1 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/constraints.md (S): New constraint.
4 * config/nds32/nds32.md (call_internal): Use constraint S.
5 (call_value_internal): Likewise.
6 (sibcall_internal): Likewise.
7 (sibcall_value_internal): Likewise.
9 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10 Chung-Ju Wu <jasonwucj@gmail.com>
12 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
16 Chung-Ju Wu <jasonwucj@gmail.com>
18 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
19 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
20 (nds32_rtx_costs_impl): Simplify.
21 (nds32_address_cost_impl): Simplify.
22 (nds32_init_rtx_costs): New function.
23 (nds32_rtx_costs_speed_prefer): Likewise.
24 (nds32_rtx_costs_size_prefer): Likewise.
25 (nds32_address_cost_speed_prefer): Likewise.
26 (nds32_address_cost_speed_fwprop): Likewise.
27 (nds32_address_cost_size_prefer): Likewise.
28 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
29 * config/nds32/nds32.c (nds32_option_override): Use
30 nds32_init_rtx_costs function.
32 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
34 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
35 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
36 (TARGET_PIPELINE_N8): Likewise.
37 (TARGET_PIPELINE_N10): Likewise.
38 (TARGET_PIPELINE_N13): Likewise.
39 (TARGET_PIPELINE_GRAYWOLF): Likewise.
41 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
43 * config/nds32/nds32-fpu.md: Update copyright year.
45 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
47 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
49 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
51 * config/nds32/nds32.c
52 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
53 * config/nds32/nds32.opt (minline-asm-r15): New option.
55 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
57 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
59 * config/nds32/nds32.md (abssi2): New pattern.
61 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
63 * config/i386/i386.md (rex64namesuffix): New mode attribute.
64 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
65 Merge insn pattern from sse_cvtsi2ss<round_name> and
66 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
67 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
68 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
69 using SWI48 mode iterator.
70 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
71 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
72 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
73 pattern from sse_cvttss2si<round_saeonly_name>
74 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
75 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
76 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
77 using SWI48 mode iterator.
78 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
79 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
80 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
81 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
82 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
83 using SWI48 mode iterator.
84 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
85 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
86 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
87 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
88 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
90 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
91 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
92 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
93 pattern from sse_cvttsd2si<round_saeonly_name>
94 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
96 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
98 * config/nds32/nds32-md-auxiliary.c
99 (nds32_valid_smw_lwm_base_p): Refine.
100 (nds32_output_smw_single_word): Refine.
101 (nds32_output_smw_double_word): New.
102 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
104 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
106 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
107 (nds32_output_stack_pop): Refine.
108 (nds32_expand_unaligned_load): Refine.
109 (nds32_expand_unaligned_store): Refine.
111 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
112 Chung-Ju Wu <jasonwucj@gmail.com>
114 * config/nds32/constants.md: Add TP_REGNUM constant.
115 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
116 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
118 * config/nds32/nds32-doubleword.md: Consider flag_pic.
119 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
120 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
121 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
122 and PIC code generation.
123 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
125 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
127 * config/nds32/nds32.md: Support TLS and PIC.
128 * config/nds32/nds32.c: Support TLS and PIC.
129 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
130 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
133 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
135 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
138 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
139 Chung-Ju Wu <jasonwucj@gmail.com>
141 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
142 * config/nds32/nds32-md-auxiliary.c
143 (symbolic_reference_mentioned_p): New.
144 (nds32_legitimize_ict_address): New.
145 (nds32_expand_ict_move): New.
146 (nds32_indirect_call_referenced_p): New.
147 (nds32_symbol_binds_local_p): Delete.
148 (nds32_long_call_p): Modify.
149 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
150 * config/nds32/nds32-protos.h
151 (symbolic_reference_mentioned_p): Declare.
152 (nds32_legitimize_ict_address): Declare.
153 (nds32_expand_ict_move): Declare.
154 (nds32_indirect_call_referenced_p): Declare.
155 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
156 (nds32_relax_group): Use nds32_ict_const_p as condition.
157 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
158 (nds32_asm_file_start): Output ict_model directive in asm code.
159 (nds32_legitimate_address_p): Consider indirect call.
160 (nds32_print_operand): Consider indirect call.
161 (nds32_print_operand_address): Consider indirect call.
162 (nds32_insert_attributes): Handle "indirect_call" attribute.
163 (TARGET_LEGITIMATE_ADDRESS_P): Define.
164 (TARGET_LEGITIMATE_CONSTANT_P): Define.
165 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
166 (TARGET_DELEGITIMIZE_ADDRESS): Define.
167 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
168 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
169 (TARGET_ICT_MODEL_SMALL): Define.
170 (TARGET_ICT_MODEL_LARGE): Define.
171 * config/nds32/nds32.md (movsi): Consider ict model.
172 (call, call_value): Consider ict model.
173 (sibcall, sibcall_value): Consider ict model.
174 * config/nds32/nds32.opt (mict-model): New option.
175 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
178 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
179 Monk Chiang <sh.chiang04@gmail.com>
180 Jim Wilson <jimw@sifive.com>
182 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
183 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
184 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
185 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
186 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
187 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
188 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
189 compute save_libcall_adjustment properly.
190 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
191 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
192 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
193 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
194 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
195 (ABI_SPEC): Handle mabi=ilp32e.
196 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
198 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
199 <-march>: Add rv32e as an example.
201 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
204 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
205 (intra_create_variable_infos): Handle C++ constructors.
207 2018-05-18 Martin Liska <mliska@suse.cz>
209 * passes.def: Remove a redundant pass.
211 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
214 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
216 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
218 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
220 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
222 (armv2, armv2a, armv3, armv3m): Delete architectures.
223 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
224 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
225 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
227 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
228 (*mulsidi3adddi): Likewise.
229 (mulsidi3): Likewise.
230 (*mulsidi3_nov6): Likewise.
231 (umulsidi3): Likewise.
232 (umulsidi3_nov6): Likewise.
233 (umaddsidi4): Likewise.
234 (*umulsidi3adddi): Likewise.
235 (smulsi3_highpart): Likewise.
236 (*smulsi3_highpart_nov6): Likewise.
237 (umulsi3_highpart): Likewise.
238 (*umulsi3_highpart_nov6): Likewise.
239 * config/arm/arm.h (arm_arch3m): Delete.
240 * config/arm/arm.c (arm_arch3m): Delete.
241 (arm_option_override_internal): Update armv3-related comment.
242 (arm_configure_build_target): Delete use of isa_bit_mode32.
243 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
244 (arm_rtx_costs_internal): Delete check of arm_arch3m.
245 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
248 * config/arm/arm-protos.h (arm_arch3m): Delete.
249 * config/arm/arm-tables.opt: Regenerate.
250 * config/arm/arm-tune.md: Likewise.
251 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
252 deleted architectures.
254 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
256 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
257 (armv5t, armv5te): New features.
258 (ARMv5, ARMv5e): Delete fgroups.
259 (ARMv5t, ARMv5te): Adjust for above changes.
261 (armv5, armv5e): Delete arches.
262 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
264 (*call_reg_arm): Likewise.
265 (*call_value_reg_armv5): Likewise.
266 (*call_value_reg_arm): Likewise.
267 (*call_symbol): Likewise.
268 (*call_value_symbol): Likewise.
269 (*sibcall_insn): Likewise.
270 (*sibcall_value_insn): Likewise.
272 (prefetch): Likewise.
273 (define_split and define_peephole2 dependent on arm_arch5):
275 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
277 (TARGET_ARM_QBIT): Likewise.
278 (TARGET_DSP_MULTIPLY): Likewise.
279 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
280 (arm_arch5, arm_arch5e): Delete.
281 (arm_arch5t, arm_arch5te): Declare.
282 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
283 (arm_arch5t): Declare.
284 (arm_option_reconfigure_globals): Update for the above.
285 (arm_options_perform_arch_sanity_checks): Update comment, replace
286 use of arm_arch5 with arm_arch5t.
287 (use_return_insn): Likewise.
288 (arm_emit_call_insn): Likewise.
289 (output_return_instruction): Likewise.
290 (arm_final_prescan_insn): Likewise.
291 (arm_coproc_builtin_available): Likewise.
292 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
293 arm_arch5e with arm_arch5t and arm_arch5te.
294 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
295 (arm_arch5t, arm_arch5te): Declare.
296 * config/arm/arm-tables.opt: Regenerate.
297 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
298 * config/arm/t-multilib: Likewise.
299 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
300 instead of arm_arch5.
301 (*call_reg_thumb1): Likewise.
302 (*call_value_reg_thumb1_v5): Likewise.
303 (*call_value_reg_thumb1): Likewise.
304 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
306 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
308 2018-05-18 Martin Liska <mliska@suse.cz>
310 PR gcov-profile/84846
311 * doc/gcov.texi: Document -t option of gcov tool.
313 2018-05-18 Martin Liska <mliska@suse.cz>
315 PR gcov-profile/84846
316 * gcov.c (print_usage): Add new -t option.
317 (process_args): Handle the option.
318 (generate_results): Use stdout as output when requested by
321 2018-05-18 Martin Liska <mliska@suse.cz>
323 PR gcov-profile/84846
324 * coverage.c (coverage_init): Write PWD to .gcno file.
325 * doc/gcov.texi: Document how working directory is printed.
326 * gcov-dump.c (dump_gcov_file): Print PWD.
327 * gcov.c (output_intermediate_file): Likewise.
328 (read_graph_file): Read PWD string.
329 (output_lines): Print PWD.
331 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
334 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
335 for retval and return false if all args to phi are zero.
337 2018-05-18 Richard Biener <rguenther@suse.de>
339 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
341 (evrp_dom_walker::before_dom_children): Call it.
343 2018-05-18 Richard Biener <rguenther@suse.de>
345 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
346 results when processing array refs with variable index.
348 2018-05-18 Toon Moene <toon@moene.org>
350 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
351 directly after that of -floop-interchange. Indicate that both
352 options are enabled by default when specifying -O3.
354 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
356 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
357 iterator. Delete separate integer-mode vec_set<mode> expander.
358 (aarch64_simd_vec_setv2di): Delete.
359 (vec_setv2di): Delete.
360 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
361 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
362 the "w, r" alternative.
364 2018-05-18 Martin Liska <mliska@suse.cz>
366 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
367 * tree-pass.h (make_pass_lower_switch_O0): New function.
368 * tree-switch-conversion.c (node_has_low_bound): Remove.
369 (node_has_high_bound): Likewise.
370 (node_is_bounded): Likewise.
371 (class pass_lower_switch): Make it a template type and create
373 (pass_lower_switch::execute): Add template argument.
374 (make_pass_lower_switch): New function.
375 (make_pass_lower_switch_O0): New function.
376 (do_jump_if_equal): Remove.
377 (emit_case_nodes): Simplify to just handle all 3 cases and leave
378 all the hard work to tree optimization passes.
380 2018-05-18 Martin Liska <mliska@suse.cz>
382 * dbgcnt.c (limit_low): Renamed from limit.
383 (limit_high): New variable.
384 (dbg_cnt_is_enabled): Check for upper limit.
385 (dbg_cnt): Adjust dumping.
386 (dbg_cnt_set_limit_by_index): Add new argument for high
388 (dbg_cnt_set_limit_by_name): Likewise.
389 (dbg_cnt_process_single_pair): Parse new format.
390 (dbg_cnt_process_opt): Use strtok.
391 (dbg_cnt_list_all_counters): Remove 'value' and add
393 * doc/invoke.texi: Document changes.
395 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
397 * doc/sourcebuild.texi (scalar_all_fma): Document.
398 * tree.def (FMA_EXPR): Delete.
399 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
400 * internal-fn.c (ternary_direct): New macro.
401 (expand_ternary_optab_fn): Likewise.
402 (direct_ternary_optab_supported_p): Likewise.
403 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
404 * builtins.c (fold_builtin_fma): Delete.
405 (fold_builtin_3): Don't call it.
406 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
407 * expr.c (expand_expr_real_2): Likewise.
408 * fold-const.c (operand_equal_p): Likewise.
409 (fold_ternary_loc): Likewise.
410 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
411 * gimple.c (DEFTREECODE): Likewise.
412 * gimplify.c (gimplify_expr): Likewise.
413 * optabs-tree.c (optab_for_tree_code): Likewise.
414 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
415 * tree-eh.c (operation_could_trap_p): Likewise.
416 (stmt_could_throw_1_p): Likewise.
417 * tree-inline.c (estimate_operator_cost): Likewise.
418 * tree-pretty-print.c (dump_generic_node): Likewise.
419 (op_code_prio): Likewise.
420 * tree-ssa-loop-im.c (stmt_cost): Likewise.
421 * tree-ssa-operands.c (get_expr_operands): Likewise.
422 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
423 * fold-const-call.h (fold_fma): Delete.
424 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
425 CFN_FNMA and CFN_FNMS.
427 * genmatch.c (combined_fn): New enum.
428 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
429 (commutative_op): New function.
430 (commutate): Use it. Handle more than 2 operands.
431 (dt_operand::gen_gimple_expr): Use commutative_op.
432 (parser::parse_expr): Allow :c to be used with non-binary
433 operators if the commutative operand is known.
434 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
435 CFN_FMS, CFN_FNMA and CFN_FNMS.
436 (backprop::process_assign_use): Remove FMA_EXPR handling.
