operand.
2510 2018-04-20 Martin Jambor <mjambor@suse.cz>
2513 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2514 recursion dependency to only apply to non-clones.
2516 2018-04-20 Martin Jambor <mjambor@suse.cz>
2519 * ipa-cp.c (create_specialized_node): Check that clones of
2520 self-recursive edges exist during IPA-CP.
2522 2018-04-19 Toon Moene <toon@moene.org>
2524 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2527 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2529 PR tree-optimization/85467
2530 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2531 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2532 VECTOR_CST element to type.
2534 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2537 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2538 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2539 (builtin_longjmp): Likewise.
2540 (save_stack_nonlocal): New pattern.
2541 (restore_stack_nonlocal): Likewise.
2543 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2546 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2547 Replace ASM_OUTPUT_LABEL with fprintf.
2549 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2552 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2553 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2554 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2555 define __IBT__ and __SHSTK__ for -fcf-protection.
2556 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2558 (ix86_trampoline_init): Likewise.
2559 (x86_output_mi_thunk): Likewise.
2560 (ix86_notrack_prefixed_insn_p): Likewise.
2561 (ix86_option_override_internal): Don't disallow -fcf-protection.
2562 * config/i386/i386.md (rdssp<mode>): Also enable for
2564 (incssp<mode>): Likewise.
2565 (nop_endbr): Likewise.
2566 * config/i386/i386.opt (mcet): Change help message to built-in
2570 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2571 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2572 enable CET built-in functions.
2574 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2576 * common/config/i386/i386-common.c
2577 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2578 OPTION_MASK_ISA_MOVDIRI_UNSET,
2579 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2580 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2581 * config.gcc (movdirintrin.h): New header.
2582 * config/i386/cpuid.h (bit_MOVDIRI,
2583 bit_MOVDIR64B): New bits.
2584 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2586 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2587 (VOID, PVOID, PCVOID)): New function types.
2588 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2589 __builtin_ia32_directstoreu_u64,
2590 __builtin_ia32_movdir64b): New builtins.
2591 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2592 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2594 (ix86_valid_target_attribute_inner_p): Ditto.
2595 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2596 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2597 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2598 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2599 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2600 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2601 (movdiri<mode>, movdir64b_<mode>): New.
2602 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2603 * config/i386/immintrin.h: Include movdirintrin.h.
2604 * config/i386/movdirintrin.h: New file.
2605 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2607 2018-04-19 Richard Biener <rguenther@suse.de>
2610 * cfg.c (clear_bb_flags): When loop state says we have
2611 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2613 2018-04-19 Richard Biener <rguenther@suse.de>
2615 PR tree-optimization/84737
2616 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2617 copying restrict info.
2618 (vect_setup_realignment): Use it.
2619 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2620 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2621 the first DR to all generated stores.
2622 (vectorizable_load): Likewise for loads.
2624 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2626 PR tree-optimization/85446
2627 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2628 the integral and pointer types to have the same precision.
2630 * doc/install.texi: Document --disable-cet being the default and
2633 2018-04-18 Martin Liska <mliska@suse.cz>
2635 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2638 2018-04-18 Martin Liska <mliska@suse.cz>
2641 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2644 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2645 arguments if they are comparable.
2647 2018-04-18 Martin Liska <mliska@suse.cz>
2650 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2653 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2656 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2659 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2660 ENDBR after calling __morestack.
2662 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2665 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2666 by using gcc_base_ver to generate a gcc_driver_version, and use
2667 it when generating GCC_DRIVER_NAME.
2668 * configure: Regenerate.
2670 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2673 * config.gcc: Obsolete powerpc*-*-*spe*.
2675 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2678 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2679 (stabstr_D): Change type of unum from unsigned int to
2680 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2683 2018-04-17 Jim Wilson <jimw@sifive.com>
2686 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2687 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2688 Set arg_pointer_offset after using pretend_args_size.
2690 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2692 PR rtl-optimization/85431
2693 * dse.c (record_store): Ignore zero width stores.
2696 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2697 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2698 __builtin_stack_restore rather than after it.
2699 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2700 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2701 argument instead of virtual_dynamic_stack_rtx.
2703 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2705 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2707 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2708 Add note to error message to explain internal mapping of overloaded
2709 built-in function name to non-overloaded built-in function name.
2710 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2713 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2716 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2717 where the inputs overlap with the output.
2719 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2721 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2722 (=v, v) alternative and explicit "memory" attribute.
2723 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2724 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2726 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2727 "sselog1" type instead of "sselog".
2728 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2729 "sselog". Remove explicit "memory" attribute.
2730 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2731 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2733 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2734 "sselog1" type instead of "sselog". Remove explicit "memory"
2736 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2737 use "sselog1" type instead of "sselog". Remove explicit "memory"
2739 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2740 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2742 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2743 "sselog1" type instead of "sselog". Remove explicit "memory"
2745 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2746 use "sselog1" type instead of "sselog". Remove explicit "memory"
2750 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2753 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2756 2018-04-17 Martin Jambor <mjambor@suse.cz>
2759 * ipa-cp.c (create_specialized_node): Call
2760 expand_all_artificial_thunks if necessary.
2762 2018-04-17 Martin Liska <mliska@suse.cz>
2765 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2766 in message, remote space in between '_G' and '('.
2768 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2771 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2772 avx512f_vmcmp<mode>3<round_saeonly_name>,
2773 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2774 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2775 avx512f_rndscale<mode><round_saeonly_name>,
2776 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2777 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2778 Use %<iptr>2 instead of %2 for -masm=intel.
2779 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2780 avx512f_vcvttss2usi<round_saeonly_name>,
2781 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2783 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2784 avx512f_vcvttsd2usi<round_saeonly_name>,
2785 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2786 Use %q1 instead of %1 for -masm=intel.
2787 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2788 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2789 of %3 for -masm=intel.
2790 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2792 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2794 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2796 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2797 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2799 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2801 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2802 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2803 %g1 and one with %0 and %1.
2804 (avx512er_vmrcp28<mode><round_saeonly_name>,
2805 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2807 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2808 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2809 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2810 of %0 and %{%4%} for -masm=intel.
2811 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2812 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2813 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2814 order of %0 and %{%5%}%{z%} for -masm=intel.
2816 2018-04-17 Jan Hubicka <jh@suse.cz>
2819 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2821 2018-04-17 Martin Liska <mliska@suse.cz>
2824 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2825 for target_clone error message. Make default implementation
2826 clone to be a local declaration.
2827 (separate_attrs): Add new argument and check for an empty
2829 (expand_target_clones): Handle it.
2830 (ipa_target_clone): Make redirection just for target_clones
2833 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2834 Tom de Vries <tom@codesourcery.com>
2837 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2838 tiled basic blocks without omp continue statements.
2840 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2843 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2844 vec_extract expression as having side effects to make sure it gets
2847 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2850 * config/i386/i386.c (get_builtin_code_for_version): Check
2853 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2856 * gcc/config.gcc: Support "skylake".
2857 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2859 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2860 (processor_target_table): Add "skylake".
2861 (ix86_option_override_internal): Add "skylake".
2862 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2863 PROCESSOR_CANNONLAKE.
2864 (get_builtin_code_for_version): Fix priority for
2865 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2866 PROCESSOR_SKYLAKE-AVX512.
2867 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2868 (processor_type): Add PROCESSOR_SKYLAKE.
2870 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2871 Jason Merrill <jason@redhat.com>
2874 * convert.c (convert_to_integer_1): Use direct recursion for
2875 enumeral types and types with a precision less than the number
2876 of bits in their mode.
2878 2018-04-16 Julia Koval <julia.koval@intel.com>
2881 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2882 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2884 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2887 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2888 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2889 and -mno-direct-move.
2891 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2894 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2895 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2896 Ensure negative shifts result in {0}.
2898 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2900 PR rtl-optimization/79916
2901 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2902 regs (if any) to define how to gnerate SD moves when LRA is in
2905 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2907 PR rtl-optimization/85393
2908 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2909 * except.c (expand_dw2_landing_pad_for_region): Make static.
2910 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2911 a label and unconditional jump to old_bb, rather than
2912 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2915 PR rtl-optimization/85376
2916 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2917 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2918 instead of a specific value.
2920 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2921 Bin Cheng <bin.cheng@arm.com>
2923 PR tree-optimization/82965
2924 PR tree-optimization/83991
2925 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2926 by_profile_only parameter.
2927 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2928 information if the loop was predicted to iterate too many times.
2929 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2931 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2934 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2937 2018-04-13 Martin Liska <mliska@suse.cz>
2938 Jakub Jelinek <jakub@redhat.com>
2941 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2942 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2943 * builtins.c (expand_builtin_memory_copy_args): Use
2944 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2945 handle dest_addr == pc_rtx.
2947 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2950 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2951 asked to not generate direct moves.
2952 (fix_trunc<mode>si2_stfiwx): Similar.
2953 (fix_trunc<mode>si2_internal): Similar.
2955 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2958 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2959 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2960 lookup if dest in some wider mode is known to be const0_rtx and
2961 if so, record permanent equivalence for it to be ZERO_EXTEND of
2962 the narrower mode destination.
2964 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2966 * lto-streamer-out.c (output_function): Revert 259346.
2967 * omp-expand.c (expand_oacc_for): Likewise.
2969 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2971 PR rtl-optimization/85354
2972 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2973 * sel-sched.c (sel_global_init): ... here.
2975 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2978 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2979 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2980 mode for PE-COFF targets.
2981 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2982 (i386_pe_asm_lto_end): Likewise.
2983 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2984 (TARGET_ASM_LTO_END): Likewise.
2985 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2986 (i386_pe_asm_lto_start): New function.
2987 (i386_pe_asm_lto_end): Likewise.
2989 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2990 Richard Biener <rguenther@suse.de>
2993 * lto-streamer-out.c (output_function): Fix CFG loop state before
2995 * omp-expand.c (expand_oacc_for): Handle calls to internal
2996 functions like regular functions.
2998 2018-04-12 Richard Biener <rguenther@suse.de>
3001 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3002 for the early LTO debug to properly generate references to it
3003 during DIE emission. Do not re-use that for the skeleton for
3005 (dwarf2out_early_finish): Likewise.
3007 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3010 * config/i386/sse.md
3011 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3012 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3013 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3014 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3015 and output is a reg, avoid creating invalid lowpart subreg, but
3016 instead split into a 512-bit move. Don't split if not AVX512VL,
3017 input is xmm16+ reg and output is a mem.
3018 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3019 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3020 xmm16+ reg and output is a mem.
3022 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3024 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3025 also for flag_dwarf2_cfi_asm.
3027 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3029 PR rtl-optimization/85342
3030 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3031 a bool scalar var inside of the loop instead. Don't try to update
3032 recog_data.operand after failed apply_change_group.
3034 2018-04-12 Tom de Vries <tom@codesourcery.com>
3037 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3038 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3039 array with flexible array member as array without given dimension.
3040 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3041 argument for undefined param to true.
3043 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3046 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3047 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3048 from PowerPC section.
3049 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3050 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3051 help text that is too long.
3052 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3053 help text that is too long.
3054 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3055 help text that is too long.
3057 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3059 * config/alpha/alpha.md (stack_probe_internal): Rename
3060 from "probe_stack". Update all callers.
3062 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3064 PR rtl-optimization/84566
3065 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3066 sched_macro_fuse_insns.
3068 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3071 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3072 (compute_block_dependences): ... from here.
3074 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3076 PR tree-optimization/85331
3077 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3078 from int to HOST_WIDE_INT.
3080 2018-04-11 Martin Jambor <mjambor@suse.cz>
3083 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3084 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3085 not the same as the source val.
3086 (cgraph_edge_brings_value_p): New parameter.
3087 (gather_edges_for_value): Pass destination value to
3088 cgraph_edge_brings_value_p.
3089 (perhaps_add_new_callers): Likewise.
3090 (get_info_about_necessary_edges): Likewise and exclude values brought
3091 only by self-recursive edges.
3092 (create_specialized_node): Redirect only clones of self-calling edges.
3093 (+self_recursive_pass_through_p): New function.
3094 (find_more_scalar_values_for_callers_subset): Use it.
3095 (find_aggregate_values_for_callers_subset): Likewise.
3096 (known_aggs_to_agg_replacement_list): Removed.
3097 (decide_whether_version_node): Re-calculate known constants for all
3098 remaining context clones.
3100 2018-04-11 Richard Biener <rguenther@suse.de>
3103 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3104 from early DWARF output.
3105 (dwarf2out_early_finish): Output line info unconditionally into
3106 early DWARF and add reference to it.
3108 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3111 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3112 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3113 other than V2DFmode using iptr mode attribute.
3114 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3116 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3118 PR rtl-optimization/84659
3119 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3121 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3124 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3126 (output_loc_list): Pass address of a dummy size variable even in the
3127 locview handling loop.
3128 (index_location_lists): Add comment on why skip_loc_list_entry can't
3131 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3134 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3137 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3140 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3141 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3142 and -mstring-compare-inline-limit.
3144 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3147 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3148 for stack clash protection in a register whenever we need it to be in
3151 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3153 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3154 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3156 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3159 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3161 (mlong-double-): Ditto.
3162 * config/rs6000/sysv4.opt (msdata=): Ditto.
3163 (mtls-size=): Ditto.
3165 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3167 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3168 erroneous entries for
3169 "vector int vec_ldl (int, long int *)", and
3170 "vector unsigned int vec_ldl (int, unsigned long int *)".
3171 Add comments and entries for
3172 "vector bool char vec_ldl (int, bool char *)",
3173 "vector bool short vec_ldl (int, bool short *)",
3174 "vector bool int vec_ldl (int, bool int *)",
3175 "vector bool long long vec_ldl (int, bool long long *)",
3176 "vector pixel vec_ldl (int, pixel *)",
3177 "vector long long vec_ldl (int, long long *)",
3178 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3179 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3180 type tree bool_long_long_type_node and correct definition of
3181 bool_V2DI_type_node to make reference to this new type tree.
3182 (rs6000_mangle_type): Replace erroneous reference to
3183 bool_long_type_node with bool_long_long_type_node.
3184 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3185 comments to emphasize sign distinctions for char and int types and
3186 replace RS6000_BTI_bool_long constant with
3187 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3188 use of RS6000_BTI_pixel.
3189 (bool_long_type_node): Remove this macro definition.
3190 (bool_long_long_type_node): New macro definition
3192 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3194 PR rtl-optimization/85300
3195 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3196 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3197 simplify_unary_operation fails.
3199 2018-04-10 Martin Liska <mliska@suse.cz>
3201 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3202 cgraph_edge and ipa_ref.
3204 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3208 * config/i386/sse.md
3209 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3210 computation of the VEC_MERGE selector from mask.
3211 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3212 Fix decoding of the VEC_MERGE selector into mask.
3214 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3216 PR tree-optimization/85286
3217 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3219 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3221 * final.c (final_1): Set insn_last_address as well as
3222 insn_current_address.
3224 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3227 * explow.c (emit_stack_probe): Call validize_mem on memory location
3228 before passing it to gen_probe_stack. Create address operand and
3229 legitimize it for the probe_stack_address case.
3231 2018-04-09 Jan Hubicka <jh@suse.cz>
3234 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3235 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3236 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3238 (free_lang_data): Rebuild type inheritance graph.
3240 2018-04-09 Martin Sebor <msebor@redhat.com>
3242 * invoke.texi (-finline-small-functions): Mention other optimization
3244 (-findirect-inlining, -fpartial-inlining): Same.
3245 (-finline-functions-called-once): Same.
3246 (-freorder-blocks-and-partition): Same.
3248 2018-04-09 Jan Hubicka <jh@suse.cz>
3251 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3252 jumps; choose last target that matches the criteria (i.e.
3253 no partition changes for non-crossing jumps).
3254 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3255 support for redirecting crossing jumps to non-crossing.
3257 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3259 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3260 also for naked functions.
3262 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3264 * config/arc/arc.md (add_shift): New pattern.
3265 (add_shift2): Likewise.
3266 (sub_shift): Likewise.
3267 (sub_shift_cmp0_noout): Likewise.
3268 (compare_si_ashiftsi): Likewise.
3269 (xbfu_cmp0_noout): New combine pattern.
3270 (xbfu_cmp0"): Likewise.
3271 (movsi_set_cc_insn): Place the predicable variant first.
3272 (commutative_binary_cmp0_noout): Remove clobber.
3273 (commutative_binary_cmp0): New pattern.
3274 (noncommutative_binary_cmp0): Likewise.
3275 (noncommutative_binary_cmp0_noout): Likewise.
3276 (noncommutative_binary_comparison_result_used): Removed.
3277 (rsub_cmp0): New pattern.
3278 (rsub_cmp0_noout): Likewise.
3279 (extzvsi): Changed, keep only meaningful variants.
3280 (SQH, SEZ): New iterators.
3281 (SQH_postfix): New mode attribute.
3282 (SEZ_prefix): New code attribute.
3283 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3284 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3285 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3287 (noncommutative_operator): Check the availability of barrel
3290 2018-04-09 Richard Biener <rguenther@suse.de>
3292 PR tree-optimization/85284
3293 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3294 Only use the niter constraining form of simple_iv when the exit
3297 2018-04-09 Tom de Vries <tom@codesourcery.com>
3300 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3301 (define_expand "*memory_barrier"): New define_expand.
3302 (define_insn "memory_barrier"): New insn.
3304 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3306 PR rtl-optimization/80463
3307 PR rtl-optimization/83972
3308 PR rtl-optimization/83480
3310 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3311 correct producer for the insn.
3312 (tidy_control_flow): Fixup seqnos in case of debug insns.
3314 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3316 PR rtl-optimization/83913
3318 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3319 different sched-times when merging exprs.
3321 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3323 PR rtl-optimization/83962
3325 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3326 tidy_fallthru_edge and tidy_control_flow.
3328 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3330 PR rtl-optimization/83530
3332 * sel-sched.c (force_next_insn): New global variable.
3333 (remove_insn_for_debug): When force_next_insn is true, also leave only
3334 next insn in the ready list.
3335 (sel_sched_region): When the region wasn't scheduled, make another pass
3336 over it with force_next_insn set to 1.
3338 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3340 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3342 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3343 for interrupt control.
3344 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3345 functions for interrupt control.
3346 * config/nds32/nds32-intrinsic.md: Likewise.
3347 * config/nds32/nds32_intrinsic.h: Likewise.
3348 * config/nds32/nds32.h (nds32_builtins): Likewise.
3350 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3352 * config/nds32/nds32.c (nds32_init_machine_status,
3353 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3354 strict_aligned_p field.
3355 (nds32_expand_to_rtl_hook): New function.
3356 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3357 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3359 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3360 Chung-Ju Wu <jasonwucj@gmail.com>
3362 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3363 * config/nds32/nds32-n7.md: New file.
3364 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3365 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3367 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3368 * config/nds32/nds32.md (pipeline_model): Add n7.
3369 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3370 * config/nds32/pipelines.md: Include n7 settings.
3372 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3373 Chung-Ju Wu <jasonwucj@gmail.com>
3375 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3376 * config/nds32/nds32-e8.md: New file.
3377 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3378 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3380 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3381 * config/nds32/nds32.md (pipeline_model): Add e8.
3382 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3383 * config/nds32/pipelines.md: Include e8 settings.
3385 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3386 Chung-Ju Wu <jasonwucj@gmail.com>
3388 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3389 * config/nds32/nds32-n8.md: New file.
3390 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3391 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3393 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3394 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3395 * config/nds32/nds32.md (pipeline_model): Add n8.
3396 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3397 * config/nds32/pipelines.md: Include n8 settings.
3399 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3400 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3403 * config/nds32/nds32-n9-2r1w.md: New file.
3404 * config/nds32/nds32-n9-3r2w.md: New file.
3405 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3406 nds32_register_ports): New or modify for cpu n9.
3407 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3409 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3410 * config/nds32/nds32-utils.c: New file.
3411 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3412 TARGET_MUL_SLOW): Define.
3413 * config/nds32/nds32.md (pipeline_model): New attribute.
3414 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3415 New options that support cpu n9.
3416 * config/nds32/pipelines.md: Include n9 settings.
3417 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3419 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3421 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3422 information if necessary.
3423 (output_cond_branch_compare_zero): Likewise.
3424 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3425 (nds32_target_alignment): Refine for alignment.
3426 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3427 (FUNCTION_BOUNDARY): Modify.
3428 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3430 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3432 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3434 * config/nds32/constants.md (unspec_volatile_element): Add values for
3435 TLB operation and data prefetch.
3436 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3437 functions for TLB operation and data prefetch.
3438 * config/nds32/nds32-intrinsic.md: Likewise.
3439 * config/nds32/nds32_intrinsic.h: Likewise.
3440 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3441 (nds32_print_operand): Likewise.
3442 * config/nds32/nds32.h (nds32_builtins): Likewise.
3444 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3445 Andrew Pinski <pinsika@gcc.gnu.org>
3448 * match.pd: Use constant_boolean_node of correct type instead of
3449 boolean_true_node or boolean_false_node for simplifying
3450 pointer comparisons to zero.
3452 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3454 PR tree-optimization/80021
3455 * tree.c (verify_type_variant): Make error call in verify_variant_match
3456 translatable and remove final full stop.
3458 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3460 * config/nds32/constants.md (unspec_volatile_element): Add
3461 UNSPEC_VOLATILE_EH_RETURN.
3462 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3463 nds32_output_stack_pop): Support dwarf exception handling process.
3464 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3465 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3466 exception handling process.
3467 (nds32_compute_stack_frame): Likewise.
3468 (nds32_return_addr_rtx): Likewise.
3469 (nds32_initial_elimination_offset): Likewise.
3470 (nds32_expand_prologue): Likewise.
3471 (nds32_expand_epilogue): Likewise.
3472 (nds32_dynamic_chain_address): New function.
3473 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3475 (DYNAMIC_CHAIN_ADDRESS): Define.
3476 (EH_RETURN_DATA_REGNO): Define.
3477 (EH_RETURN_STACKADJ_RTX): Define.
3478 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3479 patterns for dwarf exception handling.
3481 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3483 * config/nds32/nds32.h: Clean up obsolete macros.
3485 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3487 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3488 Add enum values for particular instructions.
3489 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3490 particular intrinsic functions.
3491 * config/nds32/nds32-intrinsic.md: Likewise.
3492 * config/nds32/nds32_intrinsic.h: Likewise.
3493 * config/nds32/nds32.h (nds32_builtins): Likewise.
3494 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3495 (btst, ave): New patterns for particular instructions.
3497 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3499 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3500 Add enum values for atomic load/store and memory sync.
3501 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3503 * config/nds32/nds32-intrinsic.md: Likewise.
3504 * config/nds32/nds32_intrinsic.h: Likewise.
3505 * config/nds32/nds32.h (nds32_builtins): Likewise.
3507 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3509 PR tree-optimization/85257
3510 * fold-const.c (native_encode_vector): If not all elts could fit
3511 and off is -1, return 0 rather than offset.
3512 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3513 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3514 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3515 adjust buffer in native_interpret_expr call.
3517 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3519 * config/nds32/constants.md (unspec_volatile_element): Add cache
3520 control enum values.
3521 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3522 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3523 * config/nds32/nds32.c (nds32_cctl_names): New.
3524 (nds32_print_operand): Handle cache control register names.
3525 * config/nds32/nds32.h (nds32_builtins): New enum values.
3526 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3528 * config/nds32/nds32.md (type): Add mmu.
3529 * config/nds32/pipelines.md (simple_insn): Add mmu.
3531 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3533 * config/nds32/nds32.md (type): Remove call.
3534 * config/nds32/pipelines.md (simple_insn): Likewise.
3536 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3538 * config/nds32/constants.md (unspec_volatile_element): Add
3539 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3540 UNSPEC_VOLATILE_FMFCFG.
3541 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3542 description for fmfcfg and fmfcsr.
3543 (bdesc_1arg): Add fmtcsr.
3544 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3545 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3546 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3547 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3548 unspec_fmfcfg): New patterns.
3549 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3550 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3551 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3552 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3553 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3554 __nds32__fmfcfg): Define.
3556 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3558 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3559 intrinsic register names.
3560 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3561 intrinsic register enum values and macros.
3563 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3565 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3566 for load/store addressing form.
3567 (nds32_print_operand_address): Likewise.
3569 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3572 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3573 based on LABEL_REF. Remove useless assertion.
3574 (pic_address_needs_scratch): Fix formatting.
3575 (sparc_legitimize_pic_address): Minor tweaks.
3576 (sparc_delegitimize_address): Adjust assertion accordingly.
3577 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3578 into symbolic_operand.
3579 (movsi_high_pic_label_ref): Likewise.
3580 (movsi_lo_sum_pic_label_ref): Likewise.
3581 (movdi_pic_label_ref): Likewise.
3582 (movdi_high_pic_label_ref): Likewise.
3583 (movdi_lo_sum_pic_label_ref): Likewise.
3585 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3587 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3588 custom LIB_SPEC setup.
3590 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3591 Kito Cheng <kito.cheng@gmail.com>
3593 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3594 * config/riscv/freebsd.h: New.
3596 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3598 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3599 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3602 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3603 Kito Cheng <kito.cheng@gmail.com>
3605 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3606 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3607 * config/nds32/nds32-protos.h (nds32_output_call,
3608 nds32_output_return): Declare.
3609 * config/nds32/nds32.md: Refine all the call and return patterns.
3611 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3614 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3615 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3617 PR rtl-optimization/84872
3618 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3619 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3622 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3624 * expr.c (copy_blkmode_to_reg): Revert 254862.
3625 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3627 2018-04-06 Richard Biener <rguenther@suse.de>
3630 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3631 after seeing a component reference with an adjacent field. Treat
3632 refs to arrays at struct end of external decls similar to
3633 refs to unconstrained commons.
3635 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3638 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3639 look through SAVE_EXPRs with non-side-effects argument. Adjust
3641 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3644 2018-04-06 Richard Biener <rguenther@suse.de>
3647 * alias.c (find_base_term): New wrapper around find_base_term
3648 unwinding CSELIB_VAL_PTR changes.
3649 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3652 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3654 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3656 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3657 constant definitions.
3658 ("nop"): lr 0,0 -> nopr r0
3659 ("nop_lr0", "nop_lr1"): New insn definitions.
3661 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3663 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3664 NDS32_V3PUSH_AVAILABLE_P macro.
3666 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3667 Chung-Ju Wu <jasonwucj@gmail.com>
3669 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3670 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3671 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3672 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3673 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3674 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3675 * config/nds32/constraints.md: New constraints and checking for hard
3676 float configuration.
3677 * config/nds32/iterators.md: New mode iterator and attribute for hard
3678 float configuration.
3679 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3681 * config/nds32/nds32-fpu.md: New file.
3682 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3683 deal with hard float code generation.
3684 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3686 (abi_type, float_reg_number): New enum type.
3687 * config/nds32/nds32-predicates.c: New predicates for hard float.
3688 * config/nds32/nds32-protos.h: Declare functions for hard float.
3689 * config/nds32/nds32.c: Implementation for hard float configuration.
3690 * config/nds32/nds32.h: Definitions for hard float configuration.
3691 * config/nds32/nds32.md: Include hard float machine description and
3692 modify patterns for hard float configuration.
3693 * config/nds32/nds32.opt: New options for hard float configuration.
3694 * config/nds32/predicates.md: New predicates for hard float
3697 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3699 * common/config/nds32/nds32-common.c
3700 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3702 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3705 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3706 CONSTRUCTOR_ELT (ctor, ...)->value.
3708 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3711 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3713 2018-04-05 Tom de Vries <tom@codesourcery.com>
3716 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3719 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3720 Kito Cheng <kito.cheng@gmail.com>
3722 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3723 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3724 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3725 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3727 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3728 Kito Cheng <kito.cheng@gmail.com>
3730 * config/nds32/constraints.md (Ufe): New memory constraint.
3731 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3732 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3733 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3735 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3736 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3738 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3740 * config/nds32/nds32.md: Use optimize_size in the condition for
3741 alu-shift instructions.
3743 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3745 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3747 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3749 * config/nds32/nds32.md (negsi2): Refine pattern.
3751 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3752 Chung-Ju Wu <jasonwucj@gmail.com>
3754 * config/nds32/iterators.md (shift_rotate): New code iterator.
3755 (shift): New code attribute.
3756 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3757 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3758 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3759 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3760 bit-wise operations.
3761 (andsi3, *andsi3): Ditto.
3762 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3763 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3764 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3765 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3766 nds32_ior_operand, nds32_xor_operand): New predicates.
3768 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3770 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3771 (addsi3, subsi3): ... this.
3773 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3775 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3777 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3779 * config/nds32/nds32.md: Adjust indention.
3781 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3783 * config/nds32/nds32.md (feature): New attribute.
3785 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3787 * config/nds32/nds32.md (subtype): New attribute.
3789 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3792 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3793 expansion to perform a bitwise AND of the argument followed by a
3794 boolean negation of the result.
3796 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3798 PR rtl-optimization/84878
3799 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3800 the basic block. Assert the use reference is not artificial and that
3801 it has an associated insn.
3803 2018-04-04 Michael Matz <matz@suse.de>
3805 * builtins.c (compute_objsize): Pass correct operand
3806 to array_at_struct_end_p.
3808 2018-04-04 Richard Biener <rguenther@suse.de>
3811 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3812 from contexts for DINFO_LEVEL_TERSE and below.
3814 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3816 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3817 resiter_operand condition.
3818 * config/nds32/nds32.md (*move<mode>): Ditto.
3820 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3821 Monk Chiang <sh.chiang04@gmail.com>
3823 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3825 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3827 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3829 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3830 Kito Cheng <kito.cheng@gmail.com>
3832 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3833 nds32_cond_code_str, output_cond_branch,
3834 output_cond_branch_compare_zero, nds32_expand_cbranch,
3835 nds32_expand_cstore, nds32_expand_movcc,
3836 nds32_output_cbranchsi4_equality_zero,
3837 nds32_output_cbranchsi4_equality_reg,
3838 nds32_output_cbranchsi4_equality_reg_or_const_int,
3839 nds32_output_cbranchsi4_greater_less_zero: New functions.
3840 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3841 nds32_expand_cstore, nds32_expand_movcc,
3842 nds32_output_cbranchsi4_equality_zero,
3843 nds32_output_cbranchsi4_equality_reg,
3844 nds32_output_cbranchsi4_equality_reg_or_const_int,
3845 nds32_output_cbranchsi4_greater_less_zero): Declare.
3846 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3847 nds32_rimm11s_operand): New predicates.
3848 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3849 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3852 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3854 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3855 * config/nds32/nds32.md: Ditto.
3856 * config/nds32/pipelines.md: Ditto.
3858 2018-04-04 Richard Biener <rguenther@suse.de>
3860 PR tree-optimization/85168
3861 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3862 propagating abnormals.
3864 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3866 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3868 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3869 Kito Cheng <kito.cheng@gmail.com>
3871 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3872 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3873 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3874 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3875 * config/nds32/nds32.md (sibcall_internal): New.
3876 (sibcall_register): Remove.
3877 (sibcall_immediate): Remove.
3878 (sibcall_value_internal): New.
3879 (sibcall_value_register): Remove.
3880 (sibcall_value_immediate): Remove.
3881 * config/nds32/predicates.md (nds32_general_register_operand): New.
3882 (nds32_call_address_operand): New.
3884 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3886 PR rtl-optimization/85167
3887 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3888 bb_defs if *split_p, instead preinitialize it to NULL.
3890 PR tree-optimization/85156
3891 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3892 evaluating the argument multiple times.
3894 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3896 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3898 (_mm_cvtpd_ps): Likewise.
3899 (_mm_cvttpd_epi32): Likewise.
3900 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3901 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3902 vector, pixel, and bool following altivec.h include.
3904 2018-04-03 Martin Sebor <msebor@redhat.com>
3906 * doc/extend.texi (Common Function Attributes): Clarify.
3907 (const attribute): Likewise.
3908 (pure attribute): Likewise.
3910 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3913 * config/i386/i386.c (ix86_expand_vector_set): Use
3914 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3916 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3918 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3919 instructions when changing rounding bits to preserve precision bits
3920 in the x87 control word.
3922 2018-04-03 Martin Liska <mliska@suse.cz>
3924 PR tree-optimization/82491
3925 * rtl.h (strip_offset_and_add): Replace += suboffset with
3926 poly_uint64 () + suboffset.
3928 2018-03-29 Martin Liska <mliska@suse.cz>
3929 Martin Jambor <mjambor@suse.cz>
3932 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3933 param_type is not an integral or pointer type.
3935 2018-04-03 Richard Biener <rguenther@suse.de>
3937 * sese.h (recompute_all_dominators): Remove.
3939 2018-04-02 Martin Sebor <msebor@redhat.com>
3941 * doc/invoke.texi (-Wrestrict): Fix typos.
3943 2018-04-02 Jim Wilson <jimw@sifive.com>
3945 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3946 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3947 (<optab>di3, <optab>si3_extend): Likewise.
3948 (<optab>si3_mask, <optab>si3_mask_1): New.
3949 (<optab>di3_mask, <optab>di3_mask_1): New.
3950 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3951 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3952 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3954 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3956 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3959 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3961 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3962 (nds32_canonicalize_comparison): New function.
3964 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3965 Kito Cheng <kito.cheng@gmail.com>
3966 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3968 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3969 * config/nds32/constants.md (unspec_volatile_element): Add
3970 UNSPEC_VOLATILE_RELAX_GROUP.
3971 * config/nds32/nds32-relax-opt.c: New file.
3972 * config/nds32/nds32-predicates.c
3973 (nds32_symbol_load_store_p): New function.
3974 * config/nds32/nds32-protos.h
3975 (nds32_symbol_load_store_p): Declare function.
3976 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3977 * config/nds32/nds32.c
3978 (nds32_register_pass): New function to register pass.
3979 (nds32_register_passes): New function to register passes.
3980 * config/nds32/nds32.md (relax_group): New pattern.
3981 * config/nds32/nds32.opt (mrelax-hint): New option.
3982 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3984 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3986 * config/nds32/t-nds32: Modify files dependency.
3988 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3990 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3991 (PROFILE_HOOK): Define its implementation.
3993 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3995 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3996 type and 32-bit size.
3998 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4001 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4002 (V_128_256): New mode iterator.
4003 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4004 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4005 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4007 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4008 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4010 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4013 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4014 NaN inputs correctly.
4016 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4019 * config/rs6000/vsx.md (??r): New mode attribute.
4020 (*vsx_mov<mode>_64bit): Use it.
4021 (*vsx_mov<mode>_32bit): Likewise.
4023 2018-03-30 Martin Sebor <msebor@redhat.com>
4025 PR tree-optimization/84818
4026 * builtins.c (check_access): Use warning_n.
4028 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4031 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4033 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4036 2018-03-30 Julia Koval <julia.koval@intel.com>
4039 * x86-tune.def (movx, partial_reg_dependency): Enable for
4042 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4045 * lra-constraints.c (process_alt_operands): Move setting
4046 this_alternative_matches below.
4048 2018-03-29 Martin Liska <mliska@suse.cz>
4051 * doc/invoke.texi: Document how LTO works with debug info.
4052 Describe auto-load support of binutils. Mention 'x86-64'
4053 as valid option value of -march option.
4055 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4057 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4060 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4061 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4062 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4065 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4068 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4069 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4070 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4071 for RS6000_BTM_POWERPC64.
4072 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4073 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4074 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4079 2018-03-28 Carl Love <cel@us.ibm.com>
4082 2017-09-27 Carl Love <cel@us.ibm.com>
4084 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4085 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4086 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4089 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4091 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4092 instead of __vector bool.
4093 (_mm_max_pu8): Likewise.
4094 (_mm_min_pi16): Likewise.
4096 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4099 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4100 (DIVWEUO): Likewise.
4102 (DIVDEUO): Likewise.
4103 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4104 DIVWEUO and DIVDEUO.
4105 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4106 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4107 (div_extend): Likewise.
4108 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4110 (__builtin_divweuo): Likewise.
4111 (__builtin_divdeo): Likewise.
4112 (__builtin_divdeuo): Likewise.
4114 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4117 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4118 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4120 PR tree-optimization/82004
4121 * gimple-match-head.c (optimize_pow_to_exp): New function.
4122 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4123 Don't fold to exp if optimize_pow_to_exp is false.
4125 2018-03-28 Martin Liska <mliska@suse.cz>
4128 * calls.c (initialize_argument_information): Fix trailing space.
4129 * common.opt: Fix typo and provide better explanation for
4130 -fsanitize-coverage option.
4131 * config/i386/i386.opt: Fix typo.
4133 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4134 Martin Liska <mliska@suse.cz>
4137 * gimplify.c (asan_poison_variable): Don't do the check for
4138 gimplify_omp_ctxp here.