437 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
438 (gen_hsa_fma): New function.
439 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
440 IFN_FNMA and IFN_FNMS.
441 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
442 * gimple-fold.h (follow_all_ssa_edges): Declare.
443 * gimple-fold.c (follow_all_ssa_edges): New function.
444 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
445 gimple_build interface and use follow_all_ssa_edges to fold the result.
446 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
447 instead of checking for optabs directly.
448 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
449 rather than FMA_EXPRs.
450 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
451 call to IFN_FMA instead of an FMA_EXPR.
453 2018-05-17 Jim Wilson <jimw@sifive.com>
455 * expr.c (do_tablejump): When converting index to Pmode, if we have a
456 sign extended promoted subreg, and the range does not have the sign bit
457 set, then do a sign extend.
459 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
460 test, check for sign extended subreg and/or constant operands, and
461 do a sign extend in that case.
463 2018-05-17 Steve Ellcey <sellcey@cavium.com>
465 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
466 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
468 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
469 Change logics_shift_reg to logics_shift_imm.
470 (thunderx2t99_fp_loadpair_basic): Delete.
471 (thunderx2t99_fp_storepair_basic): Delete.
472 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
473 (thunderx2t99_asimd_polynomial): Delete.
474 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
475 and neon_fp_mul_d_scalar_q.
476 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
477 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
478 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
479 (thunderx2t99_asimd_lut): Add missing tbl types.
480 (thunderx2t99_asimd_ext): Delete.
481 (thunderx2t99_asimd_load1_1_mult): Delete.
482 (thunderx2t99_asimd_load1_2_mult): Delete.
483 (thunderx2t99_asimd_load1_ldp): New.
484 (thunderx2t99_asimd_load1): New.
485 (thunderx2t99_asimd_load2): Add missing *load2* types.
486 (thunderx2t99_asimd_load3): New.
487 (thunderx2t99_asimd_load4): New.
488 (thunderx2t99_asimd_store1_1_mult): Delete.
489 (thunderx2t99_asimd_store1_2_mult): Delete.
490 (thunderx2t99_asimd_store2_mult): Delete.
491 (thunderx2t99_asimd_store2_onelane): Delete.
492 (thunderx2t99_asimd_store_stp): New.
493 (thunderx2t99_asimd_store1): New.
494 (thunderx2t99_asimd_store2): New.
495 (thunderx2t99_asimd_store3): New.
496 (thunderx2t99_asimd_store4): New.
498 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
500 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
501 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
503 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
504 Segher Boessenkool <segher@kernel.crashing.org>
507 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
510 2018-05-17 Richard Biener <rguenther@suse.de>
512 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
513 for pruning loop and prune defs feeding only already visited PHIs.
515 2018-05-17 Richard Biener <rguenther@suse.de>
517 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
519 2018-05-17 Bin Cheng <bin.cheng@arm.com>
520 Richard Biener <rguenther@suse.de>
522 PR tree-optimization/85793
523 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
524 for VMAT_ELEMENTWISE.
526 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
528 * internal-fn.h (lookup_internal_fn): Declare
529 * internal-fn.c (lookup_internal_fn): New function.
530 * gimple.c (gimple_build_call_from_tree): Handle calls to
532 * gimple-pretty-print.c (dump_gimple_call): Print "." before
533 internal function names.
534 * tree-pretty-print.c (dump_generic_node): Likewise.
535 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
537 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
539 * gimple-fold.h (gimple_build): Make the function forms take
540 combined_fn rather than built_in_function.
541 (gimple_simplify): Likewise.
542 * gimple-match-head.c (gimple_simplify): Likewise.
543 * gimple-fold.c (gimple_build): Likewise.
544 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
545 rather than gimple_build_call_internal.
546 (get_initial_defs_for_reduction): Likewise.
547 (vect_create_epilog_for_reduction): Likewise.
548 (vectorizable_live_operation): Likewise.
550 2018-05-17 Martin Liska <mliska@suse.cz>
552 * gimple-ssa-sprintf.c (format_directive): Do not use
553 space in between 'G_' and '('.
555 2018-05-17 Jakub Jelinek <jakub@redhat.com>
558 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
559 even if the mask is not all ones.
562 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
564 (ix86_gimple_fold_builtin): Likewise.
567 * config/i386/i386.c: Include tree-vector-builder.h.
568 (ix86_vector_shift_count): New function.
569 (ix86_fold_builtin): Fold shift builtins by scalar count.
570 (ix86_gimple_fold_builtin): Likewise.
572 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
573 _mm512_setzero): New intrinsics.
575 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
576 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
578 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
579 code generation for cases where splatting a value is not useful.
580 * simplify-rtx.c (simplify_ternary_operation): Simplify
581 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
582 mode to a vec_concat.
584 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
586 * config.gcc: Support "goldmont-plus".
587 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
589 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
590 PROCESSOR_GOLDMONT_PLUS.
591 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
592 (processor_target_table): Add "goldmont-plus".
593 (PTA_GOLDMONT_PLUS): Define.
594 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
595 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
596 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
597 (fold_builtin_cpu): Add "goldmont-plus".
598 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
599 (ix86_option_override_internal): Add "goldmont-plus".
600 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
601 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
602 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
603 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
605 2018-05-17 Richard Biener <rguenther@suse.de>
607 PR tree-optimization/85757
608 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
609 remove defs that only feed that PHI from further processing.
611 2018-05-16 Jim Wilson <jimw@sifive.com>
613 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
615 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
617 2018-05-16 Mark Wielaard <mark@klomp.org>
619 * dwarf2out.c (count_index_strings): New function.
620 (output_indirect_strings): Call count_index_strings and generate
621 header for dwarf_version >= 5.
623 2018-05-16 Mark Wielaard <mark@klomp.org>
625 * dwarf2out.c (dwarf_FORM): New function.
626 (set_indirect_string): Use dwarf_FORM.
627 (reset_indirect_string): Likewise.
628 (size_of_die): Likewise.
629 (value_format): Likewise.
630 (output_die): Likewise.
631 (add_skeleton_AT_string): Likewise.
632 (output_macinfo_op): Likewise.
633 (index_string): Likewise.
634 (output_index_string_offset): Likewise.
635 (output_index_string): Likewise.
636 (count_index_strings): Likewise.
638 2018-05-16 Carl Love <cel@us.ibm.com>
640 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
641 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
643 2018-05-16 Martin Jambor <mjambor@suse.cz>
645 * ipa-prop.c (ipa_free_all_edge_args): Remove.
646 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
648 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
650 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
651 (fnma<mode>4): Likewise.
652 (fms<mode>4): Likewise.
653 (fnms<mode>4): Likewise.
654 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
655 (aarch64_fnma<mode>4): Likewise.
656 (aarch64_fms<mode>4): Likewise.
657 (aarch64_fnms<mode>4): Likewise.
658 (aarch64_fnmadd<mode>4): Likewise.
660 2018-05-16 Jason Merrill <jason@redhat.com>
662 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
664 2018-05-16 Richard Biener <rguenther@suse.de>
666 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
667 (dump_stmt_cost): Declare.
668 (add_stmt_cost): Dump cost we add.
669 (add_stmt_costs): New function.
670 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
672 (vect_analyze_stmt): Adjust prototype.
673 (vectorizable_condition): Likewise.
674 (vectorizable_live_operation): Likewise.
675 (vectorizable_reduction): Likewise.
676 (vectorizable_induction): Likewise.
677 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
678 cost vector to pass to vectorizable_ and record afterwards.
679 (vect_model_reduction_cost): Take cost vector argument and adjust.
680 (vect_model_induction_cost): Likewise.
681 (vectorizable_reduction): Likewise.
682 (vectorizable_induction): Likewise.
683 (vectorizable_live_operation): Likewise.
684 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
685 SLP_TREE_NUMBER_OF_VEC_STMTS.
686 (vect_analyze_slp_cost_1): Remove.
687 (vect_analyze_slp_cost): Likewise.
688 (vect_slp_analyze_node_operations): Take visited args and
689 a target cost vector. Avoid processing already visited stmt sets.
690 (vect_slp_analyze_operations): Use a local cost vector to gather
691 costs and register those of non-discarded instances.
692 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
693 (vect_schedule_slp_instance): Remove copying of
694 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
696 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
697 adding cost. Record cost entry location.
698 (vect_prologue_cost_for_slp_op): Function to compute cost of
699 a constant or invariant generated for SLP vect in the prologue,
700 split out from vect_analyze_slp_cost_1.
701 (vect_model_simple_cost): Make static. Adjust for SLP costing.
702 (vect_model_promotion_demotion_cost): Likewise.
703 (vect_model_store_cost): Likewise, make static.
704 (vect_model_load_cost): Likewise.
705 (vectorizable_bswap): Add cost vector arg and adjust.
706 (vectorizable_call): Likewise.
707 (vectorizable_simd_clone_call): Likewise.
708 (vectorizable_conversion): Likewise.
709 (vectorizable_assignment): Likewise.
710 (vectorizable_shift): Likewise.
711 (vectorizable_operation): Likewise.
712 (vectorizable_store): Likewise.
713 (vectorizable_load): Likewise.
714 (vectorizable_condition): Likewise.
715 (vectorizable_comparison): Likewise.
716 (can_vectorize_live_stmts): Likewise.
717 (vect_analyze_stmt): Likewise.
718 (vect_transform_stmt): Adjust calls to vectorizable_*.
719 * tree-vectorizer.c: Include gimple-pretty-print.h.
720 (dump_stmt_cost): New function.
722 2018-05-16 Richard Biener <rguenther@suse.de>
724 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
725 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
726 * tree-ssa-dse.c: Include tree-ssa-loop.h.
727 (check_name): New callback.
728 (dse_classify_store): Track cycles via a visited bitmap of PHI
729 defs and simplify handling of in-loop and across loop dead stores
730 and properly fail for loop-variant refs. Handle byte-tracking with
731 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
734 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
736 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
737 (vect_get_mask_type_for_stmt): Likewise.
738 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
740 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
741 to determine the statement's vector type and the vector type that
742 should be used for calculating nunits. Deal with cases in which
743 the type has to be deferred.
744 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
745 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
746 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
747 (vect_determine_vf_for_stmt): New functions, split out from...
748 (vect_determine_vectorization_factor): ...here.
749 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
750 (vect_get_mask_type_for_stmt): New functions, split out from
751 vect_determine_vectorization_factor.
753 2018-05-16 Richard Biener <rguenther@suse.de>
755 * tree-cfg.c (verify_gimple_assign_ternary): Properly
756 verify the [VEC_]COND_EXPR embedded comparison.
758 2018-05-15 Martin Sebor <msebor@redhat.com>
760 PR tree-optimization/85753
761 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
762 RECORD_TYPE in addition to ARRAY_TYPE.
764 2018-05-15 Martin Sebor <msebor@redhat.com>
767 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
769 2018-05-15 Richard Biener <rguenther@suse.de>
771 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
772 add by_clobber_p one. Change algorithm to collect all defs
773 representing uses we need to walk and try reducing them to
774 a single one before failing.
775 (dse_dom_walker::dse_optimize_stmt): Adjust.
777 2018-05-13 Mark Wielaard <mark@klomp.org>
779 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
780 (size_of_loc_descr): Likewise.
781 (output_loc_operands): Likewise.
782 (output_loc_operands_raw): Likewise.
783 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
784 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
785 (hash_loc_operands): Likewise.
786 (compare_loc_operands): Likewise.
788 2018-05-14 Mark Wielaard <mark@klomp.org>
790 * dwarf2out.c (count_index_addrs): New function.
791 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
793 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
795 PR tree-optimization/83648
796 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
797 return value as malloc candidate.
799 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
802 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
803 as true in call to suggest_attribute.
805 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
807 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
810 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
812 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
814 (aarch64_crypto_aesd_fused): Likewise.
816 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
818 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
819 (movsi_aarch64): Likewise.
820 (load_pairsi): Likewise.
821 (load_pairdi): Likewise.
822 (store_pairsi): Likewise.
823 (store_pairdi): Likewise.
824 (load_pairsf): Likewise.
825 (load_pairdf): Likewise.
826 (store_pairsf): Likewise.
827 (store_pairdf): Likewise.
828 (zero_extend): Likewise.
829 (trunc): Swap alternatives.
830 (fcvt_target): Add '?' to prefer w over r.
832 2018-05-14 Jakub Jelinek <jakub@redhat.com>
835 * config/i386/i386.md: Disallow non-commutative arithmetics in
836 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
837 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
838 in the peephole2 before it.
840 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
842 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
843 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
844 (ix86_handle_option): Handle -mcldemote.
845 * config.gcc: New header.
846 * config/i386/cldemoteintrin.h: New file.
847 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
848 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
850 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
851 OPTION_MASK_ISA_CLDEMOTE.
852 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
853 (ix86_valid_target_attribute_inner_p): Ditto.
854 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
855 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
856 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
857 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
858 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
860 * config/i386/i386.opt: Add -mcldemote.
861 * config/i386/x86intrin.h: New header.
862 * doc/invoke.texi: Add -mcldemote.
864 2018-05-14 Richard Biener <rguenther@suse.de>
866 * doc/match-and-simplify.texi: Adjust :s documentation.
868 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
870 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
871 intended memcpy size.
872 (REORDER_45): Likewise.
874 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
877 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
878 * vec.c (qsort_chk): Use gcc_qsort.
879 * Makefile.in (OBJS-libcommon): Add sort.o.
880 (build/sort.o): New target. Use it...
881 (BUILD_RTL): ... here, and...
882 (build/gencfn-macros): ... here, and...
883 (build/genmatch): ... here.
885 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
886 Chung-Ju Wu <jasonwucj@gmail.com>
888 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
889 * config/nds32/nds32-graywolf.md: New file.
890 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
891 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
893 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
894 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
895 * config/nds32/nds32.md (pipeline_model): Add graywolf.
896 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
897 * config/nds32/pipelines.md: Include n15 settings.
899 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
900 Chung-Ju Wu <jasonwucj@gmail.com>
902 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
903 * config/nds32/nds32-n13.md: New file.
904 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
905 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
907 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
908 * config/nds32/nds32.md (pipeline_model): Add n13.
909 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
910 * config/nds32/pipelines.md: Include n13 settings.
912 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
913 Chung-Ju Wu <jasonwucj@gmail.com>
915 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
916 * config/nds32/nds32-n10.md: New file.
917 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
918 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
920 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
921 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
922 * config/nds32/nds32.md (pipeline_model): Add n10.
923 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
924 * config/nds32/pipelines.md: Include n10 settings.
926 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
927 Kito Cheng <kito.cheng@gmail.com>
928 Chung-Ju Wu <jasonwucj@gmail.com>
930 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
931 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
932 Add enum values for DSP extension instructions.
933 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
935 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
936 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
938 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
939 * config/nds32/nds32-dspext.md: New file for DSP implementation.
940 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
941 * config/nds32/nds32-intrinsic.md: Likewise.
942 * config/nds32/nds32_intrinsic.h: Likewise.
943 * config/nds32/nds32-md-auxiliary.c: Likewise.
944 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
945 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
946 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
947 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
948 * config/nds32/nds32-protos.h: New declarations for DSP extension.
949 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
950 TYPE_DMAC in switch statement.
951 * config/nds32/nds32.c: New checking and implementation for DSP
952 extension instructions.
953 * config/nds32/nds32.h: Likewise.
954 * config/nds32/nds32.md: Likewise.
955 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
956 * config/nds32/predicates.md: Implement new predicates for DSP
959 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
961 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
962 Reformat alternatives and attributes so it is easier to identify
963 which constraints/attributes go with which instruction.
964 (mov<mode>_hardfloat32, FMOVE64): Likewise.
965 (mov<mode>_softfloat32, FMOVE64): Likewise.
966 (mov<mode>_hardfloat64, FMOVE64): Likewise.
967 (mov<mode>_softfloat64, FMOVE64): Likewise.
969 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
971 * doc/extend.texi (PowerPC Built-in Functions): Rename this
973 (Basic PowerPC Built-in Functions): The new name of the
974 subsection previously known as "PowerPC Built-in Functions".
975 (Basic PowerPC Built-in Functions Available on all Configurations):
977 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
978 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
979 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
980 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
982 2018-05-11 Martin Jambor <mjambor@suse.cz>
985 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
988 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
991 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
993 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
995 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
996 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
997 (ix86_handle_option): Handle -mwaitpkg.
998 * config.gcc: New header.
999 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1000 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1001 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1003 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1004 OPTION_MASK_ISA_WAITPKG.
1005 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1006 (ix86_option_override_internal): Add PTA_WAITPKG.
1007 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1008 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1009 IX86_BUILTIN_TPAUSE.
1010 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1011 __builtin_ia32_umwait and __builtin_ia32_tpause.
1012 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1013 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1014 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1015 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1016 UNSPECV_TPAUSE): New.
1017 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1018 * config/i386/i386.opt: Add -mwaitpkg.
1019 * config/i386/waitpkgintrin.h: New file.
1020 * config/i386/x86intrin.h: New header.
1021 * doc/invoke.texi: Add -mwaitpkg.
1023 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1026 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1028 (cortex-m0): Use armv6s-m isa.
1029 (cortex-m0plus): Likewise.
1030 (cortex-m1): Likewise.
1031 (cortex-m0.small-multiply): Likewise.
1032 (cortex-m0plus.small-multiply): Likewise.
1033 (cortex-m1.small-multiply): Likewise.
1035 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1036 Jakub Jelinek <jakub@redhat.com>
1038 PR tree-optimization/85692
1039 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1040 source permute as well.
1042 2018-05-11 Martin Liska <mliska@suse.cz>
1045 * doc/extend.texi: Document LLVM style format for no_sanitize
1048 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1050 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1051 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1052 (mode_supports_vsx_dform_quad): Likewise.
1053 (mode_supports_vmx_dform): Move these functions to be next to the
1054 other mode_supports functions.
1055 (mode_supports_dq_form): Likewise.
1056 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1057 mode_supports_dq_form.
1058 (reg_offset_addressing_ok_p): Likewise.
1059 (offsettable_ok_by_alignment): Likewise.
1060 (rs6000_legitimate_offset_address_p): Likewise.
1061 (legitimate_lo_sum_address_p): Likewise.
1062 (rs6000_legitimize_address): Likewise.
1063 (rs6000_legitimize_reload_address): Likewise.
1064 (rs6000_secondary_reload_inner): Likewise.
1065 (rs6000_preferred_reload_class): Likewise.
1066 (rs6000_output_move_128bit): Likewise.
1068 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1070 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1071 Generate SImode target register for null target.
1072 <case IX86_BUILTIN_XGETBV>: Ditto.
1073 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1074 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1076 2018-05-10 Carl Love <cel@us.ibm.com>
1078 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1079 dcbtt and dcbtstt if operands[2] is 0.
1081 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1084 * config/i386/sse.md (usadv64qi): New expander.
1086 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1088 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1089 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1090 -maltivec=be support.
1091 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1092 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1093 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1094 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1095 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1096 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1097 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1098 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1099 altivec_vsumsws): Adjust.
1100 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1101 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1102 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1104 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1105 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1106 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1107 (altivec_lve<VI_char>x): Delete expand.
1108 (*altivec_lve<VI_char>x_internal): Rename to...
1109 (altivec_lve<VI_char>x): ... this.
1110 (altivec_lvxl_<mode>): Delete expand.
1111 (*altivec_lvxl_<mode>_internal): Rename to ...
1112 (altivec_lvxl_<mode>): ... this.
1113 (altivec_stvxl_<mode>): Delete expand.
1114 (*altivec_stvxl_<mode>_internal): Rename to ...
1115 (altivec_stvxl_<mode>): ... this.
1116 (altivec_stve<VI_char>x): Delete expand.
1117 (*altivec_stve<VI_char>x_internal): Rename to ...
1118 (altivec_stve<VI_char>x): ... this.
1119 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1120 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1121 reduc_plus_scal_<mode>): Adjust.
1122 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1124 (rs6000_cpu_cpp_builtins): Adjust.
1125 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1126 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1127 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1128 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1129 -maltivec=be support.
1130 (rs6000_split_vec_extract_var): Adjust.
1131 (rs6000_split_v4si_init): Adjust.
1132 (swap_selector_for_mode): Delete.
1133 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1134 altivec_expand_stvex_be): Delete.
1135 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1136 -maltivec=be support.
1137 (rs6000_gimple_fold_builtin): Ditto.
1138 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1140 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1141 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1142 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1143 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1144 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1145 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1146 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1147 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1148 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1149 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1150 anonymous split): Adjust.
1151 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1152 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1154 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1156 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1157 when --with-gxx-include-dir is also specified.
1158 * configure: Regenerate.
1160 2018-05-09 Jim Wilson <jimw@sifive.com>
1163 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1164 * config/riscv/t-withmultilib: New.
1165 * config/riscv/withmultilib.h: New.
1166 * doc/install.texi: Document RISC-V --with-multilib-list support.
1168 2018-05-09 Richard Biener <rguenther@suse.de>
1170 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1172 (vect_bb_vectorization_profitable_p): Adjust. Compute
1173 actual scalar cost using the cost vector and the add_stmt_cost
1176 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1178 PR rtl-optimization/85645
1179 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1180 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1182 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1184 PR rtl-optimization/85645
1185 * shrink-wrap.c (spread_components): Return a boolean saying if
1186 anything was changed.
1187 (try_shrink_wrapping_separate): Iterate spread_components until
1188 nothing changes anymore.
1190 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1192 PR rtl-optimization/85645
1193 * regrename.c (build_def_use): Also kill the chains that include the
1194 destination of a REG_CFA_REGISTER note.
1196 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1198 PR rtl-optimization/85645
1199 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1200 insn that has a REG_CFA_REGISTER note.
1202 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1204 * cfgexpand.c (expand_clobber): New function.
1205 (expand_gimple_stmt_1): Use it.
1206 * tree-vect-stmts.c (vect_clobber_variable): New function,
1208 (vectorizable_simd_clone_call): ...here.
1209 (vectorizable_store): Emit a clobber either side of an
1210 IFN_STORE_LANES sequence.
1211 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1213 2018-05-09 Tom de Vries <tom@codesourcery.com>
1216 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1217 (define_insn "trap_if_false"): Add exit after trap.
1219 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1221 PR rtl-optimization/85638
1222 * bb-reorder.c: Include common/common-target.h.
1223 (create_forwarder_block): New function extracted from...
1224 (fix_up_crossing_landing_pad): ...here. Rename into...
1225 (dw2_fix_up_crossing_landing_pad): ...this.
1226 (sjlj_fix_up_crossing_landing_pad): New function.
1227 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1228 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1229 from both partitions and exit the loop after one iteration.
1231 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1234 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1236 (Basic PowerPC Built-in Functions): The new name of the
1237 subsection previously known as "PowerPC Built-in Functions".
1238 (Basic PowerPC Built-in Functions Available on all Configurations):
1240 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1242 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1244 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1246 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1249 2018-05-08 Jim Wilson <jimw@sifive.com>
1251 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1252 (LD_EMUL_SUFFIX): New.
1253 (LINK_SPEC): Use it.
1255 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1257 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1259 (Basic PowerPC Built-in Functions): The new name of the
1260 subsection previously known as "PowerPC Built-in Functions".
1261 (Basic PowerPC Built-in Functions Available on all Configurations):
1263 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1265 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1267 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1269 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1272 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1275 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1276 after cmpelim optimization.
1278 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1280 * config.gcc: Support "goldmont".
1281 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1282 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1284 * config/i386/i386.c (m_GOLDMONT): Define.
1285 (processor_target_table): Add "goldmont".
1286 (PTA_GOLDMONT): Define.
1287 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1288 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1289 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1290 (fold_builtin_cpu): Add "goldmont".
1291 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1292 (ix86_option_override_internal): Add "goldmont".
1293 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1294 (processor_type): Add PROCESSOR_GOLDMONT.
1295 * config/i386/i386.md: Add CPU "glm".
1296 * config/i386/glm.md: New file.
1297 * config/i386/x86-tune.def: Add m_GOLDMONT.
1298 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1300 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1303 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1305 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1306 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1307 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1310 * config/i386/i386.c (ix86_fold_builtin): Handle
1311 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1314 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1315 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1317 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1320 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1321 (check_arch): Likewise.
1322 (check_fpu): Return the result rather than printing it.
1323 (end arch): Fix operator precedence.
1324 (end cpu): Likewise.
1325 (END): Print the result from check_fpu.
1327 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1328 Alan Hayward <alan.hayward@arm.com>
1329 David Sherwood <david.sherwood@arm.com>
1331 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1332 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1333 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1334 (*fcmuo<mode>_and): New patterns.
1336 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1338 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1339 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1340 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1341 (cmp_op, sve_imm_con): New code attributes.
1342 (SVE_COND_INT_CMP, imm_con): Delete.
1343 (cmp_op): Remove above unspecs from int attribute.
1344 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1346 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1347 comparison-specific unspecs.
1348 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1349 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1350 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1351 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1352 (*vec_fcm<cmp_op><mode>): Rename to...
1353 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1354 (*vec_fcmuo<mode>): Rename to...
1355 (*fcmuo<mode>): ...this and adjust likewise.
1356 (*pred_fcm<cmp_op><mode>): New pattern.
1357 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1358 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1360 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1362 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1363 (aarch64_emit_sve_predicated_cond): New function.
1364 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1365 (aarch64_emit_unspec_cond_or): Replace with...
1366 (aarch64_emit_sve_or_conds): ...this new function. Use
1367 aarch64_emit_sve_ptrue_op for the individual comparisons and
1368 aarch64_emit_binop to OR them together.
1369 (aarch64_emit_inverted_unspec_cond): Replace with...
1370 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1371 aarch64_emit_sve_ptrue_op for the comparison and
1372 aarch64_emit_unop to invert the result.
1373 (aarch64_expand_sve_vec_cmp_float): Update after the above
1374 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1376 2018-05-07 Nathan Sidwell <nathan@acm.org>
1378 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1379 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1380 (Backwards Compatibility): Likewise.
1382 2018-05-07 Luis Machado <luis.machado@linaro.org>
1386 2018-05-07 Luis Machado <luis.machado@linaro.org>
1388 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1389 <prefetch_dynamic_strides>: New const bool field.
1390 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1391 prefetch_dynamic_strides.
1392 (exynosm1_prefetch_tune): Likewise.
1393 (thunderxt88_prefetch_tune): Likewise.
1394 (thunderx_prefetch_tune): Likewise.