4139 (gimplify_decl_expr): Do it here.
4140 (gimplify_target_expr): Likewise.
4142 2018-03-28 Martin Liska <mliska@suse.cz>
4145 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4146 chkp_type_bounds_count if MPX is not enabled.
4148 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4150 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4152 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4155 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4156 function to create the function decl for complex long double
4157 multiply and divide for -mabi=ieeelongdouble.
4158 (init_float128_ieee): Call it.
4160 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4163 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4164 -fcf-protection=branch -mibt.
4165 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4167 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4170 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4172 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4175 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4176 extern array declarations.
4178 2018-03-27 Richard Biener <rguenther@suse.de>
4181 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4182 explicit single_use checks.
4184 2018-03-27 Richard Biener <rguenther@suse.de>
4186 PR tree-optimization/85082
4187 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4190 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4192 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4193 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4194 Turn on fasynchronous-unwind-tables and funwind-tables.
4196 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4199 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4200 (*bmi_blsr_<mode>_ccz): Ditto.
4202 2018-03-26 Tom de Vries <tom@codesourcery.com>
4204 PR tree-optimization/85063
4205 * omp-general.c (offloading_function_p): New function. Factor out
4207 * omp-offload.c (pass_omp_target_link::gate): ... here.
4208 * omp-general.h (offloading_function_p): Declare.
4209 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4210 with attribute omp declare target for offloading functions.
4212 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4214 PR tree-optimization/84005
4215 * tree-data-ref.h (get_base_for_alignment): Declare.
4216 * tree-data-ref.c (get_base_for_alignment_1): New function.
4217 (get_base_for_alignment): Likewise.
4218 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4219 get_base_for_alignment to find a suitable base object, instead
4220 of always using drb->base_address.
4222 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4225 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4226 known size by default.
4228 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4231 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4232 and non BLKmode operands.
4234 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4237 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4238 Clean up attributes.
4240 2018-03-23 Richard Biener <rguenther@suse.de>
4243 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4244 we are going to emit early debug for LTO.
4246 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4249 * function.c (match_asm_constraints_1): Don't optimize if input
4250 doesn't satisfy general_operand predicate for output's mode.
4253 * alias.c (write_dependence_p): Don't require for x_canonicalized
4254 non-VOIDmode if x has VOIDmode.
4257 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4258 just don't try to optimize it rather than assert it never happens.
4260 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4262 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4263 macro expansions for definition of ST_INTERNAL_<mode> and
4264 LD_INTERNAL_<mode> builtins.
4265 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4267 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4269 (altivec_expand_st_builtin): Likewise.
4270 (altivec_expand_builtin): Remove calls to deleted functions.
4271 (rs6000_address_for_altivec): Delete this function.
4272 * config/rs6000/vector.md: Remove expands for
4273 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4275 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4278 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4279 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4280 re-computing once computed.
4281 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4282 (arm_init_machine_status): Initialize
4283 machine->static_chain_stack_bytes.
4285 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4288 * doc/extend.texi: Add four new prototypes for vec_ld.
4289 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4290 definitions for more logical presentation.
4291 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4292 entries for V1TI variants of __builtin_altivec_ld builtin.
4293 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4294 handling of V1TI variant of LVX icode pattern.
4295 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4296 (rs6000_gimple_fold_builtin): Likewise.
4297 (altivec_init_builtins): Add code to define
4298 __builtin_altivec_lvx_v1ti function.
4300 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4303 * function.c (match_asm_constraints_1): Don't do the optimization
4304 if input isn't a REG, SUBREG, MEM or constant.
4306 2018-03-22 Tom de Vries <tom@codesourcery.com>
4308 PR tree-optimization/84956
4309 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4310 bb_has_abnormal_pred.
4312 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4315 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4316 DECL_INITIAL (decl) to decl at the end.
4317 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4320 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4322 * doc/extend.texi (__builtin_tgmath): Document when complex
4323 integer types are treated as _Complex _Float64.
4325 2018-03-21 Tom de Vries <tom@codesourcery.com>
4327 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4329 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4331 PR tree-optimization/84960
4332 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4333 if it is ENTRY block, move them into single succ of ENTRY in that case.
4335 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4337 PR tree-optimization/84811
4338 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4339 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4340 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4341 (known_subrange_p): Update accordingly. Cast each value involved
4342 in the size comparison, rather than casting the result of the
4345 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4347 PR tree-optimization/84982
4348 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4349 by flipping the least significant bit rather than all bits from
4350 bitpos to bitpos + bitsize - 1.
4352 2018-03-21 Nathan Sidwell <nathan@acm.org>
4354 * doc/extend.texi (Deprecated Features): Remove mention of
4355 long-deleted deprecations.
4357 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4360 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4361 * configure: Regenerate.
4363 2018-03-21 Tom de Vries <tom@codesourcery.com>
4365 PR tree-optimization/83126
4366 * tree-parloops.c (num_phis): New function.
4367 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4369 2018-03-21 Nathan Sidwell <nathan@acm.org>
4371 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4372 mention anon-struct/union members and trailing attributes.
4374 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4376 PR tree-optimization/84969
4377 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4378 builtin memset partitions if they set different rhs values.
4380 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4382 PR rtl-optimization/84989
4383 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4384 VEC_DUPLICATE with scalar result mode.
4386 2018-03-21 Martin Liska <mliska@suse.cz>
4389 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4390 not intended return statement.
4392 2018-03-21 Martin Liska <mliska@suse.cz>
4395 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4396 (chkp_find_bound_slots_1): Limit number of iterations.
4398 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4401 * Minor grammar fixes for x86 options.
4403 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4406 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4407 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4409 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4412 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4413 (altivec_lvx_<mode>_1op): Likewise.
4414 (altivec_stvx_<mode>_2op): Likewise.
4415 (altivec_stvx_<mode>_1op): Likewise.
4416 (altivec_lvx_<VM2:mode>): New define_expand.
4417 (altivec_stvx_<VM2:mode>): Likewise.
4418 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4419 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4420 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4421 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4422 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4423 (rs6000_gen_lvx): Likewise.
4424 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4425 (altivec_expand_stv_builtin): Likewise.
4426 (altivec_expand_builtin): Likewise.
4427 * config/rs6000/vector.md: Likewise.
4429 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4432 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4435 2018-03-20 Richard Biener <rguenther@suse.de>
4438 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4439 sign-conversions as zero, fall back to standard scalar_stmt
4442 2018-03-20 Martin Liska <mliska@suse.cz>
4445 * predict.c (rebuild_frequencies): Handle case when we have
4446 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4448 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4451 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4452 flag_section_anchors.
4453 * varasm.c (use_blocks_for_decl_p): Remove hack for
4454 dw2_force_const_mem.
4457 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4459 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4460 be created, use lowpart_subreg of operands[0] rather than operands[0]
4462 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4463 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4464 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4465 and n constraint instead of aarch64_shift_imm_di and Usd.
4466 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4467 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4469 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4472 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4473 to favor GPR over NEON registers.
4474 (<shift>di3_neon): Likewise.
4476 2018-03-20 Tom de Vries <tom@codesourcery.com>
4479 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4480 (nvptx_process_pars): Emit bar.sync asap and alap.
4482 2018-03-20 Tom de Vries <tom@codesourcery.com>
4485 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4486 seen_label if seen_label is already set.
4488 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4491 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4492 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4493 Use 1U instead of 1. Formatting fixes.
4496 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4497 instead of TREE_TYPE (s1) for the return value.
4499 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4501 PR tree-optimization/84946
4502 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4503 bitsize + bitsize in poly_uint64 rather than poly_int64.
4506 * dwarf2asm.c: Include fold-const.c.
4507 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4508 of decl rather than decl itself.
4510 PR rtl-optimization/84643
4511 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4513 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4516 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4517 calling assemble_variable.
4519 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4522 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4523 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4525 2018-03-19 Jim Wilson <jimw@sifive.com>
4528 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4529 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4530 (riscv_first_stack_step): Likewise.
4531 (riscv_option_override): Use STACK_BOUNDARY instead of
4533 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4535 (BIGGEST_ALIGNMENT): Set to 128.
4536 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4537 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4540 2018-03-19 Richard Biener <rguenther@suse.de>
4542 PR tree-optimization/84933
4543 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4544 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4546 2018-03-19 Richard Biener <rguenther@suse.de>
4548 PR tree-optimization/84859
4549 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4550 (cond_if_else_store_replacement): Perform sinking operation on
4551 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4552 Generalize what a BB with a single eligible store is.
4554 2018-03-19 Richard Biener <rguenther@suse.de>
4556 PR tree-optimization/84929
4557 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4558 chrec_is_positive against non-chrec arg.
4560 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4563 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4565 2018-03-18 Martin Liska <mliska@suse.cz>
4567 PR rtl-optimization/84635
4568 * regrename.c (build_def_use): Use matches_mode only when
4571 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4573 PR tree-optimization/84913
4574 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4575 vectorize chains of COND_EXPRs.
4577 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4579 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4581 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4583 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4585 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4587 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4589 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4590 Kito Cheng <kito.cheng@gmail.com>
4592 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4593 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4594 (nds32_adjust_reg_alloc_order): New function.
4595 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4597 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4599 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4600 nds32_print_operand, nds32_print_operand_address): Use
4601 HOST_WIDE_INT_PRINT_DEC instead.
4603 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4605 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4607 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4610 * config/i386/i386.c (initial_ix86_tune_features,
4611 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4613 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4614 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4615 rather than 1u << ix86_tune. Formatting fix.
4616 (ix86_option_override_internal): Change ix86_arch_mask from
4617 unsigned int to unsigned HOST_WIDE_INT, initialize to
4618 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4619 (ix86_function_specific_restore): Likewise.
4621 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4624 * postreload.c (reload_combine_recognize_pattern): Perform
4625 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4626 truncate_int_for_mode the result for the destination's mode.
4629 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4630 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4632 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4635 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4638 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4639 regno_allocno_class_array and sorted_pseudos.
4640 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4641 insns where regno is used.
4643 2018-03-16 Martin Liska <mliska@suse.cz>
4646 * multiple_target.c (create_dispatcher_calls): Redirect
4647 reference in the symbol table.
4649 2018-03-16 Martin Liska <mliska@suse.cz>
4652 * multiple_target.c (create_dispatcher_calls): Redirect also
4655 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4659 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4660 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4661 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4663 2018-03-16 Julia Koval <julia.koval@intel.com>
4665 * doc/invoke.texi (Skylake Server): Add CLWB.
4666 Cannonlake): Remove CLWB.
4668 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4670 PR tree-optimization/84841
4671 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4673 (FLOAT_ONE_CONST_TYPE): Define.
4674 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4675 (sort_by_operand_rank): Put entries with higher constant_type last
4676 rather than first to match comments.
4678 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4680 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4683 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4686 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4687 check and use address of target always.
4689 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4692 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4693 (indirect_thunk_bnd_needed): Likewise.
4694 (indirect_thunks_used): Likewise.
4695 (indirect_thunks_bnd_used): Likewise.
4696 (indirect_return_needed): New.
4697 (indirect_return_bnd_needed): Likewise.
4698 (output_indirect_thunk_function): Add a bool argument for
4700 (output_indirect_thunk_function): Don't generate alias for
4701 function return thunk.
4702 (ix86_code_end): Call output_indirect_thunk_function to generate
4703 function return thunks.
4704 (ix86_output_function_return): Set indirect_return_bnd_needed
4705 and indirect_return_needed instead of indirect_thunk_bnd_needed
4706 and indirect_thunk_needed.
4708 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4710 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4711 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4712 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4714 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4715 Paul Hua <paul.hua.gm@gmail.com>
4718 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4720 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4722 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4723 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4726 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4729 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4730 instead of GET_MODE_SIZE when comparing Units.
4732 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4735 * varasm.c (hash_section): Return an unchangeble hash value
4736 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4737 Return !aarch64_can_use_per_function_literal_pools_p ().
4739 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4742 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4743 rather than address of cmode as last argument to prepare_cmp_insn.
4745 2018-03-15 Julia Koval <julia.koval@intel.com>
4747 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4748 F_AVX512VNNI, F_AVX512BITALG): New.
4750 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4753 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4754 insn for floating-point loads and stores.
4756 2018-03-14 Carl Love <cel@us.ibm.com>
4758 * config/rs6000/rs6000-c.c: Add macro definitions for
4759 ALTIVEC_BUILTIN_VEC_PERMXOR.
4760 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4761 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4762 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4763 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4765 * config/doc/extend.texi: Add prototypes for vec_permxor.
4767 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4770 * diagnostic-show-locus.c (class layout_point): Convert m_line
4771 from int to linenum_type.
4772 (line_span::comparator): Use linenum "compare" function when
4773 comparing line numbers.
4774 (test_line_span): New function.
4775 (layout_range::contains_point): Convert param "row" from int to
4777 (layout_range::intersects_line_p): Likewise.
4778 (layout::will_show_line_p): Likewise.
4779 (layout::print_source_line): Likewise.
4780 (layout::should_print_annotation_line_p): Likewise.
4781 (layout::print_annotation_line): Likewise.
4782 (layout::print_leading_fixits): Likewise.
4783 (layout::annotation_line_showed_range_p): Likewise.
4784 (struct line_corrections): Likewise for field m_row.
4785 (line_corrections::line_corrections): Likewise for param "row".
4786 (layout::print_trailing_fixits): Likewise.
4787 (layout::get_state_at_point): Likewise.
4788 (layout::get_x_bound_for_row): Likewise.
4789 (layout::print_line): Likewise.
4790 (diagnostic_show_locus): Likewise for locals "last_line" and
4792 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4793 * input.c (selftest::test_linenum_comparisons): New function.
4794 (selftest::input_c_tests): Call it.
4795 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4796 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4797 * selftest.h (ASSERT_GT): New macro.
4798 (ASSERT_GT_AT): New macro.
4799 (ASSERT_LT): New macro.
4800 (ASSERT_LT_AT): New macro.
4802 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4804 PR rtl-optimization/84780
4805 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4807 2018-03-14 Martin Liska <mliska@suse.cz>
4809 * tree.c (record_node_allocation_statistics): Use
4810 get_stats_node_kind.
4811 (get_stats_node_kind): New function extracted from
4812 record_node_allocation_statistics.
4813 (free_node): Use get_stats_node_kind.
4815 2018-03-14 Richard Biener <rguenther@suse.de>
4817 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4818 that the value-set of ANTIC_IN doesn't grow.
4821 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4823 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4824 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4826 2018-03-14 Julia Koval <julia.koval@intel.com>
4828 * config.gcc (icelake-client, icelake-server): New.
4830 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4831 (initial_ix86_arch_features): Ditto.
4832 (PTA_SKYLAKE): Add SGX.
4833 (PTA_ICELAKE): Remove.
4834 (PTA_ICELAKE_CLIENT): New.
4835 (PTA_ICELAKE_SERVER): New.
4836 (ix86_option_override_internal): Split up icelake on icelake client and
4838 (get_builtin_code_for_version): Ditto.
4839 (fold_builtin_cpu): Ditto.
4840 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4841 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4842 * config/i386/i386.h (processor_type): Ditto.
4843 * doc/invoke.texi: Ditto.
4845 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4848 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4849 INTEGER_CST offset, add it together with bitpos / 8 and
4850 sign extend based on POINTER_SIZE.
4854 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4857 * config/i386/constraints.md (Yc): New register constraint.
4858 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4859 Use Yc constraint for alternative 2 of operand 0. Remove
4860 preferred_for_speed attribute.
4862 2018-03-14 Richard Biener <rguenther@suse.de>
4864 PR tree-optimization/84830
4865 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4866 with the old one to avoid oscillations.
4868 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4871 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4873 (assign_by_spills): Return a flag of reload assignment failure.
4874 Do not process the reload assignment failures. Do not spill other
4875 reload pseudos if they has the same reg class. Update n if
4877 (lra_assign): Add a return arg. Set up from the result of
4878 assign_by_spills call.
4879 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4880 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4881 usage_insns if it is not NULL.
4882 (spill_hard_reg_in_range): New function.
4883 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4884 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4885 function prototypes.
4886 (lra_assign): Change prototype.
4887 * lra.c (lra): Add code to deal with fails by splitting hard reg
4890 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4892 * config/riscv/riscv.opt (mrelax): New option.
4893 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4894 "norelax" when riscv_mrelax is disabled.
4895 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4897 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4900 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4901 reassociation for int modes.
4903 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4905 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4906 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4908 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4909 * config/aarch64/aarch64-sve.md
4910 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4911 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4912 (*extend<mode><Vwide>2): Rename to...
4913 (aarch64_sve_extend<mode><Vwide>2): ...this.
4914 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4915 renaming the old pattern to...
4916 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4918 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4919 define_expand, renaming the old pattern to...
4920 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4921 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4922 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4923 account when deciding which SVE instruction the optab should use.
4924 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4926 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4928 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4929 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4930 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4931 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4932 (tlsdesc_small_<mode>): Turn a define_expand and use
4933 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4934 (tlsdesc_small_advsimd_<mode>): ...this.
4935 (tlsdesc_small_sve_<mode>): New pattern.
4937 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4939 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4940 (UNSPEC_UMUL_HIGHPART): New constants.
4941 (MUL_HIGHPART): New int iteraor.
4942 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4943 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4945 (*<su>mul<mode>3_highpart): New define_insn.
4947 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4950 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4953 2018-03-13 Martin Liska <mliska@suse.cz>
4956 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4958 (sem_item_optimizer::~sem_item_optimizer): Release it.
4959 (sem_item_optimizer::merge_classes): Register variable aliases.
4960 (sem_item_optimizer::fixup_pt_set): New function.
4961 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4962 * ipa-icf.h: Declare new variables and functions.
4964 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4967 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4968 integer_pow2p@2 and test integer_pow2p in condition.
4969 (A < 0 ? C : 0): Similarly for @1.
4972 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4973 characters starting at p contain '\0' character, don't look beyond
4977 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4978 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4981 * reg-stack.c (change_stack): Change update_end var from int to
4982 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4983 also call set_block_for_insn on the newly added insns and rescan.
4986 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4987 on the last operand.
4990 * tree.c (stabilize_reference_1): Return save_expr (e) for
4991 STATEMENT_LIST even if it doesn't have side-effects.
4993 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4995 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4997 2018-03-12 Renlin Li <renlin.li@arm.com>
4999 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5000 aarch64_output_scalar_simd_mov_immediate.
5002 2018-03-12 Martin Sebor <msebor@redhat.com>
5004 PR tree-optimization/83456
5005 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5006 for perfectly overlapping calls to memcpy.
5007 (gimple_fold_builtin_memory_chk): Same.
5008 (gimple_fold_builtin_strcpy): Handle no-warning.
5009 (gimple_fold_builtin_stxcpy_chk): Same.
5010 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5012 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5014 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5015 parameter. Use it for SFmode.
5016 (rs6000_function_arg_advance_1): Adjust.
5017 (rs6000_function_arg): Adjust.
5018 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5020 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5022 PR rtl-optimization/84169
5023 PR rtl-optimization/84780
5024 * combine.c (can_combine_p): Check for a 2-insn combination whether
5025 the destination register is used between the two insns, too.
5027 2018-03-12 Richard Biener <rguenther@suse.de>
5029 PR tree-optimization/84803
5030 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5031 for refs DR analysis didn't process.
5033 2018-03-12 Richard Biener <rguenther@suse.de>
5035 PR tree-optimization/84777
5036 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5037 force-vectorize loops ignore whether we are optimizing for size.
5039 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5041 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5042 (TARGET_MD_ASM_ADJUST): Define.
5044 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5045 Kito Cheng <kito.cheng@gmail.com>
5046 Chung-Ju Wu <jasonwucj@gmail.com>
5048 * config/nds32/nds32.c (nds32_compute_stack_frame,
5049 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5050 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5051 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5052 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5053 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5054 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5055 * config/nds32/nds32.md (prologue, epilogue): Use macro
5056 NDS32_V3PUSH_AVAILABLE_P to do checking.
5058 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5061 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5062 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5063 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5064 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5065 addition of most attributes on !orig_type_die or the attribute not
5066 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5068 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5069 Chung-Ju Wu <jasonwucj@gmail.com>
5071 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5073 * config/nds32/nds32.opt (mvh): New option.
5075 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5076 Chung-Ju Wu <jasonwucj@gmail.com>
5078 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5080 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5081 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5084 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5085 Chung-Ju Wu <jasonwucj@gmail.com>
5087 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5089 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5090 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5092 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5093 Kito Cheng <kito.cheng@gmail.com>
5094 Chung-Ju Wu <jasonwucj@gmail.com>
5096 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5097 UNSPEC_FFMISM and UNSPEC_FLMISM.
5098 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5099 for ffb, ffmism and flmism.
5100 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5101 (unspec_ffmism): Ditto.
5102 (unspec_flmism): Ditto.
5103 (nds32_expand_builtin_impl): Check if string extension is available.
5104 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5105 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5107 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5110 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5113 * lra-assigns.c (assign_by_spills): Return a flag of reload
5114 assignment failure. Do not process the reload assignment
5115 failures. Do not spill other reload pseudos if they has the same
5117 (lra_assign): Add a return arg. Set up from the result of
5118 assign_by_spills call.
5119 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5120 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5121 usage_insns if it is not NULL.
5122 (spill_hard_reg_in_range): New function.
5123 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5124 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5125 function prototypes.
5126 (lra_assign): Change prototype.
5127 * lra.c (lra): Add code to deal with fails by splitting hard reg
5130 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5133 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5135 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5138 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5139 (dw_val_node): Add val_symbolic_view.
5140 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5141 (symview_upper_bound): New.
5142 (new_line_info_table): Initialize symviews_since_reset.
5143 (dwarf2out_source_line): Count symviews_since_reset and set
5144 symview_upper_bound.
5145 (dw_val_equal_p): Handle symview.
5146 (add_AT_symview): New.
5147 (print_dw_val): Handle symview.
5148 (attr_checksum, attr_checksum_ordered): Likewise.
5149 (same_dw_val_p, size_of_die): Likewise.
5150 (value_format, output_die): Likewise.
5151 (add_high_low_attributes): Use add_AT_symview for entry_view.
5152 (dwarf2out_finish): Reset symview_upper_bound, clear
5155 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5158 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5159 Add strict argument and use it.
5160 (rs6000_split_multireg_move): Update for new strict argument.
5161 (mem_operand_gpr): Disallow all non-offsettable addresses.
5162 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5164 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5167 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5168 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5169 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5172 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5173 decl, use remap_type if we want to use the type.
5175 2018-03-09 Martin Sebor <msebor@redhat.com>
5177 PR tree-optimization/84526
5178 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5180 (builtin_access::generic_overlap): Be prepared to handle non-array
5183 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5185 PR rtl-optimization/84682
5186 * lra-constraints.c (process_address_1): Check is_address flag
5187 for address constraints.
5188 (process_alt_operands): Likewise.
5189 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5190 preprocess_constraints.
5191 * recog.h (preprocess_constraints): Add oploc parameter.
5193 * recog.c (preprocess_constraints): Test address_operand for
5194 CT_ADDRESS constraints.
5196 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5199 * lra-assigns.c (assign_by_spills): Return a flag of reload
5200 assignment failure. Do not process the reload assignment
5201 failures. Do not spill other reload pseudos if they has the same
5203 (lra_assign): Add a return arg. Set up from the result of
5204 assign_by_spills call.
5205 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5206 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5207 usage_insns if it is not NULL.
5208 (spill_hard_reg_in_range): New function.
5209 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5210 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5211 function prototypes.
5212 (lra_assign): Change prototype.
5213 * lra.c (lra): Add code to deal with fails by splitting hard reg
5216 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5219 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5220 Accept complain bool parameter. Only emit errors if it is true.
5221 (arm_parse_cpu_option_name): Likewise.
5222 (arm_target_thumb_only): Adjust callers of the above.
5223 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5224 prototype to take a default true bool parameter.
5225 (arm_parse_arch_option_name): Likewise.
5227 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5228 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5232 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5233 * configure: Regenerate.
5234 * configure.ac ("linker --version-script option"): New.
5235 ("linker soname option"): New.
5237 2018-03-09 Richard Biener <rguenther@suse.de>
5239 PR tree-optimization/84775
5240 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5241 immediate uses of predicate stmts and mark them modified.
5244 PR tree-optimization/84178
5245 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5247 (version_loop_for_if_conversion): Delay update_ssa call.
5248 (tree_if_conversion): Delay update_ssa until after predicate
5251 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5254 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5255 when the function accesses prior frames.
5257 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5260 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5261 gen_llsym, otherwise call maybe_gen_llsym.
5264 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5265 has ',' character inside of it.
5267 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5270 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5271 as clobbering CC_REGNUM.
5273 2018-03-08 Richard Biener <rguenther@suse.de>
5276 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5277 (follow_copies_to_constant): Do not follow SSA names registered
5280 2018-03-08 Richard Biener <rguenther@suse.de>
5282 PR tree-optimization/84178
5283 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5285 (version_loop_for_if_conversion): Delay update_ssa call.
5286 (tree_if_conversion): Delay update_ssa until after predicate
5289 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5291 PR tree-optimization/84178
5292 * tree-if-conv.c (release_bb_predicate): Remove the
5293 the assertion that the stmts have NULL use_ops.
5294 Discard the statements, asserting that they haven't
5295 yet been added to a BB.
5297 2018-03-08 Richard Biener <rguenther@suse.de>
5299 PR tree-optimization/84746
5300 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5301 (phi_translate): Pass in destination ANTIC_OUT set.
5302 (phi_translate_1): Likewise. For a simplified result lookup
5303 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5304 (phi_translate_set): Adjust.
5305 (do_pre_regular_insertion): Likewise.
5306 (do_pre_partial_partial_insertion): Likewise.
5308 2018-03-08 Martin Liska <mliska@suse.cz>
5310 PR gcov-profile/84735
5311 * doc/gcov.texi: Document usage of profile files.
5312 * gcov-io.h: Document changes in the format.
5314 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5318 * dwarf2out.c (struct dw_line_info_table): Update comments for
5320 (FORCE_RESET_NEXT_VIEW): New.
5321 (FORCE_RESETTING_VIEW_P): New.
5322 (RESETTING_VIEW_P): Check for -1 too.
5323 (ZERO_VIEW_P): Likewise.
5324 (new_line_info_table): Force-reset next view.
5325 (dwarf2out_begin_function): Likewise.
5326 (dwarf2out_source_line): Simplify zero_view_p initialization.
5327 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5328 view directly. Omit view when omitting .loc at line 0.
5330 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5332 PR tree-optimization/84740
5333 * tree-switch-conversion.c (process_switch): Call build_constructors
5334 only if info.phi_count is non-zero.
5336 PR tree-optimization/84739
5337 * tree-tailcall.c (find_tail_calls): Check call arguments against
5338 DECL_ARGUMENTS (current_function_decl) rather than
5339 DECL_ARGUMENTS (func) when checking for tail recursion.
5341 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5343 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5344 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5345 Volker Reichelt's entry and add entries for people that perform
5346 GCC fuzzy testing and report numerous bugs.
5348 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5351 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5352 readonly data in sdata, if that is disabled.
5353 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5354 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5355 -mreadonly-in-sdata option.
5357 2018-03-07 Martin Sebor <msebor@redhat.com>
5359 PR tree-optimization/84468
5360 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5361 basic block when looking for nul assignment.
5363 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5366 * except.h (output_function_exception_table): Adjust prototype.
5367 * except.c (output_function_exception_table): Remove FNNAME parameter
5368 and add SECTION parameter. Ouput one part of the table at a time.
5369 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5370 the first part of the exception table and emit unwind directives.
5371 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5372 (i386_pe_seh_cold_init): Likewise.
5373 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5374 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5375 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5376 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5377 * config/i386/winnt.c: Include except.h.
5378 (struct seh_frame_state): Add reg_offset, after_prologue and
5379 in_cold_section fields.
5380 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5381 (i386_pe_seh_cold_init): New function.
5382 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5383 to seh->in_cold_section.
5384 (seh_emit_push): Record the offset of the push.
5385 (seh_emit_save): Record the offet of the save.
5386 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5387 Test seh->after_prologue to disregard the epilogue.
5388 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5389 (i386_pe_end_cold_function): New function.
5391 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5394 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5395 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5398 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5400 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5401 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5404 * multiple_target.c: Include tree-inline.h and intl.h.
5405 (expand_target_clones): Diagnose and fail if node->definition and
5406 !tree_versionable_function_p (node->decl).
5408 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5410 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5412 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5413 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5414 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5416 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5419 * combine.c (try_combine): Use reg_or_subregno instead of handling
5420 just paradoxical SUBREGs and REGs.
5422 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5424 * config/arc/arc.c (arc_finalize_pic): Remove function.
5425 (arc_must_save_register): We use single base PIC register, remove
5426 checks to save/restore the PIC register.
5427 (arc_expand_prologue): Likewise.
5428 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5430 (arc_verify_short): Likewise.
5431 (arc_attr_type): Likewise.
5432 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5433 (walk_stores): Likewise.
5434 (arc_address_cost): Make it static.
5435 (arc_verify_short): Likewise.
5436 (branch_dest): Likewise.
5437 (arc_attr_type): Likewise.
5438 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5439 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5440 (arc_final_prescan_insn): Remove inserting the nops due to
5441 hardware hazards. It is done in reorg step.
5442 (insn_length_variant_t): Remove.
5443 (insn_length_parameters_t): Likewise.
5444 (arc_insn_length_parameters): Likewise.
5445 (arc_get_insn_variants): Likewise.
5446 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5448 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5451 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5454 PR tree-optimization/84687
5455 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5457 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5459 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5461 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5462 Rename to ppc_speculation_barrier.
5463 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5464 __builtin_ppc_speculation_barrier.
5466 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5469 * combine.c (combine_simplify_rtx): Don't try to simplify if
5470 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5473 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5475 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5476 to 32 bytes when compiling for POWER9.
5478 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5481 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5482 regparm >= 3 with no arg reg available also for calls with
5483 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5486 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5488 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5490 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5493 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5495 2018-03-05 Richard Biener <rguenther@suse.de>
5497 PR tree-optimization/84486
5498 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5499 When inserting a __builtin_assume_aligned call set the LHS
5500 SSA name alignment info accordingly.
5502 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5504 PR tree-optimization/84114
5505 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5506 Avoid reassociation of FLOAT_MODE addition.
5508 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5510 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5511 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5512 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5513 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5514 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5515 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5516 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5518 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5519 __builtin_ia32_wbinvd): New builtins.
5520 (SPECIAL_ARGS2): New.
5521 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5522 (SPECIAL_ARGS2): New.
5523 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5524 (ix86_valid_target_attribute_inner_p): Ditto.
5525 (ix86_init_mmx_sse_builtins): Add special_args2.
5526 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5527 TARGET_WBNOINVD_P): New.
5528 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5529 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5530 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5531 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5532 * config/i386/pconfigintrin.h: New file.
5533 * config/i386/wbnoinvdintrin.h: Ditto.
5534 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5535 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5537 2018-03-05 Richard Biener <rguenther@suse.de>
5539 PR tree-optimization/84670
5540 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5542 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5543 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5544 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5545 doesn't grow if all successors (recursively) were visited at least
5548 2018-03-05 Richard Biener <rguenther@suse.de>
5550 PR tree-optimization/84650
5551 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5552 if executed in the loop pipeline.
5554 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5556 * doc/configfiles.texi (Configuration Files): Move info about
5557 conditionalizing $target-protos.h to...
5558 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5559 differs from $target-protos.h.
5561 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5562 Chung-Ju Wu <jasonwucj@gmail.com>
5564 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5565 * config/nds32/nds32-multiple.md (setmemsi): Define.
5566 * config/nds32/nds32-memory-manipulation.c
5567 (nds32_gen_dup_4_byte_to_word_value): New.
5568 (emit_setmem_word_loop): New.
5569 (emit_setmem_byte_loop): New.
5570 (nds32_expand_setmem_loop): New.
5571 (nds32_expand_setmem_loop_v3m): New.
5572 (nds32_expand_setmem_unroll): New.
5573 (nds32_expand_setmem): New.
5575 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5576 Chung-Ju Wu <jasonwucj@gmail.com>
5578 * config/nds32/nds32-memory-manipulation.c
5579 (nds32_emit_load_store): New.
5580 (nds32_emit_post_inc_load_store): New.
5581 (nds32_emit_mem_move): New.
5582 (nds32_emit_mem_move_block): New.
5583 (nds32_expand_movmemsi_loop_unknown_size): New.
5584 (nds32_expand_movmemsi_loop_known_size): New.
5585 (nds32_expand_movmemsi_loop): New.
5586 (nds32_expand_movmemsi_unroll): New.
5587 (nds32_expand_movmemqi): Rename ...
5588 (nds32_expand_movmemsi): ... to this.
5589 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5590 (movmemsi): ... to this.
5591 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5592 (nds32_expand_movmemsi): ... to this.
5594 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5595 Monk Chiang <sh.chiang04@gmail.com>
5596 Chung-Ju Wu <jasonwucj@gmail.com>
5598 * config/nds32/nds32-protos.h
5599 (nds32_expand_load_multiple): New arguments.
5600 (nds32_expand_store_multiple): Ditto.
5601 (nds32_valid_multiple_load_store): Rename ...
5602 (nds32_valid_multiple_load_store_p): ... to this.
5603 * config/nds32/nds32-memory-manipulation.c
5604 (nds32_expand_load_multiple): Refine implementation.
5605 (nds32_expand_store_multiple): Ditto.
5606 * config/nds32/nds32-multiple.md
5607 (load_multiple): Update nds32_expand_load_multiple interface.
5608 (store_multiple): Update nds32_expand_store_multiple interface.
5609 * config/nds32/nds32-predicates.c
5610 (nds32_valid_multiple_load_store): Rename ...
5611 (nds32_valid_multiple_load_store_p): ... to this and refine
5613 * config/nds32/predicates.md
5614 (nds32_load_multiple_and_update_address_operation): New predicate.
5615 (nds32_store_multiple_and_update_address_operation): New predicate.
5617 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5618 Chung-Ju Wu <jasonwucj@gmail.com>
5620 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5621 (combo): New attribute.
5622 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5624 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5626 * config/nds32/nds32.opt: Change -mcmodel= default value.
5628 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5629 Monk Chiang <sh.chiang04@gmail.com>
5630 Chung-Ju Wu <jasonwucj@gmail.com>
5632 * config/nds32/constants.md (unspec_element): New enum.
5633 * config/nds32/constraints.md (Umw): New constraint.
5634 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5635 * config/nds32/nds32-intrinsic.md: Likewise.
5636 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5637 (nds32_valid_smw_lwm_base_p): New.
5638 (nds32_output_smw_single_word): New.
5639 (nds32_output_lmw_single_word): New.
5640 (nds32_expand_unaligned_load): New.
5641 (nds32_expand_unaligned_store): New.
5642 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5643 (nds32_output_smw_single_word): Declare.
5644 (nds32_output_lmw_single_word): Declare.
5645 (nds32_expand_unaligned_load): Declare.
5646 (nds32_expand_unaligned_store): Declare.
5647 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5648 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5649 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5650 NDS32_BUILTIN_UASTORE_DW.
5651 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5654 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5655 Kito Cheng <kito.cheng@gmail.com>
5656 Chung-Ju Wu <jasonwucj@gmail.com>
5658 * config/nds32/nds32-intrinsic.c
5659 (nds32_expand_builtin_null_ftype_reg): Delete.
5660 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5661 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5662 (nds32_read_argument): New.
5663 (nds32_legitimize_target): Ditto.
5664 (nds32_legitimize_argument): Ditto.
5665 (nds32_check_constant_argument): Ditto.
5666 (nds32_expand_unop_builtin): Ditto.
5667 (nds32_expand_unopimm_builtin): Ditto.
5668 (nds32_expand_binop_builtin): Ditto.
5669 (nds32_builtin_decl_impl): Ditto.
5670 (builtin_description): Ditto.
5671 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5672 (nds32_init_builtins_impl): Ditto.
5673 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5674 (nds32_builtin_decl): New.
5675 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5676 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5678 2018-03-02 Jeff Law <law@redhat.com>
5680 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5681 (redundant_insn, fill_simple_delay_slots): Likewise.
5682 (fill_slots_from_thread): Likewise.
5683 * resource.c (mark_referenced_resources): Likewise.
5684 (mark_set_resources, find_dead_or_set_registers): Likewise.
5686 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5688 * substring-locations.h (format_warning_va): Formatting fix for
5690 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5692 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5693 * substring-locations.c: Include intl.h.
5694 (format_warning_va): Turned into small wrapper around
5695 format_warning_n_va, renamed to ...
5696 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5697 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5699 (format_warning_at_substring_n): New function.
5700 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5701 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5702 format_warning_at_substring with just a shorter name instead of
5703 const function pointer.
5704 (fmtwarn_n): New function.
5705 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5706 appropriate, get rid of all the fmtstr temporaries, move conditionals
5707 with G_() wrapped string literals directly into fmtwarn arguments,
5708 cast dir.len to (int), formatting fixes.
5710 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5712 * doc/invoke.texi: Remove "Cilk Plus" references.
5714 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5715 Richard Biener <rguenther@suse.de>
5718 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5719 for error or warning attributes if CALL_FROM_THUNK_P is set.
5722 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5725 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5726 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5729 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5730 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5732 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5733 instead of -1U in last predictors element's probability member.
5735 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5738 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5739 arguments if they are comparable.
5741 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5743 PR tree-optimization/84634
5744 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5745 masks and masked_loop_p with a single loop_masks, making sure it's
5746 null for bb vectorization.
5748 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5750 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5751 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5752 loop->force_vectorize to check whether there is no alias.
5754 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5757 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5759 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5761 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5762 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5763 instead of a loop around prev_real_insn.
5764 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5768 * config/i386/i386.c (ix86_print_operand): Use conditional
5769 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5772 2018-03-02 Richard Biener <rguenther@suse.de>
5774 PR tree-optimization/84427
5775 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5776 (bitmap_set_subtract_values): Rewrite to handle multiple
5779 (prune_clobbered_mems): Likewise.
5780 (phi_translate): Take edge instead of pred/phiblock.
5781 (phi_translate_1): Likewise.
5782 (phi_translate_set): Likewise. Insert all translated
5783 exprs for a value into the set, keeping possibly multiple
5784 expressions per value.
5785 (compute_antic_aux): Adjust for phi_translate changes.
5786 When intersecting union the expressions and prune those
5787 not in the final value set, keeping possibly multiple
5788 expressions per value. Do not use value-insertion
5789 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5790 all expressions. Add verification that the value-sets
5791 only shrink during iteration.
5792 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5793 (do_pre_regular_insertion): Likewise.
5794 (do_pre_partial_partial_insertion): Likewise.
5796 2018-03-02 Richard Biener <rguenther@suse.de>
5799 * config/darwin.c (saved_debug_info_level): New static global.
5800 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5801 (darwin_asm_lto_end): Restore debug info generation settings.
5803 2018-03-01 Martin Liska <mliska@suse.cz>
5806 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5809 2018-03-01 Richard Biener <rguenther@suse.de>
5812 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5813 type annotation with LTO when debug was disabled at compile-time.
5815 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5817 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5819 (mips_final_postscan_insn): Likewise.
5821 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5823 PR rtl-optimization/84528
5824 * alias.c (init_alias_target): Add commentary.
5825 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5826 a unique base value if the frame pointer is not eliminated
5827 to the stack pointer.
5829 2018-03-01 Tom de Vries <tom@codesourcery.com>
5831 PR rtl-optimization/83327
5832 * lra-int.h (hard_regs_spilled_into): Declare.
5833 * lra.c (hard_regs_spilled_into): Define.
5834 (init_reg_info): Init hard_regs_spilled_into.
5835 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5836 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5837 (process_bb_lives): Handle hard_regs_spilled_into.
5838 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5839 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5841 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5843 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5844 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5845 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5846 * config/rs6000/aix72.h: New file.
5848 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5850 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5851 instead of warning_at with conditional singular and plural messages
5855 * stor-layout.c (update_alignment_for_field): For
5856 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5857 && !DECL_PACKED (field), do the alignment update, just use
5858 only desired_align instead of MAX (type_align, desired_align)
5860 (place_field): Don't do known_align < desired_align handling
5861 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5862 is non-NULL, instead do it after rli->prev_field handling and
5863 only if not within a bitfield word. For DECL_PACKED (field)
5864 use type_align of BITS_PER_UNIT.
5866 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5868 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5869 superfluous parentheses and trailing spaces.
5871 2018-02-28 Richard Biener <rguenther@suse.de>
5873 PR tree-optimization/84584
5874 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5875 SCoPs with fake exit edge.
5877 2018-02-28 Martin Liska <mliska@suse.cz>
5880 * timevar.c (timer::print): Fix format to properly print 100%
5883 2018-02-28 Richard Biener <rguenther@suse.de>
5886 * genmatch.c (capture_info::walk_match): Do not mark
5887 captured expressions without operands as expr_p given
5888 they act more like predicates and should be subject to
5889 "lost tail" side-effect preserving.
5891 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5893 PR rtl-optimization/81611
5894 * auto-inc-dec.c (attempt_change): Move dead note from
5895 mem_insn if it's the next use of regno
5896 (find_address): Take address use of reg holding
5897 non-incremented value. Add parm to limit search to the named
5899 (merge_in_block): Attempt to use a mem insn that is the next
5900 use of the original regno.
5902 2018-02-27 Martin Sebor <msebor@redhat.com>
5905 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5906 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5908 2018-02-27 Martin Sebor <msebor@redhat.com>
5910 PR translation/84207
5911 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5912 n argument to unsigned HOST_WIDE_INT.
5913 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5914 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5915 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5916 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5918 2018-02-27 Richard Biener <rguenther@suse.de>
5920 PR tree-optimization/84512
5921 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5922 Do not use the estimate returned from record_stmt_cost for
5923 the scalar iteration cost but sum properly using add_stmt_cost.
5925 2018-02-27 Richard Biener <rguenther@suse.de>
5927 PR tree-optimization/84466
5928 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5929 Adjust last change to less strictly validate use operands.
5931 2018-02-27 Martin Liska <mliska@suse.cz>
5933 PR gcov-profile/84548
5934 * gcov.c (process_file): Allow partial overlap and consider it
5935 also as group functions.
5936 (output_lines): Properly calculate range of lines for a group.
5938 2018-02-27 Martin Liska <mliska@suse.cz>
5940 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5941 'ggc' suffixes. Change first column width.
5942 (timer::print): Fix formatting of the column.
5944 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5946 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5947 preserve inline entry blocks for the sake of debug inline
5948 entry point markers alone.
5949 (remove_unused_locals): Suggest in comments a better place to
5950 force the preservation of inline entry blocks that are
5951 otherwise unused, but do not preserve them.
5953 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5955 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5957 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5960 * config/i386/constraints.md (Bs): Replace
5961 ix86_indirect_branch_register with
5962 TARGET_INDIRECT_BRANCH_REGISTER.
5964 * config/i386/i386.md (indirect_jump): Likewise.
5965 (tablejump): Likewise.
5966 (*sibcall_memory): Likewise.
5967 (*sibcall_value_memory): Likewise.
5968 Peepholes of indirect call and jump via memory: Likewise.
5969 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5970 (*sibcall_value_GOT_32): Likewise.
5971 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5972 (GOT_memory_operand): Likewise.
5973 (call_insn_operand): Likewise.
5974 (sibcall_insn_operand): Likewise.
5975 (GOT32_symbol_operand): Likewise.
5976 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5978 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5980 PR rtl-optimization/83496
5981 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5982 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5983 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5984 redundant insn, if any.
5985 (relax_delay_slots): Likewise.
5986 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5988 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5990 PR tree-optimization/83965
5991 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5992 that grouped statements are part of a reduction chain. Return
5993 true if the statement is not marked as a reduction itself but
5995 (vect_recog_dot_prod_pattern): Don't check whether the statement
5996 is part of a group here.
5997 (vect_recog_sad_pattern): Likewise.
5998 (vect_recog_widen_sum_pattern): Likewise.
6000 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6003 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6005 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6008 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6010 (ix86_output_indirect_function_return): New prototype.
6011 (ix86_split_simple_return_pop_internal): Likewise.
6012 * config/i386/i386.c (indirect_return_via_cx): New.
6013 (indirect_return_via_cx_bnd): Likewise.
6014 (indirect_thunk_name): Handle return va CX_REG.
6015 (output_indirect_thunk_function): Create alias for
6016 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6017 (ix86_output_indirect_jmp): Remove the bool argument.
6018 (ix86_output_indirect_function_return): New function.
6019 (ix86_split_simple_return_pop_internal): Likewise.
6020 * config/i386/i386.md (*indirect_jump): Don't pass false
6021 to ix86_output_indirect_jmp.
6022 (*tablejump_1): Likewise.
6023 (simple_return_pop_internal): Change it to define_insn_and_split.
6024 Call ix86_split_simple_return_pop_internal to split it for
6026 (simple_return_indirect_internal): Call
6027 ix86_output_indirect_function_return instead of
6028 ix86_output_indirect_jmp.
6030 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6033 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6034 memset and value initialization afterwards.
6036 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6038 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6040 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6043 * common/config/aarch64/aarch64-common.c
6044 (aarch_option_optimization_table[]): Switch
6045 off fomit-frame-pointer
6047 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6048 Chung-Ju Wu <jasonwucj@gmail.com>
6050 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6052 (store_multiple): Ditto.
6054 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6056 * config.gcc: Add --with-cpu support for nds32 target.
6057 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6058 * config/nds32/nds32.opt: Add -mcpu= option.
6060 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6062 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6063 isel=yes): Warn for these deprecated options.
6065 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6067 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6068 ISA_2_5_MASKS_EMBEDDED.
6070 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6072 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6073 p->max as pointers rather than using iterative_hash_expr.
6075 2018-02-23 Carl Love <cel@us.ibm.com>
6077 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6078 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6080 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6081 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6082 P8V_BUILTIN_VEC_VUNSIGNED2.
6084 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6087 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6088 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6090 * lra-constraints.c (curr_insn_transform): Set up
6091 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6093 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6094 * lra-eliminations.c (spill_pseudos): Ditto.
6095 (process_insn_for_elimination): Ditto.
6096 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6097 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6098 LRA_NON_CLOBBERED_ALT.
6100 2018-02-22 Martin Sebor <msebor@redhat.com>
6102 PR tree-optimization/84480
6103 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6104 to maybe_diag_stxncpy_trunc. Call it.
6105 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6106 from gimple_fold_builtin_strcpy. Print inlining stack.
6107 (handle_builtin_stxncpy): Print inlining stack.
6108 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6110 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6113 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6114 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6115 and -fcheck-pointer-bounds are used together.
6116 (indirect_thunk_prefix): New enum.
6117 (indirect_thunk_need_prefix): New function.
6118 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6119 "_nt" instead of "_bnd" for NOTRACK prefix.
6120 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6121 (output_indirect_thunk_function): Likewise.
6123 (ix86_code_end): Update output_indirect_thunk_function calls.
6124 (ix86_output_indirect_branch_via_reg): Replace
6125 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6126 (ix86_output_indirect_branch_via_push): Likewise.
6127 (ix86_output_function_return): Likewise.
6128 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6129 incompatible with -fcf-protection=branch and
6130 -fcheck-pointer-bounds.
6132 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6135 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6136 Change gcc_assert call to output_operand_lossage.
6138 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6140 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6142 2018-02-22 DJ Delorie <dj@redhat.com>
6143 Sebastian Perta <sebastian.perta@renesas.com>
6144 Oleg Endo <olegendo@gcc.gnu.org>
6146 * config/rx/rx.c (rx_rtx_costs): New function.
6147 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6149 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6151 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6153 2018-02-22 Martin Liska <mliska@suse.cz>
6156 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6157 Add "native" as a possible value.
6159 2018-02-22 Martin Liska <mliska@suse.cz>
6162 * config/i386/i386.c (ix86_option_override_internal):
6163 Add "native" as a possible value for -march and -mtune.
6165 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6168 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6169 to all type variants.
6171 PR tree-optimization/84503
6172 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6173 width as info->bitpos + info->bitsize - start.
6174 (merged_store_group::merge_overlapping): Simplify width computation.
6175 (check_no_overlap): New function.
6176 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6177 start + width and last_order of the group, fail if check_no_overlap
6179 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6180 to group if check_no_overlap fails.
6182 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6184 * config/rs6000/altivec.md: Delete contraint arguments to
6185 define_expand, define_split, and define_peephole2, and in
6186 define_insn_and_split if always unused.
6187 * config/rs6000/darwin.md: Ditto.
6188 * config/rs6000/dfp.md: Ditto.
6189 * config/rs6000/rs6000.md: Ditto.
6190 * config/rs6000/sync.md: Ditto.
6191 * config/rs6000/vector.md: Ditto.
6192 * config/rs6000/vsx.md: Ditto.
6194 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6196 * config/rs6000/altivec.md: Write output control strings as braced
6197 blocks instead of double-quoted strings.
6198 * config/rs6000/darwin.md: Ditto.
6199 * config/rs6000/rs6000.md: Ditto.
6200 * config/rs6000/vector.md: Ditto.
6201 * config/rs6000/vsx.md: Ditto.
6203 2018-02-21 Jason Merrill <jason@redhat.com>
6205 PR c++/84314 - ICE with templates and fastcall attribute.
6206 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6208 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6210 * ipa-cp.c (determine_versionability): Fix comment typos.
6212 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6215 * ipa-cp.c (determine_versionability): Do not version functions caling
6218 2018-02-21 Martin Liska <mliska@suse.cz>
6221 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6222 Add "native" as a possible value.
6223 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6224 the macro when native cpu detection is available.
6226 2018-02-21 Martin Liska <mliska@suse.cz>
6229 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6230 Add "native" as a possible value.
6231 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6232 when native cpu detection is available.
6234 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6235 Martin Sebor <msebor@redhat.com>
6237 PR tree-optimization/84478
6238 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6240 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6241 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6242 support which is conservatively correct, for 2 only stay conservative
6243 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6244 argument to the 2 argument get_range_strlen, adjust 6 arg
6245 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6247 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6248 (gimple_fold_builtin_strlen): Pass true as last argument to
6251 2018-02-20 Martin Sebor <msebor@redhat.com>
6254 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6255 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6256 (builtin_memref::builtin_memref): Factor out parts into
6257 set_base_and_offset and call it.
6259 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6262 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6263 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6264 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6265 search at the associated MODE_INT.
6267 2018-02-20 Jeff Law <law@redhat.com>
6270 PR tree-optimization/81592
6272 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6273 for range data rather than using global data.
6274 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6275 range data rather than using global data.
6276 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6277 pass it to children as needed.
6278 (struct directive::fmtresult): Similarly.
6279 (struct directive::set_width): Similarly.
6280 (struct directive::set_precision): Similarly.
6281 (format_integer, format_directive, parse_directive): Similarly.
6282 (format_none): Accept unnamed vr_values parameter.
6283 (format_percent, format_floating, format_character): Similarly.
6284 (format_string, format_plain): Similarly.
6285 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6286 the EVRP range analyzer for range data rather than using global data.
6287 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6288 gimple-ssa-evrp-analyze.h
6289 (class sprintf_dom_walker): Add after_dom_children member function.
6290 Add evrp_range_analyzer member.
6291 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6292 range analyzer as needed.
6293 (sprintf_dom_walker::after_dom_children): New member function.
6294 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6296 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6297 (evrp_range_analyzer::pop_to_marker): Likewise.
6299 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6301 PR tree-optimization/84419
6302 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6303 with the required type if its current type is compatible but
6306 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6309 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6310 after vectorization.
6312 2018-02-20 Martin Liska <mliska@suse.cz>
6315 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6316 possible values if we don't have a hint.
6318 2018-02-20 Martin Liska <mliska@suse.cz>
6322 * final.c (shorten_branches): Build align_tab array with one
6324 * opts.c (finish_options): Add alignment option limit check.
6325 (MAX_CODE_ALIGN): Likewise.
6326 (MAX_CODE_ALIGN_VALUE): Likewise.
6327 * doc/invoke.texi: Document maximum allowed option value for
6328 all -falign-* options.
6330 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6333 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6334 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6335 * var-tracking.c (emit_note_insn_var_location): Remove all references
6336 to NOTE_INSN_CALL_ARG_LOCATION.
6337 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6338 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6339 Use copy_rtx_if_shared.
6340 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6341 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6342 (dwarf2out_var_location): Remove handling of
6343 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6345 * final.c (final_scan_insn): Remove all references to
6346 NOTE_INSN_CALL_ARG_LOCATION.
6347 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6348 before dumping final insns.
6349 * except.c (emit_note_eh_region_end): Remove all references to
6350 NOTE_INSN_CALL_ARG_LOCATION.
6351 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6352 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6353 * config/arc/arc.c (hwloop_optimize): Likewise.
6354 * config/arm/arm.c (create_fix_barrier): Likewise.
6355 * config/s390/s390.c (s390_chunkify_start): Likewise.
6356 * config/sh/sh.c (find_barrier): Likewise.
6357 * config/i386/i386.c (rest_of_insert_endbranch,
6358 ix86_seh_fixup_eh_fallthru): Likewise.
6359 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6360 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6361 * config/frv/frv.c (frv_function_prologue): Likewise.
6362 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6364 (note_outside_basic_block_p): Remove all references to
6365 NOTE_INSN_CALL_ARG_LOCATION.
6366 * gengtype.c (adjust_field_rtx_def): Likewise.
6367 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6369 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6370 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6373 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6376 PR tree-optimization/84452
6377 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6378 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6381 2018-02-19 Martin Liska <mliska@suse.cz>
6384 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6386 2018-02-19 Martin Liska <mliska@suse.cz>
6387 Richard Sandiford <richard.sandiford@linaro.org>
6389 PR tree-optimization/82491
6390 * gimple-fold.c (get_base_constructor): Make earlier bail out
6393 2018-02-19 Carl Love <cel@us.ibm.com>
6395 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6396 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6398 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6399 P8V_BUILTIN_VEC_NEG.
6401 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6403 * config/rl78/rl78.md (movdf): New define expand.
6405 2018-02-19 Martin Liska <mliska@suse.cz>
6408 * doc/invoke.texi: Fix typo.
6409 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6411 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6413 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6414 handle rs6000_single_float and rs6000_double_float specially for
6417 2018-02-16 Jeff Law <law@redhat.com>
6419 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6422 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6425 * ipa-inline.c (inline_small_functions): Fix a typo.
6427 2018-02-16 Nathan Sidwell <nathan@acm.org>
6429 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6431 2018-02-16 Carl Love <cel@us.ibm.com>
6433 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6434 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6435 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6436 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6437 expansion to P8V_BUILTIN_VEC_FLOAT2.
6439 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6441 PR rtl-optimization/70023
6442 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6443 src_regno into account.
6445 2018-02-16 Carl Love <cel@us.ibm.com>
6447 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6448 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6449 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6450 * config/rs6000/rs6000.c: Remove case statements for
6451 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6452 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6453 and P9V_BUILTIN_VEC_VINSERT4B.
6454 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6455 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6456 * config/rs6000/vsx.md:
6457 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6460 2018-02-16 Carl Love <cel@us.ibm.com>
6462 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6464 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6466 * config/rs6000/rs6000-c.c: Add the definitions for
6467 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6468 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6469 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6470 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6471 definition for insert4b and define insn *insert3b_internal.
6472 * doc/extend.texi: Add documentation for vec_extract4b.
6474 2018-02-16 Nathan Sidwell <nathan@acm.org>
6476 * doc/extend.texi (Backwards Compatibility): Mention friend
6477 injection. Note for-scope is deprecated.
6478 * doc/invoke.texi (-ffriend-injection): Deprecate.
6480 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6482 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6483 that moved to I2, also allow destinations that are a paradoxical
6484 subreg (instead of a normal reg).
6486 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6489 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6492 2018-02-16 Richard Biener <rguenther@suse.de>
6494 PR tree-optimization/84037
6495 PR tree-optimization/84016
6497 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6498 Adjust vec_construct for the fact we need additional higher latency
6499 128bit inserts for AVX256 and AVX512 vector builds.
6500 (ix86_add_stmt_cost): Scale vector construction cost for
6503 2018-02-16 Richard Biener <rguenther@suse.de>
6505 PR tree-optimization/84417
6506 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6507 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6508 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6510 2018-02-16 Martin Liska <mliska@suse.cz>
6513 * internal-fn.def (ASAN_CHECK): Set proper flags.
6514 (ASAN_MARK): Likewise.
6516 2018-02-16 Julia Koval <julia.koval@intel.com>
6518 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6519 from PTA_CANNONLAKE.
6521 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6524 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6525 Use ++iter rather than iter++ for std::list iterators.
6526 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6527 defer deleting them until all nodes in the forest are processed. Do
6528 free even leaf nodes. Change to_process into auto_vec.
6531 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6532 * vec.h (vec_default_construct): Use memset instead of placement new
6533 if BROKEN_VALUE_INITIALIZATION is defined.
6534 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6535 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6538 PR rtl-optimization/83723
6539 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6540 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6541 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6543 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6545 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6547 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6549 PR rtl-optimization/81443
6550 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6551 from inner REGs to paradoxical SUBREGs.
6553 2018-02-16 Richard Biener <rguenther@suse.de>
6555 PR tree-optimization/84399
6556 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6557 For operands we can analyze at their definition make sure we can
6558 analyze them at each use as well.
6560 2018-02-16 Richard Biener <rguenther@suse.de>
6562 PR tree-optimization/84190
6563 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6564 volatile accesses if the decl isn't volatile.
6566 2018-02-15 Jason Merrill <jason@redhat.com>
6568 PR c++/84314 - ICE with templates and fastcall attribute.
6569 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6570 TYPE_CANONICAL on an existing type.
6572 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6574 PR tree-optimization/84383
6575 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6576 dstoff nor call operand_equal_p if dstbase is NULL.
6578 PR tree-optimization/84334
6579 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6580 also a CONSTANT_CLASS_P, punt.
6582 2018-02-14 Jim Wilson <jimw@sifive.com>
6584 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6585 first SMALL_OPERAND check. New local min_second_step. Move assert
6586 to where locals are set. Add TARGET_RVC support.
6587 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6589 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6591 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6593 2018-02-14 Martin Sebor <msebor@redhat.com>
6595 PR tree-optimization/83698
6596 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6597 arrays constrain the offset range to their bounds.
6598 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6599 (builtin_access::overlap): Avoid setting the size of overlap if it's
6601 (maybe_diag_overlap): Also consider arrays when deciding what values
6602 of offsets to include in diagnostics.
6604 2018-02-14 Martin Sebor <msebor@redhat.com>
6607 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6608 that correspond to the kind of a declaration.
6610 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6613 * config/pa/pa.md: Load address of PIC label using the linkage table
6614 if the label is nonlocal.
6616 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6618 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6619 warning message if user requests -maltivec=be.
6620 * doc/invoke.texi: Document deprecation of -maltivec=be.
6622 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6625 * config/rs6000/rs6000-c.c: Update definitions for
6626 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6627 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6629 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6632 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6633 add _get_ssp intrinsics. Remove argument from
6634 __builtin_ia32_rdssp[d|q].
6635 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6636 * config/i386/i386-builtin.def: Remove argument from
6637 __builtin_ia32_rdssp[d|q].
6638 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6639 ix86_expand_special_args_builtin for _rdssp[d|q].
6640 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6641 Clear register before usage.
6642 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6643 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6645 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6647 PR tree-optimization/84357
6648 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6649 operand 1 of an ARRAY_REF too.
6651 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6654 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6655 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6657 (set_of_reg): New struct.
6658 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6659 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6660 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6662 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6663 Split into bitclr, bitset, bitinvert patterns if appropriate.
6664 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6665 use rx_fuse_in_memory_bitop.
6666 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6667 to named insn, correct maximum insn length.
6669 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6672 * machmode.def: Define a complex mode for PARTIAL_INT.
6673 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6675 * doc/rtl.texi: Document CSPImode.
6676 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6678 (msp430_hard_regno_nregs_with_padding): Likewise.
6680 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6683 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6685 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6687 PR rtl-optimization/84169
6688 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6689 we generated a parallel as new i3 and we split that to new i2 and i3
6690 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6691 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6692 those to i2, not i1. Partially rewrite this scan code.
6694 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6697 * stor-layout.c (place_field): For variable length fields, adjust
6698 offset_align afterwards not just based on the field's alignment,
6699 but also on the size.
6702 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6703 of exps and logs in the use_exp2 case.
6705 2018-02-13 Jeff Law <law@redhat.com>
6707 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6710 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6713 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6717 * common.opt (gas-loc-support, gas-locview-support): New.
6718 (ginline-points, ginternal-reset-location-views): New.
6719 * doc/invoke.texi: Document them. Use @itemx where intended.
6720 (gvariable-location-views): Adjust.
6721 * target.def (reset_location_view): New.
6722 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6723 (TARGET_RESET_LOCATION_VIEW): New.
6724 * doc/tm.texi: Rebuilt.
6725 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6726 (dwarf2out_default_as_locview_support): New.
6727 (output_asm_line_debug_info): Use option variables.
6728 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6729 (output_loc_list): Likewise.
6730 (add_high_low_attributes): Check option variables.
6731 Don't output entry view attribute in strict mode.
6732 (gen_inlined_subroutine_die): Check option variables.
6733 (dwarf2out_inline_entry): Likewise.
6734 (init_sections_and_labels): Likewise.
6735 (dwarf2out_early_finish): Likewise.
6736 (maybe_reset_location_view): New, from...
6737 (dwarf2out_var_location): ... here. Call it.
6738 * debug.h (dwarf2out_default_as_loc_support): Declare.
6739 (dwarf2out_default_as_locview_support): Declare.
6740 * hooks.c (hook_int_rtx_insn_0): New.
6741 * hooks.h (hook_int_rtx_insn_0): Declare.
6742 * toplev.c (process_options): Take -gas-loc-support and
6743 -gas-locview-support from dwarf2out. Enable
6744 -gvariable-location-views by default only with locview
6745 assembler support. Enable -ginternal-reset-location-views by
6746 default only if the target defines the corresponding hook.
6747 Enable -ginline-points by default if location views are
6748 enabled; force it disabled if statement frontiers are
6750 * tree-inline.c (expand_call_inline): Check option variables.
6751 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6753 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6755 PR tree-optimization/84321
6756 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6757 handling. Also check whether the anti-range contains any values
6758 that satisfy the mask; switch to a VR_RANGE if not.
6760 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6763 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6765 2018-02-13 Martin Jambor <mjambor@suse.cz>
6768 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6769 of call statements, also set location of a load to a temporary.
6771 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6773 * config/rl78/rl78.c (add_vector_labels): New function.
6774 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6775 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6776 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6777 which checks that no arguments are passed.
6778 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6779 * doc/extend.texi: Documentation for the new attribute.
6781 2018-02-13 Andreas Schwab <schwab@suse.de>
6783 * config/riscv/linux.h (CPP_SPEC): Define.
6785 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6788 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6789 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6790 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6791 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6792 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6793 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6794 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6795 temporarily for AES and PCLMUL builtins.
6797 PR tree-optimization/84339
6798 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6799 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6803 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6804 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6805 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6807 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6809 * omp-simd-clone.h: New file.
6810 * omp-simd-clone.c: Include omp-simd-clone.h.
6811 (expand_simd_clones): No longer static.
6812 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6813 cgraph.h and omp-simd-clone.h.
6814 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6815 (vect_recog_widen_shift_pattern): Formatting fix.
6816 (vect_pattern_recog_1): Don't check optab for calls.
6819 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6820 operands[2] into a REG before using gen_lowpart on it.
6822 2018-02-12 Jeff Law <law@redhat.com>
6825 * config/sh/sh.c (find_barrier): Consider a sibling call
6828 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6829 successfully back substituting a reg.
6831 2018-02-12 Richard Biener <rguenther@suse.de>
6833 PR tree-optimization/84037
6834 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6835 parameter, move visited init to caller.
6836 (vect_slp_analyze_operations): Separate cost from validity
6837 check, initialize visited once for all instances.
6838 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6840 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6842 (vect_model_promotion_demotion_cost): Likewise.
6843 (vectorizable_bswap): Guard cost modeling with !slp_node
6844 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6846 (vectorizable_call): Likewise.
6847 (vectorizable_conversion): Likewise.
6848 (vectorizable_assignment): Likewise.
6849 (vectorizable_shift): Likewise.
6850 (vectorizable_operation): Likewise.
6851 (vectorizable_store): Likewise.
6852 (vectorizable_load): Likewise.
6853 (vectorizable_condition): Likewise.
6854 (vectorizable_comparison): Likewise.
6856 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6859 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6860 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6861 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6863 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6866 * params.def (inline-min-speedup): Increase from 8 to 15.
6867 (max-inline-insns-auto): Decrease from 40 to 30.
6868 * ipa-split.c (consider_split): Add some buffer for function to
6869 be considered inlining candidate.
6870 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6873 2018-02-12 Richard Biener <rguenther@suse.de>
6875 PR tree-optimization/84037
6876 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6877 matched stmts if we cannot swap the non-matched ones.
6879 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6881 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6882 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6883 _mm_maskz_scalef_round_ss): New intrinsics.
6884 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6885 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6886 __builtin_ia32_scalefss_round): Remove.
6887 (__builtin_ia32_scalefsd_mask_round,
6888 __builtin_ia32_scalefss_mask_round): New intrinsics.
6889 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6890 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6891 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6892 "<round_constraint>")): Changed to ...
6893 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6894 "<round_scalar_constraint>")): ... this.
6895 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6896 %0, %1, %2<round_op3>}"): Changed to ...
6897 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6898 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6899 %2<round_scalar_mask_op3>}"): ... this.
6900 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6902 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6904 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6905 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6906 (_mm_maskz_sqrt_round_ss): New intrinsics.
6907 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6908 (__builtin_ia32_sqrtsd_mask_round)
6909 (__builtin_ia32_sqrtss_mask_round): New builtins.
6910 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6911 (__builtin_ia32_sqrtss_round): Remove.
6912 (__builtin_ia32_sqrtsd_mask_round)
6913 (__builtin_ia32_sqrtss_mask_round): New builtins.
6914 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6915 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6916 ((match_operand:VF_128 1 "vector_operand"
6917 "xBm,<round_constraint>")): Changed to ...
6918 ((match_operand:VF_128 1 "vector_operand"
6919 "xBm,<round_scalar_constraint>")): ... this.
6920 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6921 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6922 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6923 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6924 %<iptr>1<round_scalar_mask_op3>}): ... this.
6925 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6926 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6928 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6931 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6932 Cast vec_cmpeq result to correct type.
6933 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6934 Cast vec_cmpgt result to correct type.
6936 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6938 * final.c (final_scan_insn_1): Renamed from...
6939 (final_scan_insn): ... this. New wrapper, to recover
6940 seen from the outermost call in recursive ones.
6941 * config/sparc/sparc.c (output_return): Drop seen from call.
6942 (output_sibcall): Likewise.
6943 * config/visium/visium.c (output_branch): Likewise.
6945 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6947 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6950 2018-02-10 Alan Modra <amodra@gmail.com>
6953 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6954 Specify LR as an input.
6956 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6959 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6960 remove_member_access_dummy_vars): New functions.
6961 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6962 lower_omp_1, execute_lower_omp): Use them.
6964 PR rtl-optimization/84308
6965 * shrink-wrap.c (spread_components): Release todo vector.
6967 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6969 PR rtl-optimization/57193
6970 * ira-color.c (struct allocno_color_data): Add member
6971 conflict_allocno_hard_prefs.
6972 (update_conflict_allocno_hard_prefs): New.
6973 (bucket_allocno_compare_func): Add a preference based on
6974 conflict_allocno_hard_prefs.
6975 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6976 (color_allocnos): Remove a dead code. Initiate
6977 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6979 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6982 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6983 constraint from =wa to wa. Avoid a subreg on the output operand,
6984 instead use a pseudo and subreg it in a move.
6985 (p9_xxbrd_<mode>): Changed to ...
6986 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6987 (p9_xxbrd_v2df): New expander.
6988 (p9_xxbrw_<mode>): Changed to ...
6989 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6990 (p9_xxbrw_v4sf): New expander.
6992 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6994 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6996 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6999 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7000 multiply in 32-bit mode.
7001 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7002 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7005 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7007 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7008 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7009 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7010 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7012 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7015 * dwarf2out.c (is_trivial_indirect_ref): New function.
7016 (dwarf2out_late_global_decl): Do not generate a location
7017 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7018 and that are not defined in the current unit.
7020 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7022 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7023 instead of a libcall for UNORDERED.
7025 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7028 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7029 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7031 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7034 * config/s390/s390.c (s390_set_current_function): Invoke
7035 s390_indirect_branch_settings also if fndecl didn't change.
7037 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7039 * config/rs6000/rs6000.md (blockage): Set length to zero.
7041 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7043 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7045 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7048 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7049 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7053 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7054 PARALLEL incoming that failed vt_get_decl_and_offset check.
7057 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7058 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7060 (get_variable_section): For decls in named .bss* sections pass true as
7061 second argument to bss_initializer_p.
7063 2018-02-09 Marek Polacek <polacek@redhat.com>
7064 Jakub Jelinek <jakub@redhat.com>
7067 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7068 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7069 Sync some changes from cxx_fold_indirect_ref.
7071 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7073 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7075 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7076 (BLOCK_INLINE_ENTRY_LABEL): New.
7077 (dwarf2out_var_location): Disregard inline entry markers.
7078 (inline_entry_data): New struct.
7079 (inline_entry_data_hasher): New hashtable type.
7080 (inline_entry_data_hasher::hash): New.
7081 (inline_entry_data_hasher::equal): New.
7082 (inline_entry_data_table): New variable.
7083 (add_high_low_attributes): Add DW_AT_entry_pc and
7084 DW_AT_GNU_entry_view attributes if a pending entry is found
7085 in inline_entry_data_table. Add old entry_pc attribute only
7086 if debug nonbinding markers are disabled.
7087 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7088 markers are enabled.
7089 (block_within_block_p, dwarf2out_inline_entry): New.
7090 (dwarf2out_finish): Check that no entries remained in
7091 inline_entry_data_table.
7092 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7093 (final_scan_insn, notice_source_line): Likewise.
7094 (rest_of_clean_state): Skip inline entry markers.
7095 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7097 * gimple.c (gimple_build_debug_inline_entry): New.
7098 * gimple.h (enum gimple_debug_subcode): Add
7099 GIMPLE_DEBUG_INLINE_ENTRY.
7100 (gimple_build_debug_inline_entry): Declare.
7101 (gimple_debug_inline_entry_p): New.
7102 (gimple_debug_nonbind_marker_p): Adjust.
7103 * insn-notes.def (INLINE_ENTRY): New.
7104 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7105 inline entry marker notes.
7106 (print_insn): Likewise.
7107 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7108 (INSN_DEBUG_MARKER_KIND): Likewise.
7109 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7110 * tree-inline.c (expand_call_inline): Build and insert
7111 debug_inline_entry stmt.
7112 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7113 inline entry blocks early, if nonbind markers are enabled.
7114 (dump_scope_block): Dump fragment info.
7115 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7116 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7117 (gimple_build_debug_inline_entry): New.
7118 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7119 Enable/disable inline entry points too.
7120 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7121 (DEBUG_INSN): Describe inline entry markers.
7123 * common.opt (gvariable-location-views): New.
7124 (gvariable-location-views=incompat5): New.
7125 * config.in: Rebuilt.
7126 * configure: Rebuilt.
7127 * configure.ac: Test assembler for view support.
7128 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7129 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7130 * dwarf2out.c (var_loc_view): New typedef.
7131 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7132 (dwarf2out_locviews_in_attribute): New.
7133 (dwarf2out_locviews_in_loclist): New.
7134 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7135 (enum dw_line_info_opcode): Add LI_adv_address.
7136 (struct dw_line_info_table): Add view.
7137 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7138 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7139 (zero_view_p): New variable.
7140 (ZERO_VIEW_P): New macro.
7141 (output_asm_line_debug_info): New.
7142 (struct var_loc_node): Add view.
7143 (add_AT_view_list, AT_loc_list): New.
7144 (add_var_loc_to_decl): Add view param. Test it against last.
7145 (new_loc_list): Add view params. Record them.
7146 (AT_loc_list_ptr): Handle loc and view lists.
7147 (view_list_to_loc_list_val_node): New.
7148 (print_dw_val): Handle dw_val_class_view_list.
7149 (size_of_die): Likewise.
7150 (value_format): Likewise.
7151 (loc_list_has_views): New.
7152 (gen_llsym): Set vl_symbol too.
7153 (maybe_gen_llsym, skip_loc_list_entry): New.
7154 (dwarf2out_maybe_output_loclist_view_pair): New.
7155 (output_loc_list): Output view list or entries too.
7156 (output_view_list_offset): New.
7157 (output_die): Handle dw_val_class_view_list.
7158 (output_dwarf_version): New.
7159 (output_compilation_unit_header): Use it.
7160 (output_skeleton_debug_sections): Likewise.
7161 (output_rnglists, output_line_info): Likewise.
7162 (output_pubnames, output_aranges): Update version comments.
7163 (output_one_line_info_table): Output view numbers in asm comments.
7164 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7165 Call maybe_gen_llsym.