1395 (thunderx2t99_prefetch_tune): Likewise.
1396 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1397 (aarch64_override_options_internal): Update to set
1398 PARAM_PREFETCH_DYNAMIC_STRIDES.
1399 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1400 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1401 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1402 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1403 prefetch-dynamic-strides setting.
1405 2018-05-07 Luis Machado <luis.machado@linaro.org>
1407 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1408 <minimum_stride>: New const int field.
1409 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1410 minimum_stride field.
1411 (exynosm1_prefetch_tune): Likewise.
1412 (thunderxt88_prefetch_tune): Likewise.
1413 (thunderx_prefetch_tune): Likewise.
1414 (thunderx2t99_prefetch_tune): Likewise.
1415 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1416 (aarch64_override_options_internal): Update to set
1417 PARAM_PREFETCH_MINIMUM_STRIDE.
1418 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1419 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1420 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1421 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1422 stride is constant and is below the minimum stride threshold.
1424 2018-05-07 Luis Machado <luis.machado@linaro.org>
1426 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1429 2018-05-07 Luis Machado <luis.machado@linaro.org>
1431 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1432 <prefetch_dynamic_strides>: New const bool field.
1433 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1434 prefetch_dynamic_strides.
1435 (exynosm1_prefetch_tune): Likewise.
1436 (thunderxt88_prefetch_tune): Likewise.
1437 (thunderx_prefetch_tune): Likewise.
1438 (thunderx2t99_prefetch_tune): Likewise.
1439 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1440 (aarch64_override_options_internal): Update to set
1441 PARAM_PREFETCH_DYNAMIC_STRIDES.
1442 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1443 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1444 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1445 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1446 prefetch-dynamic-strides setting.
1448 2018-05-07 Luis Machado <luis.machado@linaro.org>
1450 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1451 <minimum_stride>: New const int field.
1452 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1453 minimum_stride field.
1454 (exynosm1_prefetch_tune): Likewise.
1455 (thunderxt88_prefetch_tune): Likewise.
1456 (thunderx_prefetch_tune): Likewise.
1457 (thunderx2t99_prefetch_tune): Likewise.
1458 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1459 (aarch64_override_options_internal): Update to set
1460 PARAM_PREFETCH_MINIMUM_STRIDE.
1461 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1462 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1463 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1464 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1465 stride is constant and is below the minimum stride threshold.
1467 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1470 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1471 the type is addressable. Don't force op into register if it has
1474 2018-05-05 Roland McGrath <mcgrathr@google.com>
1477 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1478 any section for which we don't know a specific type it should have,
1479 regardless of name. Previously this was done only for the exact
1480 names ".init_array", ".fini_array", and ".preinit_array".
1481 (default_elf_asm_named_section): Add comment about
1482 relationship with default_section_type_flags and SECTION_NOTYPE.
1483 (get_section): Don't consider it a type conflict if one side has
1484 SECTION_NOTYPE and the other doesn't, as long as neither has the
1485 SECTION_BSS et al used in the default_section_type_flags logic.
1487 2018-05-05 Tom de Vries <tom@codesourcery.com>
1490 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1491 (workaround_barsyncs): New function.
1492 (nvptx_reorg): Use workaround_barsyncs.
1493 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1494 (define_expand "nvptx_membar_cta"): New define_expand.
1495 (define_insn "*nvptx_membar_cta"): New insn.
1497 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1499 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1500 To improve optimization opportunities.
1501 * builtin-types.def: The new needed builtin types for the above.
1503 2018-05-04 Richard Biener <rguenther@suse.de>
1505 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1506 * gimple-ssa-store-merging.c
1507 (imm_store_chain_info::output_merged_store): Remove redundant create,
1508 release split_store vector contents on failure.
1509 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1510 scalar stmt vector on cache hit.
1512 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1514 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1516 * config.gcc (powerpc-xilinx-eabi*): Remove.
1517 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1519 (fusion_addis_mem_combo_load): Ditto.
1520 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1522 (rs6000_cpu_cpp_builtins): Ditto.
1523 * config/rs6000/rs6000-linux.c
1524 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1525 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1526 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1528 (rs6000_setup_reg_addr_masks): Ditto.
1529 (rs6000_init_hard_regno_mode_ok): Ditto.
1530 (rs6000_option_override_internal): Ditto.
1531 (legitimate_lo_sum_address_p): Ditto.
1532 (rs6000_legitimize_address): Ditto.
1533 (rs6000_legitimize_reload_address): Ditto.
1534 (rs6000_legitimate_address_p): Ditto.
1535 (abi_v4_pass_in_fpr): Ditto.
1536 (setup_incoming_varargs): Ditto.
1537 (rs6000_gimplify_va_arg): Ditto.
1538 (rs6000_split_multireg_move): Ditto.
1539 (rs6000_savres_strategy): Ditto.
1540 (rs6000_emit_prologue_components): Ditto.
1541 (rs6000_emit_epilogue_components): Ditto.
1542 (rs6000_emit_prologue): Ditto.
1543 (rs6000_emit_epilogue): Ditto.
1544 (rs6000_elf_file_end): Ditto.
1545 (rs6000_function_value): Ditto.
1546 (rs6000_libcall_value): Ditto.
1547 * config/rs6000/rs6000.h: Ditto.
1548 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1549 (TARGET_MINMAX): ... this. New.
1550 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1551 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1552 (*movsi_internal1_single): Delete.
1553 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1554 mfpu=, mxilinx-fpu): Delete.
1555 * config/rs6000/singlefp.h: Delete.
1556 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1557 * config/rs6000/t-rs6000: Ditto.
1558 * config/rs6000/t-xilinx: Delete.
1559 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1560 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1561 (VStype_simple): Delete.
1562 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1563 * config/rs6000/xfpu.h: Delete.
1564 * config/rs6000/xfpu.md: Delete.
1565 * config/rs6000/xilinx.h: Delete.
1566 * config/rs6000/xilinx.opt: Delete.
1567 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1568 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1570 2018-05-04 Tom de Vries <tom@codesourcery.com>
1573 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1576 2018-05-04 Richard Biener <rguenther@suse.de>
1579 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1580 (expand_complex_div_wide): Likewise.
1581 (expand_complex_operations_1): Likewise.
1582 (expand_complex_libcall): Preserve EH info of the original stmt.
1583 (tree_lower_complex): Handle removed blocks.
1584 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1585 on complex multiplication and division libcall builtins.
1587 2018-05-04 Richard Biener <rguenther@suse.de>
1590 * fold-const.c (negate_expr_p): Restrict negation of operand
1591 zero of a division to when we know that can happen without
1593 (fold_negate_expr_1): Likewise.
1595 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1598 * real.h (real_nextafter): Declare.
1599 * real.c (real_nextafter): New function.
1600 * fold-const-call.c (fold_const_nextafter): New function.
1601 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1602 CASE_CFN_NEXTTOWARD.
1603 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1604 even when arg1_mode is different from arg0_mode.
1606 2018-05-03 Nathan Sidwell <nathan@acm.org>
1608 * doc/extend.texi (Deprecated Features): Remove
1610 (Backwards Compatibility): Likewise.
1611 * doc/invoke.texi (C++ Language Options): Likewise.
1612 (C++ Dialect Options): Likewise.
1614 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1617 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1618 _mm512_mask_mullox_epi64): New intrinsics.
1620 2018-05-03 Tom de Vries <tom@codesourcery.com>
1623 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1624 dump files): Add offload-tree.
1626 2018-05-03 Richard Biener <rguenther@suse.de>
1628 PR tree-optimization/85615
1629 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1630 to loops not nested in BBs loop father to avoid creating multi-entry
1633 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1635 PR tree-optimization/70291
1636 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1637 arguments. Change return type to tree. Emit libcall as a new
1638 statement rather than replacing existing one when inplace_p is true.
1639 (expand_complex_multiplication_components): New function.
1640 (expand_complex_multiplication): Expand floating-point complex
1641 multiplication using the above.
1642 (expand_complex_division): Rename inner_type parameter to type.
1643 Update expand_complex_libcall call-site.
1644 (expand_complex_operations_1): Update expand_complex_multiplication
1645 and expand_complex_division call-sites.
1647 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1650 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1651 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1652 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1653 the highest significant bit of the shift count mask is clear. In
1654 check whether and[sq]i3 is needed verify that all significant bits
1655 of the shift count other than the highest are set.
1657 2018-05-02 Tom de Vries <tom@codesourcery.com>
1660 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1661 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1662 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1663 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1664 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1665 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1666 __builtin_goacc_parlevel_size.
1668 2018-05-02 Richard Biener <rguenther@suse.de>
1670 PR tree-optimization/85597
1671 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1672 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1675 2018-05-02 Tom de Vries <tom@codesourcery.com>
1678 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1679 dump files): Add ltrans-tree.
1681 2018-05-02 Tom de Vries <tom@codesourcery.com>
1684 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1685 dump files): Add wpa-ipa.
1687 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1689 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1690 powerpc*-*-linux*paired* target.
1691 * config/rs6000/750cl.h: Delete.
1692 * config/rs6000/paired.h: Delete.
1693 * config/rs6000/paired.md: Delete.
1694 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1696 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1697 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1698 comment. Remove paired float support.
1699 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1700 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1702 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1703 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1705 * config/rs6000/rs6000.c: Remove paired float support.
1706 (paired_expand_vector_init, paired_expand_vector_move,
1707 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1708 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1709 paired_expand_builtin, paired_expand_predicate_builtin,
1710 paired_init_builtins): Delete.
1711 * config/rs6000/rs6000.h: Remove paired float support.
1712 * config/rs6000/rs6000.md: Remove paired float support.
1713 (move_from_CR_ov_bit): Delete.
1714 * config/rs6000/rs6000.opt (mpaired): Delete.
1715 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1716 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1718 2018-05-02 Richard Biener <rguenther@suse.de>
1721 * gimplify.c (gimplify_save_expr): When in SSA form allow
1722 SAVE_EXPRs to compute to SSA vars.
1724 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1727 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1728 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1729 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1730 clobber operands[2], instead use a new pseudo. Formatting fixes.
1732 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1734 PR tree-optimization/85586
1735 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1736 exit early for statements in the same group if the accesses are
1739 2018-05-02 Tom de Vries <tom@codesourcery.com>
1742 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1745 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1747 PR tree-optimization/85143
1748 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1750 2018-05-01 Tom de Vries <tom@codesourcery.com>
1753 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1754 not found" error message.
1756 2018-05-01 Tom de Vries <tom@codesourcery.com>
1759 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1760 * vec.c (test_ordered_remove_if): New function.
1761 (vec_c_tests): Call test_ordered_remove_if.
1762 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1763 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1764 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1765 VEC_ORDERED_REMOVE_IF.
1767 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1769 PR tree-optimization/82665
1770 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1771 pointer subtraction where arguments come from a memchr call.
1773 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1775 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1776 --push-state --as-needed and --pop-state instead of --as-needed and
1777 --no-as-needed if ld supports it.
1778 * configure: Regenerated.
1781 * doc/install.texi2html: Replace _002d with - and _002a with * in
1782 generated html files using sed.
1784 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1787 * gcc-rich-location.c (blank_line_before_p): New function.
1788 (use_new_line): New function.
1789 (gcc_rich_location::add_fixit_insert_formatted): New function.
1790 * gcc-rich-location.h
1791 (gcc_rich_location::add_fixit_insert_formatted): New function.
1793 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1795 * selftest.c (assert_streq): Rename "expected" and "actual" to
1796 "val1" and "val2". Extend NULL-handling to cover both inputs
1797 symmetrically, while still requiring both to be non-NULL for a pass.
1798 * selftest.h (assert_streq): Rename "expected" and "actual" to
1800 (ASSERT_EQ): Likewise.
1801 (ASSERT_EQ_AT): Likewise.
1802 (ASSERT_KNOWN_EQ): Likewise.
1803 (ASSERT_KNOWN_EQ_AT): Likewise.
1804 (ASSERT_NE): Likewise.
1805 (ASSERT_MAYBE_NE): Likewise.
1806 (ASSERT_MAYBE_NE_AT): Likewise.
1807 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1808 the assertion to pass.
1809 (ASSERT_STREQ_AT): Likewise.
1811 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1813 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1814 interaction with -pie.
1816 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1818 * selftest.h: Fix alphabetization of per-source-file selftest
1821 2018-04-30 Jason Merrill <jason@redhat.com>
1823 PR c++/61982 - dead stores to destroyed objects.
1824 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1827 2018-04-30 Jason Merrill <jason@redhat.com>
1829 * tree.c (build_clobber): New.
1830 * tree.h: Declare it.
1831 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1833 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1835 * diagnostic-show-locus.c (layout::layout): Update for
1836 location_get_source_line returning a char_span.
1837 (struct char_span): Move to input.h.
1838 (struct correction): Update for fields in char_span becoming
1840 (struct source_line): Update for location_get_source_line
1841 returning a char_span.
1842 (layout::print_line): Likewise.
1843 * edit-context.c (edited_file::print_content): Likewise.
1844 (edited_file::print_diff_hunk): Likewise.
1845 (edited_file::print_run_of_changed_lines): Likewise.
1846 (edited_file::get_num_lines): Likewise.
1847 (edited_line::edited_line): Likewise.
1848 * final.c (asm_show_source): Likewise.
1849 * input.c (location_get_source_line): Convert return type
1850 from const char * to char_span, losing the final "line_len"
1852 (dump_location_info): Update for the above.
1853 (get_substring_ranges_for_loc): Likewise. Use a char_span
1854 when handling the literal within the line.
1855 (test_reading_source_line): Update for location_get_source_line
1856 returning a char_span.
1857 * input.h (class char_span): Move here from
1858 diagnostic-show-locus.c, converting from a struct to a class.
1859 Make data members private.
1860 (char_span::operator bool): New.
1861 (char_span::length): New.
1862 (char_span::get_buffer): New.
1863 (char_span::operator[]): New.
1864 (char_span::subspan): Make const.
1865 (char_span::xstrdup): New.
1866 (location_get_source_line): Convert return type from const char *
1867 to char_span, losing the final "line_size" param.
1869 2018-04-30 Jan Hubicka <jh@suse.cz>
1871 * lto-wrapper.c (ltrans_priorities): New static var.
1872 (cmp_priority): New.
1873 (run_gcc): Read priorities and if doing parallel build order
1874 the Makefile by them.
1876 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1878 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1880 2018-04-30 Richard Biener <rguenther@suse.de>
1882 * tree-cfg.c (verify_address): Remove base argument, add
1883 flag whether to check TREE_ADDRESSABLE and do that.
1884 (verify_expr): Remove.
1885 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1886 (verify_gimple_assign_single): Likewise.
1887 (verify_gimple_switch): Likewise.
1888 (verify_expr_location_1): Dereference tp once. Add (disabled)
1889 piece from verify_expr.
1890 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1892 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1894 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1896 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1898 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1899 (small_data_pattern): Likewise.
1900 (arc_rewrite_small_data): Likewise.
1901 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1902 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1903 (get_symbol_alignment): New function.
1904 (legitimate_small_data_address_p): Likewise.
1905 (legitimate_scaled_address): Update, call
1906 legitimate_small_data_address_p.
1907 (output_sdata): New static variable.
1908 (arc_print_operand): Update how we handle small data operands.
1909 (arc_print_operand_address): Likewise.
1910 (arc_legitimate_address_p): Update, use
1911 legitimate_small_data_address_p.
1912 (arc_rewrite_small_data_p): Remove.
1913 (arc_rewrite_small_data_1): Likewise.
1914 (arc_rewrite_small_data): Likewise.
1915 (small_data_pattern): Likewise.
1916 (compact_sda_memory_operand): Update to use
1917 legitimate_small_data_address_p and get_symbol_alignment.
1918 (prepare_move_operands): Don't rewite sdata pattern.
1919 (prepare_extend_operands): Remove.
1920 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1922 (zero_extendqisi2): Likewise.
1923 (zero_extendhisi2): Likewise.
1924 (extendqihi2): Likewise.
1925 (extendqisi2): Likewise.
1926 (extendhisi2): Likewise.
1930 * config/arc/constraints.md (Usd): Change it to memory constraint.
1932 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1934 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1935 as source of std instructions.
1936 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1937 allow 6-bit constants as source for store instructions.
1938 (movdi_insn): Update instruction pattern to allow 6-bit constants
1939 as source for store instructions.
1941 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1943 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1945 2018-04-30 Nathan Sidwell <nathan@acm.org>
1946 Sandra Loosemore <sandra@codesourcery.com>
1948 * dumpfile.c (dump_open): Allow '-' for stdout.
1949 * doc/invoke.texi (Developer Options): Document dump filename
1950 determination early. Document stdin/stdout selection.
1952 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1954 Microblaze Target: PIC data text relative
1956 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1957 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1959 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1960 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1961 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1962 New addressing mode for data-text relative position indepenedent code.
1963 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1964 'ADDRESS_SYMBOLIC_TXT_REL'.
1965 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1966 (microblaze_legitimate_pic_operand): Exclude function calls from
1967 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1968 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1970 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1971 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1972 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1973 for 'address + offset'.
1974 (microblaze_expand_prologue): Add new function prologue call for
1976 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1977 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1978 table in case of TARGET_PIC_DATA_TEXT_REL.
1979 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1980 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1981 Add new macros 'UNSPEC_TEXT',
1982 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1983 + exclude function calls from 'UNSPEC_PLT' in case of data text
1985 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1986 new target hook for generating address diff vector tables in case of
1988 * doc/tm.texi : Regenerate.
1989 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1990 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1991 of addr diff vector generation.
1992 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1993 target hook definition.
1994 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1995 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1996 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1997 option for data text relative.
1999 2018-04-30 Richard Biener <rguenther@suse.de>
2001 * tree-chrec.h (evolution_function_is_constant_p): Remove
2003 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2005 2018-04-30 Richard Biener <rguenther@suse.de>
2008 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2010 2018-04-30 Richard Biener <rguenther@suse.de>
2012 PR tree-optimization/28364
2013 PR tree-optimization/85275
2014 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2015 copying first exit test.
2017 2018-04-28 Mark Wielaard <mark@klomp.org>
2019 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2021 (dwarf_AT): Handle DW_AT_addr_base.
2022 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2024 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2027 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2028 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2029 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2030 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2032 2018-04-28 Richard Biener <rguenther@suse.de>
2034 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2035 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2036 to reflect use. Only add interesting stmts.
2038 2018-04-27 Martin Jambor <mjambor@suse.cz>
2041 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2042 the jump function allows for passing through aggregate values.
2044 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2046 * input.h (in_system_header_at): Convert from macro to inline
2048 (from_macro_expansion_at): Likewise.
2049 (from_macro_definition_at): Likewise.
2051 2018-04-27 Jeff Law <law@redhat.com>
2053 * config.gcc: Mark tile* targets as deprecated/obsolete.
2055 2018-04-27 Richard Biener <rguenther@suse.de>
2057 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2060 2018-04-27 Richard Biener <rguenther@suse.de>
2062 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2064 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2066 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2067 with Yd constraint. Set "preferred_for_speed" attribute from
2068 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2070 (*movdi_internal): Ditto.
2071 (movti_interunit splitters): Remove
2072 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2073 (movdi_interunit splitters): Ditto.
2074 * config/i386/constraints.md (Ye): Remove.
2075 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2077 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2080 * config/aarch64/constraints.md (Usg): Limit to 31.
2083 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2085 PR tree-optimization/85529
2086 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2087 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2088 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2089 zero extension or masking of the MSB bit.
2090 (optimize_range_tests): Add FIRST_BB argument, pass it through
2091 to optimize_range_tests_var_bound.
2092 (maybe_optimize_range_tests, reassociate_bb): Adjust
2093 optimize_range_tests callers.
2095 2018-04-26 Richard Biener <rguenther@suse.de>
2096 Jakub Jelinek <jakub@redhat.com>
2098 * cgraph.h (symbol_table): Just declare debug method here.
2099 * symtab.c (symbol_table::debug): Define.
2101 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2103 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2105 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2107 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2108 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2109 (*movdi_internal): Substitute Yi and Yj constraint with x
2110 and Ym and Yn constraint with y constraint. Update "isa"
2111 attribute and set "preferred_for_speed" attribute from
2112 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2113 (*movsi_internal): Ditto.
2114 (*movdf_internal): Ditto.
2115 (*movsf_internal): Ditto.
2116 (*zero_extendsidi2): Ditto.
2117 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2118 (sse2_loadld): Ditto.
2119 (*vec_extract<ssevecmodelower>_0): Ditto.
2120 (*vec_extractv4si_0_zext_sse4): Ditto.
2121 (vec_concatv2di): Ditto.
2122 (*vec_dup<mode>): Ditto.
2123 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2124 * config/i386/constraints.md (Yi): Remove.
2129 2018-04-26 Nathan Sidwell <nathan@acm.org>
2131 * dumpfile.c (dump_open): New.
2132 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2133 (dump_finish): Detect stdio/stderr by value not name.
2135 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2137 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2139 2018-04-26 Tom de Vries <tom@codesourcery.com>
2142 * config/nvptx/nvptx.c (verify_neutering_jumps)
2143 (verify_neutering_labels): New function
2144 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2146 2018-04-26 Tom de Vries <tom@codesourcery.com>
2149 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2150 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2153 2018-04-26 Richard Biener <rguenther@suse.de>
2154 Tom de Vries <tom@codesourcery.com>
2157 * lto-streamer-out.c (output_function): Fixup loops if required to match
2158 discovery done in the reader.
2160 2018-04-26 Richard Biener <rguenther@suse.de>
2162 PR tree-optimization/85116
2163 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2164 have a loop exit from the single latch predecessor. Remove
2165 case of header with just condition.
2166 (ch_base::copy_headers): Exclude infinite loops from any
2168 (pass_ch::execute): Record exits.
2170 2018-04-26 Richard Biener <rguenther@suse.de>
2172 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2173 prologue cost vector and pass it to vect_get_load_cost.
2174 (vect_get_peeling_costs_all_drs): Likewise.
2175 (vect_peeling_hash_get_lowest_cost): Likewise.
2176 (vect_enhance_data_refs_alignment): Likewise.
2178 2018-04-26 Richard Biener <rguenther@suse.de>
2181 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2182 checking of integer<->pointer conversions.
2183 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2184 sign-/zero-extending pointer types.
2185 (expand_omp_for_static_chunk): Likewise.
2187 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2188 Jean Lee <xiaoyur347@gmail.com>
2190 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2191 (TARGET_ASAN_SHADOW_OFFSET): Define.
2192 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2193 true for -fsanitize=address.
2195 2018-04-25 Mark Wielaard <mark@klomp.org>
2197 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2200 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2202 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2203 than "alu", remove explicit "memory" and "imm_disp" attributes.
2204 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2207 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2208 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2209 gen_lowpart_no_emit.
2211 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2214 * config/i386/i386.c (ix86_expand_builtin): Change memory
2215 operand to XI, extend p0 to Pmode.
2216 * config/i386/i386.md: Change unspec volatile and operand
2217 1 mode to XI, change operand 0 mode to P.
2219 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2221 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2222 GET_MODE_MASK before any checking.
2223 (nds32_can_use_bset_p): Likewise.
2224 (nds32_can_use_btgl_p): Likewise.
2226 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2228 * config/nds32/nds32-doubleword.md: New define_split pattern for
2229 illegal register number.
2231 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2233 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2235 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2237 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2239 2018-04-25 Richard Biener <rguenther@suse.de>
2241 * lto-streamer.h (LTO_major_version): Bump to 8.
2243 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2245 * BASE-VER: Set to 9.0.0.
2247 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2249 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2250 in __abskf2 and __powikf2.
2252 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2255 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2256 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2257 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2258 Use the above on operand 2. Reindent.
2259 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2261 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2264 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2266 * config/i386/i386.opt (mcet): Removed.
2267 * doc/install.texi: Remove -mcet documentation.
2268 * doc/invoke.texi: Likewise.
2270 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2273 * doc/install.texi: Remove -mcet from bootstrap-cet.
2275 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2278 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2279 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2283 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2284 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2285 containing a CONST_VECTOR.
2287 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2289 * doc/install.texi: Update newlib dependency for nvptx.
2291 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2294 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2295 instead of INTVAL when shifting x left.
2297 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2299 PR tree-optimization/85478
2300 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2301 vect_grouped_store_supported for single element vectors.
2303 2018-04-24 Richard Biener <rguenther@suse.de>
2306 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2307 load cost increase to the case of non-constant step.
2309 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2312 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2313 destination if any_malformed_asm.
2315 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2318 * expr.c (store_field): In the bitfield case, if the value comes from
2319 a function call and is returned in registers by means of a PARALLEL,
2320 do not change the mode of the temporary unless BLKmode and VOIDmode.
2322 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2324 PR rtl-optimization/85423
2325 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2326 dependencies to debug insns when the previous insn is non-debug.
2328 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2330 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2331 enums into a single definition.
2332 (fls): Fix predicates and printing.
2335 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2337 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2338 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2339 and short u6 immediate.
2340 (check_if_valid_sleep_operand): Remove.
2341 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2343 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2345 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2346 flag_always_save_lp condition.
2347 * config/nds32/nds32.opt (malways-save-lp): New option.
2349 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2351 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2352 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2353 * config/nds32/nds32.h
2354 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2355 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2357 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2359 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2360 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2362 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2363 Chung-Ju Wu <jasonwucj@gmail.com>
2365 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2367 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2368 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2370 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2372 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2374 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2376 * config/nds32/nds32-protos.h (nds32_data_alignment,
2377 nds32_local_alignment): Declare.
2378 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2379 nds32_local_alignment): New functions.
2380 (TARGET_CONSTANT_ALIGNMENT): Define.
2381 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2383 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2385 * config/nds32/nds32.c
2386 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2387 (TARGET_MODES_TIEABLE_P): Likewise.
2389 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2391 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2394 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2395 Chung-Ju Wu <jasonwucj@gmail.com>
2397 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2398 for unaligned access.
2399 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2401 * config/nds32/nds32-intrinsic.md: Likewise.
2402 * config/nds32/nds32_intrinsic.h: Likewise.
2403 * config/nds32/nds32.h (nds32_builtins): Likewise.
2404 * config/nds32/nds32.opt (munaligned-access): New option.
2405 * config/nds32/nds32.c (nds32_asm_file_start): Display
2406 flag_unaligned_access status.
2408 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2410 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2411 -mno-relax is present.
2412 * config/riscv/linux.h (LINK_SPEC): Ditto.