7166 (loc_list_from_tree_1): Adjust.
7167 (add_AT_location_description): Create view list attribute if
7168 needed, check it's absent otherwise.
7169 (convert_cfa_to_fb_loc_list): Adjust.
7170 (maybe_emit_file): Call output_asm_line_debug_info for test.
7171 (dwarf2out_var_location): Reset views as needed. Precompute
7172 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7173 attribute. Set view.
7174 (new_line_info_table): Reset next view.
7175 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7176 (dwarf2out_source_line): Likewise. Output view resets and labels to
7177 the assembler, or select appropriate line info opcodes.
7178 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7179 (optimize_string_length): Catch it. Adjust.
7180 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7181 dw_val_class_view_list, and remove it if no longer needed.
7182 (hash_loc_list): Hash view numbers.
7183 (loc_list_hasher::equal): Compare them.
7184 (optimize_location_lists): Check whether a view list symbol is
7185 needed, and whether the locview attribute is present, and
7186 whether they match. Remove the locview attribute if no longer
7188 (index_location_lists): Call skip_loc_list_entry for test.
7189 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7190 Use output_dwarf_version.
7191 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7192 (struct dw_val_node): Add val_view_list.
7193 * final.c (SEEN_NEXT_VIEW): New.
7194 (set_next_view_needed): New.
7195 (clear_next_view_needed): New.
7196 (maybe_output_next_view): New.
7197 (final_start_function): Rename to...
7198 (final_start_function_1): ... this. Take pointer to FIRST,
7199 add SEEN parameter. Emit param bindings in the initial view.
7200 (final_start_function): Reintroduce SEEN-less interface.
7201 (final): Rename to...
7202 (final_1): ... this. Take SEEN parameter. Output final pending
7203 next view at the end.
7204 (final): Reintroduce seen-less interface.
7205 (final_scan_insn): Output pending next view before switching
7206 sections or ending a block. Mark the next view as needed when
7207 outputting variable locations. Notify debug backend of section
7208 changes, and of location view changes.
7209 (rest_of_handle_final): Adjust.
7210 * toplev.c (process_options): Autodetect value for debug variable
7211 location views option. Warn on incompat5 without -gdwarf-5.
7212 * doc/invoke.texi (gvariable-location-views): New.
7213 (gvariable-location-views=incompat5): New.
7214 (gno-variable-location-views): New.
7216 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7218 PR tree-optimization/84136
7219 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7220 that the result of find_edge is non-NULL.
7222 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7225 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7226 storing integer register in SImode. Fix cost of 256 and 512
7227 byte aligned SSE register store.
7229 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7231 * config/i386/i386.c (ix86_multiplication_cost): Fix
7232 multiplication cost for TARGET_AVX512DQ.
7234 2018-02-08 Marek Polacek <polacek@redhat.com>
7236 PR tree-optimization/84238
7237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7240 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7242 PR tree-optimization/84265
7243 * tree-vect-stmts.c (vectorizable_store): Don't treat
7244 VMAT_CONTIGUOUS accesses as grouped.
7245 (vectorizable_load): Likewise.
7247 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7249 PR tree-optimization/81635
7250 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7251 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7252 (test_round_for_mask): New functions.
7253 (wide_int_cc_tests): Call test_round_for_mask.
7254 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7255 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7256 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7257 range returned by get_range_info.
7259 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7262 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7263 * symtab.c: Include builtins.h
7264 (symtab_node::output_to_lto_symbol_table_p): Move here
7265 from lto-streamer-out.c:output_symbol_p.
7266 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7267 (output_symbol_p): Move all logic to symtab.c
7268 (produce_symtab): Update.
7270 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7272 * config/s390/s390-opts.h (enum indirect_branch): Define.
7273 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7274 (s390_indirect_branch_via_thunk)
7275 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7276 (enum s390_indirect_branch_type): Define.
7277 * config/s390/s390.c (struct s390_frame_layout, struct
7278 machine_function): Remove.
7279 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7280 (indirect_branch_table_label_no, indirect_branch_table_name):
7282 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7283 (enum s390_indirect_branch_option): Define.
7284 (s390_return_addr_from_memory): New function.
7285 (s390_handle_string_attribute): New function.
7286 (s390_attribute_table): Add new attribute handler.
7287 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7288 (s390_indirect_branch_via_thunk): New function.
7289 (s390_indirect_branch_via_inline_thunk): New function.
7290 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7291 sibling call optimization for non z10 compiles.
7292 (s390_emit_call): Force indirect branch target to be a single
7293 register. Add r1 clobber for non-z10 compiles.
7294 (s390_emit_epilogue): Emit return jump via return_use expander.
7295 (s390_reorg): Handle JUMP_INSNs as execute targets.
7296 (s390_option_override_internal): Perform validity checks for the
7297 new command line options.
7298 (s390_indirect_branch_attrvalue): New function.
7299 (s390_indirect_branch_settings): New function.
7300 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7301 (s390_output_indirect_thunk_function): New function.
7302 (s390_code_end): Implement target hook.
7303 (s390_case_values_threshold): Implement target hook.
7304 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7306 * config/s390/s390.h (struct s390_frame_layout)
7307 (struct machine_function): Move here from s390.c.
7308 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7309 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7310 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7311 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7312 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7313 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7314 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7315 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7316 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7317 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7318 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7319 (mnemonic attribute): Add values which aren't recognized
7321 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7322 pattern for branch conversion. Fix mnemonic attribute.
7323 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7324 indirect branch via thunk if requested.
7325 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7326 ("*indirect_jump"): Disable for branch conversion using out of
7328 ("indirect_jump_via_thunk<mode>_z10")
7329 ("indirect_jump_via_thunk<mode>")
7330 ("indirect_jump_via_inlinethunk<mode>_z10")
7331 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7332 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7333 ("casesi_jump_via_inlinethunk<mode>_z10")
7334 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7335 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7336 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7337 ("*indirect2_jump"): Disable for branch conversion.
7338 ("casesi_jump"): Turn into expander and expand patterns for branch
7340 ("return_use"): New expander.
7341 ("*return"): Emit return via thunk and rename it to ...
7342 ("*return<mode>"): ... this one.
7343 * config/s390/s390.opt: Add new options and and enum for the
7346 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7348 * lra-constraints.c (match_reload): Unconditionally use
7349 gen_lowpart_SUBREG, rather than selecting between that
7350 and equivalent gen_rtx_SUBREG code.
7352 2018-02-08 Richard Biener <rguenther@suse.de>
7354 PR tree-optimization/84233
7355 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7356 changed flag instead of boguously re-using phi_inserted.
7358 2018-02-08 Martin Jambor <mjambor@suse.cz>
7360 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7361 static local variables.
7363 2018-02-08 Richard Biener <rguenther@suse.de>
7365 PR tree-optimization/84278
7366 * tree-vect-stmts.c (vectorizable_store): When looking for
7367 smaller vector types to perform grouped strided loads/stores
7368 make sure the mode is supported by the target.
7369 (vectorizable_load): Likewise.
7371 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7373 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7374 Increase LDP/STP opportunities by adding adjacent callee-saves.
7376 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7378 PR rtl-optimization/84068
7379 PR rtl-optimization/83459
7380 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7382 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7384 PR tree-optimization/84224
7385 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7386 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7389 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7392 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7393 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7394 count, remove one USE.
7396 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7398 * doc/install.texi (Configuration): Document the
7399 --with-long-double-format={ibm,ieee} PowerPC configuration
7403 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7404 Convert from define_expand to be define_insn_and_split. Rework
7405 float/double/_Float128 conversions to QI/HI/SImode to work with
7406 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7407 conversions to QI/HImode types did a store and then a load to
7408 truncate the value. For conversions to VSX registers, don't split
7409 the insn, instead emit the code directly. Use the code iterator
7410 any_fix to combine signed and unsigned conversions.
7411 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7412 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7413 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7414 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7415 (fix_<mode>di2_hw): Likewise.
7416 (fixuns_<mode>di2_hw): Likewise.
7417 (fix_<mode>si2_hw): Likewise.
7418 (fixuns_<mode>si2_hw): Likewise.
7419 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7420 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7421 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7422 fix<uns>_trunc<SFDF:mode>si2_p8.
7423 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7425 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7426 (fix<uns>_<mode>_mem): Likewise.
7427 (fctiw<u>z_<mode>_mem): Likewise.
7428 (fix<uns>_<mode>_mem): Likewise.
7429 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7430 the register allocator from doing a direct move to the GPRs to do
7431 a store, and instead use the ISA 3.0 store byte/half-word from
7432 vector register instruction. For IEEE 128-bit floating point,
7433 also optimize stores of 32-bit ints.
7434 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7436 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7438 * genextract.c (push_pathstr_operand): New function to support
7440 (walk_rtx): Call push_pathstr_operand.
7441 (print_path): Support [a-zA-Z].
7443 2018-02-07 Richard Biener <rguenther@suse.de>
7445 PR tree-optimization/84037
7446 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7447 (cse_and_gimplify_to_preheader): Declare.
7448 (vect_get_place_in_interleaving_chain): Likewise.
7449 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7451 (_loop_vec_info::~_loop_vec_info): Delete it.
7452 (cse_and_gimplify_to_preheader): New function.
7453 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7454 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7455 (vectorizable_load): Likewise. For grouped stores always base
7456 the IV on the first element.
7457 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7458 condition before gimplifying.
7460 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7462 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7463 *DIV_EXPR and *MOD_EXPR.
7465 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7468 * config/i386/i386.c (ix86_option_override_internal): Mask out
7469 the CF_SET bit when checking -fcf-protection.
7471 2018-02-07 Tom de Vries <tom@codesourcery.com>
7474 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7477 2018-02-07 Richard Biener <rguenther@suse.de>
7479 PR tree-optimization/84204
7480 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7483 PR tree-optimization/84205
7484 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7485 special-case isl_ast_op_zdiv_r.
7487 PR tree-optimization/84223
7488 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7489 Only add conditions from within the region.
7490 (gather_bbs::after_dom_children): Adjust.
7492 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7495 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7496 * config/avr/avr.md: Only post-reload split REG-REG moves if
7497 either register is GENERAL_REG_P.
7499 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7501 PR tree-optimization/84235
7502 * tree-ssa-scopedtables.c
7503 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7504 if the subtraction is performed in floating point type where NaNs are
7505 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7506 build 1. Formatting fix.
7508 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7511 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7512 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7513 and skip it regardless of bb boundaries. Use CALL_P macro,
7514 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7515 unnecessarily, formatting fix.
7517 2018-02-06 Michael Collison <michael.collison@arm.com>
7519 * config/arm/thumb2.md:
7520 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7521 (*thumb_mov_notscc): Ditto.
7523 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7526 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7527 unsigned_fix, not "s".
7529 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7531 * configure.ac (gcc_fn_eh_frame_ro): New function.
7532 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7533 correct .eh_frame permissions.
7534 * configure: Regenerate.
7536 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7538 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7541 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7543 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7544 Display warning message for -mno-speculate-indirect-jumps.
7546 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7548 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7550 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7552 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7554 PR tree-optimization/84225
7555 * tree-eh.c (find_trapping_overflow): Only call
7556 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7558 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7561 * config/i386/i386.c: Reimplement the check of possible options
7562 -mibt/-mshstk conbination. Change error messages.
7563 * doc/invoke.texi: Fix a typo: remove extra '='.
7565 2018-02-06 Marek Polacek <polacek@redhat.com>
7567 PR tree-optimization/84228
7568 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7570 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7573 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7574 emitted arch directives.
7575 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7576 __ARM_FEATURE_COPROC before changing architectures.
7578 2018-02-06 Richard Biener <rguenther@suse.de>
7580 * config/i386/i386.c (print_reg): Fix typo.
7581 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7583 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7585 * configure: Regenerate.
7587 2018-02-05 Martin Sebor <msebor@redhat.com>
7589 PR tree-optimization/83369
7590 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7593 2018-02-05 Martin Liska <mliska@suse.cz>
7595 * doc/invoke.texi: Cherry-pick upstream r323995.
7597 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7599 * ira.c (ira_init_register_move_cost): Adjust comment.
7601 2018-02-05 Martin Liska <mliska@suse.cz>
7603 PR gcov-profile/84137
7604 * doc/gcov.texi: Fix typo in documentation.
7606 2018-02-05 Martin Liska <mliska@suse.cz>
7608 PR gcov-profile/83879
7609 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7610 using dlopen functionality.
7612 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7614 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7615 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7616 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7617 _mm_maskz_range_round_ss): New intrinsics.
7618 (__builtin_ia32_rangesd128_round)
7619 (__builtin_ia32_rangess128_round): Remove.
7620 (__builtin_ia32_rangesd128_mask_round,
7621 __builtin_ia32_rangess128_mask_round): New builtins.
7622 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7623 __builtin_ia32_rangess128_round): Remove.
7624 (__builtin_ia32_rangesd128_mask_round,
7625 __builtin_ia32_rangess128_mask_round): New builtins.
7626 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7627 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7628 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7629 "<round_saeonly_constraint>")): Changed to ...
7630 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7631 "<round_saeonly_scalar_constraint>")): ... this.
7632 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7633 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7634 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7635 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7636 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7638 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7640 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7642 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7643 Remove all values except native, 8540 and 8548.
7645 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7647 * config/i386/i386.c (ix86_output_function_return): Pass
7648 INVALID_REGNUM, instead of -1, as invalid register number to
7649 indirect_thunk_name and output_indirect_thunk.
7651 2018-02-02 Julia Koval <julia.koval@intel.com>
7653 * config.gcc: Add -march=icelake.
7654 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7655 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7656 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7657 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7658 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7659 (processor_target_table): Add icelake.
7660 (ix86_option_override_internal): Handle new PTAs.
7661 (get_builtin_code_for_version): Handle icelake.
7662 (M_INTEL_COREI7_ICELAKE): New.
7663 (fold_builtin_cpu): Handle icelake.
7664 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7665 * doc/invoke.texi: Add -march=icelake.
7667 2018-02-02 Julia Koval <julia.koval@intel.com>
7669 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7670 to wide_int_bitmask.
7671 * wide-int-bitmask.h: New.
7673 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7676 * config/i386/i386.md: Replace Pmode with word_mode in
7677 builtin_setjmp_setup and builtin_longjmp to support x32.
7679 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7683 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7685 (rs6000_supported_cpu_names): New static variable.
7686 (linux_cpu_translation_table): Likewise.
7687 (elf_platform) <cpu>: Define new static variable and use it.
7688 Translate kernel AT_PLATFORM name to canonical name if needed.
7689 Error if platform name is unknown.
7691 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7694 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7696 2018-02-01 Jeff Law <law@redhat.com>
7699 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7700 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7701 the scratch if RELEASE_VIA_POP is false.
7702 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7703 If we have to save a temporary register, decrement SIZE appropriately.
7704 Pass new arguments to release_scratch_register_on_entry.
7705 (ix86_adjust_stack_and_probe): Likewise.
7706 (ix86_emit_probe_stack_range): Pass new arguments to
7707 release_scratch_register_on_entry.
7709 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7711 PR rtl-optimization/84157
7712 * combine.c (change_zero_ext): Use REG_P predicate in
7713 front of HARD_REGISTER_P predicate.
7715 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7717 * config/avr/avr.c (avr_option_override): Move disabling of
7718 -fdelete-null-pointer-checks to...
7719 * common/config/avr/avr-common.c (avr_option_optimization_table):
7722 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7724 PR tree-optimization/81635
7725 * tree-data-ref.c (split_constant_offset_1): For types that
7726 wrap on overflow, try to use range info to prove that wrapping
7729 2018-02-01 Renlin Li <renlin.li@arm.com>
7732 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7734 (aarch64_register_move_cost): Likewise.
7735 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7737 (REG_CLASS_NAMES): Likewise.
7738 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7739 TAILCALL_ADDR_REGS. Remove IP registers.
7740 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7742 2018-02-01 Richard Biener <rguenther@suse.de>
7744 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7745 for specifying RPO order and allow NULL for that.
7746 * domwalk.c (dom_walker::dom_walker): Likewise.
7747 (dom_walker::walk): Handle NULL RPO order.
7748 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7750 (rewrite_update_dom_walker): Likewise.
7751 (mark_def_dom_walker): Likewise.
7753 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7755 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7756 (aarch64_maybe_expand_sve_subreg_move): Declare.
7757 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7758 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7760 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7761 that are semantically a reverse operation.
7762 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7763 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7764 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7766 (aarch64_can_change_mode_class): For big-endian, forbid changes
7767 between two SVE modes if they have different element sizes.
7769 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7771 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7772 the TImode handling for big-endian targets.
7774 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7776 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7777 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7779 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7780 Remove BSWAP handing for big-endian targets and use the form of
7781 LD1RQ appropariate for the mode.
7783 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7785 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7786 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7789 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7792 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7793 check for operands that need to go through aarch64_sve_reload_be.
7795 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7797 PR tree-optimization/81661
7798 PR tree-optimization/84117
7799 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7800 * tree-eh.c: Include gimplify.h.
7801 (find_trapping_overflow, replace_trapping_overflow,
7802 rewrite_to_non_trapping_overflow): New functions.
7803 * tree-vect-loop.c: Include tree-eh.h.
7804 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7805 * tree-data-ref.c: Include tree-eh.h.
7806 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7808 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7810 PR rtl-optimization/84123
7811 * combine.c (change_zero_ext): Check if hard register satisfies
7812 can_change_dest_mode before calling gen_lowpart_SUBREG.
7814 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7817 * ira.c (ira_init_register_move_cost): Remove assert.
7819 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7821 PR rtl-optimization/84071
7822 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7823 * doc/tm.texi: Regenerate.
7825 2018-01-31 Richard Biener <rguenther@suse.de>
7827 PR tree-optimization/84132
7828 * tree-data-ref.c (analyze_miv_subscript): Properly
7829 check whether evolution_function_is_affine_multivariate_p
7830 before calling gcd_of_steps_may_divide_p.
7832 2018-01-31 Julia Koval <julia.koval@intel.com>
7835 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7836 * config/i386/i386.md (rdpid_rex64) New.
7837 (rdpid): Make 32bit only.
7839 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7842 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7843 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7845 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7848 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7850 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7852 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7854 PR rtl-optimization/84071
7855 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7856 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7858 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7860 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7861 (arc_attribute_table): Add 'aux' attribute.
7862 (arc_in_small_data_p): Consider aux like variables.
7863 (arc_is_aux_reg_p): New function.
7864 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7865 (arc_get_aux_arg): New function.
7866 (prepare_move_operands): Handle aux-register access.
7867 (arc_handle_aux_attribute): New function.
7868 * doc/extend.texi (ARC Variable attributes): Add subsection.
7870 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7872 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7873 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7874 (arc_attribute_table): Add 'uncached' attribute.
7875 (arc_print_operand): Print '.di' flag for uncached memory
7877 (arc_in_small_data_p): Do not consider for small data the uncached
7879 (arc_is_uncached_mem_p): New function.
7880 * config/arc/predicates.md (compact_store_memory_operand): Check
7881 for uncached memory accesses.
7882 (nonvol_nonimm_operand): Likewise.
7883 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7885 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7888 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7889 falign-loops=): Add Optimization flag.
7891 2018-01-30 Jeff Law <law@redhat.com>
7894 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7895 INT_REGISTERS_SAVED. Check it prior to calling
7896 get_scratch_register_on_entry.
7897 (ix86_adjust_stack_and_probe): Similarly.
7898 (ix86_emit_probe_stack_range): Similarly.
7899 (ix86_expand_prologue): Corresponding changes.
7901 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7904 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7905 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7907 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7910 * lra-constraints.c (curr_insn_transform): Process AND in the
7913 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7915 PR rtl-optimization/83986
7916 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7917 dependence against last_pending_memory_flush in addition to
7920 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7922 PR tree-optimization/81611
7923 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7926 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7929 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7932 2018-01-30 Richard Biener <rguenther@suse.de>
7933 Jakub Jelinek <jakub@redhat.com>
7935 PR tree-optimization/84111
7936 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7937 inner loops added during recursion, as they don't have up-to-date
7940 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7943 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7944 (can_inline_edge_by_limits_p): ... here.
7945 (can_early_inline_edge_p, check_callers,
7946 update_caller_keys, update_callee_keys, recursive_inlining,
7947 add_new_edges_to_heap, speculation_useful_p,
7948 inline_small_functions,
7949 inline_small_functions, flatten_function,
7950 inline_to_all_callers_1): Update.
7952 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7954 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7957 2018-01-30 Richard Biener <rguenther@suse.de>
7959 PR tree-optimization/83008
7960 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7961 invariant and constant vector uses in stmts when they need
7964 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7967 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7968 * configure: Regenerate.
7970 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7972 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7974 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7975 Use gen_rtx_REG rather than gen_lowpart.
7977 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7979 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7980 rather than 0 when creating partial subregs.
7982 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7984 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7987 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7990 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7991 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7992 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7993 flags. This restores the settings used before the 2017-07-24.
7994 Turning off pre increment/decrement/modify allows IVOPTS to
7995 optimize DF/SF loops where the index is an int.
7997 2018-01-29 Richard Biener <rguenther@suse.de>
7998 Kelvin Nilsen <kelvin@gcc.gnu.org>
8001 * tree-vect-stmts.c (vectorizable_call): Don't call
8002 targetm.vectorize_builtin_md_vectorized_function if callee is
8005 2018-01-22 Carl Love <cel@us.ibm.com>
8007 * doc/extend.tex: Fix typo in second arg in
8008 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8010 2018-01-29 Richard Biener <rguenther@suse.de>
8012 PR tree-optimization/84086
8013 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8014 (flush_ssaname_freelist): When SSA names were released reset
8015 the SCEV hash table.
8017 2018-01-29 Richard Biener <rguenther@suse.de>
8019 PR tree-optimization/84057
8020 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8021 removed paths when removing edges.
8023 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8025 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8026 -mfunction-return=@var{choice}.
8028 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8031 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8033 (layout::print_source_line): Likewise.
8034 (test_get_line_width_without_trailing_whitespace): Add test cases.
8036 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8039 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8042 2018-01-26 Jim Wilson <jimw@sifive.com>
8044 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8046 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8049 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8051 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8052 and CMP + SUB-immediate -> SUBS.
8054 2018-01-26 Martin Sebor <msebor@redhat.com>
8056 PR tree-optimization/83896
8057 * tree-ssa-strlen.c (get_string_len): Rename...
8058 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8059 Avoid assuming length is constant.
8060 (handle_char_store): Use HOST_WIDE_INT for string length.
8062 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8065 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8066 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8068 2018-01-26 Richard Biener <rguenther@suse.de>
8070 PR rtl-optimization/84003
8071 * dse.c (record_store): Only record redundant stores when
8072 the earlier store aliases at least all accesses the later one does.
8074 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8076 PR rtl-optimization/83985
8077 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8078 REG_CFA_RESTORE insns.
8079 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8080 REG_CFA_RESTORE insns here.
8083 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8084 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8086 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8088 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8089 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8090 (arc_init): Likewise.
8091 (arc_override_options): Likewise.
8092 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8094 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8096 * config/arc/arc.h (TARGET_DBNZ): Define.
8097 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8098 properly set the tune attribute.
8099 (dbnz): Use TARGET_DBNZ guard.
8100 * config/arc/arc.opt (mtune): Add core3 option.
8102 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8104 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8105 recognize new pic like addresses.
8106 (arc_delegitimize_address): Clean up.
8108 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8110 * config/arc/arc-arches.def: Option mrf16 valid for all
8112 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8113 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8114 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8115 * config/arc/arc-tables.opt: Regenerate.
8116 * config/arc/arc.c (arc_conditional_register_usage): Handle
8117 reduced register file case.
8118 (arc_file_start): Set must have build attributes.
8119 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8121 * config/arc/arc.opt (mrf16): Add new option.
8122 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8123 * config/arc/genmultilib.awk: Handle new mrf16 option.
8124 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8125 * config/arc/t-multilib: Regenerate.
8126 * doc/invoke.texi (ARC Options): Document mrf16 option.
8128 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8130 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8131 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8132 (arc_attribute_table): Add 'secure_call' attribute.
8133 (arc_print_operand): Print secure call operand.
8134 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8136 (arc_is_secure_call_p): New function. * config/arc/arc.md
8137 (call_i): Add support for sjli instruction.
8138 (call_value_i): Likewise.
8139 * config/arc/constraints.md (Csc): New constraint.
8141 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8142 John Eric Martin <John.Martin@emmicro-us.com>
8144 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8145 * config/arc/arc.c (_arc_jli_section): New struct.
8146 (arc_jli_section): New type.
8147 (rc_jli_sections): New static variable.
8148 (arc_handle_jli_attribute): New function.
8149 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8150 (arc_file_end): New function.
8151 (TARGET_ASM_FILE_END): Define.
8152 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8153 (arc_add_jli_section): New function.
8154 (jli_call_scan): Likewise.
8155 (arc_reorg): Call jli_call_scan.
8156 (arc_output_addsi): Remove 'S' from printing asm operand.
8157 (arc_is_jli_call_p): New function.
8158 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8160 (movhi_insn): Likewise.
8161 (movsi_insn): Likewise.
8162 (movsi_set_cc_insn): Likewise.
8163 (loadqi_update): Likewise.
8164 (load_zeroextendqisi_update): Likewise.
8165 (load_signextendqisi_update): Likewise.
8166 (loadhi_update): Likewise.
8167 (load_zeroextendhisi_update): Likewise.
8168 (load_signextendhisi_update): Likewise.
8169 (loadsi_update): Likewise.
8170 (loadsf_update): Likewise.
8171 (movsicc_insn): Likewise.
8172 (bset_insn): Likewise.
8173 (bxor_insn): Likewise.
8174 (bclr_insn): Likewise.
8175 (bmsk_insn): Likewise.
8176 (bicsi3_insn): Likewise.
8177 (cmpsi_cc_c_insn): Likewise.
8178 (movsi_ne): Likewise.
8179 (movsi_cond_exec): Likewise.
8180 (clrsbsi2): Likewise.
8191 (call_i): Remove 'S' asm letter, add jli instruction.
8192 (call_value_i): Likewise.
8193 * config/arc/arc.op (mjli-always): New option.
8194 * config/arc/constraints.md (Cji): New constraint.
8195 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8197 (subsf3_fpx): Likewise.
8198 (mulsf3_fpx): Likewise.
8199 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8201 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8203 * doc/invoke.texi (ARC): Document mjli-always option.
8205 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8207 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8208 avoid addition with 0 and use incw and decw where possible.
8210 2018-01-26 Richard Biener <rguenther@suse.de>
8212 PR tree-optimization/81082
8213 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8214 association if it requires casting to unsigned.
8215 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8216 from fold_plusminus_mult_expr to catch important cases late when
8217 range info is available.
8219 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8221 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8222 * configure.ac (hidden_linkonce): New test.
8223 * configure: Regenerate.
8224 * config.in: Regenerate.
8226 2018-01-26 Julia Koval <julia.koval@intel.com>
8228 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8229 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8230 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8231 _mm_mask_bitshuffle_epi64_mask): Fix type.
8232 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8233 USI_FTYPE_V4DI_V4DI_USI): Remove.
8234 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8235 __builtin_ia32_vpshufbitqmb256_mask,
8236 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8237 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8238 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8240 2018-01-26 Alan Modra <amodra@gmail.com>
8243 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8244 UNSPEC_VBPERMQ. Sort other unspecs.
8246 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8248 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8250 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8253 * predict.c (drop_profile): Do not push/pop cfun; update also
8255 (handle_missing_profiles): Fix logic looking for zero profiles.
8257 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8260 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8261 on functions with #pragma omp declare simd or functions with simd
8263 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8264 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8265 Remove trailing \n from warning_at calls.
8267 2018-01-25 Tom de Vries <tom@codesourcery.com>
8270 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8271 for neutered workers.
8273 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8276 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8277 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8279 2018-01-24 Jeff Law <law@redhat.com>
8282 * i386.c (get_probe_interval): Move to earlier point.
8283 (ix86_compute_frame_layout): If -fstack-clash-protection and
8284 the frame is larger than the probe interval, then use pushes
8285 to save registers rather than reg->mem moves.
8286 (ix86_expand_prologue): Remove conditional for int_registers_saved
8289 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8292 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8293 min/max for never referenced object.
8295 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8298 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8300 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8301 attributes from DECL_ATTRIBUTES (decl) without affecting
8302 DECL_ATTRIBUTES (current_function_decl).
8303 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8304 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8306 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8308 PR tree-optimization/83979
8309 * fold-const.c (fold_comparison): Use constant_boolean_node
8310 instead of boolean_{true,false}_node.
8312 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8314 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8317 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8319 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8320 Simplify the clause that sets the length attribute.
8321 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8322 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8323 clause that sets the length attribute.
8324 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8326 2018-01-24 Tom de Vries <tom@codesourcery.com>
8329 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8330 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8331 Add strict parameter.
8332 (prevent_branch_around_nothing): Insert dummy insn between branch to
8333 label and label with no ptx insn inbetween.
8334 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8336 2018-01-24 Tom de Vries <tom@codesourcery.com>
8339 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8340 for neutered threads in warp.
8341 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8343 2018-01-24 Richard Biener <rguenther@suse.de>
8345 PR tree-optimization/83176
8346 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8349 2018-01-24 Richard Biener <rguenther@suse.de>
8351 PR tree-optimization/82819
8352 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8353 code generating pluses that are no-ops in the target precision.
8355 2018-01-24 Richard Biener <rguenther@suse.de>
8358 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8360 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8362 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8363 to merge probabilities.
8364 * predict.c (probably_never_executed): Also mark as cold functions
8365 with global 0 profile and guessed local profile.
8366 * profile-count.c (profile_probability::combine_with_count): New
8368 * profile-count.h (profile_probability::operator*,
8369 profile_probability::operator*=, profile_probability::operator/,
8370 profile_probability::operator/=): Reduce precision to adjusted
8371 and set value to guessed on contradictory divisions.
8372 (profile_probability::combine_with_freq): Remove.
8373 (profile_probability::combine_wiht_count): Declare.
8374 (profile_count::force_nonzero):: Set to adjusted.
8375 (profile_count::probability_in):: Set quality to adjusted.
8376 * tree-ssa-tail-merge.c (replace_block_by): Use
8379 2018-01-23 Andrew Waterman <andrew@sifive.com>
8380 Jim Wilson <jimw@sifive.com>
8382 * config/riscv/riscv.c (riscv_stack_boundary): New.
8383 (riscv_option_override): Set riscv_stack_boundary. Handle
8384 riscv_preferred_stack_boundary_arg.
8385 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8386 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8387 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8388 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8389 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8390 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8392 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8395 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8396 of struct ix86_frame.
8397 (ix86_expand_epilogue): Likewise. Add a local variable for
8398 the reg_save_offset field in struct ix86_frame.
8400 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8402 PR tree-optimization/82604
8403 * tree-loop-distribution.c (enum partition_kind): New enum item
8404 PKIND_PARTIAL_MEMSET.
8405 (partition_builtin_p): Support above new enum item.
8406 (generate_code_for_partition): Ditto.
8407 (compute_access_range): Differentiate cases that equality can be
8408 proven at all loops, the innermost loops or no loops.
8409 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8410 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8411 (finalize_partitions, distribute_loop): Don't fuse partition of
8412 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8413 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8416 2018-01-23 Martin Liska <mliska@suse.cz>
8418 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8419 order to ignore the predictor.
8420 (PRED_POLYMORPHIC_CALL): Likewise.
8421 (PRED_RECURSIVE_CALL): Likewise.
8423 2018-01-23 Martin Liska <mliska@suse.cz>
8425 * tree-profile.c (tree_profiling): Print function header to
8426 aware reader which function we are working on.
8427 * value-prof.c (gimple_find_values_to_profile): Do not print
8428 not interesting value histograms.
8430 2018-01-23 Martin Liska <mliska@suse.cz>
8432 * profile-count.h (enum profile_quality): Add
8433 profile_uninitialized as the first value. Do not number values
8434 as they are zero based.
8435 (profile_count::verify): Update sanity check.
8436 (profile_probability::verify): Likewise.
8438 2018-01-23 Nathan Sidwell <nathan@acm.org>
8440 * doc/invoke.texi (ffor-scope): Deprecate.
8442 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8444 PR tree-optimization/83510
8445 * domwalk.c (set_all_edges_as_executable): New function.
8446 (dom_walker::dom_walker): Convert bool param
8447 "skip_unreachable_blocks" to enum reachability. Move setup of
8448 edge flags to set_all_edges_as_executable and only do it when
8449 reachability is REACHABLE_BLOCKS.
8450 * domwalk.h (enum dom_walker::reachability): New enum.
8451 (dom_walker::dom_walker): Convert bool param
8452 "skip_unreachable_blocks" to enum reachability.
8453 (set_all_edges_as_executable): New decl.
8454 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8455 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8457 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8458 but converting true to REACHABLE_BLOCKS.
8459 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8461 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8462 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8463 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8465 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8466 if check_all_array_refs will be called.
8468 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8470 * tree.c (selftest::test_location_wrappers): Add more test
8473 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8475 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8476 (selftest::test_bit_in_range): Likewise.
8478 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8481 * doc/sourcebuild.texi (vect_float): Say that the selector
8482 only describes the situation when -funsafe-math-optimizations is on.
8483 (vect_float_strict): Document.
8485 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8487 PR tree-optimization/83965
8488 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8489 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8490 instead of checking only for a reduction.
8491 (vect_recog_widen_sum_pattern): Likewise.
8493 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8495 * predict.c (probably_never_executed): Only use precise profile info.
8496 (compute_function_frequency): Skip after inlining hack since we now
8497 have quality checking.
8499 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8501 * profile-count.h (profile_probability::very_unlikely,
8502 profile_probability::unlikely, profile_probability::even): Set
8503 precision to guessed.
8505 2018-01-23 Richard Biener <rguenther@suse.de>
8507 PR tree-optimization/83963
8508 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8509 Properly terminate dominator walk when crossing the exit edge not
8510 when visiting its source block.
8512 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8515 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8516 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8518 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8520 PR tree-optimization/83957
8521 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8522 semicolon after for body surrounded by braces.
8524 PR tree-optimization/83081
8525 * profile-count.h (profile_probability::split): New method.
8526 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8527 Use profile_probability::split.
8528 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8529 when splitting a single conditional jump into 2.
8531 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8533 PR tree-optimization/69452
8534 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8539 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8540 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8541 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8543 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8545 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8546 * config/rl78/rl78.md (movdi): New define_expand.
8547 * config/rl78/rl78.c (rl78_split_movdi): New function.
8549 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8552 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8554 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8555 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8556 128-bit to produce an UNSPEC move to get the double word with the
8557 signbit and then a shift directly to do signbit.
8558 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8559 implementation with a new version that just does either a direct
8560 move or a regular move. Move memory interface to separate insns.
8561 Move insns so they are next to the expander.
8562 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8563 with signbit move. Split big and little endian case.
8564 (signbit<mode>2_dm_mem_le): Likewise.
8565 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8566 (signbit<mode>2_dm2): Likewise.
8568 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8570 * config/rl78/rl78.md (anddi3): New define_expand.
8572 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8574 * config/rl78/rl78.md (umindi3): New define_expand.
8576 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8578 * config/rl78/rl78.md (smindi3): New define_expand.
8580 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8582 * config/rl78/rl78.md (smaxdi3): New define_expand.
8584 2018-01-22 Carl Love <cel@us.ibm.com>
8586 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8587 LVX_V1TI): Add macro expansion.
8588 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8589 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8590 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8591 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8592 Change check to determine if the instruction is a byte reversing
8593 entry. Fix typo in comment.
8594 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8595 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8596 Add def_builtin calls for new builtins.
8597 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8598 Add define_insn expansion.
8600 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8602 * config/rl78/rl78.md (umaxdi3): New define_expand.
8604 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8606 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8607 for non-QImode registers.
8609 2018-01-22 Richard Biener <rguenther@suse.de>
8611 PR tree-optimization/83963
8612 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8613 including the loop exit block.
8614 (scop_detection::merge_sese): Likewise.
8615 (scop_detection::add_scop): Do it here instead.
8617 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8619 * doc/sourcebuild.texi (arm_softfloat): Document.
8621 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8624 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8625 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8626 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8628 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8629 David Edelsohn <dje.gcc@gmail.com>
8632 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8633 Change "crset eq" to "crset 2".
8634 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8635 (*call_indirect_aix<mode>_nospec): Likewise.
8636 (*call_value_indirect_aix<mode>_nospec): Likewise.
8637 (*call_indirect_elfv2<mode>_nospec): Likewise.
8638 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8639 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8640 change assembly output from . to $.
8641 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8642 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8643 (*tablejump<mode>_internal1_nospec): Likewise.