2414 2018-04-20 Martin Sebor <msebor@redhat.com>
2417 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2419 (gimple_fold_builtin_stxcpy_chk): Same.
2420 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2422 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2425 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2426 __powikf2 when long double is IEEE 128-bit.
2428 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2430 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2431 step to make sure stack always aligned.
2433 2018-04-20 Carl Love <cel@us.ibm.com>
2436 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2437 size check for arg0.
2439 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2440 Tom de Vries <tom@codesourcery.com>
2443 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2444 Emit insns for calls too.
2445 (nvptx_find_par): Always look for worker-level predecessor insn.
2446 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2448 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2449 (nvptx_process_pars): Propagate frames for calls.
2451 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2454 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2456 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2457 (ix86_handle_option): Don't handle OPT_mibt.
2458 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2460 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2462 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2463 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2464 (ix86_target_macros): Define __CET__ with flag_cf_protection
2465 for -fcf-protection.
2466 * config/i386/i386.c (isa2_opts): Remove -mibt.
2467 * config/i386/i386.h (TARGET_IBT): Removed.
2468 (TARGET_IBT_P): Likewise.
2469 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2470 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2471 * config/i386/i386.opt (mcet): Update help message.
2474 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2475 -mcet as an alias for -mshstk.
2477 2018-04-20 Richard Biener <rguenther@suse.de>
2480 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2481 complexity by forcing a single use of the multiply operand.
2483 2018-04-20 Martin Jambor <mjambor@suse.cz>
2486 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2487 recursion dependency to only apply to non-clones.
2489 2018-04-20 Martin Jambor <mjambor@suse.cz>
2492 * ipa-cp.c (create_specialized_node): Check that clones of
2493 self-recursive edges exist during IPA-CP.
2495 2018-04-19 Toon Moene <toon@moene.org>
2497 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2500 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2502 PR tree-optimization/85467
2503 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2504 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2505 VECTOR_CST element to type.
2507 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2510 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2511 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2512 (builtin_longjmp): Likewise.
2513 (save_stack_nonlocal): New pattern.
2514 (restore_stack_nonlocal): Likewise.
2516 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2519 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2520 Replace ASM_OUTPUT_LABEL with fprintf.
2522 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2525 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2526 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2527 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2528 define __IBT__ and __SHSTK__ for -fcf-protection.
2529 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2531 (ix86_trampoline_init): Likewise.
2532 (x86_output_mi_thunk): Likewise.
2533 (ix86_notrack_prefixed_insn_p): Likewise.
2534 (ix86_option_override_internal): Don't disallow -fcf-protection.
2535 * config/i386/i386.md (rdssp<mode>): Also enable for
2537 (incssp<mode>): Likewise.
2538 (nop_endbr): Likewise.
2539 * config/i386/i386.opt (mcet): Change help message to built-in
2543 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2544 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2545 enable CET built-in functions.
2547 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2549 * common/config/i386/i386-common.c
2550 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2551 OPTION_MASK_ISA_MOVDIRI_UNSET,
2552 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2553 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2554 * config.gcc (movdirintrin.h): New header.
2555 * config/i386/cpuid.h (bit_MOVDIRI,
2556 bit_MOVDIR64B): New bits.
2557 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2559 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2560 (VOID, PVOID, PCVOID)): New function types.
2561 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2562 __builtin_ia32_directstoreu_u64,
2563 __builtin_ia32_movdir64b): New builtins.
2564 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2565 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2567 (ix86_valid_target_attribute_inner_p): Ditto.
2568 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2569 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2570 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2571 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2572 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2573 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2574 (movdiri<mode>, movdir64b_<mode>): New.
2575 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2576 * config/i386/immintrin.h: Include movdirintrin.h.
2577 * config/i386/movdirintrin.h: New file.
2578 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2580 2018-04-19 Richard Biener <rguenther@suse.de>
2583 * cfg.c (clear_bb_flags): When loop state says we have
2584 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2586 2018-04-19 Richard Biener <rguenther@suse.de>
2588 PR tree-optimization/84737
2589 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2590 copying restrict info.
2591 (vect_setup_realignment): Use it.
2592 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2593 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2594 the first DR to all generated stores.
2595 (vectorizable_load): Likewise for loads.
2597 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2599 PR tree-optimization/85446
2600 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2601 the integral and pointer types to have the same precision.
2603 * doc/install.texi: Document --disable-cet being the default and
2606 2018-04-18 Martin Liska <mliska@suse.cz>
2608 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2611 2018-04-18 Martin Liska <mliska@suse.cz>
2614 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2617 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2618 arguments if they are comparable.
2620 2018-04-18 Martin Liska <mliska@suse.cz>
2623 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2626 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2629 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2632 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2633 ENDBR after calling __morestack.
2635 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2638 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2639 by using gcc_base_ver to generate a gcc_driver_version, and use
2640 it when generating GCC_DRIVER_NAME.
2641 * configure: Regenerate.
2643 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2646 * config.gcc: Obsolete powerpc*-*-*spe*.
2648 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2651 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2652 (stabstr_D): Change type of unum from unsigned int to
2653 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2656 2018-04-17 Jim Wilson <jimw@sifive.com>
2659 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2660 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2661 Set arg_pointer_offset after using pretend_args_size.
2663 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2665 PR rtl-optimization/85431
2666 * dse.c (record_store): Ignore zero width stores.
2669 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2670 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2671 __builtin_stack_restore rather than after it.
2672 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2673 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2674 argument instead of virtual_dynamic_stack_rtx.
2676 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2678 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2680 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2681 Add note to error message to explain internal mapping of overloaded
2682 built-in function name to non-overloaded built-in function name.
2683 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2686 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2689 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2690 where the inputs overlap with the output.
2692 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2694 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2695 (=v, v) alternative and explicit "memory" attribute.
2696 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2697 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2699 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2700 "sselog1" type instead of "sselog".
2701 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2702 "sselog". Remove explicit "memory" attribute.
2703 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2704 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2706 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2707 "sselog1" type instead of "sselog". Remove explicit "memory"
2709 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2710 use "sselog1" type instead of "sselog". Remove explicit "memory"
2712 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2713 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2715 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2716 "sselog1" type instead of "sselog". Remove explicit "memory"
2718 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2719 use "sselog1" type instead of "sselog". Remove explicit "memory"
2723 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2726 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2729 2018-04-17 Martin Jambor <mjambor@suse.cz>
2732 * ipa-cp.c (create_specialized_node): Call
2733 expand_all_artificial_thunks if necessary.
2735 2018-04-17 Martin Liska <mliska@suse.cz>
2738 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2739 in message, remote space in between '_G' and '('.
2741 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2744 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2745 avx512f_vmcmp<mode>3<round_saeonly_name>,
2746 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2747 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2748 avx512f_rndscale<mode><round_saeonly_name>,
2749 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2750 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2751 Use %<iptr>2 instead of %2 for -masm=intel.
2752 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2753 avx512f_vcvttss2usi<round_saeonly_name>,
2754 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2756 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2757 avx512f_vcvttsd2usi<round_saeonly_name>,
2758 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2759 Use %q1 instead of %1 for -masm=intel.
2760 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2761 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2762 of %3 for -masm=intel.
2763 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2765 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2767 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2769 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2770 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2772 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2774 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2775 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2776 %g1 and one with %0 and %1.
2777 (avx512er_vmrcp28<mode><round_saeonly_name>,
2778 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2780 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2781 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2782 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2783 of %0 and %{%4%} for -masm=intel.
2784 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2785 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2786 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2787 order of %0 and %{%5%}%{z%} for -masm=intel.
2789 2018-04-17 Jan Hubicka <jh@suse.cz>
2792 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2794 2018-04-17 Martin Liska <mliska@suse.cz>
2797 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2798 for target_clone error message. Make default implementation
2799 clone to be a local declaration.
2800 (separate_attrs): Add new argument and check for an empty
2802 (expand_target_clones): Handle it.
2803 (ipa_target_clone): Make redirection just for target_clones
2806 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2807 Tom de Vries <tom@codesourcery.com>
2810 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2811 tiled basic blocks without omp continue statements.
2813 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2816 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2817 vec_extract expression as having side effects to make sure it gets
2820 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2823 * config/i386/i386.c (get_builtin_code_for_version): Check
2826 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2829 * gcc/config.gcc: Support "skylake".
2830 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2832 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2833 (processor_target_table): Add "skylake".
2834 (ix86_option_override_internal): Add "skylake".
2835 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2836 PROCESSOR_CANNONLAKE.
2837 (get_builtin_code_for_version): Fix priority for
2838 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2839 PROCESSOR_SKYLAKE-AVX512.
2840 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2841 (processor_type): Add PROCESSOR_SKYLAKE.
2843 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2844 Jason Merrill <jason@redhat.com>
2847 * convert.c (convert_to_integer_1): Use direct recursion for
2848 enumeral types and types with a precision less than the number
2849 of bits in their mode.
2851 2018-04-16 Julia Koval <julia.koval@intel.com>
2854 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2855 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2857 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2860 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2861 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2862 and -mno-direct-move.
2864 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2867 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2868 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2869 Ensure negative shifts result in {0}.
2871 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2873 PR rtl-optimization/79916
2874 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2875 regs (if any) to define how to gnerate SD moves when LRA is in
2878 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2880 PR rtl-optimization/85393
2881 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2882 * except.c (expand_dw2_landing_pad_for_region): Make static.
2883 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2884 a label and unconditional jump to old_bb, rather than
2885 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2888 PR rtl-optimization/85376
2889 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2890 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2891 instead of a specific value.
2893 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2894 Bin Cheng <bin.cheng@arm.com>
2896 PR tree-optimization/82965
2897 PR tree-optimization/83991
2898 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2899 by_profile_only parameter.
2900 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2901 information if the loop was predicted to iterate too many times.
2902 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2904 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2907 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2910 2018-04-13 Martin Liska <mliska@suse.cz>
2911 Jakub Jelinek <jakub@redhat.com>
2914 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2915 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2916 * builtins.c (expand_builtin_memory_copy_args): Use
2917 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2918 handle dest_addr == pc_rtx.
2920 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2923 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2924 asked to not generate direct moves.
2925 (fix_trunc<mode>si2_stfiwx): Similar.
2926 (fix_trunc<mode>si2_internal): Similar.
2928 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2931 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2932 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2933 lookup if dest in some wider mode is known to be const0_rtx and
2934 if so, record permanent equivalence for it to be ZERO_EXTEND of
2935 the narrower mode destination.
2937 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2939 * lto-streamer-out.c (output_function): Revert 259346.
2940 * omp-expand.c (expand_oacc_for): Likewise.
2942 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2944 PR rtl-optimization/85354
2945 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2946 * sel-sched.c (sel_global_init): ... here.
2948 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2951 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2952 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2953 mode for PE-COFF targets.
2954 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2955 (i386_pe_asm_lto_end): Likewise.
2956 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2957 (TARGET_ASM_LTO_END): Likewise.
2958 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2959 (i386_pe_asm_lto_start): New function.
2960 (i386_pe_asm_lto_end): Likewise.
2962 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2963 Richard Biener <rguenther@suse.de>
2966 * lto-streamer-out.c (output_function): Fix CFG loop state before
2968 * omp-expand.c (expand_oacc_for): Handle calls to internal
2969 functions like regular functions.
2971 2018-04-12 Richard Biener <rguenther@suse.de>
2974 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2975 for the early LTO debug to properly generate references to it
2976 during DIE emission. Do not re-use that for the skeleton for
2978 (dwarf2out_early_finish): Likewise.
2980 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2983 * config/i386/sse.md
2984 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2985 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2986 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2987 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2988 and output is a reg, avoid creating invalid lowpart subreg, but
2989 instead split into a 512-bit move. Don't split if not AVX512VL,
2990 input is xmm16+ reg and output is a mem.
2991 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2992 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2993 xmm16+ reg and output is a mem.
2995 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2997 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2998 also for flag_dwarf2_cfi_asm.
3000 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3002 PR rtl-optimization/85342
3003 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3004 a bool scalar var inside of the loop instead. Don't try to update
3005 recog_data.operand after failed apply_change_group.
3007 2018-04-12 Tom de Vries <tom@codesourcery.com>
3010 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3011 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3012 array with flexible array member as array without given dimension.
3013 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3014 argument for undefined param to true.
3016 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3019 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3020 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3021 from PowerPC section.
3022 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3023 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3024 help text that is too long.
3025 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3026 help text that is too long.
3027 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3028 help text that is too long.
3030 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3032 * config/alpha/alpha.md (stack_probe_internal): Rename
3033 from "probe_stack". Update all callers.
3035 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3037 PR rtl-optimization/84566
3038 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3039 sched_macro_fuse_insns.
3041 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3044 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3045 (compute_block_dependences): ... from here.
3047 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3049 PR tree-optimization/85331
3050 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3051 from int to HOST_WIDE_INT.
3053 2018-04-11 Martin Jambor <mjambor@suse.cz>
3056 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3057 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3058 not the same as the source val.
3059 (cgraph_edge_brings_value_p): New parameter.
3060 (gather_edges_for_value): Pass destination value to
3061 cgraph_edge_brings_value_p.
3062 (perhaps_add_new_callers): Likewise.
3063 (get_info_about_necessary_edges): Likewise and exclude values brought
3064 only by self-recursive edges.
3065 (create_specialized_node): Redirect only clones of self-calling edges.
3066 (+self_recursive_pass_through_p): New function.
3067 (find_more_scalar_values_for_callers_subset): Use it.
3068 (find_aggregate_values_for_callers_subset): Likewise.
3069 (known_aggs_to_agg_replacement_list): Removed.
3070 (decide_whether_version_node): Re-calculate known constants for all
3071 remaining context clones.
3073 2018-04-11 Richard Biener <rguenther@suse.de>
3076 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3077 from early DWARF output.
3078 (dwarf2out_early_finish): Output line info unconditionally into
3079 early DWARF and add reference to it.
3081 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3084 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3085 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3086 other than V2DFmode using iptr mode attribute.
3087 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3089 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3091 PR rtl-optimization/84659
3092 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3094 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3097 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3099 (output_loc_list): Pass address of a dummy size variable even in the
3100 locview handling loop.
3101 (index_location_lists): Add comment on why skip_loc_list_entry can't
3104 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3107 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3110 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3113 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3114 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3115 and -mstring-compare-inline-limit.
3117 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3120 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3121 for stack clash protection in a register whenever we need it to be in
3124 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3126 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3127 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3129 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3132 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3134 (mlong-double-): Ditto.
3135 * config/rs6000/sysv4.opt (msdata=): Ditto.
3136 (mtls-size=): Ditto.
3138 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3140 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3141 erroneous entries for
3142 "vector int vec_ldl (int, long int *)", and
3143 "vector unsigned int vec_ldl (int, unsigned long int *)".
3144 Add comments and entries for
3145 "vector bool char vec_ldl (int, bool char *)",
3146 "vector bool short vec_ldl (int, bool short *)",
3147 "vector bool int vec_ldl (int, bool int *)",
3148 "vector bool long long vec_ldl (int, bool long long *)",
3149 "vector pixel vec_ldl (int, pixel *)",
3150 "vector long long vec_ldl (int, long long *)",
3151 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3152 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3153 type tree bool_long_long_type_node and correct definition of
3154 bool_V2DI_type_node to make reference to this new type tree.
3155 (rs6000_mangle_type): Replace erroneous reference to
3156 bool_long_type_node with bool_long_long_type_node.
3157 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3158 comments to emphasize sign distinctions for char and int types and
3159 replace RS6000_BTI_bool_long constant with
3160 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3161 use of RS6000_BTI_pixel.
3162 (bool_long_type_node): Remove this macro definition.
3163 (bool_long_long_type_node): New macro definition
3165 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3167 PR rtl-optimization/85300
3168 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3169 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3170 simplify_unary_operation fails.
3172 2018-04-10 Martin Liska <mliska@suse.cz>
3174 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3175 cgraph_edge and ipa_ref.
3177 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3181 * config/i386/sse.md
3182 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3183 computation of the VEC_MERGE selector from mask.
3184 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3185 Fix decoding of the VEC_MERGE selector into mask.
3187 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3189 PR tree-optimization/85286
3190 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3192 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3194 * final.c (final_1): Set insn_last_address as well as
3195 insn_current_address.
3197 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3200 * explow.c (emit_stack_probe): Call validize_mem on memory location
3201 before passing it to gen_probe_stack. Create address operand and
3202 legitimize it for the probe_stack_address case.
3204 2018-04-09 Jan Hubicka <jh@suse.cz>
3207 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3208 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3209 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3211 (free_lang_data): Rebuild type inheritance graph.
3213 2018-04-09 Martin Sebor <msebor@redhat.com>
3215 * invoke.texi (-finline-small-functions): Mention other optimization
3217 (-findirect-inlining, -fpartial-inlining): Same.
3218 (-finline-functions-called-once): Same.
3219 (-freorder-blocks-and-partition): Same.
3221 2018-04-09 Jan Hubicka <jh@suse.cz>
3224 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3225 jumps; choose last target that matches the criteria (i.e.
3226 no partition changes for non-crossing jumps).
3227 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3228 support for redirecting crossing jumps to non-crossing.
3230 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3232 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3233 also for naked functions.
3235 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3237 * config/arc/arc.md (add_shift): New pattern.
3238 (add_shift2): Likewise.
3239 (sub_shift): Likewise.
3240 (sub_shift_cmp0_noout): Likewise.
3241 (compare_si_ashiftsi): Likewise.
3242 (xbfu_cmp0_noout): New combine pattern.
3243 (xbfu_cmp0"): Likewise.
3244 (movsi_set_cc_insn): Place the predicable variant first.
3245 (commutative_binary_cmp0_noout): Remove clobber.
3246 (commutative_binary_cmp0): New pattern.
3247 (noncommutative_binary_cmp0): Likewise.
3248 (noncommutative_binary_cmp0_noout): Likewise.
3249 (noncommutative_binary_comparison_result_used): Removed.
3250 (rsub_cmp0): New pattern.
3251 (rsub_cmp0_noout): Likewise.
3252 (extzvsi): Changed, keep only meaningful variants.
3253 (SQH, SEZ): New iterators.
3254 (SQH_postfix): New mode attribute.
3255 (SEZ_prefix): New code attribute.
3256 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3257 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3258 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3260 (noncommutative_operator): Check the availability of barrel
3263 2018-04-09 Richard Biener <rguenther@suse.de>
3265 PR tree-optimization/85284
3266 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3267 Only use the niter constraining form of simple_iv when the exit
3270 2018-04-09 Tom de Vries <tom@codesourcery.com>
3273 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3274 (define_expand "*memory_barrier"): New define_expand.
3275 (define_insn "memory_barrier"): New insn.
3277 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3279 PR rtl-optimization/80463
3280 PR rtl-optimization/83972
3281 PR rtl-optimization/83480
3283 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3284 correct producer for the insn.
3285 (tidy_control_flow): Fixup seqnos in case of debug insns.
3287 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3289 PR rtl-optimization/83913
3291 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3292 different sched-times when merging exprs.
3294 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3296 PR rtl-optimization/83962
3298 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3299 tidy_fallthru_edge and tidy_control_flow.
3301 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3303 PR rtl-optimization/83530
3305 * sel-sched.c (force_next_insn): New global variable.
3306 (remove_insn_for_debug): When force_next_insn is true, also leave only
3307 next insn in the ready list.
3308 (sel_sched_region): When the region wasn't scheduled, make another pass
3309 over it with force_next_insn set to 1.
3311 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3313 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3315 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3316 for interrupt control.
3317 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3318 functions for interrupt control.
3319 * config/nds32/nds32-intrinsic.md: Likewise.
3320 * config/nds32/nds32_intrinsic.h: Likewise.
3321 * config/nds32/nds32.h (nds32_builtins): Likewise.
3323 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3325 * config/nds32/nds32.c (nds32_init_machine_status,
3326 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3327 strict_aligned_p field.
3328 (nds32_expand_to_rtl_hook): New function.
3329 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3330 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3332 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3333 Chung-Ju Wu <jasonwucj@gmail.com>
3335 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3336 * config/nds32/nds32-n7.md: New file.
3337 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3338 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3340 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3341 * config/nds32/nds32.md (pipeline_model): Add n7.
3342 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3343 * config/nds32/pipelines.md: Include n7 settings.
3345 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3346 Chung-Ju Wu <jasonwucj@gmail.com>
3348 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3349 * config/nds32/nds32-e8.md: New file.
3350 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3351 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3353 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3354 * config/nds32/nds32.md (pipeline_model): Add e8.
3355 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3356 * config/nds32/pipelines.md: Include e8 settings.
3358 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3359 Chung-Ju Wu <jasonwucj@gmail.com>
3361 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3362 * config/nds32/nds32-n8.md: New file.
3363 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3364 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3366 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3367 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3368 * config/nds32/nds32.md (pipeline_model): Add n8.
3369 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3370 * config/nds32/pipelines.md: Include n8 settings.
3372 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3373 Chung-Ju Wu <jasonwucj@gmail.com>
3375 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3376 * config/nds32/nds32-n9-2r1w.md: New file.
3377 * config/nds32/nds32-n9-3r2w.md: New file.
3378 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3379 nds32_register_ports): New or modify for cpu n9.
3380 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3382 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3383 * config/nds32/nds32-utils.c: New file.
3384 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3385 TARGET_MUL_SLOW): Define.
3386 * config/nds32/nds32.md (pipeline_model): New attribute.
3387 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3388 New options that support cpu n9.
3389 * config/nds32/pipelines.md: Include n9 settings.
3390 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3392 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3394 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3395 information if necessary.
3396 (output_cond_branch_compare_zero): Likewise.
3397 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3398 (nds32_target_alignment): Refine for alignment.
3399 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3400 (FUNCTION_BOUNDARY): Modify.
3401 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3403 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3405 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3407 * config/nds32/constants.md (unspec_volatile_element): Add values for
3408 TLB operation and data prefetch.
3409 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3410 functions for TLB operation and data prefetch.
3411 * config/nds32/nds32-intrinsic.md: Likewise.
3412 * config/nds32/nds32_intrinsic.h: Likewise.
3413 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3414 (nds32_print_operand): Likewise.
3415 * config/nds32/nds32.h (nds32_builtins): Likewise.
3417 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3418 Andrew Pinski <pinsika@gcc.gnu.org>
3421 * match.pd: Use constant_boolean_node of correct type instead of
3422 boolean_true_node or boolean_false_node for simplifying
3423 pointer comparisons to zero.
3425 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3427 PR tree-optimization/80021
3428 * tree.c (verify_type_variant): Make error call in verify_variant_match
3429 translatable and remove final full stop.
3431 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3433 * config/nds32/constants.md (unspec_volatile_element): Add
3434 UNSPEC_VOLATILE_EH_RETURN.
3435 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3436 nds32_output_stack_pop): Support dwarf exception handling process.
3437 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3438 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3439 exception handling process.
3440 (nds32_compute_stack_frame): Likewise.
3441 (nds32_return_addr_rtx): Likewise.
3442 (nds32_initial_elimination_offset): Likewise.
3443 (nds32_expand_prologue): Likewise.
3444 (nds32_expand_epilogue): Likewise.
3445 (nds32_dynamic_chain_address): New function.
3446 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3448 (DYNAMIC_CHAIN_ADDRESS): Define.
3449 (EH_RETURN_DATA_REGNO): Define.
3450 (EH_RETURN_STACKADJ_RTX): Define.
3451 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3452 patterns for dwarf exception handling.
3454 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3456 * config/nds32/nds32.h: Clean up obsolete macros.
3458 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3460 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3461 Add enum values for particular instructions.
3462 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3463 particular intrinsic functions.
3464 * config/nds32/nds32-intrinsic.md: Likewise.
3465 * config/nds32/nds32_intrinsic.h: Likewise.
3466 * config/nds32/nds32.h (nds32_builtins): Likewise.
3467 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3468 (btst, ave): New patterns for particular instructions.
3470 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3472 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3473 Add enum values for atomic load/store and memory sync.
3474 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3476 * config/nds32/nds32-intrinsic.md: Likewise.
3477 * config/nds32/nds32_intrinsic.h: Likewise.
3478 * config/nds32/nds32.h (nds32_builtins): Likewise.
3480 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3482 PR tree-optimization/85257
3483 * fold-const.c (native_encode_vector): If not all elts could fit
3484 and off is -1, return 0 rather than offset.
3485 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3486 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3487 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3488 adjust buffer in native_interpret_expr call.
3490 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3492 * config/nds32/constants.md (unspec_volatile_element): Add cache
3493 control enum values.
3494 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3495 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3496 * config/nds32/nds32.c (nds32_cctl_names): New.
3497 (nds32_print_operand): Handle cache control register names.
3498 * config/nds32/nds32.h (nds32_builtins): New enum values.
3499 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3501 * config/nds32/nds32.md (type): Add mmu.
3502 * config/nds32/pipelines.md (simple_insn): Add mmu.
3504 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3506 * config/nds32/nds32.md (type): Remove call.
3507 * config/nds32/pipelines.md (simple_insn): Likewise.
3509 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3511 * config/nds32/constants.md (unspec_volatile_element): Add
3512 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3513 UNSPEC_VOLATILE_FMFCFG.
3514 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3515 description for fmfcfg and fmfcsr.
3516 (bdesc_1arg): Add fmtcsr.
3517 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3518 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3519 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3520 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3521 unspec_fmfcfg): New patterns.
3522 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3523 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3524 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3525 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3526 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3527 __nds32__fmfcfg): Define.
3529 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3531 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3532 intrinsic register names.
3533 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3534 intrinsic register enum values and macros.
3536 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3538 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3539 for load/store addressing form.
3540 (nds32_print_operand_address): Likewise.
3542 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3545 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3546 based on LABEL_REF. Remove useless assertion.
3547 (pic_address_needs_scratch): Fix formatting.
3548 (sparc_legitimize_pic_address): Minor tweaks.
3549 (sparc_delegitimize_address): Adjust assertion accordingly.
3550 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3551 into symbolic_operand.
3552 (movsi_high_pic_label_ref): Likewise.
3553 (movsi_lo_sum_pic_label_ref): Likewise.
3554 (movdi_pic_label_ref): Likewise.
3555 (movdi_high_pic_label_ref): Likewise.
3556 (movdi_lo_sum_pic_label_ref): Likewise.
3558 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3560 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3561 custom LIB_SPEC setup.