8645 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8648 * config/sh/sh_optimize_sett_clrt.cc:
8649 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8651 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8653 PR tree-optimization/83940
8654 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8655 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8656 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8657 use it to pass back the definition type.
8658 (vect_check_store_rhs): Likewise rhs_dt_out.
8659 (vect_build_gather_load_calls): Add a mask_dt argument and use
8660 it instead of a call to vect_is_simple_use.
8661 (vectorizable_store): Update calls to vect_check_load_store_mask
8662 and vect_check_store_rhs. Use the dt returned by the latter instead
8663 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8664 instead of calls to vect_is_simple_use. Pass the scalar rather
8665 than the vector operand to vect_is_simple_use when handling
8666 second and subsequent copies of an rhs value.
8667 (vectorizable_load): Update calls to vect_check_load_store_mask
8668 and vect_build_gather_load_calls. Use the cached mask_dt and
8669 gs_info.offset_dt instead of calls to vect_is_simple_use.
8671 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8674 * tree-emutls.c: Include gimplify.h.
8675 (lower_emutls_2): New function.
8676 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8677 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8678 it before further processing.
8681 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8682 UINTVAL (trueop1) instead of INTVAL (op1).
8684 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8688 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8689 INCOMING_FRAME_SP_OFFSET if not defined.
8690 (scan_trace): Add ENTRY argument. If true and
8691 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8692 emit a note to adjust the CFA offset.
8693 (create_cfi_notes): Adjust scan_trace callers.
8694 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8695 INCOMING_FRAME_SP_OFFSET in the CIE.
8696 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8697 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8699 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8700 * doc/tm.texi: Regenerated.
8702 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8704 PR rtl-optimization/83147
8705 * lra-constraints.c (remove_inheritance_pseudos): Use
8706 lra_substitute_pseudo_within_insn.
8708 2018-01-19 Tom de Vries <tom@codesourcery.com>
8709 Cesar Philippidis <cesar@codesourcery.com>
8712 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8714 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8717 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8718 spaces for function labels.
8720 2018-01-19 Martin Liska <mliska@suse.cz>
8722 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8723 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8724 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8725 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8726 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8727 (PRED_CONST_RETURN): Change from 69 to 65.
8728 (PRED_NULL_RETURN): Change from 91 to 71.
8729 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8730 (PRED_LOOP_GUARD): Change from 66 to 73.
8732 2018-01-19 Martin Liska <mliska@suse.cz>
8734 * predict.c (predict_insn_def): Add new assert.
8735 (struct branch_predictor): Change type to signed integer.
8736 (test_prediction_value_range): Amend test to cover
8738 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8739 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8740 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8741 (PRED_LOOP_IV_COMPARE): Likewise.
8742 * predict.h (PROB_UNINITIALIZED): Define new constant.
8744 2018-01-19 Martin Liska <mliska@suse.cz>
8746 * predict.c (dump_prediction): Add new format for
8747 analyze_brprob.py script which is enabled with -details
8749 * profile-count.h (precise_p): New function.
8751 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8753 PR tree-optimization/83922
8754 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8755 there are no statements that need masking.
8756 (vect_active_double_reduction_p): New function.
8757 (vect_analyze_loop_operations): Use it when handling phis that
8758 are not in the loop header.
8760 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8762 PR tree-optimization/83914
8763 * tree-vect-loop.c (vectorizable_induction): Don't convert
8764 init_expr or apply the peeling adjustment for inductions
8765 that are nested within the vectorized loop.
8767 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8769 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8772 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8776 * function.h (gimplify_parameters): Add gimple_seq * argument.
8777 * function.c: Include gimple.h and options.h.
8778 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8779 for the added local temporaries if needed.
8780 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8781 if there are any parameter cleanups, wrap whole body into a
8782 try/finally with the cleanups.
8784 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8787 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8788 Use GET_MODE_CLASS for scalar floating point.
8790 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8794 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8795 Fix call of call_cgraph_insertion_hooks.
8797 2018-01-18 Martin Sebor <msebor@redhat.com>
8799 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8801 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8804 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8807 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8810 * common.opt: (-ffile-prefix-map): New option.
8811 * opts.c (common_handle_option): Defer it.
8812 * opts-global.c (handle_common_deferred_options): Handle it.
8813 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8814 * file-prefix-map.h: New file.
8815 (remap_debug_filename, add_debug_prefix_map): ...here.
8816 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8817 * final.c (debug_prefix_map, add_debug_prefix_map
8818 remap_debug_filename): Move to...
8819 * file-prefix-map.c: New file.
8820 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8821 generalize, get rid of alloca(), use strrchr() instead of strchr().
8822 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8823 Implement in terms of add_prefix_map().
8824 (remap_macro_filename, remap_debug_filename): Implement in term of
8826 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8827 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8828 * dbxout.c: Include file-prefix-map.h.
8829 * varasm.c: Likewise.
8830 * vmsdbgout.c: Likewise.
8831 * xcoffout.c: Likewise.
8832 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8833 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8834 * doc/invoke.texi (-ffile-prefix-map): Document.
8835 (-fdebug-prefix-map): Update description.
8837 2018-01-18 Martin Liska <mliska@suse.cz>
8839 * config/i386/i386.c (indirect_thunk_name): Document that also
8841 (output_indirect_thunk): Document why both instructions
8842 (pause and lfence) are generated.
8844 2018-01-18 Richard Biener <rguenther@suse.de>
8846 PR tree-optimization/83887
8847 * graphite-scop-detection.c
8848 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8849 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8850 (scop_detection::merge_sese): Re-implement with a flood-fill
8851 algorithm that properly finds a SESE region if it exists.
8853 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8856 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8857 pointer_diff optimizations use view_convert instead of convert.
8859 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8861 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8862 Generate different code for -mno-speculate-indirect-jumps.
8863 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8864 (*call_indirect_aix<mode>): Disable for
8865 -mno-speculate-indirect-jumps.
8866 (*call_indirect_aix<mode>_nospec): New define_insn.
8867 (*call_value_indirect_aix<mode>): Disable for
8868 -mno-speculate-indirect-jumps.
8869 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8870 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8871 -mno-speculate-indirect-jumps.
8872 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8874 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8876 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8877 long double type, set the flags for noting the default long double
8878 type, even if we don't pass or return a long double type.
8880 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8883 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8886 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8888 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8889 support for merge[hl].
8890 (fold_mergehl_helper): New helper function.
8891 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8892 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8893 (altivec_vmrglw_direct): Add xxmrglw insn.
8895 2018-01-17 Andrew Waterman <andrew@sifive.com>
8897 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8898 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8900 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8903 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8904 call the lto_location_cache before reading the
8905 DECL_SOURCE_LOCATION of the types.
8907 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8908 Richard Sandiford <richard.sandiford@linaro.org>
8910 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8911 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8912 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8913 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8914 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8916 * config/aarch64/constraints.md (aarch64_movti_operand):
8918 * config/aarch64/predicates.md (Uti): Add new constraint.
8920 2018-01-17 Carl Love <cel@us.ibm.com>
8922 * config/rs6000/vsx.md (define_expand xl_len_r,
8923 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8924 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8926 (define_expand, define_insn): Move the shift left from the
8927 define_insn to the define_expand for lxvl and stxvl instructions.
8928 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8929 and XL_LEN_R definitions to PURE.
8931 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8933 * config/i386/i386.c (indirect_thunk_name): Declare regno
8934 as unsigned int. Compare regno with INVALID_REGNUM.
8935 (output_indirect_thunk): Ditto.
8936 (output_indirect_thunk_function): Ditto.
8937 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8938 in the call to output_indirect_thunk_function.
8940 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8943 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8944 rather than the size of inner_type to determine the stack slot size
8945 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8947 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8950 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8953 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8955 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8956 endian Linux systems to optionally enable multilibs for selecting
8957 the long double type if the user configured an explicit type.
8958 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8959 have no long double multilibs if not defined.
8960 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8961 warn if the user used -mabi={ieee,ibm}longdouble and we built
8962 multilibs for long double.
8963 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8964 appropriate multilib option.
8965 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8967 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8968 for building long double multilibs.
8969 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8971 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8973 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8976 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8978 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8981 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8984 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8987 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8989 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8990 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8992 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8994 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8995 different rtl trees depending on TARGET_64BIT.
8996 (rs6000_gen_lvx): Likewise.
8998 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9000 * config/visium/visium.md (nop): Tweak comment.
9001 (hazard_nop): Likewise.
9003 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9005 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9006 -mspeculate-indirect-jumps.
9007 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9008 for -mno-speculate-indirect-jumps.
9009 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9010 (*call_value_indirect_elfv2<mode>): Disable for
9011 -mno-speculate-indirect-jumps.
9012 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9013 (indirect_jump): Emit different RTL for
9014 -mno-speculate-indirect-jumps.
9015 (*indirect_jump<mode>): Disable for
9016 -mno-speculate-indirect-jumps.
9017 (*indirect_jump<mode>_nospec): New define_insn.
9018 (tablejump): Emit different RTL for
9019 -mno-speculate-indirect-jumps.
9020 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9021 (tablejumpsi_nospec): New define_expand.
9022 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9023 (tablejumpdi_nospec): New define_expand.
9024 (*tablejump<mode>_internal1): Disable for
9025 -mno-speculate-indirect-jumps.
9026 (*tablejump<mode>_internal1_nospec): New define_insn.
9027 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9030 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9032 * caller-save.c (insert_save): Drop unnecessary parameter. All
9035 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9036 Richard Biener <rguenth@suse.de>
9039 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9040 return early, inline manually is_gimple_sizepos. Make sure if we
9041 call gimplify_expr we don't end up with a gimple constant.
9042 * tree.c (variably_modified_type_p): Don't return true for
9043 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9044 * gimplify.h (is_gimple_sizepos): Remove.
9046 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9048 PR tree-optimization/83857
9049 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9050 vectorizable_live_operation for pure SLP statements.
9051 (vectorizable_live_operation): Handle PHIs.
9053 2018-01-16 Richard Biener <rguenther@suse.de>
9055 PR tree-optimization/83867
9056 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9057 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9059 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9062 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9063 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9064 If off is not INTEGER_CST, issue a may not be aligned warning
9065 rather than isn't aligned. Use isn%'t rather than isn't.
9066 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9068 <case MULT_EXPR>: Improve the case when bottom and one of the
9069 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9070 operand, in that case check if the other operand is multiple of
9071 bottom divided by the INTEGER_CST operand.
9073 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9076 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9077 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9078 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9079 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9080 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9081 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9082 (pa_function_arg_size): New function.
9084 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9086 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9087 in a separate statement.
9089 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9091 PR tree-optimization/83847
9092 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9093 group gathers and scatters.
9095 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9097 PR rtl-optimization/86620
9098 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9100 PR rtl-optimization/83213
9101 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9102 to last if both are JUMP_INSNs.
9104 PR tree-optimization/83843
9105 * gimple-ssa-store-merging.c
9106 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9107 store_immediate_info for bswap/nop orig_stores.
9109 2018-01-15 Andrew Waterman <andrew@sifive.com>
9111 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9113 <UDIV>: Increase cost if !TARGET_DIV.
9115 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9117 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9118 (define_attr "cr_logical_3op"): New.
9119 (cceq_ior_compare): Adjust.
9120 (cceq_ior_compare_complement): Adjust.
9121 (*cceq_rev_compare): Adjust.
9122 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9123 (is_cracked_insn): Adjust.
9124 (insn_must_be_first_in_group): Adjust.
9125 * config/rs6000/40x.md: Adjust.
9126 * config/rs6000/440.md: Adjust.
9127 * config/rs6000/476.md: Adjust.
9128 * config/rs6000/601.md: Adjust.
9129 * config/rs6000/603.md: Adjust.
9130 * config/rs6000/6xx.md: Adjust.
9131 * config/rs6000/7450.md: Adjust.
9132 * config/rs6000/7xx.md: Adjust.
9133 * config/rs6000/8540.md: Adjust.
9134 * config/rs6000/cell.md: Adjust.
9135 * config/rs6000/e300c2c3.md: Adjust.
9136 * config/rs6000/e500mc.md: Adjust.
9137 * config/rs6000/e500mc64.md: Adjust.
9138 * config/rs6000/e5500.md: Adjust.
9139 * config/rs6000/e6500.md: Adjust.
9140 * config/rs6000/mpc.md: Adjust.
9141 * config/rs6000/power4.md: Adjust.
9142 * config/rs6000/power5.md: Adjust.
9143 * config/rs6000/power6.md: Adjust.
9144 * config/rs6000/power7.md: Adjust.
9145 * config/rs6000/power8.md: Adjust.
9146 * config/rs6000/power9.md: Adjust.
9147 * config/rs6000/rs64.md: Adjust.
9148 * config/rs6000/titan.md: Adjust.
9150 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9152 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9153 ix86_indirect_branch_register logic.
9155 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9157 * config/i386/constraints.md (Bs): Update
9158 ix86_indirect_branch_register check. Don't check
9159 ix86_indirect_branch_register with GOT_memory_operand.
9161 * config/i386/predicates.md (GOT_memory_operand): Don't check
9162 ix86_indirect_branch_register here.
9163 (GOT32_symbol_operand): Likewise.
9165 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9167 * config/i386/predicates.md (constant_call_address_operand):
9168 Rewrite ix86_indirect_branch_register logic.
9169 (sibcall_insn_operand): Likewise.
9171 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9173 * config/i386/constraints.md (Bs): Replace
9174 ix86_indirect_branch_thunk_register with
9175 ix86_indirect_branch_register.
9177 * config/i386/i386.md (indirect_jump): Likewise.
9178 (tablejump): Likewise.
9179 (*sibcall_memory): Likewise.
9180 (*sibcall_value_memory): Likewise.
9181 Peepholes of indirect call and jump via memory: Likewise.
9182 * config/i386/i386.opt: Likewise.
9183 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9184 (GOT_memory_operand): Likewise.
9185 (call_insn_operand): Likewise.
9186 (sibcall_insn_operand): Likewise.
9187 (GOT32_symbol_operand): Likewise.
9189 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9192 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9193 type rather than type addr's type points to.
9194 (expand_omp_atomic_mutex): Likewise.
9195 (expand_omp_atomic): Likewise.
9197 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9200 * config/i386/i386.c (output_indirect_thunk_function): Use
9201 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9202 for __x86_return_thunk.
9204 2018-01-15 Richard Biener <rguenther@suse.de>
9207 * expmed.c (extract_bit_field_1): Fix typo.
9209 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9212 * config/arm/iterators.md (VF): New mode iterator.
9213 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9214 Remove integer-related logic from pattern.
9215 (neon_vabd<mode>_3): Likewise.
9217 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9220 * common.opt (fstrict-overflow): No longer an alias.
9221 (fwrapv-pointer): New option.
9222 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9223 also for pointer types based on flag_wrapv_pointer.
9224 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9225 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9226 opts->x_flag_wrapv got set.
9227 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9228 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9229 POINTER_TYPE_OVERFLOW_UNDEFINED.
9230 * match.pd: Likewise in address comparison pattern.
9231 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9233 2018-01-15 Richard Biener <rguenther@suse.de>
9236 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9237 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9238 Reset type names to their identifier if their TYPE_DECL doesn't
9239 have linkage (and thus is used for ODR and devirt).
9240 (save_debug_info_for_decl): Remove.
9241 (save_debug_info_for_type): Likewise.
9242 (add_tree_to_fld_list): Adjust.
9243 * tree-pretty-print.c (dump_generic_node): Make dumping of
9244 type names more robust.
9246 2018-01-15 Richard Biener <rguenther@suse.de>
9248 * BASE-VER: Bump to 8.0.1.
9250 2018-01-14 Martin Sebor <msebor@redhat.com>
9253 * builtins.c (check_access): Avoid warning when the no-warning bit
9256 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9258 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9259 * ira-color (allocno_hard_regs_compare): Likewise.
9261 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9264 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9265 Use .pushsection/.popsection.
9267 2018-01-14 Martin Sebor <msebor@redhat.com>
9270 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9272 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9274 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9275 entry from extra_headers.
9276 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9277 extra_headers, make the list bitwise identical to the i?86-*-* one.
9279 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9281 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9282 -mcmodel=large with -mindirect-branch=thunk,
9283 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9284 -mfunction-return=thunk-extern.
9285 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9286 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9287 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9289 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9291 * config/i386/i386.c (print_reg): Print the name of the full
9292 integer register without '%'.
9293 (ix86_print_operand): Handle 'V'.
9294 * doc/extend.texi: Document 'V' modifier.
9296 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9298 * config/i386/constraints.md (Bs): Disallow memory operand for
9299 -mindirect-branch-register.
9301 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9302 (GOT_memory_operand): Likewise.
9303 (call_insn_operand): Likewise.
9304 (sibcall_insn_operand): Likewise.
9305 (GOT32_symbol_operand): Likewise.
9306 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9307 for -mindirect-branch-register.
9308 (tablejump): Likewise.
9309 (*sibcall_memory): Likewise.
9310 (*sibcall_value_memory): Likewise.
9311 Disallow peepholes of indirect call and jump via memory for
9312 -mindirect-branch-register.
9313 (*call_pop): Replace m with Bw.
9314 (*call_value_pop): Likewise.
9315 (*sibcall_pop_memory): Replace m with Bs.
9316 * config/i386/i386.opt (mindirect-branch-register): New option.
9317 * doc/invoke.texi: Document -mindirect-branch-register option.
9319 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9321 * config/i386/i386-protos.h (ix86_output_function_return): New.
9322 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9323 set function_return_type.
9324 (indirect_thunk_name): Add ret_p to indicate thunk for function
9326 (output_indirect_thunk_function): Pass false to
9327 indirect_thunk_name.
9328 (ix86_output_indirect_branch_via_reg): Likewise.
9329 (ix86_output_indirect_branch_via_push): Likewise.
9330 (output_indirect_thunk_function): Create alias for function
9331 return thunk if regno < 0.
9332 (ix86_output_function_return): New function.
9333 (ix86_handle_fndecl_attribute): Handle function_return.
9334 (ix86_attribute_table): Add function_return.
9335 * config/i386/i386.h (machine_function): Add
9336 function_return_type.
9337 * config/i386/i386.md (simple_return_internal): Use
9338 ix86_output_function_return.
9339 (simple_return_internal_long): Likewise.
9340 * config/i386/i386.opt (mfunction-return=): New option.
9341 (indirect_branch): Mention -mfunction-return=.
9342 * doc/extend.texi: Document function_return function attribute.
9343 * doc/invoke.texi: Document -mfunction-return= option.
9345 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9347 * config/i386/i386-opts.h (indirect_branch): New.
9348 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9349 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9350 with local indirect jump when converting indirect call and jump.
9351 (ix86_set_indirect_branch_type): New.
9352 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9353 (indirectlabelno): New.
9354 (indirect_thunk_needed): Likewise.
9355 (indirect_thunk_bnd_needed): Likewise.
9356 (indirect_thunks_used): Likewise.
9357 (indirect_thunks_bnd_used): Likewise.
9358 (INDIRECT_LABEL): Likewise.
9359 (indirect_thunk_name): Likewise.
9360 (output_indirect_thunk): Likewise.
9361 (output_indirect_thunk_function): Likewise.
9362 (ix86_output_indirect_branch_via_reg): Likewise.
9363 (ix86_output_indirect_branch_via_push): Likewise.
9364 (ix86_output_indirect_branch): Likewise.
9365 (ix86_output_indirect_jmp): Likewise.
9366 (ix86_code_end): Call output_indirect_thunk_function if needed.
9367 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9369 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9370 (ix86_attribute_table): Add indirect_branch.
9371 * config/i386/i386.h (machine_function): Add indirect_branch_type
9372 and has_local_indirect_jump.
9373 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9375 (tablejump): Likewise.
9376 (*indirect_jump): Use ix86_output_indirect_jmp.
9377 (*tablejump_1): Likewise.
9378 (simple_return_indirect_internal): Likewise.
9379 * config/i386/i386.opt (mindirect-branch=): New option.
9380 (indirect_branch): New.
9383 (thunk-inline): Likewise.
9384 (thunk-extern): Likewise.
9385 * doc/extend.texi: Document indirect_branch function attribute.
9386 * doc/invoke.texi: Document -mindirect-branch= option.
9388 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9391 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9393 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9395 * ipa-inline.c (want_inline_small_function_p): Return false if
9396 inlining has already failed with CIF_FINAL_ERROR.
9397 (update_caller_keys): Call want_inline_small_function_p before
9399 (update_callee_keys): Likewise.
9401 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9403 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9405 (rs6000_quadword_masked_address_p): Likewise.
9406 (quad_aligned_load_p): Likewise.
9407 (quad_aligned_store_p): Likewise.
9408 (const_load_sequence_p): Add comment to describe the outer-most loop.
9409 (mimic_memory_attributes_and_flags): New function.
9410 (rs6000_gen_stvx): Likewise.
9411 (replace_swapped_aligned_store): Likewise.
9412 (rs6000_gen_lvx): Likewise.
9413 (replace_swapped_aligned_load): Likewise.
9414 (replace_swapped_load_constant): Capitalize argument name in
9415 comment describing this function.
9416 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9417 and stores that access quad-word aligned addresses and replace
9418 with stvx or lvx instructions when appropriate.
9419 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9420 New function prototype.
9421 (rs6000_quadword_masked_address_p): Likewise.
9422 (rs6000_gen_lvx): Likewise.
9423 (rs6000_gen_stvx): Likewise.
9424 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9425 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9426 when memory address is aligned.
9427 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9428 this split to select lvx instruction when memory address is aligned.
9429 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9430 instruction when memory address is aligned.
9431 (*vsx_le_perm_load_v16qi): Likewise.
9432 (four unnamed splitters): Modify to select the stvx instruction
9433 when memory is aligned.
9435 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9437 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9438 which appears in the queue multiple times.
9440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9441 Alan Hayward <alan.hayward@arm.com>
9442 David Sherwood <david.sherwood@arm.com>
9444 * tree-vectorizer.h (vec_lower_bound): New structure.
9445 (_loop_vec_info): Add check_nonzero and lower_bounds.
9446 (LOOP_VINFO_CHECK_NONZERO): New macro.
9447 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9448 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9449 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9450 fields. Make seg_len the distance travelled, not including the
9452 (dr_direction_indicator): Declare.
9453 (dr_zero_step_indicator): Likewise.
9454 (dr_known_forward_stride_p): Likewise.
9455 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9457 (runtime_alias_check_p): Allow runtime alias checks with
9459 (operator ==): Compare access_size and align.
9460 (prune_runtime_alias_test_list): Rework for new distinction between
9461 the access_size and seg_len.
9462 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9464 (get_segment_min_max): New function.
9465 (create_intersect_range_checks): Use it.
9466 (dr_step_indicator): New function.
9467 (dr_direction_indicator): Likewise.
9468 (dr_zero_step_indicator): Likewise.
9469 (dr_known_forward_stride_p): Likewise.
9470 * tree-loop-distribution.c (data_ref_segment_size): Return
9471 DR_STEP * (niters - 1).
9472 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9474 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9475 (vect_preserves_scalar_order_p): New function, split out from...
9476 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9477 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9478 (vect_vfa_access_size): New function.
9479 (vect_vfa_align): Likewise.
9480 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9481 (dump_lower_bound): New function.
9482 (vect_check_lower_bound): Likewise.
9483 (vect_small_gap_p): Likewise.
9484 (vectorizable_with_step_bound_p): Likewise.
9485 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9486 depencies if the vectorization factor is 1. Convert the checks
9487 for nonzero steps into checks on the bounds of DR_STEP. Try using
9488 a bunds check for variable steps if the minimum required step is
9489 relatively small. Update calls to the dr_with_seg_len
9490 constructor and to vect_compile_time_alias.
9491 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9493 (vect_loop_versioning): Call it.
9494 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9496 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9499 Alan Hayward <alan.hayward@arm.com>
9500 David Sherwood <david.sherwood@arm.com>
9502 * doc/sourcebuild.texi (vect_scatter_store): Document.
9503 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9505 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9507 * genopinit.c (main): Add supports_vec_scatter_store and
9508 supports_vec_scatter_store_cached to target_optabs.
9509 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9510 IFN_MASK_SCATTER_STORE.
9511 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9513 * internal-fn.h (internal_store_fn_p): Declare.
9514 (internal_fn_stored_value_index): Likewise.
9515 * internal-fn.c (scatter_store_direct): New macro.
9516 (expand_scatter_store_optab_fn): New function.
9517 (direct_scatter_store_optab_supported_p): New macro.
9518 (internal_store_fn_p): New function.
9519 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9520 IFN_MASK_SCATTER_STORE.
9521 (internal_fn_mask_index): Likewise.
9522 (internal_fn_stored_value_index): New function.
9523 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9525 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9526 * optabs-query.c (supports_vec_scatter_store_p): New function.
9527 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9528 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9529 true for scatter stores.
9530 (vect_gather_scatter_fn_p): Handle scatter stores too.
9531 (vect_check_gather_scatter): Consider using scatter stores if
9532 supports_vec_scatter_store_p.
9533 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9535 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9536 internal_fn_stored_value_index.
9537 (check_load_store_masking): Handle scatter stores too.
9538 (vect_get_store_rhs): Make public.
9539 (vectorizable_call): Use internal_store_fn_p.
9540 (vectorizable_store): Handle scatter store internal functions.
9541 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9542 when deciding whether the end of the group has been reached.
9543 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9544 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9545 (mask_scatter_store<mode>): New insns.
9547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9548 Alan Hayward <alan.hayward@arm.com>
9549 David Sherwood <david.sherwood@arm.com>
9551 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9552 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9553 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9555 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9556 Use vect_truncate_gather_scatter_offset if we can't treat the
9557 operation as a normal gather load or scatter store.
9558 (get_group_load_store_type): Take the gather_scatter_info
9559 as argument. Try using a gather load or scatter store for
9560 single-element groups.
9561 (get_load_store_type): Update calls to get_group_load_store_type
9562 and vect_use_strided_gather_scatters_p.
9564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9565 Alan Hayward <alan.hayward@arm.com>
9566 David Sherwood <david.sherwood@arm.com>
9568 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9569 optional tree argument.
9570 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9572 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9573 but continue to use the current value as a fallback.
9574 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9575 to compare the updates.
9576 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9577 (get_load_store_type): Use it when handling a strided access.
9578 (vect_get_strided_load_store_ops): New function.
9579 (vect_get_data_ptr_increment): Likewise.
9580 (vectorizable_load): Handle strided gather loads. Always pass
9581 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9584 Alan Hayward <alan.hayward@arm.com>
9585 David Sherwood <david.sherwood@arm.com>
9587 * doc/md.texi (gather_load@var{m}): Document.
9588 (mask_gather_load@var{m}): Likewise.
9589 * genopinit.c (main): Add supports_vec_gather_load and
9590 supports_vec_gather_load_cached to target_optabs.
9591 * optabs-tree.c (init_tree_optimization_optabs): Use
9592 ggc_cleared_alloc to allocate target_optabs.
9593 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9594 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9596 * internal-fn.h (internal_load_fn_p): Declare.
9597 (internal_gather_scatter_fn_p): Likewise.
9598 (internal_fn_mask_index): Likewise.
9599 (internal_gather_scatter_fn_supported_p): Likewise.
9600 * internal-fn.c (gather_load_direct): New macro.
9601 (expand_gather_load_optab_fn): New function.
9602 (direct_gather_load_optab_supported_p): New macro.
9603 (direct_internal_fn_optab): New function.
9604 (internal_load_fn_p): Likewise.
9605 (internal_gather_scatter_fn_p): Likewise.
9606 (internal_fn_mask_index): Likewise.
9607 (internal_gather_scatter_fn_supported_p): Likewise.
9608 * optabs-query.c (supports_at_least_one_mode_p): New function.
9609 (supports_vec_gather_load_p): Likewise.
9610 * optabs-query.h (supports_vec_gather_load_p): Declare.
9611 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9612 and memory_type field.
9613 (NUM_PATTERNS): Bump to 15.
9614 * tree-vect-data-refs.c: Include internal-fn.h.
9615 (vect_gather_scatter_fn_p): New function.
9616 (vect_describe_gather_scatter_call): Likewise.
9617 (vect_check_gather_scatter): Try using internal functions for
9618 gather loads. Recognize existing calls to a gather load function.
9619 (vect_analyze_data_refs): Consider using gather loads if
9620 supports_vec_gather_load_p.
9621 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9622 (vect_get_gather_scatter_offset_type): Likewise.
9623 (vect_convert_mask_for_vectype): Likewise.
9624 (vect_add_conversion_to_patterm): Likewise.
9625 (vect_try_gather_scatter_pattern): Likewise.
9626 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9627 (vect_vect_recog_func_ptrs): Add it.
9628 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9629 internal_fn_mask_index and internal_gather_scatter_fn_p.
9630 (check_load_store_masking): Take the gather_scatter_info as an
9631 argument and handle gather loads.
9632 (vect_get_gather_scatter_ops): New function.
9633 (vectorizable_call): Check internal_load_fn_p.
9634 (vectorizable_load): Likewise. Handle gather load internal
9636 (vectorizable_store): Update call to check_load_store_masking.
9637 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9638 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9639 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9640 (aarch64_gather_scale_operand_d): New predicates.
9641 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9642 (mask_gather_load<mode>): New insns.
9644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9645 Alan Hayward <alan.hayward@arm.com>
9646 David Sherwood <david.sherwood@arm.com>
9648 * optabs.def (fold_left_plus_optab): New optab.
9649 * doc/md.texi (fold_left_plus_@var{m}): Document.
9650 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9651 * internal-fn.c (fold_left_direct): Define.
9652 (expand_fold_left_optab_fn): Likewise.
9653 (direct_fold_left_optab_supported_p): Likewise.
9654 * fold-const-call.c (fold_const_fold_left): New function.
9655 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9656 * tree-parloops.c (valid_reduction_p): New function.
9657 (gather_scalar_reductions): Use it.
9658 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9659 (vect_finish_replace_stmt): Declare.
9660 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9661 (needs_fold_left_reduction_p): New function, split out from...
9662 (vect_is_simple_reduction): ...here. Accept reductions that
9663 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9664 (vect_force_simple_reduction): Also store the reduction type in
9665 the assignment's STMT_VINFO_REDUC_TYPE.
9666 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9667 (merge_with_identity): New function.
9668 (vect_expand_fold_left): Likewise.
9669 (vectorize_fold_left_reduction): Likewise.
9670 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9671 scalar phi in place for it. Check for target support and reject
9672 cases that would reassociate the operation. Defer the transform
9673 phase to vectorize_fold_left_reduction.
9674 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9675 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9676 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9680 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9681 call to ifc_temp_var.
9683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9684 Alan Hayward <alan.hayward@arm.com>
9685 David Sherwood <david.sherwood@arm.com>
9687 * target.def (legitimize_address_displacement): Take the original
9688 offset as a poly_int.
9689 * targhooks.h (default_legitimize_address_displacement): Update
9691 * targhooks.c (default_legitimize_address_displacement): Likewise.
9692 * doc/tm.texi: Regenerate.
9693 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9694 as an argument, moving assert of ad->disp == ad->disp_term to...
9695 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9696 Try calling targetm.legitimize_address_displacement before expanding
9697 the address rather than afterwards, and adjust for the new interface.
9698 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9699 Match the new hook interface. Handle SVE addresses.
9700 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9705 * Makefile.in (OBJS): Add early-remat.o.
9706 * target.def (select_early_remat_modes): New hook.
9707 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9708 * doc/tm.texi: Regenerate.
9709 * targhooks.h (default_select_early_remat_modes): Declare.
9710 * targhooks.c (default_select_early_remat_modes): New function.
9711 * timevar.def (TV_EARLY_REMAT): New timevar.
9712 * passes.def (pass_early_remat): New pass.
9713 * tree-pass.h (make_pass_early_remat): Declare.
9714 * early-remat.c: New file.
9715 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9717 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9720 Alan Hayward <alan.hayward@arm.com>
9721 David Sherwood <david.sherwood@arm.com>
9723 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9724 vfm1 with a bound_epilog parameter.
9725 (vect_do_peeling): Update calls accordingly, and move the prologue
9726 call earlier in the function. Treat the base bound_epilog as 0 for
9727 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9728 this base when peeling for gaps.
9729 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9730 with fully-masked loops.
9731 (vect_estimate_min_profitable_iters): Handle the single peeled
9732 iteration in that case.
9734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9735 Alan Hayward <alan.hayward@arm.com>
9736 David Sherwood <david.sherwood@arm.com>
9738 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9739 single-element interleaving even if the size is not a power of 2.
9740 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9741 accesses for single-element interleaving if the group size is
9744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9745 Alan Hayward <alan.hayward@arm.com>
9746 David Sherwood <david.sherwood@arm.com>
9748 * doc/md.texi (fold_extract_last_@var{m}): Document.
9749 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9750 * optabs.def (fold_extract_last_optab): New optab.
9751 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9752 * internal-fn.c (fold_extract_direct): New macro.
9753 (expand_fold_extract_optab_fn): Likewise.
9754 (direct_fold_extract_optab_supported_p): Likewise.
9755 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9756 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9757 EXTRACT_LAST_REDUCTION.
9758 (get_initial_def_for_reduction): Do not create an initial vector
9759 for EXTRACT_LAST_REDUCTION reductions.
9760 (vectorizable_reduction): Leave the scalar phi in place for
9761 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9762 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9763 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9764 transform phase to vectorizable_condition.
9765 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9767 (vect_finish_stmt_generation): ...here.
9768 (vect_finish_replace_stmt): New function.
9769 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9770 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9772 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9775 Alan Hayward <alan.hayward@arm.com>
9776 David Sherwood <david.sherwood@arm.com>
9778 * doc/md.texi (extract_last_@var{m}): Document.
9779 * optabs.def (extract_last_optab): New optab.
9780 * internal-fn.def (EXTRACT_LAST): New internal function.
9781 * internal-fn.c (cond_unary_direct): New macro.
9782 (expand_cond_unary_optab_fn): Likewise.
9783 (direct_cond_unary_optab_supported_p): Likewise.
9784 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9785 loops using EXTRACT_LAST.
9786 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9787 (extract_last_<mode>): ...this optab.
9788 (vec_extract<mode><Vel>): Update accordingly.
9790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9791 Alan Hayward <alan.hayward@arm.com>
9792 David Sherwood <david.sherwood@arm.com>
9794 * target.def (empty_mask_is_expensive): New hook.
9795 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9796 * doc/tm.texi: Regenerate.
9797 * targhooks.h (default_empty_mask_is_expensive): Declare.
9798 * targhooks.c (default_empty_mask_is_expensive): New function.
9799 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9800 if the target says that empty masks are expensive.
9801 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9803 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9806 Alan Hayward <alan.hayward@arm.com>
9807 David Sherwood <david.sherwood@arm.com>
9809 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9810 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9811 (vect_use_loop_mask_for_alignment_p): New function.
9812 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9813 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9814 niters_skip argument. Make sure that the first niters_skip elements
9815 of the first iteration are inactive.
9816 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9817 Update call to vect_set_loop_masks_directly.
9818 (get_misalign_in_elems): New function, split out from...
9819 (vect_gen_prolog_loop_niters): ...here.
9820 (vect_update_init_of_dr): Take a code argument that specifies whether
9821 the adjustment should be added or subtracted.
9822 (vect_update_init_of_drs): Likewise.
9823 (vect_prepare_for_masked_peels): New function.
9824 (vect_do_peeling): Skip prologue peeling if we're using a mask
9825 instead. Update call to vect_update_inits_of_drs.
9826 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9828 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9829 alignment. Do not include the number of peeled iterations in
9830 the minimum threshold in that case.
9831 (vectorizable_induction): Adjust the start value down by
9832 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9833 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9834 Take the number of skipped iterations into account when calculating
9836 * tree-vect-stmts.c (vect_gen_while_not): New function.
9838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9839 Alan Hayward <alan.hayward@arm.com>
9840 David Sherwood <david.sherwood@arm.com>
9842 * doc/sourcebuild.texi (vect_fully_masked): Document.
9843 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9845 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9847 (vect_analyze_loop_2): ...here. Don't check the vectorization
9848 factor against the number of loop iterations if the loop is
9851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9852 Alan Hayward <alan.hayward@arm.com>
9853 David Sherwood <david.sherwood@arm.com>
9855 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9856 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9857 (dump_groups): Update accordingly.
9858 (iv_use::mem_type): New member variable.
9859 (address_p): New function.
9860 (record_use): Add a mem_type argument and initialize the new
9862 (record_group_use): Add a mem_type argument. Use address_p.
9863 Remove obsolete null checks of base_object. Update call to record_use.
9864 (find_interesting_uses_op): Update call to record_group_use.
9865 (find_interesting_uses_cond): Likewise.
9866 (find_interesting_uses_address): Likewise.
9867 (get_mem_type_for_internal_fn): New function.
9868 (find_address_like_use): Likewise.
9869 (find_interesting_uses_stmt): Try find_address_like_use before
9870 calling find_interesting_uses_op.
9871 (addr_offset_valid_p): Use the iv mem_type field as the type
9872 of the addressed memory.
9873 (add_autoinc_candidates): Likewise.
9874 (get_address_cost): Likewise.
9875 (split_small_address_groups_p): Use address_p.
9876 (split_address_groups): Likewise.
9877 (add_iv_candidate_for_use): Likewise.
9878 (autoinc_possible_for_pair): Likewise.
9879 (rewrite_groups): Likewise.
9880 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9881 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9882 (get_alias_ptr_type_for_ptr_address): New function.
9883 (rewrite_use_address): Rewrite address uses in calls that were
9884 identified by find_address_like_use.
9886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9887 Alan Hayward <alan.hayward@arm.com>
9888 David Sherwood <david.sherwood@arm.com>
9890 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9892 * gimple-expr.h (is_gimple_addressable: Likewise.
9893 * gimple-expr.c (is_gimple_address): Likewise.
9894 * internal-fn.c (expand_call_mem_ref): New function.
9895 (expand_mask_load_optab_fn): Use it.
9896 (expand_mask_store_optab_fn): Likewise.
9898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9899 Alan Hayward <alan.hayward@arm.com>
9900 David Sherwood <david.sherwood@arm.com>
9902 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9903 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9904 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9905 (cond_umax@var{mode}): Document.
9906 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9907 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9908 (cond_umin_optab, cond_umax_optab): New optabs.
9909 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9910 (COND_IOR, COND_XOR): New internal functions.
9911 * internal-fn.h (get_conditional_internal_fn): Declare.
9912 * internal-fn.c (cond_binary_direct): New macro.
9913 (expand_cond_binary_optab_fn): Likewise.
9914 (direct_cond_binary_optab_supported_p): Likewise.
9915 (get_conditional_internal_fn): New function.
9916 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9917 Cope with reduction statements that are vectorized as calls rather
9919 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9920 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9921 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9922 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9923 (UNSPEC_COND_EOR): New unspecs.
9924 (optab): Add mappings for them.
9925 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9926 (sve_int_op, sve_fp_op): New int attributes.
9928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9929 Alan Hayward <alan.hayward@arm.com>
9930 David Sherwood <david.sherwood@arm.com>
9932 * optabs.def (while_ult_optab): New optab.
9933 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9934 * internal-fn.def (WHILE_ULT): New internal function.
9935 * internal-fn.h (direct_internal_fn_supported_p): New override
9936 that takes two types as argument.
9937 * internal-fn.c (while_direct): New macro.
9938 (expand_while_optab_fn): New function.
9939 (convert_optab_supported_p): Likewise.
9940 (direct_while_optab_supported_p): New macro.
9941 * wide-int.h (wi::udiv_ceil): New function.
9942 * tree-vectorizer.h (rgroup_masks): New structure.
9943 (vec_loop_masks): New typedef.
9944 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9946 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9947 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9948 (vect_max_vf): New function.
9949 (slpeel_make_loop_iterate_ntimes): Delete.
9950 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9951 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9952 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9953 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9954 internal-fn.h, stor-layout.h and optabs-query.h.
9955 (vect_set_loop_mask): New function.
9956 (add_preheader_seq): Likewise.
9957 (add_header_seq): Likewise.
9958 (interleave_supported_p): Likewise.
9959 (vect_maybe_permute_loop_masks): Likewise.
9960 (vect_set_loop_masks_directly): Likewise.
9961 (vect_set_loop_condition_masked): Likewise.
9962 (vect_set_loop_condition_unmasked): New function, split out from
9963 slpeel_make_loop_iterate_ntimes.
9964 (slpeel_make_loop_iterate_ntimes): Rename to..
9965 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9966 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9967 (vect_do_peeling): Update call accordingly.
9968 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9970 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9971 mask_compare_type, can_fully_mask_p and fully_masked_p.
9972 (release_vec_loop_masks): New function.
9973 (_loop_vec_info): Use it to free the loop masks.
9974 (can_produce_all_loop_masks_p): New function.
9975 (vect_get_max_nscalars_per_iter): Likewise.
9976 (vect_verify_full_masking): Likewise.
9977 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9978 retries, and free the mask rgroups before retrying. Check loop-wide
9979 reasons for disallowing fully-masked loops. Make the final decision
9980 about whether use a fully-masked loop or not.
9981 (vect_estimate_min_profitable_iters): Do not assume that peeling
9982 for the number of iterations will be needed for fully-masked loops.
9983 (vectorizable_reduction): Disable fully-masked loops.
9984 (vectorizable_live_operation): Likewise.
9985 (vect_halve_mask_nunits): New function.
9986 (vect_double_mask_nunits): Likewise.
9987 (vect_record_loop_mask): Likewise.
9988 (vect_get_loop_mask): Likewise.
9989 (vect_transform_loop): Handle the case in which the final loop
9990 iteration might handle a partial vector. Call vect_set_loop_condition
9991 instead of slpeel_make_loop_iterate_ntimes.
9992 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9993 (check_load_store_masking): New function.
9994 (prepare_load_store_mask): Likewise.
9995 (vectorizable_store): Handle fully-masked loops.
9996 (vectorizable_load): Likewise.
9997 (supportable_widening_operation): Use vect_halve_mask_nunits for
9999 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10000 (vect_gen_while): New function.
10001 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10002 (aarch64_uqdec<mode>): New insn.
10004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10005 Alan Hayward <alan.hayward@arm.com>
10006 David Sherwood <david.sherwood@arm.com>
10008 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10009 (reduc_xor_scal_optab): New optabs.
10010 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10011 (reduc_xor_scal_@var{m}): Document.
10012 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10013 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10014 internal functions.
10015 * fold-const-call.c (fold_const_call): Handle them.
10016 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10017 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10018 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10019 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10020 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10021 (UNSPEC_XORV): New unspecs.
10022 (optab): Add entries for them.
10023 (BITWISEV): New int iterator.
10024 (bit_reduc_op): New int attributes.
10026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10027 Alan Hayward <alan.hayward@arm.com>
10028 David Sherwood <david.sherwood@arm.com>
10030 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10031 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10032 * optabs.def (vec_shl_insert_optab): New optab.
10033 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10034 (duplicate_and_interleave): Likewise.
10035 * tree-vect-loop.c: Include internal-fn.h.
10036 (neutral_op_for_slp_reduction): New function, split out from
10037 get_initial_defs_for_reduction.
10038 (get_initial_def_for_reduction): Handle option 2 for variable-length
10039 vectors by loading the neutral value into a vector and then shifting
10040 the initial value into element 0.
10041 (get_initial_defs_for_reduction): Replace the code argument with
10042 the neutral value calculated by neutral_op_for_slp_reduction.
10043 Use gimple_build_vector for constant-length vectors.
10044 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10045 but the first group_size elements have a neutral value.
10046 Use duplicate_and_interleave otherwise.
10047 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10048 Update call to get_initial_defs_for_reduction. Handle SLP
10049 reductions for variable-length vectors by creating one vector
10050 result for each scalar result, with the elements associated
10051 with other scalar results stubbed out with the neutral value.
10052 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10053 Require IFN_VEC_SHL_INSERT for double reductions on
10054 variable-length vectors, or SLP reductions that have
10055 a neutral value. Require can_duplicate_and_interleave_p
10056 support for variable-length unchained SLP reductions if there
10057 is no neutral value, such as for MIN/MAX reductions. Also require
10058 the number of vector elements to be a multiple of the number of
10059 SLP statements when doing variable-length unchained SLP reductions.
10060 Update call to vect_create_epilog_for_reduction.
10061 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10062 and remove initial values.
10063 (duplicate_and_interleave): Make public.
10064 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10065 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10068 Alan Hayward <alan.hayward@arm.com>
10069 David Sherwood <david.sherwood@arm.com>
10071 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10072 (can_duplicate_and_interleave_p): New function.
10073 (vect_get_and_check_slp_defs): Take the vector of statements
10074 rather than just the current one. Remove excess parentheses.
10075 Restriction rejectinon of vect_constant_def and vect_external_def
10076 for variable-length vectors to boolean types, or types for which
10077 can_duplicate_and_interleave_p is false.
10078 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10079 (duplicate_and_interleave): New function.
10080 (vect_get_constant_vectors): Use gimple_build_vector for
10081 constant-length vectors and suitable variable-length constant
10082 vectors. Use duplicate_and_interleave for other variable-length
10083 vectors. Don't defer the update when inserting new statements.
10085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10086 Alan Hayward <alan.hayward@arm.com>
10087 David Sherwood <david.sherwood@arm.com>
10089 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10090 min_profitable_iters doesn't go negative.
10092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10093 Alan Hayward <alan.hayward@arm.com>
10094 David Sherwood <david.sherwood@arm.com>
10096 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10097 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10098 * optabs.def (vec_mask_load_lanes_optab): New optab.
10099 (vec_mask_store_lanes_optab): Likewise.
10100 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10101 (MASK_STORE_LANES): Likewise.
10102 * internal-fn.c (mask_load_lanes_direct): New macro.
10103 (mask_store_lanes_direct): Likewise.
10104 (expand_mask_load_optab_fn): Handle masked operations.
10105 (expand_mask_load_lanes_optab_fn): New macro.
10106 (expand_mask_store_optab_fn): Handle masked operations.
10107 (expand_mask_store_lanes_optab_fn): New macro.
10108 (direct_mask_load_lanes_optab_supported_p): Likewise.
10109 (direct_mask_store_lanes_optab_supported_p): Likewise.
10110 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10112 (vect_load_lanes_supported): Likewise.
10113 * tree-vect-data-refs.c (strip_conversion): New function.
10114 (can_group_stmts_p): Likewise.
10115 (vect_analyze_data_ref_accesses): Use it instead of checking
10116 for a pair of assignments.
10117 (vect_store_lanes_supported): Take a masked_p parameter.
10118 (vect_load_lanes_supported): Likewise.
10119 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10120 vect_store_lanes_supported and vect_load_lanes_supported.
10121 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10122 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10123 parameter. Don't allow gaps for masked accesses.
10124 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10125 and vect_load_lanes_supported.
10126 (get_load_store_type): Take a masked_p parameter and update
10127 call to get_group_load_store_type.
10128 (vectorizable_store): Update call to get_load_store_type.
10129 Handle IFN_MASK_STORE_LANES.
10130 (vectorizable_load): Update call to get_load_store_type.
10131 Handle IFN_MASK_LOAD_LANES.
10133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10134 Alan Hayward <alan.hayward@arm.com>
10135 David Sherwood <david.sherwood@arm.com>
10137 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10139 * config/aarch64/aarch64-protos.h
10140 (aarch64_sve_struct_memory_operand_p): Declare.
10141 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10142 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10143 (VPRED, vpred): Handle SVE structure modes.
10144 * config/aarch64/constraints.md (Utx): New constraint.
10145 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10146 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10147 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10148 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10149 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10150 structure modes. Split into pieces after RA.
10151 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10152 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10154 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10155 SVE structure modes.
10156 (aarch64_classify_address): Likewise.
10157 (sizetochar): Move earlier in file.
10158 (aarch64_print_operand): Handle SVE register lists.
10159 (aarch64_array_mode): New function.
10160 (aarch64_sve_struct_memory_operand_p): Likewise.
10161 (TARGET_ARRAY_MODE): Redefine.
10163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10164 Alan Hayward <alan.hayward@arm.com>
10165 David Sherwood <david.sherwood@arm.com>
10167 * target.def (array_mode): New target hook.
10168 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10169 * doc/tm.texi: Regenerate.
10170 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10171 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10172 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10173 targetm.array_mode.
10174 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10178 Alan Hayward <alan.hayward@arm.com>
10179 David Sherwood <david.sherwood@arm.com>
10181 * fold-const.c (fold_binary_loc): Check the argument types
10182 rather than the result type when testing for a vector operation.
10184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10186 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10187 * doc/tm.texi: Regenerate.
10189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10190 Alan Hayward <alan.hayward@arm.com>
10191 David Sherwood <david.sherwood@arm.com>
10193 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10194 (sve): Document new AArch64 extension.
10195 * doc/md.texi (w): Extend the description of the AArch64
10196 constraint to include SVE vectors.
10197 (Upl, Upa): Document new AArch64 predicate constraints.
10198 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10200 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10201 (msve-vector-bits=): New option.
10202 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10203 SVE when these are disabled.
10204 (sve): New extension.
10205 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10206 modes. Adjust their number of units based on aarch64_sve_vg.
10207 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10208 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10209 aarch64_addr_query_type.
10210 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10211 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10212 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10213 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10214 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10215 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10216 (aarch64_simd_imm_zero_p): Delete.
10217 (aarch64_check_zero_based_sve_index_immediate): Declare.
10218 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10219 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10220 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10221 (aarch64_sve_float_mul_immediate_p): Likewise.
10222 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10223 rather than an rtx.
10224 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10225 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10226 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10227 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10228 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10229 (aarch64_regmode_natural_size): Likewise.
10230 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10231 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10233 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10234 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10235 for VG and the SVE predicate registers.
10236 (V_ALIASES): Add a "z"-prefixed alias.
10237 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10238 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10239 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10240 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10241 (REG_CLASS_NAMES): Add entries for them.
10242 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10243 and the predicate registers.
10244 (aarch64_sve_vg): Declare.
10245 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10246 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10247 (REGMODE_NATURAL_SIZE): Define.
10248 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10250 * config/aarch64/aarch64.c: Include cfgrtl.h.
10251 (simd_immediate_info): Add a constructor for series vectors,
10252 and an associated step field.
10253 (aarch64_sve_vg): New variable.
10254 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10255 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10256 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10257 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10258 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10259 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10260 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10261 (aarch64_get_mask_mode): New functions.
10262 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10263 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10264 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10265 predicate modes and predicate registers. Explicitly restrict
10266 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10267 to store a vector mode if it is recognized by
10268 aarch64_classify_vector_mode.
10269 (aarch64_regmode_natural_size): New function.
10270 (aarch64_hard_regno_caller_save_mode): Return the original mode
10272 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10273 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10274 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10275 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10277 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10278 does not overlap dest if the function is frame-related. Handle
10280 (aarch64_split_add_offset): New function.
10281 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10282 them aarch64_add_offset.
10283 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10284 and update call to aarch64_sub_sp.
10285 (aarch64_add_cfa_expression): New function.
10286 (aarch64_expand_prologue): Pass extra temporary registers to the
10287 functions above. Handle the case in which we need to emit new
10288 DW_CFA_expressions for registers that were originally saved
10289 relative to the stack pointer, but now have to be expressed
10290 relative to the frame pointer.
10291 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10293 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10294 IP0 and IP1 values for SVE frames.
10295 (aarch64_expand_vec_series): New function.
10296 (aarch64_expand_sve_widened_duplicate): Likewise.
10297 (aarch64_expand_sve_const_vector): Likewise.
10298 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10299 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10300 into the register, rather than emitting a SET directly.
10301 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10302 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10303 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10304 (offset_9bit_signed_scaled_p): New functions.
10305 (aarch64_replicate_bitmask_imm): New function.
10306 (aarch64_bitmask_imm): Use it.
10307 (aarch64_cannot_force_const_mem): Reject expressions involving
10308 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10309 (aarch64_classify_index): Handle SVE indices, by requiring
10310 a plain register index with a scale that matches the element size.
10311 (aarch64_classify_address): Handle SVE addresses. Assert that
10312 the mode of the address is VOIDmode or an integer mode.
10313 Update call to aarch64_classify_symbol.
10314 (aarch64_classify_symbolic_expression): Update call to
10315 aarch64_classify_symbol.
10316 (aarch64_const_vec_all_in_range_p): New function.
10317 (aarch64_print_vector_float_operand): Likewise.
10318 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10319 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10320 and the FP immediates 1.0 and 0.5.
10321 (aarch64_print_address_internal): Handle SVE addresses.
10322 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10323 (aarch64_regno_regclass): Handle predicate registers.
10324 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10326 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10327 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10328 (aarch64_convert_sve_vector_bits): New function.
10329 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10330 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10331 rather than an rtx.
10332 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10333 Handle SVE vector and predicate modes. Accept VL-based constants
10334 that need only one temporary register, and VL offsets that require
10335 no temporary registers.
10336 (aarch64_conditional_register_usage): Mark the predicate registers
10337 as fixed if SVE isn't available.
10338 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10339 Return true for SVE vector and predicate modes.
10340 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10341 rather than an unsigned int. Handle SVE modes.
10342 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10344 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10346 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10347 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10348 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10349 (aarch64_sve_float_mul_immediate_p): New functions.
10350 (aarch64_sve_valid_immediate): New function.
10351 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10352 Explicitly reject structure modes. Check for INDEX constants.
10353 Handle PTRUE and PFALSE constants.
10354 (aarch64_check_zero_based_sve_index_immediate): New function.
10355 (aarch64_simd_imm_zero_p): Delete.
10356 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10357 vector modes. Accept constants in the range of CNT[BHWD].
10358 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10359 ask for an Advanced SIMD mode.
10360 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10361 (aarch64_simd_vector_alignment): Handle SVE predicates.
10362 (aarch64_vectorize_preferred_vector_alignment): New function.
10363 (aarch64_simd_vector_alignment_reachable): Use it instead of
10365 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10366 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10368 (MAX_VECT_LEN): Delete.
10369 (expand_vec_perm_d): Add a vec_flags field.
10370 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10371 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10372 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10374 (aarch64_evpc_rev): Rename to...
10375 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10376 (aarch64_evpc_rev_global): New function.
10377 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10378 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10380 (aarch64_evpc_sve_tbl): New function.
10381 (aarch64_expand_vec_perm_const_1): Update after rename of
10382 aarch64_evpc_rev. Handle SVE permutes too, trying
10383 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10384 than aarch64_evpc_tbl.
10385 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10386 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10387 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10388 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10389 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10390 (aarch64_expand_sve_vcond): New functions.
10391 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10392 of aarch64_vector_mode_p.
10393 (aarch64_dwarf_poly_indeterminate_value): New function.
10394 (aarch64_compute_pressure_classes): Likewise.
10395 (aarch64_can_change_mode_class): Likewise.
10396 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10397 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10398 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10399 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10400 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10401 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10402 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10403 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10405 (Dn, Dl, Dr): Accept const as well as const_vector.
10406 (Dz): Likewise. Compare against CONST0_RTX.
10407 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10408 of "vector" where appropriate.
10409 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10410 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10411 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10412 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10413 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10414 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10415 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10416 (v_int_equiv): Extend to SVE modes.
10417 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10419 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10420 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10421 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10422 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10423 (SVE_COND_FP_CMP): New int iterators.
10424 (perm_hilo): Handle the new unpack unspecs.
10425 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10427 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10428 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10429 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10430 (aarch64_equality_operator, aarch64_constant_vector_operand)
10431 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10432 (aarch64_sve_nonimmediate_operand): Likewise.
10433 (aarch64_sve_general_operand): Likewise.
10434 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10435 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10436 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10437 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10438 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10439 (aarch64_sve_float_arith_immediate): Likewise.
10440 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10441 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10442 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10443 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10444 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10445 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10446 (aarch64_sve_float_arith_operand): Likewise.
10447 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10448 (aarch64_sve_float_mul_operand): Likewise.
10449 (aarch64_sve_vec_perm_operand): Likewise.
10450 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10451 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10452 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10453 as well as const_vector.
10454 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10455 in file. Use CONST0_RTX and CONSTM1_RTX.
10456 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10457 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10458 Use aarch64_simd_imm_zero.
10459 * config/aarch64/aarch64-sve.md: New file.
10460 * config/aarch64/aarch64.md: Include it.
10461 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10462 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10463 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10464 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10465 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10466 (sve): New attribute.
10467 (enabled): Disable instructions with the sve attribute unless
10469 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10470 aarch64_expand_mov_immediate.
10471 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10472 CNT[BHSD] immediates.
10473 (movti): Split CONST_POLY_INT moves into two halves.
10474 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10475 Split additions that need a temporary here if the destination
10476 is the stack pointer.
10477 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10478 (*add<mode>3_poly_1): New instruction.
10479 (set_clobber_cc): New expander.
10481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10483 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10484 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10485 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10486 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10487 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10488 Change innermode from fixed_mode_size to machine_mode.
10489 (simplify_subreg): Update call accordingly. Handle a constant-sized
10490 subreg of a variable-length CONST_VECTOR.
10492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10493 Alan Hayward <alan.hayward@arm.com>
10494 David Sherwood <david.sherwood@arm.com>
10496 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10497 (add_offset_to_base): New function, split out from...
10498 (create_mem_ref): ...here. When handling a scale other than 1,
10499 check first whether the address is valid without the offset.
10500 Add it into the base if so, leaving the index and scale as-is.
10502 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10505 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10506 fold_for_warn before checking if arg2 is INTEGER_CST.
10508 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10510 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10511 (store_multiple_operation): Delete.
10512 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10513 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10514 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10515 guarded by TARGET_STRING.
10516 (rs6000_output_load_multiple): Delete.
10517 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10518 OPTION_MASK_STRING / TARGET_STRING handling.
10519 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10520 (const rs6000_opt_masks) <"string">: Change mask to 0.
10521 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10522 (MASK_STRING): Delete.
10523 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10525 (load_multiple): Delete.
10532 (store_multiple): Delete.
10539 (movmemsi_8reg): Delete.
10540 (corresponding unnamed define_insn): Delete.
10541 (movmemsi_6reg): Delete.
10542 (corresponding unnamed define_insn): Delete.
10543 (movmemsi_4reg): Delete.
10544 (corresponding unnamed define_insn): Delete.
10545 (movmemsi_2reg): Delete.
10546 (corresponding unnamed define_insn): Delete.
10547 (movmemsi_1reg): Delete.
10548 (corresponding unnamed define_insn): Delete.
10549 * config/rs6000/rs6000.opt (mno-string): New.
10550 (mstring): Replace by deprecation warning stub.
10551 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10553 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10555 * regrename.c (regrename_do_replace): If replacing the same
10556 reg multiple times, try to reuse last created gen_raw_REG.
10559 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10560 main to workaround a bug in GDB.
10562 2018-01-12 Tom de Vries <tom@codesourcery.com>
10565 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10567 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10569 PR rtl-optimization/80481
10570 * ira-color.c (get_cap_member): New function.
10571 (allocnos_conflict_by_live_ranges_p): Use it.
10572 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10573 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10575 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10578 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10579 (*saddl_se_1): Ditto.
10580 (*ssubsi_1): Ditto.
10581 (*ssubl_se_1): Ditto.
10583 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10585 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10586 rather than wi::to_widest for DR_INITs.
10587 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10588 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10589 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10591 (vect_analyze_group_access_1): Note that here.
10593 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10595 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10596 polynomial type sizes.
10598 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10600 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10601 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10602 (gimple_add_tmp_var): Likewise.
10604 2018-01-12 Martin Liska <mliska@suse.cz>
10606 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10607 (gimple_alloc_sizes): Likewise.
10608 (dump_gimple_statistics): Use PRIu64 in printf format.
10609 * gimple.h: Change uint64_t to int.
10611 2018-01-12 Martin Liska <mliska@suse.cz>
10613 * tree-core.h: Use uint64_t instead of int.
10614 * tree.c (tree_node_counts): Likewise.
10615 (tree_node_sizes): Likewise.
10616 (dump_tree_statistics): Use PRIu64 in printf format.
10618 2018-01-12 Martin Liska <mliska@suse.cz>
10620 * Makefile.in: As qsort_chk is implemented in vec.c, add
10621 vec.o to linkage of gencfn-macros.
10622 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10623 passing the info to record_node_allocation_statistics.
10624 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10626 * ggc-common.c (struct ggc_usage): Add operator== and use
10627 it in operator< and compare function.
10628 * mem-stats.h (struct mem_usage): Likewise.
10629 * vec.c (struct vec_usage): Remove operator< and compare
10630 function. Can be simply inherited.
10632 2018-01-12 Martin Jambor <mjambor@suse.cz>
10635 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10636 * tree-ssa-math-opts.c: Include domwalk.h.
10637 (convert_mult_to_fma_1): New function.
10638 (fma_transformation_info): New type.
10639 (fma_deferring_state): Likewise.
10640 (cancel_fma_deferring): New function.
10641 (result_of_phi): Likewise.
10642 (last_fma_candidate_feeds_initial_phi): Likewise.
10643 (convert_mult_to_fma): Added deferring logic, split actual
10644 transformation to convert_mult_to_fma_1.
10645 (math_opts_dom_walker): New type.
10646 (math_opts_dom_walker::after_dom_children): New method, body moved
10647 here from pass_optimize_widening_mul::execute, added deferring logic
10649 (pass_optimize_widening_mul::execute): Moved most of code to
10650 math_opts_dom_walker::after_dom_children.
10651 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10652 * config/i386/i386.c (ix86_option_override_internal): Added
10653 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10655 2018-01-12 Richard Biener <rguenther@suse.de>
10658 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10659 inline instance vars.
10661 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10664 * config/rx/rx.c (rx_is_restricted_memory_address):
10665 Handle SUBREG case.
10667 2018-01-12 Richard Biener <rguenther@suse.de>
10669 PR tree-optimization/80846
10670 * target.def (split_reduction): New target hook.
10671 * targhooks.c (default_split_reduction): New function.
10672 * targhooks.h (default_split_reduction): Declare.
10673 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10674 target requests first reduce vectors by combining low and high
10676 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10677 (get_vectype_for_scalar_type_and_size): Export.
10678 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10679 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10680 * doc/tm.texi: Regenerate.
10681 * config/i386/i386.c (ix86_split_reduction): Implement
10682 TARGET_VECTORIZE_SPLIT_REDUCTION.
10684 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10687 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10688 in PIC mode except for TARGET_VXWORKS_RTP.
10689 * config/sparc/sparc.c: Include cfgrtl.h.
10690 (TARGET_INIT_PIC_REG): Define.
10691 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10692 (sparc_pic_register_p): New predicate.
10693 (sparc_legitimate_address_p): Use it.
10694 (sparc_legitimize_pic_address): Likewise.
10695 (sparc_delegitimize_address): Likewise.
10696 (sparc_mode_dependent_address_p): Likewise.
10697 (gen_load_pcrel_sym): Remove 4th parameter.
10698 (load_got_register): Adjust call to above. Remove obsolete stuff.
10699 (sparc_expand_prologue): Do not call load_got_register here.
10700 (sparc_flat_expand_prologue): Likewise.
10701 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10702 (sparc_use_pseudo_pic_reg): New function.
10703 (sparc_init_pic_reg): Likewise.
10704 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10705 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10707 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10709 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10710 Add item for branch_cost.
10712 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10714 PR rtl-optimization/83565
10715 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10716 not extend the result to a larger mode for rotate operations.
10717 (num_sign_bit_copies1): Likewise.
10719 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10722 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10724 Use values-Xc.o for -pedantic.
10725 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10727 2018-01-12 Martin Liska <mliska@suse.cz>
10730 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10732 (possible_polymorphic_call_targets): Use it.
10733 (ipa_devirt): Likewise.
10735 2018-01-12 Martin Liska <mliska@suse.cz>
10737 * profile-count.h (enum profile_quality): Use 0 as invalid
10738 enum value of profile_quality.
10740 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10742 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10743 -mext-string options.
10745 2018-01-12 Richard Biener <rguenther@suse.de>
10747 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10748 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10749 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10751 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10753 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10755 * configure.ac (--with-long-double-format): Add support for the
10756 configuration option to change the default long double format on
10758 * config.gcc (powerpc*-linux*-*): Likewise.
10759 * configure: Regenerate.
10760 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10761 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10762 used without modification.
10764 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10766 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10767 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10768 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10769 MISC_BUILTIN_SPEC_BARRIER.
10770 (rs6000_init_builtins): Likewise.
10771 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10773 (speculation_barrier): New define_insn.
10774 * doc/extend.texi: Document __builtin_speculation_barrier.
10776 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10779 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10780 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10781 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10783 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10784 integral modes instead of "ss" and "sd".
10785 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10786 vectors with 32-bit and 64-bit elements.
10787 (vecdupssescalarmodesuffix): New mode attribute.
10788 (vec_dup<mode>): Use it.
10790 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10793 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10794 frame if argument is passed on stack.
10796 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10799 * ree.c (combine_reaching_defs): Optimize also
10800 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10801 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10803 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10805 PR middle-end/83189
10806 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10808 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10810 PR middle-end/83718
10811 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10812 after they are computed.
10814 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10816 PR tree-optimization/83695
10817 * gimple-loop-linterchange.cc
10818 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10819 reset cached scev information after interchange.
10820 (pass_linterchange::execute): Remove call to scev_reset_htab.
10822 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10824 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10825 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10826 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10827 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10828 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10829 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10830 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10831 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10832 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10833 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10834 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10835 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10836 (V_lane_reg): Likewise.
10837 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10839 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10840 (vfmal_lane_low<mode>_intrinsic,
10841 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10842 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10843 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10844 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10845 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10846 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10848 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10850 * config/arm/arm-cpus.in (fp16fml): New feature.
10851 (ALL_SIMD): Add fp16fml.
10852 (armv8.2-a): Add fp16fml as an option.
10853 (armv8.3-a): Likewise.
10854 (armv8.4-a): Add fp16fml as part of fp16.
10855 * config/arm/arm.h (TARGET_FP16FML): Define.
10856 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10858 * config/arm/arm-modes.def (V2HF): Define.
10859 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10860 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10861 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10862 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10863 vfmsl_low, vfmsl_high): New set of builtins.
10864 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10865 (vfml_op): New code attribute.
10866 (VFMLHALVES): New int iterator.
10867 (VFML, VFMLSEL): New mode attributes.
10868 (V_reg): Define mapping for V2HF.
10869 (V_hi, V_lo): New mode attributes.
10870 (VF_constraint): Likewise.
10871 (vfml_half, vfml_half_selector): New int attributes.
10872 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10874 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10875 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10877 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10878 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10879 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10880 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10882 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10883 Document new effective target and option set.
10885 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10887 * config/arm/arm-cpus.in (armv8_4): New feature.
10888 (ARMv8_4a): New fgroup.
10889 (armv8.4-a): New arch.
10890 * config/arm/arm-tables.opt: Regenerate.
10891 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10892 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10893 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10894 Add matching rules for -march=armv8.4-a and extensions.
10895 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10897 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10900 * config/rx/rx.md (BW): New mode attribute.
10901 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10903 2018-01-11 Richard Biener <rguenther@suse.de>
10905 PR tree-optimization/83435
10906 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10907 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10908 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10910 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10911 Alan Hayward <alan.hayward@arm.com>
10912 David Sherwood <david.sherwood@arm.com>
10914 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10916 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10917 (aarch64_print_address_internal): Use it to check for a zero offset.
10919 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10920 Alan Hayward <alan.hayward@arm.com>
10921 David Sherwood <david.sherwood@arm.com>
10923 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10924 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10925 Return a poly_int64 rather than a HOST_WIDE_INT.
10926 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10927 rather than a HOST_WIDE_INT.
10928 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10929 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10930 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10931 final_offset from HOST_WIDE_INT to poly_int64.
10932 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10933 to_constant when getting the number of units in an Advanced SIMD
10935 (aarch64_builtin_vectorized_function): Check for a constant number
10937 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10939 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10940 attribute instead of GET_MODE_NUNITS.
10941 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10942 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10943 GET_MODE_SIZE for fixed-size registers.
10944 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10945 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10946 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10947 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10948 (aarch64_print_operand, aarch64_print_address_internal)
10949 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10950 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10951 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10952 Handle polynomial GET_MODE_SIZE.
10953 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10954 wider than SImode without modification.
10955 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10956 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10957 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10958 passing and returning SVE modes.
10959 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10960 rather than GEN_INT.
10961 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10962 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10963 (aarch64_allocate_and_probe_stack_space): Likewise.
10964 (aarch64_layout_frame): Cope with polynomial offsets.
10965 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10966 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10967 polynomial offsets.
10968 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10969 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10970 poly_int64 rather than a HOST_WIDE_INT.
10971 (aarch64_get_separate_components, aarch64_process_components)
10972 (aarch64_expand_prologue, aarch64_expand_epilogue)
10973 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10974 (aarch64_anchor_offset): New function, split out from...
10975 (aarch64_legitimize_address): ...here.
10976 (aarch64_builtin_vectorization_cost): Handle polynomial
10977 TYPE_VECTOR_SUBPARTS.
10978 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10980 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10981 number of elements from the PARALLEL rather than the mode.
10982 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10983 rather than GET_MODE_BITSIZE.
10984 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10985 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10986 (aarch64_expand_vec_perm_const_1): Handle polynomial
10987 d->perm.length () and d->perm elements.
10988 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10989 Apply to_constant to d->perm elements.
10990 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10991 polynomial CONST_VECTOR_NUNITS.
10992 (aarch64_move_pointer): Take amount as a poly_int64 rather
10994 (aarch64_progress_pointer): Avoid temporary variable.
10995 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10996 the mode attribute instead of GET_MODE.
10998 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10999 Alan Hayward <alan.hayward@arm.com>
11000 David Sherwood <david.sherwood@arm.com>
11002 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11003 x exists before using it.
11004 (aarch64_add_constant_internal): Rename to...
11005 (aarch64_add_offset_1): ...this. Replace regnum with separate
11006 src and dest rtxes. Handle the case in which they're different,
11007 including when the offset is zero. Replace scratchreg with an rtx.
11008 Use 2 additions if there is no spare register into which we can
11009 move a 16-bit constant.
11010 (aarch64_add_constant): Delete.
11011 (aarch64_add_offset): Replace reg with separate src and dest
11012 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11013 Use aarch64_add_offset_1.
11014 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11015 an rtx rather than an int. Take the delta as a poly_int64
11016 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11017 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11018 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11019 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11020 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11021 and aarch64_add_sp.
11022 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11023 aarch64_add_constant.
11025 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11027 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11028 Use scalar_float_mode.
11030 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11032 * config/aarch64/aarch64-simd.md
11033 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11034 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11035 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11036 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11037 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11038 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11039 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11040 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11041 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11042 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11044 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11047 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11048 targ_options->x_arm_arch_string is non NULL.
11050 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11052 * config/aarch64/aarch64.h
11053 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11055 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11058 * expmed.c (emit_store_flag_force): Swap if const op0
11059 and change VOIDmode to mode of op0.
11061 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11063 PR rtl-optimization/83761
11064 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11065 than bytes to mode_for_size.
11067 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11069 PR middle-end/83189
11070 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11071 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11074 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11076 PR middle-end/83575
11077 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11078 when in layout mode.
11079 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11080 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11083 2018-01-10 Michael Collison <michael.collison@arm.com>
11085 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11086 * config/aarch64/aarch64-option-extension.def: Add
11087 AARCH64_OPT_EXTENSION of 'fp16fml'.
11088 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11089 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11090 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11091 * config/aarch64/constraints.md (Ui7): New constraint.
11092 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11093 (VFMLA_SEL_W): Ditto.
11096 (VFMLA16_LOW): New int iterator.
11097 (VFMLA16_HIGH): Ditto.
11098 (UNSPEC_FMLAL): New unspec.
11099 (UNSPEC_FMLSL): Ditto.
11100 (UNSPEC_FMLAL2): Ditto.
11101 (UNSPEC_FMLSL2): Ditto.
11102 (f16mac): New code attribute.
11103 * config/aarch64/aarch64-simd-builtins.def
11104 (aarch64_fmlal_lowv2sf): Ditto.
11105 (aarch64_fmlsl_lowv2sf): Ditto.
11106 (aarch64_fmlalq_lowv4sf): Ditto.
11107 (aarch64_fmlslq_lowv4sf): Ditto.
11108 (aarch64_fmlal_highv2sf): Ditto.
11109 (aarch64_fmlsl_highv2sf): Ditto.
11110 (aarch64_fmlalq_highv4sf): Ditto.
11111 (aarch64_fmlslq_highv4sf): Ditto.
11112 (aarch64_fmlal_lane_lowv2sf): Ditto.
11113 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11114 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11115 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11116 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11117 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11118 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11119 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11120 (aarch64_fmlal_lane_highv2sf): Ditto.
11121 (aarch64_fmlsl_lane_highv2sf): Ditto.
11122 (aarch64_fmlal_laneq_highv2sf): Ditto.
11123 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11124 (aarch64_fmlalq_lane_highv4sf): Ditto.
11125 (aarch64_fmlsl_lane_highv4sf): Ditto.
11126 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11127 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11128 * config/aarch64/aarch64-simd.md:
11129 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11130 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11131 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11132 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11133 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11134 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11135 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11136 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11137 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11138 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11139 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11140 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11141 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11142 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11143 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11144 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11145 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11146 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11147 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11148 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11149 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11150 (vfmlsl_low_u32): Ditto.
11151 (vfmlalq_low_u32): Ditto.
11152 (vfmlslq_low_u32): Ditto.
11153 (vfmlal_high_u32): Ditto.
11154 (vfmlsl_high_u32): Ditto.
11155 (vfmlalq_high_u32): Ditto.
11156 (vfmlslq_high_u32): Ditto.
11157 (vfmlal_lane_low_u32): Ditto.
11158 (vfmlsl_lane_low_u32): Ditto.
11159 (vfmlal_laneq_low_u32): Ditto.
11160 (vfmlsl_laneq_low_u32): Ditto.
11161 (vfmlalq_lane_low_u32): Ditto.
11162 (vfmlslq_lane_low_u32): Ditto.
11163 (vfmlalq_laneq_low_u32): Ditto.
11164 (vfmlslq_laneq_low_u32): Ditto.
11165 (vfmlal_lane_high_u32): Ditto.
11166 (vfmlsl_lane_high_u32): Ditto.
11167 (vfmlal_laneq_high_u32): Ditto.
11168 (vfmlsl_laneq_high_u32): Ditto.
11169 (vfmlalq_lane_high_u32): Ditto.
11170 (vfmlslq_lane_high_u32): Ditto.
11171 (vfmlalq_laneq_high_u32): Ditto.
11172 (vfmlslq_laneq_high_u32): Ditto.
11173 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11174 (AARCH64_FL_FOR_ARCH8_4): New.
11175 (AARCH64_ISA_F16FML): New ISA flag.
11176 (TARGET_F16FML): New feature flag for fp16fml.
11177 (doc/invoke.texi): Document new fp16fml option.
11179 2018-01-10 Michael Collison <michael.collison@arm.com>
11181 * config/aarch64/aarch64-builtins.c:
11182 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11183 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11184 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11185 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11186 (AARCH64_ISA_SHA3): New ISA flag.
11187 (TARGET_SHA3): New feature flag for sha3.
11188 * config/aarch64/iterators.md (sha512_op): New int attribute.
11189 (CRYPTO_SHA512): New int iterator.
11190 (UNSPEC_SHA512H): New unspec.
11191 (UNSPEC_SHA512H2): Ditto.
11192 (UNSPEC_SHA512SU0): Ditto.
11193 (UNSPEC_SHA512SU1): Ditto.
11194 * config/aarch64/aarch64-simd-builtins.def
11195 (aarch64_crypto_sha512hqv2di): New builtin.
11196 (aarch64_crypto_sha512h2qv2di): Ditto.
11197 (aarch64_crypto_sha512su0qv2di): Ditto.
11198 (aarch64_crypto_sha512su1qv2di): Ditto.
11199 (aarch64_eor3qv8hi): Ditto.
11200 (aarch64_rax1qv2di): Ditto.
11201 (aarch64_xarqv2di): Ditto.
11202 (aarch64_bcaxqv8hi): Ditto.
11203 * config/aarch64/aarch64-simd.md:
11204 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11205 (aarch64_crypto_sha512su0qv2di): Ditto.
11206 (aarch64_crypto_sha512su1qv2di): Ditto.
11207 (aarch64_eor3qv8hi): Ditto.
11208 (aarch64_rax1qv2di): Ditto.
11209 (aarch64_xarqv2di): Ditto.
11210 (aarch64_bcaxqv8hi): Ditto.
11211 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11212 (vsha512h2q_u64): Ditto.
11213 (vsha512su0q_u64): Ditto.
11214 (vsha512su1q_u64): Ditto.
11215 (veor3q_u16): Ditto.
11216 (vrax1q_u64): Ditto.
11217 (vxarq_u64): Ditto.
11218 (vbcaxq_u16): Ditto.
11219 * config/arm/types.md (crypto_sha512): New type attribute.
11220 (crypto_sha3): Ditto.
11221 (doc/invoke.texi): Document new sha3 option.
11223 2018-01-10 Michael Collison <michael.collison@arm.com>
11225 * config/aarch64/aarch64-builtins.c:
11226 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11227 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11228 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11229 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11230 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11231 (AARCH64_ISA_SM4): New ISA flag.
11232 (TARGET_SM4): New feature flag for sm4.
11233 * config/aarch64/aarch64-simd-builtins.def
11234 (aarch64_sm3ss1qv4si): Ditto.
11235 (aarch64_sm3tt1aq4si): Ditto.
11236 (aarch64_sm3tt1bq4si): Ditto.
11237 (aarch64_sm3tt2aq4si): Ditto.
11238 (aarch64_sm3tt2bq4si): Ditto.
11239 (aarch64_sm3partw1qv4si): Ditto.
11240 (aarch64_sm3partw2qv4si): Ditto.
11241 (aarch64_sm4eqv4si): Ditto.
11242 (aarch64_sm4ekeyqv4si): Ditto.
11243 * config/aarch64/aarch64-simd.md:
11244 (aarch64_sm3ss1qv4si): Ditto.
11245 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11246 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11247 (aarch64_sm4eqv4si): Ditto.
11248 (aarch64_sm4ekeyqv4si): Ditto.
11249 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11250 (sm3part_op): Ditto.
11251 (CRYPTO_SM3TT): Ditto.
11252 (CRYPTO_SM3PART): Ditto.
11253 (UNSPEC_SM3SS1): New unspec.
11254 (UNSPEC_SM3TT1A): Ditto.
11255 (UNSPEC_SM3TT1B): Ditto.
11256 (UNSPEC_SM3TT2A): Ditto.
11257 (UNSPEC_SM3TT2B): Ditto.
11258 (UNSPEC_SM3PARTW1): Ditto.
11259 (UNSPEC_SM3PARTW2): Ditto.
11260 (UNSPEC_SM4E): Ditto.
11261 (UNSPEC_SM4EKEY): Ditto.
11262 * config/aarch64/constraints.md (Ui2): New constraint.
11263 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11264 * config/arm/types.md (crypto_sm3): New type attribute.
11265 (crypto_sm4): Ditto.
11266 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11267 (vsm3tt1aq_u32): Ditto.
11268 (vsm3tt1bq_u32): Ditto.
11269 (vsm3tt2aq_u32): Ditto.
11270 (vsm3tt2bq_u32): Ditto.
11271 (vsm3partw1q_u32): Ditto.
11272 (vsm3partw2q_u32): Ditto.
11273 (vsm4eq_u32): Ditto.
11274 (vsm4ekeyq_u32): Ditto.
11275 (doc/invoke.texi): Document new sm4 option.
11277 2018-01-10 Michael Collison <michael.collison@arm.com>
11279 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11280 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11281 (AARCH64_FL_FOR_ARCH8_4): New.
11282 (AARCH64_FL_V8_4): New flag.
11283 (doc/invoke.texi): Document new armv8.4-a option.
11285 2018-01-10 Michael Collison <michael.collison@arm.com>
11287 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11288 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11289 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11290 * config/aarch64/aarch64-option-extension.def: Add
11291 AARCH64_OPT_EXTENSION of 'sha2'.
11292 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11293 (crypto): Disable sha2 and aes if crypto disabled.
11294 (crypto): Enable aes and sha2 if enabled.
11295 (simd): Disable sha2 and aes if simd disabled.
11296 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11298 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11299 (TARGET_SHA2): New feature flag for sha2.
11300 (TARGET_AES): New feature flag for aes.
11301 * config/aarch64/aarch64-simd.md:
11302 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11303 conditional on TARGET_AES.
11304 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11305 (aarch64_crypto_sha1hsi): Make pattern conditional
11307 (aarch64_crypto_sha1hv4si): Ditto.
11308 (aarch64_be_crypto_sha1hv4si): Ditto.
11309 (aarch64_crypto_sha1su1v4si): Ditto.
11310 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11311 (aarch64_crypto_sha1su0v4si): Ditto.
11312 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11313 (aarch64_crypto_sha256su0v4si): Ditto.
11314 (aarch64_crypto_sha256su1v4si): Ditto.
11315 (doc/invoke.texi): Document new aes and sha2 options.
11317 2018-01-10 Martin Sebor <msebor@redhat.com>
11319 PR tree-optimization/83781
11320 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11323 2018-01-11 Martin Sebor <msebor@gmail.com>
11324 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11326 PR tree-optimization/83501
11327 PR tree-optimization/81703
11329 * tree-ssa-strlen.c (get_string_cst): Rename...
11330 (get_string_len): ...to this. Handle global constants.
11331 (handle_char_store): Adjust.
11333 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11334 Jim Wilson <jimw@sifive.com>
11336 * config/riscv/riscv-protos.h (riscv_output_return): New.
11337 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11338 (riscv_attribute_table, riscv_output_return),
11339 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11340 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11341 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11343 (riscv_expand_prologue): Add early return for naked function.
11344 (riscv_expand_epilogue): Likewise.
11345 (riscv_function_ok_for_sibcall): Return false for naked function.
11346 (riscv_set_current_function): New.
11347 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11348 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11349 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11350 * doc/extend.texi (RISC-V Function Attributes): New.
11352 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11354 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11355 check for 128-bit long double before checking TCmode.
11356 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11357 128-bit long doubles before checking TFmode or TCmode.
11358 (FLOAT128_IBM_P): Likewise.
11360 2018-01-10 Martin Sebor <msebor@redhat.com>
11362 PR tree-optimization/83671
11363 * builtins.c (c_strlen): Unconditionally return zero for the empty
11365 Use -Warray-bounds for warnings.
11366 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11367 for non-constant array indices with COMPONENT_REF, arrays of
11368 arrays, and pointers to arrays.
11369 (gimple_fold_builtin_strlen): Determine and set length range for
11370 non-constant character arrays.
11372 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11374 PR middle-end/81897
11375 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11378 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11380 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11382 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11385 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11386 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11387 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11388 indexed_or_indirect_operand predicate.
11389 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11390 (*vsx_le_perm_load_v8hi): Likewise.
11391 (*vsx_le_perm_load_v16qi): Likewise.
11392 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11393 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11394 (*vsx_le_perm_store_v8hi): Likewise.
11395 (*vsx_le_perm_store_v16qi): Likewise.
11396 (eight unnamed splitters): Likewise.
11398 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11400 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11401 * config/rs6000/emmintrin.h: Likewise.
11402 * config/rs6000/mmintrin.h: Likewise.
11403 * config/rs6000/xmmintrin.h: Likewise.
11405 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11408 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11410 * tree.c (tree_nop_conversion): Return true for location wrapper
11412 (maybe_wrap_with_location): New function.
11413 (selftest::check_strip_nops): New function.
11414 (selftest::test_location_wrappers): New function.
11415 (selftest::tree_c_tests): Call it.
11416 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11417 (maybe_wrap_with_location): New decl.
11418 (EXPR_LOCATION_WRAPPER_P): New macro.
11419 (location_wrapper_p): New inline function.
11420 (tree_strip_any_location_wrapper): New inline function.
11422 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11425 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11426 stack_realign_offset for the largest alignment of stack slot
11428 (ix86_find_max_used_stack_alignment): New function.
11429 (ix86_finalize_stack_frame_flags): Use it. Set
11430 max_used_stack_alignment if we don't realign stack.
11431 * config/i386/i386.h (machine_function): Add
11432 max_used_stack_alignment.
11434 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11436 * config/arm/arm.opt (-mbranch-cost): New option.
11437 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11440 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11443 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11444 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11446 2018-01-10 Richard Biener <rguenther@suse.de>
11449 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11450 early out so it also covers the case where we have a non-NULL
11453 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11455 PR tree-optimization/83753
11456 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11457 for non-strided grouped accesses if the number of elements is 1.
11459 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11462 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11463 * i386.h (TARGET_USE_GATHER): Define.
11464 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11466 2018-01-10 Martin Liska <mliska@suse.cz>
11469 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11470 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11472 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11473 CLEANUP_NO_PARTITIONING is not set.
11475 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11477 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11478 for vectors, as a partial revert of r254296.
11479 * rtl.h (const_vec_p): Delete.
11480 (const_vec_duplicate_p): Don't test for vector CONSTs.
11481 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11482 * expmed.c (make_tree): Likewise.
11485 * common.md (E, F): Use CONSTANT_P instead of checking for
11487 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11488 checking for CONST_VECTOR.
11490 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11492 PR middle-end/83575
11493 * predict.c (force_edge_cold): Handle in more sane way edges
11494 with no prediction.
11496 2018-01-09 Carl Love <cel@us.ibm.com>
11498 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11500 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11501 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11502 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11503 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11504 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11505 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11506 * config/rs6000/rs6000-protos.h: Add extern defition for
11507 rs6000_generate_float2_double_code.
11508 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11510 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11511 (float2_v2df): Add define_expand.
11513 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11516 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11517 op_mode in the force_to_mode call.
11519 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11521 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11522 instead of checking each element individually.
11523 (aarch64_evpc_uzp): Likewise.
11524 (aarch64_evpc_zip): Likewise.
11525 (aarch64_evpc_ext): Likewise.
11526 (aarch64_evpc_rev): Likewise.
11527 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11528 instead of checking each element individually. Return true without
11530 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11531 whether all selected elements come from the same input, instead of
11532 checking each element individually. Remove calls to gen_rtx_REG,
11533 start_sequence and end_sequence and instead assert that no rtl is
11536 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11538 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11539 order of HIGH and CONST checks.
11541 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11543 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11544 if the destination isn't an SSA_NAME.
11546 2018-01-09 Richard Biener <rguenther@suse.de>
11548 PR tree-optimization/83668
11549 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11551 (canonicalize_loop_form): ... here, renamed from ...
11552 (canonicalize_loop_closed_ssa_form): ... this and amended to
11553 swap successor edges for loop exit blocks to make us use
11554 the RPO order we need for initial schedule generation.
11556 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11558 PR tree-optimization/64811
11559 * match.pd: When optimizing comparisons with Inf, avoid
11560 introducing or losing exceptions from comparisons with NaN.
11562 2018-01-09 Martin Liska <mliska@suse.cz>
11565 * asan.c (shadow_mem_size): Add gcc_assert.
11567 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11569 Don't save registers in main().
11572 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11573 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11574 * config/avr/avr.c (avr_set_current_function): Don't error if
11575 naked, OS_task or OS_main are specified at the same time.
11576 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11578 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11580 * common/config/avr/avr-common.c (avr_option_optimization_table):
11581 Switch on -mmain-is-OS_task for optimizing compilations.
11583 2018-01-09 Richard Biener <rguenther@suse.de>
11585 PR tree-optimization/83572
11586 * graphite.c: Include cfganal.h.
11587 (graphite_transform_loops): Connect infinite loops to exit
11588 and remove fake edges at the end.
11590 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11592 * ipa-inline.c (edge_badness): Revert accidental checkin.
11594 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11597 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11598 symbols; not inline clones.
11600 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11603 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11604 hard registers. Formatting fixes.
11606 PR preprocessor/83722
11607 * gcc.c (try_generate_repro): Pass
11608 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11609 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11612 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11613 Kito Cheng <kito.cheng@gmail.com>
11615 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11616 (riscv_leaf_function_p): Delete.
11617 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11619 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11621 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11623 (do_ifelse): New function.
11624 (do_isel): New function.
11625 (do_sub3): New function.
11626 (do_add3): New function.
11627 (do_load_mask_compare): New function.
11628 (do_overlap_load_compare): New function.
11629 (expand_compare_loop): New function.
11630 (expand_block_compare): Call expand_compare_loop() when appropriate.
11631 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11632 option description.
11633 (-mblock-compare-inline-loop-limit): New option.
11635 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11638 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11639 Reverse order of second and third operands in first alternative.
11640 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11641 of first and second elements in UNSPEC_VPERMR vector.
11642 (altivec_expand_vec_perm_le): Likewise.
11644 2018-01-08 Jeff Law <law@redhat.com>
11646 PR rtl-optimizatin/81308
11647 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11648 (process_switch): If group_case_labels makes a change, then set
11650 (pass_convert_switch::execute): If a switch is converted, then
11651 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11653 PR rtl-optimization/81308
11654 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11657 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11659 PR target/83663 - Revert r255946
11660 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11661 generation for cases where splatting a value is not useful.
11662 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11663 across a vec_duplicate and a paradoxical subreg forming a vector
11664 mode to a vec_concat.
11666 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11668 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11669 -march=armv8.3-a variants.
11670 * config/arm/t-multilib: Likewise.
11671 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11673 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11675 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11677 (cceq_ior_compare_complement): Give it a name so I can use it, and
11678 change boolean_or_operator predicate to boolean_operator so it can
11679 be used to generate a crand.
11680 (eqne): New code iterator.
11681 (bd/bd_neg): New code_attrs.
11682 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11683 a single define_insn.
11684 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11685 decrement (bdnzt/bdnzf/bdzt/bdzf).
11686 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11687 with the new names of the branch decrement patterns, and added the
11688 names of the branch decrement conditional patterns.
11690 2018-01-08 Richard Biener <rguenther@suse.de>
11692 PR tree-optimization/83563
11693 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11696 2018-01-08 Richard Biener <rguenther@suse.de>
11698 PR middle-end/83713
11699 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11701 2018-01-08 Richard Biener <rguenther@suse.de>
11703 PR tree-optimization/83685
11704 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11705 references to abnormals.
11707 2018-01-08 Richard Biener <rguenther@suse.de>
11710 * dwarf2out.c (output_indirect_strings): Handle empty
11711 skeleton_debug_str_hash.
11712 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11714 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11716 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11717 (emit_store_direct): Likewise.
11718 (arc_trampoline_adjust_address): Likewise.
11719 (arc_asm_trampoline_template): New function.
11720 (arc_initialize_trampoline): Use asm_trampoline_template.
11721 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11722 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11723 * config/arc/arc.md (flush_icache): Delete pattern.
11725 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11727 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11728 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11731 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11734 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11735 by not USED_FOR_TARGET.
11736 (make_pass_resolve_sw_modes): Likewise.
11738 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11740 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11743 2018-01-08 Richard Biener <rguenther@suse.de>
11745 PR middle-end/83580
11746 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11748 2018-01-08 Richard Biener <rguenther@suse.de>
11750 PR middle-end/83517
11751 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11753 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11755 PR middle-end/81897
11756 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11757 basic blocks with a small number of successors.
11758 (convert_control_dep_chain_into_preds): Improve handling of
11760 (dump_predicates): Split apart into...
11761 (dump_pred_chain): ...here...
11762 (dump_pred_info): ...and here.
11763 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11764 (can_chain_union_be_invalidated_p): Improve check for invalidation
11766 (uninit_uses_cannot_happen): Avoid unnecessary if
11767 convert_control_dep_chain_into_preds yielded nothing.
11769 2018-01-06 Martin Sebor <msebor@redhat.com>
11771 PR tree-optimization/83640
11772 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11773 subtracting negative offset from size.
11774 (builtin_access::overlap): Adjust offset bounds of the access to fall
11775 within the size of the object if possible.
11777 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11779 PR rtl-optimization/83699
11780 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11781 extract_bit_field_as_subreg to cases in which the extracted
11782 value is also a vector.
11784 * lra-constraints.c (process_alt_operands): Test for the equivalence
11785 substitutions when detecting a possible reload cycle.
11787 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11790 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11791 by default if flag_selective_schedling{,2}. Formatting fixes.
11793 PR rtl-optimization/83682
11794 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11795 if it has non-VECTOR_MODE element mode.
11796 (vec_duplicate_p): Likewise.
11798 PR middle-end/83694
11799 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11800 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11802 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11805 * config/i386/i386-builtin.def
11806 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11807 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11808 Require also OPTION_MASK_ISA_AVX512F in addition to
11809 OPTION_MASK_ISA_GFNI.
11810 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11811 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11812 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11813 to OPTION_MASK_ISA_GFNI.
11814 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11815 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11816 OPTION_MASK_ISA_AVX512BW.
11817 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11818 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11819 addition to OPTION_MASK_ISA_GFNI.
11820 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11821 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11822 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11823 to OPTION_MASK_ISA_GFNI.
11824 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11825 a requirement for all ISAs rather than any of them with a few
11827 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11829 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11830 bitmasks to be enabled with 3 exceptions, instead of requiring any
11831 enabled ISA with lots of exceptions.
11832 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11833 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11834 Change avx512bw in isa attribute to avx512f.
11835 * config/i386/sgxintrin.h: Add license boilerplate.
11836 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11837 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11838 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11839 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11841 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11842 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11843 temporarily sse2 rather than sse if not enabled already.
11846 * config/i386/sse.md (VI248_VLBW): Rename to ...
11847 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11848 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11849 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11850 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11851 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11852 mode iterator instead of VI248_VLBW.
11854 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11856 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11857 (record_modified): Skip clobbers; add debug output.
11858 (param_change_prob): Use sreal frequencies.
11860 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11862 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11863 punt for user-aligned variables.
11865 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11867 * tree-chrec.c (chrec_contains_symbols): Return true for
11870 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11873 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11874 of (x|y) == x for BICS pattern.
11876 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11878 PR tree-optimization/83605
11879 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11880 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11883 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11885 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11886 * config/epiphany/rtems.h: New file.
11888 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11889 Uros Bizjak <ubizjak@gmail.com>
11892 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11893 QIreg_operand instead of register_operand predicate.
11894 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11895 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11896 comments instead of -fmitigate[-_]rop.
11898 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11901 * cgraphunit.c (symbol_table::compile): Switch to text_section
11902 before calling assembly_start debug hook.
11903 * run-rtl-passes.c (run_rtl_passes): Likewise.
11906 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11908 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11909 range_int_cst_p rather than !symbolic_range_p before calling
11910 extract_range_from_multiplicative_op_1.
11912 2018-01-04 Jeff Law <law@redhat.com>
11914 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11915 redundant test in assertion.
11917 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11919 * doc/rtl.texi: Document machine_mode wrapper classes.
11921 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11923 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11924 using tree_to_uhwi.
11926 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11928 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11929 the VEC_PERM_EXPR fold to fail.
11931 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11934 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11935 to switched_sections.
11937 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11940 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11941 test for d.testing.
11943 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11946 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11947 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11949 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11952 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11953 is BLKmode and bitpos not zero or mode change is needed.
11955 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11958 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11961 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11964 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11965 instead of MULT rtx. Update all corresponding splitters.
11966 (*saddl_se): Ditto.
11967 (*ssub<modesuffix>): Ditto.
11968 (*ssubl_se): Ditto.
11969 (*cmp_sadd_di): Update split patterns.
11970 (*cmp_sadd_si): Ditto.
11971 (*cmp_sadd_sidi): Ditto.
11972 (*cmp_ssub_di): Ditto.
11973 (*cmp_ssub_si): Ditto.
11974 (*cmp_ssub_sidi): Ditto.
11975 * config/alpha/predicates.md (const23_operand): New predicate.
11976 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11977 Look for ASHIFT, not MULT inner operand.
11978 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11980 2018-01-04 Martin Liska <mliska@suse.cz>
11982 PR gcov-profile/83669
11983 * gcov.c (output_intermediate_file): Add version to intermediate
11985 * doc/gcov.texi: Document new field 'version' in intermediate
11986 file format. Fix location of '-k' option of gcov command.
11988 2018-01-04 Martin Liska <mliska@suse.cz>
11991 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11993 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11995 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11997 2018-01-03 Martin Sebor <msebor@redhat.com>
11999 PR tree-optimization/83655
12000 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12001 checking calls with invalid arguments.
12003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12005 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12006 (vectorizable_mask_load_store): Delete.
12007 (vectorizable_call): Return false for masked loads and stores.
12008 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12009 instead of gimple_assign_rhs1.
12010 (vectorizable_load): Handle IFN_MASK_LOAD.
12011 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12015 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12017 (vectorizable_mask_load_store): ...here.
12018 (vectorizable_load): ...and here.
12020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12022 * tree-vect-stmts.c (vect_build_all_ones_mask)
12023 (vect_build_zero_merge_argument): New functions, split out from...
12024 (vectorizable_load): ...here.
12026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12028 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12030 (vectorizable_mask_load_store): ...here.
12031 (vectorizable_store): ...and here.
12033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12035 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12037 (vectorizable_mask_load_store): ...here.
12039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12041 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12042 (vect_model_store_cost): Take a vec_load_store_type instead of a
12044 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12045 (vect_model_store_cost): Take a vec_load_store_type instead of a
12047 (vectorizable_mask_load_store): Update accordingly.
12048 (vectorizable_store): Likewise.
12049 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12053 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12054 IFN_MASK_LOAD calls here rather than...
12055 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12058 Alan Hayward <alan.hayward@arm.com>
12059 David Sherwood <david.sherwood@arm.com>
12061 * expmed.c (extract_bit_field_1): For vector extracts,
12062 fall back to extract_bit_field_as_subreg if vec_extract
12065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12066 Alan Hayward <alan.hayward@arm.com>
12067 David Sherwood <david.sherwood@arm.com>
12069 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12070 they are variable or constant sized.
12071 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12072 slots for constant-sized data.
12074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12075 Alan Hayward <alan.hayward@arm.com>
12076 David Sherwood <david.sherwood@arm.com>
12078 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12079 handling COND_EXPRs with boolean comparisons, try to find a better
12080 basis for the mask type than the boolean itself.
12082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12084 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12085 is calculated and how it can be overridden.
12086 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12087 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12089 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12093 Alan Hayward <alan.hayward@arm.com>
12094 David Sherwood <david.sherwood@arm.com>
12096 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12097 Remove the mode argument.
12098 (aarch64_simd_valid_immediate): Remove the mode and inverse
12100 * config/aarch64/iterators.md (bitsize): New iterator.
12101 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12102 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12103 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12104 aarch64_simd_valid_immediate.
12105 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12106 (aarch64_reg_or_bic_imm): Likewise.
12107 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12108 with an insn_type enum and msl with a modifier_type enum.
12109 Replace element_width with a scalar_mode. Change the shift
12110 to unsigned int. Add constructors for scalar_float_mode and
12111 scalar_int_mode elements.
12112 (aarch64_vect_float_const_representable_p): Delete.
12113 (aarch64_can_const_movi_rtx_p)
12114 (aarch64_simd_scalar_immediate_valid_for_move)
12115 (aarch64_simd_make_constant): Update call to
12116 aarch64_simd_valid_immediate.
12117 (aarch64_advsimd_valid_immediate_hs): New function.
12118 (aarch64_advsimd_valid_immediate): Likewise.
12119 (aarch64_simd_valid_immediate): Remove mode and inverse
12120 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12121 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12122 and aarch64_float_const_representable_p on the result.
12123 (aarch64_output_simd_mov_immediate): Remove mode argument.
12124 Update call to aarch64_simd_valid_immediate and use of
12125 simd_immediate_info.
12126 (aarch64_output_scalar_simd_mov_immediate): Update call
12129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12130 Alan Hayward <alan.hayward@arm.com>
12131 David Sherwood <david.sherwood@arm.com>
12133 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12134 (mode_nunits): Likewise CONST_MODE_NUNITS.
12135 * machmode.def (ADJUST_NUNITS): Document.
12136 * genmodes.c (mode_data::need_nunits_adj): New field.
12137 (blank_mode): Update accordingly.
12138 (adj_nunits): New variable.
12139 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12141 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12142 listed in adj_nunits.
12143 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12144 listed in adj_nunits. Don't emit case statements for such modes.
12145 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12146 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12147 nothing if adj_nunits is nonnull.
12148 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12149 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12150 (emit_mode_fbit): Update use of print_maybe_const_decl.
12151 (emit_move_size): Likewise. Treat the array as non-const
12153 (emit_mode_adjustments): Handle adj_nunits.
12155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12157 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12158 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12159 (VECTOR_MODES): Use it.
12160 (make_vector_modes): Take the prefix as an argument.
12162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12163 Alan Hayward <alan.hayward@arm.com>
12164 David Sherwood <david.sherwood@arm.com>
12166 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12167 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12168 for MODE_VECTOR_BOOL.
12169 * machmode.def (VECTOR_BOOL_MODE): Document.
12170 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12171 (make_vector_bool_mode): New function.
12172 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12174 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12175 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12177 * stor-layout.c (int_mode_for_mode): Likewise.
12178 * tree.c (build_vector_type_for_mode): Likewise.
12179 * varasm.c (output_constant_pool_2): Likewise.
12180 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12181 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12182 for MODE_VECTOR_BOOL.
12183 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12184 of mode class checks.
12185 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12186 instead of a list of mode class checks.
12187 (expand_vector_scalar_condition): Likewise.
12188 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12191 Alan Hayward <alan.hayward@arm.com>
12192 David Sherwood <david.sherwood@arm.com>
12194 * machmode.h (mode_size): Change from unsigned short to
12196 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12197 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12198 or if measurement_type is not polynomial.
12199 (fixed_size_mode::includes_p): Check for constant-sized modes.
12200 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12201 return a poly_uint16 rather than an unsigned short.
12202 (emit_mode_size): Change the type of mode_size from unsigned short
12203 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12204 (emit_mode_adjustments): Cope with polynomial vector sizes.
12205 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12207 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12209 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12210 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12211 * caller-save.c (setup_save_areas): Likewise.
12212 (replace_reg_with_saved_mem): Likewise.
12213 * calls.c (emit_library_call_value_1): Likewise.
12214 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12215 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12216 (gen_lowpart_for_combine): Likewise.
12217 * convert.c (convert_to_integer_1): Likewise.
12218 * cse.c (equiv_constant, cse_insn): Likewise.
12219 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12220 (cselib_subst_to_values): Likewise.
12221 * dce.c (word_dce_process_block): Likewise.
12222 * df-problems.c (df_word_lr_mark_ref): Likewise.
12223 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12224 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12225 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12226 (rtl_for_decl_location): Likewise.
12227 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12228 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12229 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12230 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12231 (expand_expr_real_1): Likewise.
12232 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12233 (pad_below): Likewise.
12234 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12235 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12236 * ira.c (get_subreg_tracking_sizes): Likewise.
12237 * ira-build.c (ira_create_allocno_objects): Likewise.
12238 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12239 (ira_sort_regnos_for_alter_reg): Likewise.
12240 * ira-costs.c (record_operand_costs): Likewise.
12241 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12242 (resolve_simple_move): Likewise.
12243 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12244 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12245 (lra_constraints): Likewise.
12246 (CONST_POOL_OK_P): Reject variable-sized modes.
12247 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12248 (add_pseudo_to_slot, lra_spill): Likewise.
12249 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12250 * optabs-query.c (get_best_extraction_insn): Likewise.
12251 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12252 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12253 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12254 * recog.c (offsettable_address_addr_space_p): Likewise.
12255 * regcprop.c (maybe_mode_change): Likewise.
12256 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12257 * regrename.c (build_def_use): Likewise.
12258 * regstat.c (dump_reg_info): Likewise.
12259 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12260 (find_reloads, find_reloads_subreg_address): Likewise.
12261 * reload1.c (eliminate_regs_1): Likewise.
12262 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12263 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12264 (simplify_binary_operation_1, simplify_subreg): Likewise.
12265 * targhooks.c (default_function_arg_padding): Likewise.
12266 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12267 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12268 (verify_gimple_assign_ternary): Likewise.
12269 * tree-inline.c (estimate_move_cost): Likewise.
12270 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12271 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12272 (get_address_cost_ainc): Likewise.
12273 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12274 (vect_supportable_dr_alignment): Likewise.
12275 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12276 (vectorizable_reduction): Likewise.
12277 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12278 (vectorizable_operation, vectorizable_load): Likewise.
12279 * tree.c (build_same_sized_truth_vector_type): Likewise.
12280 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12281 * var-tracking.c (emit_note_insn_var_location): Likewise.
12282 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12283 (ADDR_VEC_ALIGN): Likewise.
12285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12286 Alan Hayward <alan.hayward@arm.com>
12287 David Sherwood <david.sherwood@arm.com>
12289 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12291 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12292 or if measurement_type is polynomial.
12293 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12294 * combine.c (make_extraction): Likewise.
12295 * dse.c (find_shift_sequence): Likewise.
12296 * dwarf2out.c (mem_loc_descriptor): Likewise.
12297 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12298 (extract_bit_field, extract_low_bits): Likewise.
12299 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12300 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12301 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12302 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12303 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12304 * reload.c (find_reloads): Likewise.
12305 * reload1.c (alter_reg): Likewise.
12306 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12307 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12308 * tree-if-conv.c (predicate_mem_writes): Likewise.
12309 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12310 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12311 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12312 * valtrack.c (dead_debug_insert_temp): Likewise.
12313 * varasm.c (mergeable_constant_section): Likewise.
12314 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12317 Alan Hayward <alan.hayward@arm.com>
12318 David Sherwood <david.sherwood@arm.com>
12320 * expr.c (expand_assignment): Cope with polynomial mode sizes
12321 when assigning to a CONCAT.
12323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12324 Alan Hayward <alan.hayward@arm.com>
12325 David Sherwood <david.sherwood@arm.com>
12327 * machmode.h (mode_precision): Change from unsigned short to
12329 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12331 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12332 or if measurement_type is not polynomial.
12333 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12334 in which the mode is already known to be a scalar_int_mode.
12335 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12336 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12338 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12339 for GET_MODE_PRECISION.
12340 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12341 for GET_MODE_PRECISION.
12342 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12344 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12345 (expand_field_assignment, make_extraction): Likewise.
12346 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12347 (get_last_value): Likewise.
12348 * convert.c (convert_to_integer_1): Likewise.
12349 * cse.c (cse_insn): Likewise.
12350 * expr.c (expand_expr_real_1): Likewise.
12351 * lra-constraints.c (simplify_operand_subreg): Likewise.
12352 * optabs-query.c (can_atomic_load_p): Likewise.
12353 * optabs.c (expand_atomic_load): Likewise.
12354 (expand_atomic_store): Likewise.
12355 * ree.c (combine_reaching_defs): Likewise.
12356 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12357 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12358 * tree.h (type_has_mode_precision_p): Likewise.
12359 * ubsan.c (instrument_si_overflow): Likewise.
12361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12362 Alan Hayward <alan.hayward@arm.com>
12363 David Sherwood <david.sherwood@arm.com>
12365 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12366 polynomial numbers of units.
12367 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12368 (valid_vector_subparts_p): New function.
12369 (build_vector_type): Remove temporary shim and take the number
12370 of units as a poly_uint64 rather than an int.
12371 (build_opaque_vector_type): Take the number of units as a
12372 poly_uint64 rather than an int.
12373 * tree.c (build_vector_from_ctor): Handle polynomial
12374 TYPE_VECTOR_SUBPARTS.
12375 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12376 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12377 (build_vector_from_val): If the number of units is variable,
12378 use build_vec_duplicate_cst for constant operands and
12379 VEC_DUPLICATE_EXPR otherwise.
12380 (make_vector_type): Remove temporary is_constant ().
12381 (build_vector_type, build_opaque_vector_type): Take the number of
12382 units as a poly_uint64 rather than an int.
12383 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12385 * cfgexpand.c (expand_debug_expr): Likewise.
12386 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12387 (store_constructor, expand_expr_real_1): Likewise.
12388 (const_scalar_mask_from_tree): Likewise.
12389 * fold-const-call.c (fold_const_reduction): Likewise.
12390 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12391 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12392 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12393 (fold_relational_const): Likewise.
12394 (native_interpret_vector): Likewise. Change the size from an
12395 int to an unsigned int.
12396 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12397 TYPE_VECTOR_SUBPARTS.
12398 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12399 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12400 duplicating a non-constant operand into a variable-length vector.
12401 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12402 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12403 * ipa-icf.c (sem_variable::equals): Likewise.
12404 * match.pd: Likewise.
12405 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12406 * print-tree.c (print_node): Likewise.
12407 * stor-layout.c (layout_type): Likewise.
12408 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12409 * tree-cfg.c (verify_gimple_comparison): Likewise.
12410 (verify_gimple_assign_binary): Likewise.
12411 (verify_gimple_assign_ternary): Likewise.
12412 (verify_gimple_assign_single): Likewise.
12413 * tree-pretty-print.c (dump_generic_node): Likewise.
12414 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12415 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12416 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12417 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12418 (vect_shift_permute_load_chain): Likewise.
12419 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12420 (expand_vector_condition, optimize_vector_constructor): Likewise.
12421 (lower_vec_perm, get_compute_type): Likewise.
12422 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12423 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12424 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12425 (vect_recog_mask_conversion_pattern): Likewise.
12426 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12427 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12428 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12429 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12430 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12431 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12432 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12433 (supportable_widening_operation): Likewise.
12434 (supportable_narrowing_operation): Likewise.
12435 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12437 * varasm.c (output_constant): Likewise.
12439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12440 Alan Hayward <alan.hayward@arm.com>
12441 David Sherwood <david.sherwood@arm.com>
12443 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12444 so that both the length == 3 and length != 3 cases set up their
12445 own permute vectors. Add comments explaining why we know the
12446 number of elements is constant.
12447 (vect_permute_load_chain): Likewise.
12449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12450 Alan Hayward <alan.hayward@arm.com>
12451 David Sherwood <david.sherwood@arm.com>
12453 * machmode.h (mode_nunits): Change from unsigned char to
12455 (ONLY_FIXED_SIZE_MODES): New macro.
12456 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12457 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12458 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12460 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12461 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12462 or if measurement_type is not polynomial.
12463 * genmodes.c (ZERO_COEFFS): New macro.
12464 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12466 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12467 Use ZERO_COEFFS when emitting initializers.
12468 * data-streamer.h (bp_pack_poly_value): New function.
12469 (bp_unpack_poly_value): Likewise.
12470 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12471 for GET_MODE_NUNITS.
12472 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12473 for GET_MODE_NUNITS.
12474 * tree.c (make_vector_type): Remove temporary shim and make
12475 the real function take the number of units as a poly_uint64
12476 rather than an int.
12477 (build_vector_type_for_mode): Handle polynomial nunits.
12478 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12479 * emit-rtl.c (const_vec_series_p_1): Likewise.
12480 (gen_rtx_CONST_VECTOR): Likewise.
12481 * fold-const.c (test_vec_duplicate_folding): Likewise.
12482 * genrecog.c (validate_pattern): Likewise.
12483 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12484 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12485 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12486 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12487 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12488 * rtlanal.c (subreg_get_info): Likewise.
12489 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12490 (vect_grouped_load_supported): Likewise.
12491 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12492 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12493 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12494 (simplify_const_unary_operation, simplify_binary_operation_1)
12495 (simplify_const_binary_operation, simplify_ternary_operation)
12496 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12497 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12498 instead of CONST_VECTOR_NUNITS.
12499 * varasm.c (output_constant_pool_2): Likewise.
12500 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12501 explicit-encoded elements in the XVEC for variable-length vectors.
12503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12505 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12508 Alan Hayward <alan.hayward@arm.com>
12509 David Sherwood <david.sherwood@arm.com>
12511 * coretypes.h (fixed_size_mode): Declare.
12512 (fixed_size_mode_pod): New typedef.
12513 * builtins.h (target_builtins::x_apply_args_mode)
12514 (target_builtins::x_apply_result_mode): Change type to
12515 fixed_size_mode_pod.
12516 * builtins.c (apply_args_size, apply_result_size, result_vector)
12517 (expand_builtin_apply_args_1, expand_builtin_apply)
12518 (expand_builtin_return): Update accordingly.
12520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12522 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12523 * cselib.c (cselib_hash_rtx): Likewise.
12524 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12525 CONST_VECTOR encoding.
12527 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12528 Jeff Law <law@redhat.com>
12531 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12532 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12533 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12534 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12537 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12538 explicitly probe *sp in a noreturn function if there were any callee
12539 register saves or frame pointer is needed.
12541 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12544 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12545 BLKmode for ternary, binary or unary expressions.
12548 * var-tracking.c (delete_vta_debug_insn): New inline function.
12549 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12550 insns from get_insns () to NULL instead of each bb separately.
12551 Use delete_vta_debug_insn. No longer static.
12552 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12553 delete_vta_debug_insns callers.
12554 * rtl.h (delete_vta_debug_insns): Declare.
12555 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12556 instead of variable_tracking_main.
12558 2018-01-03 Martin Sebor <msebor@redhat.com>
12560 PR tree-optimization/83603
12561 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12562 arguments past the endof the argument list in functions declared
12563 without a prototype.
12564 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12565 Avoid checking when arguments are null.
12567 2018-01-03 Martin Sebor <msebor@redhat.com>
12570 * doc/extend.texi (attribute const): Fix a typo.
12571 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12572 issuing -Wsuggest-attribute for void functions.
12574 2018-01-03 Martin Sebor <msebor@redhat.com>
12576 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12577 offset_int::from instead of wide_int::to_shwi.
12578 (maybe_diag_overlap): Remove assertion.
12579 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12580 * gimple-ssa-sprintf.c (format_directive): Same.
12581 (parse_directive): Same.
12582 (sprintf_dom_walker::compute_format_length): Same.
12583 (try_substitute_return_value): Same.
12585 2018-01-03 Jeff Law <law@redhat.com>
12587 PR middle-end/83654
12588 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12589 non-constant residual for zero at runtime and avoid probing in
12590 that case. Reorganize code for trailing problem to mirror handling
12593 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12595 PR tree-optimization/83501
12596 * tree-ssa-strlen.c (get_string_cst): New.
12597 (handle_char_store): Call get_string_cst.
12599 2018-01-03 Martin Liska <mliska@suse.cz>
12601 PR tree-optimization/83593
12602 * tree-ssa-strlen.c: Include tree-cfg.h.
12603 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12604 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12605 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12607 (strlen_dom_walker::before_dom_children): Call
12608 gimple_purge_dead_eh_edges. Dump tranformation with details
12610 (strlen_dom_walker::before_dom_children): Update call by adding
12611 new argument cleanup_eh.
12612 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12614 2018-01-03 Martin Liska <mliska@suse.cz>
12617 * cif-code.def (VARIADIC_THUNK): New enum value.
12618 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12621 2018-01-03 Jan Beulich <jbeulich@suse.com>
12623 * sse.md (mov<mode>_internal): Tighten condition for when to use
12624 vmovdqu<ssescalarsize> for TI and OI modes.
12626 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12628 Update copyright years.
12630 2018-01-03 Martin Liska <mliska@suse.cz>
12633 * ipa-visibility.c (function_and_variable_visibility): Skip
12634 functions with noipa attribure.
12636 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12638 * gcc.c (process_command): Update copyright notice dates.
12639 * gcov-dump.c (print_version): Ditto.
12640 * gcov.c (print_version): Ditto.
12641 * gcov-tool.c (print_version): Ditto.
12642 * gengtype.c (create_file): Ditto.
12643 * doc/cpp.texi: Bump @copying's copyright year.
12644 * doc/cppinternals.texi: Ditto.
12645 * doc/gcc.texi: Ditto.
12646 * doc/gccint.texi: Ditto.
12647 * doc/gcov.texi: Ditto.
12648 * doc/install.texi: Ditto.
12649 * doc/invoke.texi: Ditto.
12651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12653 * vector-builder.h (vector_builder::m_full_nelts): Change from
12654 unsigned int to poly_uint64.
12655 (vector_builder::full_nelts): Update prototype accordingly.
12656 (vector_builder::new_vector): Likewise.
12657 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12658 (vector_builder::operator ==): Likewise.
12659 (vector_builder::finalize): Likewise.
12660 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12661 Take the number of elements as a poly_uint64 rather than an
12663 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12664 from unsigned int to poly_uint64.
12665 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12666 (vec_perm_indices::new_vector): Likewise.
12667 (vec_perm_indices::length): Likewise.
12668 (vec_perm_indices::nelts_per_input): Likewise.
12669 (vec_perm_indices::input_nelts): Likewise.
12670 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12671 number of elements per input as a poly_uint64 rather than an
12672 unsigned int. Use the original encoding for variable-length
12673 vectors, rather than clamping each individual element.
12674 For the second and subsequent elements in each pattern,
12675 clamp the step and base before clamping their sum.
12676 (vec_perm_indices::series_p): Handle polynomial element counts.
12677 (vec_perm_indices::all_in_range_p): Likewise.
12678 (vec_perm_indices_to_tree): Likewise.
12679 (vec_perm_indices_to_rtx): Likewise.
12680 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12681 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12682 (tree_vector_builder::new_binary_operation): Handle polynomial
12683 element counts. Return false if we need to know the number
12684 of elements at compile time.
12685 * fold-const.c (fold_vec_perm): Punt if the number of elements
12686 isn't known at compile time.
12688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12690 * vec-perm-indices.h (vec_perm_builder): Change element type
12691 from HOST_WIDE_INT to poly_int64.
12692 (vec_perm_indices::element_type): Update accordingly.
12693 (vec_perm_indices::clamp): Handle polynomial element_types.
12694 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12695 (vec_perm_indices::all_in_range_p): Likewise.
12696 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12698 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12699 polynomial vec_perm_indices element types.
12700 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12701 * fold-const.c (fold_vec_perm): Likewise.
12702 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12703 * tree-vect-generic.c (lower_vec_perm): Likewise.
12704 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12705 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12706 element type to HOST_WIDE_INT.
12708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12709 Alan Hayward <alan.hayward@arm.com>
12710 David Sherwood <david.sherwood@arm.com>
12712 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12713 rather than an int. Use plus_constant.
12714 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12715 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12718 Alan Hayward <alan.hayward@arm.com>
12719 David Sherwood <david.sherwood@arm.com>
12721 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12722 a HOST_WIDE_INT to a poly_int64.
12724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12725 Alan Hayward <alan.hayward@arm.com>
12726 David Sherwood <david.sherwood@arm.com>
12728 * calls.c (load_register_parameters): Cope with polynomial
12729 mode sizes. Require a constant size for BLKmode parameters
12730 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12731 forces a parameter to be padded at the lsb end in order to
12732 fill a complete number of words, require the parameter size
12733 to be ordered wrt UNITS_PER_WORD.
12735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * reload1.c (spill_stack_slot_width): Change element type
12740 from unsigned int to poly_uint64_pod.
12741 (alter_reg): Treat mode sizes as polynomial.
12743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12744 Alan Hayward <alan.hayward@arm.com>
12745 David Sherwood <david.sherwood@arm.com>
12747 * reload.c (complex_word_subreg_p): New function.
12748 (reload_inner_reg_of_subreg, push_reload): Use it.
12750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12751 Alan Hayward <alan.hayward@arm.com>
12752 David Sherwood <david.sherwood@arm.com>
12754 * lra-constraints.c (process_alt_operands): Reject matched
12755 operands whose sizes aren't ordered.
12756 (match_reload): Refer to this check here.
12758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12759 Alan Hayward <alan.hayward@arm.com>
12760 David Sherwood <david.sherwood@arm.com>
12762 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12763 that the mode size is in the set {1, 2, 4, 8, 16}.
12765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12766 Alan Hayward <alan.hayward@arm.com>
12767 David Sherwood <david.sherwood@arm.com>
12769 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12770 Use plus_constant instead of gen_rtx_PLUS.
12772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12773 Alan Hayward <alan.hayward@arm.com>
12774 David Sherwood <david.sherwood@arm.com>
12776 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12777 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12778 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12779 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12780 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12781 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12782 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12783 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12784 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12785 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12787 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12788 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12789 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12790 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12791 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12792 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12793 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12794 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12795 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12796 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12798 * expr.c (emit_move_resolve_push): Treat the input and result
12799 of PUSH_ROUNDING as a poly_int64.
12800 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12801 (emit_push_insn): Likewise.
12802 * lra-eliminations.c (mark_not_eliminable): Likewise.
12803 * recog.c (push_operand): Likewise.
12804 * reload1.c (elimination_effects): Likewise.
12805 * rtlanal.c (nonzero_bits1): Likewise.
12806 * calls.c (store_one_arg): Likewise. Require the padding to be
12807 known at compile time.
12809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12810 Alan Hayward <alan.hayward@arm.com>
12811 David Sherwood <david.sherwood@arm.com>
12813 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12814 Use plus_constant instead of gen_rtx_PLUS.
12816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12817 Alan Hayward <alan.hayward@arm.com>
12818 David Sherwood <david.sherwood@arm.com>
12820 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12821 rather than an int.
12823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12824 Alan Hayward <alan.hayward@arm.com>
12825 David Sherwood <david.sherwood@arm.com>
12827 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12828 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12829 via stack temporaries. Treat the mode size as polynomial too.
12831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12832 Alan Hayward <alan.hayward@arm.com>
12833 David Sherwood <david.sherwood@arm.com>
12835 * expr.c (expand_expr_real_2): When handling conversions involving
12836 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12837 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12838 as a poly_uint64 too.
12840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12841 Alan Hayward <alan.hayward@arm.com>
12842 David Sherwood <david.sherwood@arm.com>
12844 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12847 Alan Hayward <alan.hayward@arm.com>
12848 David Sherwood <david.sherwood@arm.com>
12850 * combine.c (can_change_dest_mode): Handle polynomial
12851 REGMODE_NATURAL_SIZE.
12852 * expmed.c (store_bit_field_1): Likewise.
12853 * expr.c (store_constructor): Likewise.
12854 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12855 and polynomial REGMODE_NATURAL_SIZE.
12856 (gen_lowpart_common): Likewise.
12857 * reginfo.c (record_subregs_of_mode): Likewise.
12858 * rtlanal.c (read_modify_subreg_p): Likewise.
12860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12861 Alan Hayward <alan.hayward@arm.com>
12862 David Sherwood <david.sherwood@arm.com>
12864 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12865 numbers of elements.
12867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12868 Alan Hayward <alan.hayward@arm.com>
12869 David Sherwood <david.sherwood@arm.com>
12871 * match.pd: Cope with polynomial numbers of vector elements.
12873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12874 Alan Hayward <alan.hayward@arm.com>
12875 David Sherwood <david.sherwood@arm.com>
12877 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12878 in a POINTER_PLUS_EXPR.
12880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12881 Alan Hayward <alan.hayward@arm.com>
12882 David Sherwood <david.sherwood@arm.com>
12884 * omp-simd-clone.c (simd_clone_subparts): New function.
12885 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12886 (ipa_simd_modify_function_body): Likewise.
12888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12889 Alan Hayward <alan.hayward@arm.com>
12890 David Sherwood <david.sherwood@arm.com>
12892 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12893 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12894 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12895 (expand_vector_condition, vector_element): Likewise.
12896 (subparts_gt): New function.
12897 (get_compute_type): Use subparts_gt.
12898 (count_type_subparts): Delete.
12899 (expand_vector_operations_1): Use subparts_gt instead of
12900 count_type_subparts.
12902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12903 Alan Hayward <alan.hayward@arm.com>
12904 David Sherwood <david.sherwood@arm.com>
12906 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12907 (vect_compile_time_alias): ...this new function. Do the calculation
12908 on poly_ints rather than trees.
12909 (vect_prune_runtime_alias_test_list): Update call accordingly.
12911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12912 Alan Hayward <alan.hayward@arm.com>
12913 David Sherwood <david.sherwood@arm.com>
12915 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12917 (vect_schedule_slp_instance): Likewise.
12919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12920 Alan Hayward <alan.hayward@arm.com>
12921 David Sherwood <david.sherwood@arm.com>
12923 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12924 constant and extern definitions for variable-length vectors.
12925 (vect_get_constant_vectors): Note that the number of units
12926 is known to be constant.
12928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12929 Alan Hayward <alan.hayward@arm.com>
12930 David Sherwood <david.sherwood@arm.com>
12932 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12933 of units as polynomial. Choose between WIDE and NARROW based
12936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12937 Alan Hayward <alan.hayward@arm.com>
12938 David Sherwood <david.sherwood@arm.com>
12940 * tree-vect-stmts.c (simd_clone_subparts): New function.
12941 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12944 Alan Hayward <alan.hayward@arm.com>
12945 David Sherwood <david.sherwood@arm.com>
12947 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12948 vectors as polynomial. Use build_index_vector for
12949 IFN_GOMP_SIMD_LANE.
12951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12952 Alan Hayward <alan.hayward@arm.com>
12953 David Sherwood <david.sherwood@arm.com>
12955 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12956 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12957 for variable-length vectors.
12958 (vectorizable_mask_load_store): Treat the number of units as
12959 polynomial, asserting that it is constant if the condition has
12960 already been enforced.
12961 (vectorizable_store, vectorizable_load): Likewise.
12963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12964 Alan Hayward <alan.hayward@arm.com>
12965 David Sherwood <david.sherwood@arm.com>
12967 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12968 of units as polynomial. Punt if we can't tell at compile time
12969 which vector contains the final result.
12971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12972 Alan Hayward <alan.hayward@arm.com>
12973 David Sherwood <david.sherwood@arm.com>
12975 * tree-vect-loop.c (vectorizable_induction): Treat the number
12976 of units as polynomial. Punt on SLP inductions. Use an integer
12977 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12978 cast of such a series for variable-length floating-point
12981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12982 Alan Hayward <alan.hayward@arm.com>
12983 David Sherwood <david.sherwood@arm.com>
12985 * tree.h (build_index_vector): Declare.
12986 * tree.c (build_index_vector): New function.
12987 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12988 of units as polynomial, forcibly converting it to a constant if
12989 vectorizable_reduction has already enforced the condition.
12990 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12991 to create a {1,2,3,...} vector.
12992 (vectorizable_reduction): Treat the number of units as polynomial.
12993 Choose vectype_in based on the largest scalar element size rather
12994 than the smallest number of units. Enforce the restrictions
12997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12998 Alan Hayward <alan.hayward@arm.com>
12999 David Sherwood <david.sherwood@arm.com>
13001 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13002 number of units as polynomial.
13004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13005 Alan Hayward <alan.hayward@arm.com>
13006 David Sherwood <david.sherwood@arm.com>
13008 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13009 * target.def (autovectorize_vector_sizes): Return the vector sizes
13010 by pointer, using vector_sizes rather than a bitmask.
13011 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13012 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13013 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13015 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13016 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13017 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13018 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13019 * omp-general.c (omp_max_vf): Likewise.
13020 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13021 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13022 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13023 * tree-vect-slp.c (vect_slp_bb): Likewise.
13024 * doc/tm.texi: Regenerate.
13025 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13027 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13028 the vector size as a poly_uint64 rather than an unsigned int.
13029 (current_vector_size): Change from an unsigned int to a poly_uint64.
13030 (get_vectype_for_scalar_type): Update accordingly.
13031 * tree.h (build_truth_vector_type): Take the size and number of
13032 units as a poly_uint64 rather than an unsigned int.
13033 (build_vector_type): Add a temporary overload that takes
13034 the number of units as a poly_uint64 rather than an unsigned int.
13035 * tree.c (make_vector_type): Likewise.
13036 (build_truth_vector_type): Take the number of units as a poly_uint64
13037 rather than an unsigned int.
13039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13040 Alan Hayward <alan.hayward@arm.com>
13041 David Sherwood <david.sherwood@arm.com>
13043 * target.def (get_mask_mode): Take the number of units and length
13044 as poly_uint64s rather than unsigned ints.
13045 * targhooks.h (default_get_mask_mode): Update accordingly.
13046 * targhooks.c (default_get_mask_mode): Likewise.
13047 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13048 * doc/tm.texi: Regenerate.
13050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13051 Alan Hayward <alan.hayward@arm.com>
13052 David Sherwood <david.sherwood@arm.com>
13054 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13055 * omp-general.c (omp_max_vf): Likewise.
13056 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13057 (expand_omp_simd): Handle polynomial safelen.
13058 * omp-low.c (omplow_simd_context): Add a default constructor.
13059 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13060 (lower_rec_simd_input_clauses): Update accordingly.
13061 (lower_rec_input_clauses): Likewise.
13063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13064 Alan Hayward <alan.hayward@arm.com>
13065 David Sherwood <david.sherwood@arm.com>
13067 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13068 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13069 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13070 (vect_analyze_slp_cost): Likewise.
13071 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13072 (vect_model_load_cost): Likewise.
13074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13075 Alan Hayward <alan.hayward@arm.com>
13076 David Sherwood <david.sherwood@arm.com>
13078 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13079 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13080 from an unsigned int * to a poly_uint64_pod *.
13081 (calculate_unrolling_factor): New function.
13082 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13085 Alan Hayward <alan.hayward@arm.com>
13086 David Sherwood <david.sherwood@arm.com>
13088 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13089 from an unsigned int to a poly_uint64.
13090 (_loop_vec_info::slp_unrolling_factor): Likewise.
13091 (_loop_vec_info::vectorization_factor): Change from an int
13093 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13094 (vect_get_num_vectors): New function.
13095 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13096 (vect_get_num_copies): Use vect_get_num_vectors.
13097 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13098 to an unsigned int *.
13099 (vect_analyze_data_refs): Change min_vf from an int * to a
13101 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13102 than an unsigned HOST_WIDE_INT.
13103 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13104 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13105 to an unsigned int *.
13106 (vect_analyze_data_ref_dependences): Likewise.
13107 (vect_compute_data_ref_alignment): Handle polynomial vf.
13108 (vect_enhance_data_refs_alignment): Likewise.
13109 (vect_prune_runtime_alias_test_list): Likewise.
13110 (vect_shift_permute_load_chain): Likewise.
13111 (vect_supportable_dr_alignment): Likewise.
13112 (dependence_distance_ge_vf): Take the vectorization factor as a
13113 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13114 (vect_analyze_data_refs): Change min_vf from an int * to a
13116 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13117 vfm1 as a poly_uint64 rather than an int. Make the same change
13118 for the returned bound_scalar.
13119 (vect_gen_vector_loop_niters): Handle polynomial vf.
13120 (vect_do_peeling): Likewise. Update call to
13121 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13122 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13124 * tree-vect-loop.c (vect_determine_vectorization_factor)
13125 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13126 (vect_get_known_peeling_cost): Likewise.
13127 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13128 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13129 (vect_transform_loop): Likewise. Use the lowest possible VF when
13130 updating the upper bounds of the loop.
13131 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13132 rather than an int.
13133 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13134 polynomial unroll factors.
13135 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13136 (vect_make_slp_decision): Likewise.
13137 (vect_supported_load_permutation_p): Likewise, and polynomial
13139 (vect_analyze_slp_cost): Handle polynomial vf.
13140 (vect_slp_analyze_node_operations): Likewise.
13141 (vect_slp_analyze_bb_1): Likewise.
13142 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13143 than an unsigned HOST_WIDE_INT.
13144 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13145 (vectorizable_load): Handle polynomial vf.
13146 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13148 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13151 Alan Hayward <alan.hayward@arm.com>
13152 David Sherwood <david.sherwood@arm.com>
13154 * match.pd: Handle bit operations involving three constants
13155 and try to fold one pair.
13157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13159 * tree-vect-loop-manip.c: Include gimple-fold.h.
13160 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13161 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13162 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13163 Add a path that uses a step of VF instead of 1, but disable it
13165 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13166 and niters_no_overflow parameters. Update calls to
13167 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13168 Create a new SSA name if the latter choses to use a ste other
13169 than zero, and return it via niters_vector_mult_vf_var.
13170 * tree-vect-loop.c (vect_transform_loop): Update calls to
13171 vect_do_peeling, vect_gen_vector_loop_niters and
13172 slpeel_make_loop_iterate_ntimes.
13173 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13174 (vect_gen_vector_loop_niters): Update declarations after above changes.
13176 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13178 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13179 128-bit round to integer instructions.
13180 (ceil<mode>2): Likewise.
13181 (btrunc<mode>2): Likewise.
13182 (round<mode>2): Likewise.
13184 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13186 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13187 unaligned VSX load/store on P8/P9.
13188 (expand_block_clear): Allow the use of unaligned VSX
13189 load/store on P8/P9.
13191 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13193 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13195 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13196 swap associated with both a load and a store.
13198 2018-01-02 Andrew Waterman <andrew@sifive.com>
13200 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13201 * config/riscv/riscv.md (clear_cache): Use it.
13203 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13205 * web.c: Remove out-of-date comment.
13207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13209 * expr.c (fixup_args_size_notes): Check that any existing
13210 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13211 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13212 (emit_single_push_insn): ...here.
13214 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13216 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13217 (const_vector_encoded_nelts): New function.
13218 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13219 (const_vector_int_elt, const_vector_elt): Declare.
13220 * emit-rtl.c (const_vector_int_elt_1): New function.
13221 (const_vector_elt): Likewise.
13222 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13223 of CONST_VECTOR_ELT.
13225 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13227 * expr.c: Include rtx-vector-builder.h.
13228 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13229 directly on the tree encoding.
13230 (const_vector_from_tree): Likewise.
13231 * optabs.c: Include rtx-vector-builder.h.
13232 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13233 sequence of "u" values.
13234 * vec-perm-indices.c: Include rtx-vector-builder.h.
13235 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13236 directly on the vec_perm_indices encoding.
13238 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13240 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13241 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13242 * rtx-vector-builder.h: New file.
13243 * rtx-vector-builder.c: Likewise.
13244 * rtl.h (rtx_def::u2): Add a const_vector field.
13245 (CONST_VECTOR_NPATTERNS): New macro.
13246 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13247 (CONST_VECTOR_DUPLICATE_P): Likewise.
13248 (CONST_VECTOR_STEPPED_P): Likewise.
13249 (CONST_VECTOR_ENCODED_ELT): Likewise.
13250 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13251 (unwrap_const_vec_duplicate): Likewise.
13252 (const_vec_series_p): Check for a non-duplicated vector encoding.
13253 Say that the function only returns true for integer vectors.
13254 * emit-rtl.c: Include rtx-vector-builder.h.
13255 (gen_const_vec_duplicate_1): Delete.
13256 (gen_const_vector): Call gen_const_vec_duplicate instead of
13257 gen_const_vec_duplicate_1.
13258 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13259 (gen_const_vec_duplicate): Use rtx_vector_builder.
13260 (gen_const_vec_series): Likewise.
13261 (gen_rtx_CONST_VECTOR): Likewise.
13262 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13263 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13264 Build a new vector rather than modifying a CONST_VECTOR in-place.
13265 (handle_special_swappables): Update call accordingly.
13266 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13267 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13268 Build a new vector rather than modifying a CONST_VECTOR in-place.
13269 (handle_special_swappables): Update call accordingly.
13271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13273 * simplify-rtx.c (simplify_const_binary_operation): Use
13274 CONST_VECTOR_ELT instead of XVECEXP.
13276 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13278 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13279 the selector elements to be different from the data elements
13280 if the selector is a VECTOR_CST.
13281 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13282 ssizetype for the selector.
13284 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13286 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13287 before testing each element individually.
13288 * tree-vect-generic.c (lower_vec_perm): Likewise.
13290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13292 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13293 * selftest-run-tests.c (selftest::run_tests): Call it.
13294 * vector-builder.h (vector_builder::operator ==): New function.
13295 (vector_builder::operator !=): Likewise.
13296 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13297 (vec_perm_indices::all_from_input_p): New function.
13298 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13299 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13300 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13301 instead of reading the VECTOR_CST directly. Detect whether both
13302 vector inputs are the same before constructing the vec_perm_indices,
13303 and update the number of inputs argument accordingly. Use the
13304 utility functions added above. Only construct sel2 if we need to.
13306 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13308 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13309 the broadcast of the low byte.
13310 (expand_mult_highpart): Use an explicit encoding for the permutes.
13311 * optabs-query.c (can_mult_highpart_p): Likewise.
13312 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13313 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13314 (vectorizable_bswap): Likewise.
13315 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13316 explicit encoding for the power-of-2 permutes.
13317 (vect_permute_store_chain): Likewise.
13318 (vect_grouped_load_supported): Likewise.
13319 (vect_permute_load_chain): Likewise.
13321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13323 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13324 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13325 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13326 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13327 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13328 (vect_gen_perm_mask_any): Likewise.
13330 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13332 * int-vector-builder.h: New file.
13333 * vec-perm-indices.h: Include int-vector-builder.h.
13334 (vec_perm_indices): Redefine as an int_vector_builder.
13335 (auto_vec_perm_indices): Delete.
13336 (vec_perm_builder): Redefine as a stand-alone class.
13337 (vec_perm_indices::vec_perm_indices): New function.
13338 (vec_perm_indices::clamp): Likewise.
13339 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13340 (vec_perm_indices::new_vector): New function.
13341 (vec_perm_indices::new_expanded_vector): Update for new
13342 vec_perm_indices class.
13343 (vec_perm_indices::rotate_inputs): New function.
13344 (vec_perm_indices::all_in_range_p): Operate directly on the
13345 encoded form, without computing elided elements.
13346 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13347 encoding. Update for new vec_perm_indices class.
13348 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13349 the given vec_perm_builder.
13350 (expand_vec_perm_var): Update vec_perm_builder constructor.
13351 (expand_mult_highpart): Use vec_perm_builder instead of
13352 auto_vec_perm_indices.
13353 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13354 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13355 or double series encoding as appropriate.
13356 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13357 vec_perm_indices instead of auto_vec_perm_indices.
13358 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13359 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13360 (vect_permute_store_chain): Likewise.
13361 (vect_grouped_load_supported): Likewise.
13362 (vect_permute_load_chain): Likewise.
13363 (vect_shift_permute_load_chain): Likewise.
13364 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13365 (vect_transform_slp_perm_load): Likewise.
13366 (vect_schedule_slp_instance): Likewise.
13367 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13368 (vectorizable_mask_load_store): Likewise.
13369 (vectorizable_bswap): Likewise.
13370 (vectorizable_store): Likewise.
13371 (vectorizable_load): Likewise.
13372 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13373 vec_perm_indices instead of auto_vec_perm_indices. Use
13374 tree_to_vec_perm_builder to read the vector from a tree.
13375 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13376 vec_perm_builder instead of a vec_perm_indices.
13377 (have_whole_vector_shift): Use vec_perm_builder and
13378 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13379 truncation to calc_vec_perm_mask_for_shift.
13380 (vect_create_epilog_for_reduction): Likewise.
13381 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13382 from auto_vec_perm_indices to vec_perm_indices.
13383 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13384 instead of changing individual elements.
13385 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13386 the vector in d.perm.
13387 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13388 from auto_vec_perm_indices to vec_perm_indices.
13389 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13390 instead of changing individual elements.
13391 (arm_vectorize_vec_perm_const): Use new_vector to install
13392 the vector in d.perm.
13393 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13394 Update vec_perm_builder constructor.
13395 (rs6000_expand_interleave): Likewise.
13396 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13397 (rs6000_expand_interleave): Likewise.
13399 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13401 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13402 to qimode could truncate the indices.
13403 * optabs.c (expand_vec_perm_var): Likewise.
13405 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13407 * Makefile.in (OBJS): Add vec-perm-indices.o.
13408 * vec-perm-indices.h: New file.
13409 * vec-perm-indices.c: Likewise.
13410 * target.h (vec_perm_indices): Replace with a forward class
13412 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13413 * optabs.h: Include vec-perm-indices.h.
13414 (expand_vec_perm): Delete.
13415 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13416 (expand_vec_perm_const): Declare.
13417 * target.def (vec_perm_const_ok): Replace with...
13418 (vec_perm_const): ...this new hook.
13419 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13420 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13421 * doc/tm.texi: Regenerate.
13422 * optabs.def (vec_perm_const): Delete.
13423 * doc/md.texi (vec_perm_const): Likewise.
13424 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13425 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13426 expand_vec_perm for constant permutation vectors. Assert that
13427 the mode of variable permutation vectors is the integer equivalent
13428 of the mode that is being permuted.
13429 * optabs-query.h (selector_fits_mode_p): Declare.
13430 * optabs-query.c: Include vec-perm-indices.h.
13431 (selector_fits_mode_p): New function.
13432 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13433 is defined, instead of checking whether the vec_perm_const_optab
13434 exists. Use targetm.vectorize.vec_perm_const instead of
13435 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13436 fit in the vector mode before using a variable permute.
13437 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13438 vec_perm_indices instead of an rtx.
13439 (expand_vec_perm): Replace with...
13440 (expand_vec_perm_const): ...this new function. Take the selector
13441 as a vec_perm_indices rather than an rtx. Also take the mode of
13442 the selector. Update call to shift_amt_for_vec_perm_mask.
13443 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13444 Use vec_perm_indices::new_expanded_vector to expand the original
13445 selector into bytes. Check whether the indices fit in the vector
13446 mode before using a variable permute.
13447 (expand_vec_perm_var): Make global.
13448 (expand_mult_highpart): Use expand_vec_perm_const.
13449 * fold-const.c: Includes vec-perm-indices.h.
13450 * tree-ssa-forwprop.c: Likewise.
13451 * tree-vect-data-refs.c: Likewise.
13452 * tree-vect-generic.c: Likewise.
13453 * tree-vect-loop.c: Likewise.
13454 * tree-vect-slp.c: Likewise.
13455 * tree-vect-stmts.c: Likewise.
13456 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13458 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13459 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13460 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13461 (aarch64_vectorize_vec_perm_const): ...this new function.
13462 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13463 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13464 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13465 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13466 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13467 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13468 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13470 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13471 check for NEON modes.
13472 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13473 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13474 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13475 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13477 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13478 the old VEC_PERM_CONST conditions.
13479 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13480 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13481 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13482 (ia64_vectorize_vec_perm_const_ok): Merge into...
13483 (ia64_vectorize_vec_perm_const): ...this new function.
13484 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13485 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13486 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13487 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13488 * config/mips/mips.c (mips_expand_vec_perm_const)
13489 (mips_vectorize_vec_perm_const_ok): Merge into...
13490 (mips_vectorize_vec_perm_const): ...this new function.
13491 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13492 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13493 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13494 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13495 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13496 (rs6000_expand_vec_perm_const): Delete.
13497 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13499 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13500 (altivec_expand_vec_perm_const_le): Take each operand individually.
13501 Operate on constant selectors rather than rtxes.
13502 (altivec_expand_vec_perm_const): Likewise. Update call to
13503 altivec_expand_vec_perm_const_le.
13504 (rs6000_expand_vec_perm_const): Delete.
13505 (rs6000_vectorize_vec_perm_const_ok): Delete.
13506 (rs6000_vectorize_vec_perm_const): New function.
13507 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13508 an element count and rtx array.
13509 (rs6000_expand_extract_even): Update call accordingly.
13510 (rs6000_expand_interleave): Likewise.
13511 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13512 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13513 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13514 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13515 (rs6000_expand_vec_perm_const): Delete.
13516 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13517 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13518 (altivec_expand_vec_perm_const_le): Take each operand individually.
13519 Operate on constant selectors rather than rtxes.
13520 (altivec_expand_vec_perm_const): Likewise. Update call to
13521 altivec_expand_vec_perm_const_le.
13522 (rs6000_expand_vec_perm_const): Delete.
13523 (rs6000_vectorize_vec_perm_const_ok): Delete.
13524 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13525 reference to the SPE evmerge intructions.
13526 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13527 an element count and rtx array.
13528 (rs6000_expand_extract_even): Update call accordingly.
13529 (rs6000_expand_interleave): Likewise.
13530 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13531 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13533 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13537 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13538 vector mode and that that mode matches the mode of the data
13540 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13541 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13542 directly using expand_vec_perm_1 when forcing selectors into
13544 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13546 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13548 * optabs-query.h (can_vec_perm_p): Delete.
13549 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13550 * optabs-query.c (can_vec_perm_p): Split into...
13551 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13552 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13553 particular selector is valid.
13554 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13555 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13556 (vect_grouped_load_supported): Likewise.
13557 (vect_shift_permute_load_chain): Likewise.
13558 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13559 (vect_transform_slp_perm_load): Likewise.
13560 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13561 (vectorizable_bswap): Likewise.
13562 (vect_gen_perm_mask_checked): Likewise.
13563 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13564 implementations of variable permutation vectors into account
13565 when deciding which selector to use.
13566 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13567 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13568 with a false third argument.
13569 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13570 to test whether the constant selector is valid and can_vec_perm_var_p
13571 to test whether a variable selector is valid.
13573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13575 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13576 * optabs-query.c (can_vec_perm_p): Likewise.
13577 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13578 instead of vec_perm_indices.
13579 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13580 (vect_gen_perm_mask_checked): Likewise,
13581 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13582 (vect_gen_perm_mask_checked): Likewise,
13584 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13586 * optabs-query.h (qimode_for_vec_perm): Declare.
13587 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13588 (qimode_for_vec_perm): ...this new function.
13589 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13591 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13593 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13594 does not have a conditional at the top.
13596 2018-01-02 Richard Biener <rguenther@suse.de>
13598 * ipa-inline.c (big_speedup_p): Fix expression.
13600 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13603 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13606 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13610 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13611 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13612 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13613 cond_taken_branch_cost 3->4.
13615 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13617 PR tree-optimization/83581
13618 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13619 TODO_cleanup_cfg if any changes have been made.
13621 PR middle-end/83608
13622 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13623 convert_modes if target mode has the right side, but different mode
13626 PR middle-end/83609
13627 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13628 last argument when extracting from CONCAT. If either from_real or
13629 from_imag is NULL, use expansion through memory. If result is not
13630 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13631 the parts directly to inner mode, if even that fails, use expansion
13634 PR middle-end/83623
13635 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13636 check for bswap in mode rather than HImode and use that in expand_unop
13639 Copyright (C) 2018 Free Software Foundation, Inc.
13641 Copying and distribution of this file, with or without modification,
13642 are permitted in any medium without royalty provided the copyright
13643 notice and this notice are preserved.