3563 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3564 Kito Cheng <kito.cheng@gmail.com>
3566 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3567 * config/riscv/freebsd.h: New.
3569 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3571 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3572 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3575 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3576 Kito Cheng <kito.cheng@gmail.com>
3578 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3579 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3580 * config/nds32/nds32-protos.h (nds32_output_call,
3581 nds32_output_return): Declare.
3582 * config/nds32/nds32.md: Refine all the call and return patterns.
3584 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3587 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3588 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3590 PR rtl-optimization/84872
3591 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3592 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3595 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3597 * expr.c (copy_blkmode_to_reg): Revert 254862.
3598 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3600 2018-04-06 Richard Biener <rguenther@suse.de>
3603 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3604 after seeing a component reference with an adjacent field. Treat
3605 refs to arrays at struct end of external decls similar to
3606 refs to unconstrained commons.
3608 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3611 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3612 look through SAVE_EXPRs with non-side-effects argument. Adjust
3614 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3617 2018-04-06 Richard Biener <rguenther@suse.de>
3620 * alias.c (find_base_term): New wrapper around find_base_term
3621 unwinding CSELIB_VAL_PTR changes.
3622 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3625 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3627 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3629 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3630 constant definitions.
3631 ("nop"): lr 0,0 -> nopr r0
3632 ("nop_lr0", "nop_lr1"): New insn definitions.
3634 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3636 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3637 NDS32_V3PUSH_AVAILABLE_P macro.
3639 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3640 Chung-Ju Wu <jasonwucj@gmail.com>
3642 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3643 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3644 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3645 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3646 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3647 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3648 * config/nds32/constraints.md: New constraints and checking for hard
3649 float configuration.
3650 * config/nds32/iterators.md: New mode iterator and attribute for hard
3651 float configuration.
3652 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3654 * config/nds32/nds32-fpu.md: New file.
3655 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3656 deal with hard float code generation.
3657 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3659 (abi_type, float_reg_number): New enum type.
3660 * config/nds32/nds32-predicates.c: New predicates for hard float.
3661 * config/nds32/nds32-protos.h: Declare functions for hard float.
3662 * config/nds32/nds32.c: Implementation for hard float configuration.
3663 * config/nds32/nds32.h: Definitions for hard float configuration.
3664 * config/nds32/nds32.md: Include hard float machine description and
3665 modify patterns for hard float configuration.
3666 * config/nds32/nds32.opt: New options for hard float configuration.
3667 * config/nds32/predicates.md: New predicates for hard float
3670 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3672 * common/config/nds32/nds32-common.c
3673 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3675 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3678 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3679 CONSTRUCTOR_ELT (ctor, ...)->value.
3681 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3684 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3686 2018-04-05 Tom de Vries <tom@codesourcery.com>
3689 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3692 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3693 Kito Cheng <kito.cheng@gmail.com>
3695 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3696 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3697 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3698 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3700 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3701 Kito Cheng <kito.cheng@gmail.com>
3703 * config/nds32/constraints.md (Ufe): New memory constraint.
3704 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3705 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3706 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3708 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3709 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3711 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3713 * config/nds32/nds32.md: Use optimize_size in the condition for
3714 alu-shift instructions.
3716 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3718 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3720 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3722 * config/nds32/nds32.md (negsi2): Refine pattern.
3724 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3725 Chung-Ju Wu <jasonwucj@gmail.com>
3727 * config/nds32/iterators.md (shift_rotate): New code iterator.
3728 (shift): New code attribute.
3729 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3730 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3731 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3732 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3733 bit-wise operations.
3734 (andsi3, *andsi3): Ditto.
3735 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3736 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3737 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3738 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3739 nds32_ior_operand, nds32_xor_operand): New predicates.
3741 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3743 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3744 (addsi3, subsi3): ... this.
3746 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3748 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3750 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3752 * config/nds32/nds32.md: Adjust indention.
3754 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3756 * config/nds32/nds32.md (feature): New attribute.
3758 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3760 * config/nds32/nds32.md (subtype): New attribute.
3762 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3765 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3766 expansion to perform a bitwise AND of the argument followed by a
3767 boolean negation of the result.
3769 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3771 PR rtl-optimization/84878
3772 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3773 the basic block. Assert the use reference is not artificial and that
3774 it has an associated insn.
3776 2018-04-04 Michael Matz <matz@suse.de>
3778 * builtins.c (compute_objsize): Pass correct operand
3779 to array_at_struct_end_p.
3781 2018-04-04 Richard Biener <rguenther@suse.de>
3784 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3785 from contexts for DINFO_LEVEL_TERSE and below.
3787 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3789 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3790 resiter_operand condition.
3791 * config/nds32/nds32.md (*move<mode>): Ditto.
3793 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3794 Monk Chiang <sh.chiang04@gmail.com>
3796 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3798 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3800 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3802 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3803 Kito Cheng <kito.cheng@gmail.com>
3805 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3806 nds32_cond_code_str, output_cond_branch,
3807 output_cond_branch_compare_zero, nds32_expand_cbranch,
3808 nds32_expand_cstore, nds32_expand_movcc,
3809 nds32_output_cbranchsi4_equality_zero,
3810 nds32_output_cbranchsi4_equality_reg,
3811 nds32_output_cbranchsi4_equality_reg_or_const_int,
3812 nds32_output_cbranchsi4_greater_less_zero: New functions.
3813 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3814 nds32_expand_cstore, nds32_expand_movcc,
3815 nds32_output_cbranchsi4_equality_zero,
3816 nds32_output_cbranchsi4_equality_reg,
3817 nds32_output_cbranchsi4_equality_reg_or_const_int,
3818 nds32_output_cbranchsi4_greater_less_zero): Declare.
3819 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3820 nds32_rimm11s_operand): New predicates.
3821 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3822 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3825 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3827 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3828 * config/nds32/nds32.md: Ditto.
3829 * config/nds32/pipelines.md: Ditto.
3831 2018-04-04 Richard Biener <rguenther@suse.de>
3833 PR tree-optimization/85168
3834 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3835 propagating abnormals.
3837 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3839 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3841 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3842 Kito Cheng <kito.cheng@gmail.com>
3844 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3845 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3846 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3847 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3848 * config/nds32/nds32.md (sibcall_internal): New.
3849 (sibcall_register): Remove.
3850 (sibcall_immediate): Remove.
3851 (sibcall_value_internal): New.
3852 (sibcall_value_register): Remove.
3853 (sibcall_value_immediate): Remove.
3854 * config/nds32/predicates.md (nds32_general_register_operand): New.
3855 (nds32_call_address_operand): New.
3857 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3859 PR rtl-optimization/85167
3860 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3861 bb_defs if *split_p, instead preinitialize it to NULL.
3863 PR tree-optimization/85156
3864 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3865 evaluating the argument multiple times.
3867 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3869 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3871 (_mm_cvtpd_ps): Likewise.
3872 (_mm_cvttpd_epi32): Likewise.
3873 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3874 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3875 vector, pixel, and bool following altivec.h include.
3877 2018-04-03 Martin Sebor <msebor@redhat.com>
3879 * doc/extend.texi (Common Function Attributes): Clarify.
3880 (const attribute): Likewise.
3881 (pure attribute): Likewise.
3883 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3886 * config/i386/i386.c (ix86_expand_vector_set): Use
3887 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3889 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3891 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3892 instructions when changing rounding bits to preserve precision bits
3893 in the x87 control word.
3895 2018-04-03 Martin Liska <mliska@suse.cz>
3897 PR tree-optimization/82491
3898 * rtl.h (strip_offset_and_add): Replace += suboffset with
3899 poly_uint64 () + suboffset.
3901 2018-03-29 Martin Liska <mliska@suse.cz>
3902 Martin Jambor <mjambor@suse.cz>
3905 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3906 param_type is not an integral or pointer type.
3908 2018-04-03 Richard Biener <rguenther@suse.de>
3910 * sese.h (recompute_all_dominators): Remove.
3912 2018-04-02 Martin Sebor <msebor@redhat.com>
3914 * doc/invoke.texi (-Wrestrict): Fix typos.
3916 2018-04-02 Jim Wilson <jimw@sifive.com>
3918 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3919 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3920 (<optab>di3, <optab>si3_extend): Likewise.
3921 (<optab>si3_mask, <optab>si3_mask_1): New.
3922 (<optab>di3_mask, <optab>di3_mask_1): New.
3923 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3924 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3925 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3927 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3929 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3932 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3934 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3935 (nds32_canonicalize_comparison): New function.
3937 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3938 Kito Cheng <kito.cheng@gmail.com>
3939 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3941 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3942 * config/nds32/constants.md (unspec_volatile_element): Add
3943 UNSPEC_VOLATILE_RELAX_GROUP.
3944 * config/nds32/nds32-relax-opt.c: New file.
3945 * config/nds32/nds32-predicates.c
3946 (nds32_symbol_load_store_p): New function.
3947 * config/nds32/nds32-protos.h
3948 (nds32_symbol_load_store_p): Declare function.
3949 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3950 * config/nds32/nds32.c
3951 (nds32_register_pass): New function to register pass.
3952 (nds32_register_passes): New function to register passes.
3953 * config/nds32/nds32.md (relax_group): New pattern.
3954 * config/nds32/nds32.opt (mrelax-hint): New option.
3955 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3957 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3959 * config/nds32/t-nds32: Modify files dependency.
3961 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3963 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3964 (PROFILE_HOOK): Define its implementation.
3966 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3968 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3969 type and 32-bit size.
3971 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3974 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3975 (V_128_256): New mode iterator.
3976 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3977 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3978 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3980 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3981 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3983 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3986 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3987 NaN inputs correctly.
3989 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3992 * config/rs6000/vsx.md (??r): New mode attribute.
3993 (*vsx_mov<mode>_64bit): Use it.
3994 (*vsx_mov<mode>_32bit): Likewise.
3996 2018-03-30 Martin Sebor <msebor@redhat.com>
3998 PR tree-optimization/84818
3999 * builtins.c (check_access): Use warning_n.
4001 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4004 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4006 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4009 2018-03-30 Julia Koval <julia.koval@intel.com>
4012 * x86-tune.def (movx, partial_reg_dependency): Enable for
4015 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4018 * lra-constraints.c (process_alt_operands): Move setting
4019 this_alternative_matches below.
4021 2018-03-29 Martin Liska <mliska@suse.cz>
4024 * doc/invoke.texi: Document how LTO works with debug info.
4025 Describe auto-load support of binutils. Mention 'x86-64'
4026 as valid option value of -march option.
4028 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4030 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4033 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4034 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4035 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4038 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4041 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4042 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4043 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4044 for RS6000_BTM_POWERPC64.
4045 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4046 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4047 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4052 2018-03-28 Carl Love <cel@us.ibm.com>
4055 2017-09-27 Carl Love <cel@us.ibm.com>
4057 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4058 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4059 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4062 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4064 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4065 instead of __vector bool.
4066 (_mm_max_pu8): Likewise.
4067 (_mm_min_pi16): Likewise.
4069 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4072 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4073 (DIVWEUO): Likewise.
4075 (DIVDEUO): Likewise.
4076 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4077 DIVWEUO and DIVDEUO.
4078 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4079 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4080 (div_extend): Likewise.
4081 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4083 (__builtin_divweuo): Likewise.
4084 (__builtin_divdeo): Likewise.
4085 (__builtin_divdeuo): Likewise.
4087 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4090 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4091 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4093 PR tree-optimization/82004
4094 * gimple-match-head.c (optimize_pow_to_exp): New function.
4095 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4096 Don't fold to exp if optimize_pow_to_exp is false.
4098 2018-03-28 Martin Liska <mliska@suse.cz>
4101 * calls.c (initialize_argument_information): Fix trailing space.
4102 * common.opt: Fix typo and provide better explanation for
4103 -fsanitize-coverage option.
4104 * config/i386/i386.opt: Fix typo.
4106 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4107 Martin Liska <mliska@suse.cz>
4110 * gimplify.c (asan_poison_variable): Don't do the check for
4111 gimplify_omp_ctxp here.
4112 (gimplify_decl_expr): Do it here.
4113 (gimplify_target_expr): Likewise.
4115 2018-03-28 Martin Liska <mliska@suse.cz>
4118 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4119 chkp_type_bounds_count if MPX is not enabled.
4121 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4123 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4125 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4128 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4129 function to create the function decl for complex long double
4130 multiply and divide for -mabi=ieeelongdouble.
4131 (init_float128_ieee): Call it.
4133 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4136 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4137 -fcf-protection=branch -mibt.
4138 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4140 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4143 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4145 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4148 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4149 extern array declarations.
4151 2018-03-27 Richard Biener <rguenther@suse.de>
4154 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4155 explicit single_use checks.
4157 2018-03-27 Richard Biener <rguenther@suse.de>
4159 PR tree-optimization/85082
4160 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4163 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4165 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4166 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4167 Turn on fasynchronous-unwind-tables and funwind-tables.
4169 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4172 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4173 (*bmi_blsr_<mode>_ccz): Ditto.
4175 2018-03-26 Tom de Vries <tom@codesourcery.com>
4177 PR tree-optimization/85063
4178 * omp-general.c (offloading_function_p): New function. Factor out
4180 * omp-offload.c (pass_omp_target_link::gate): ... here.
4181 * omp-general.h (offloading_function_p): Declare.
4182 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4183 with attribute omp declare target for offloading functions.
4185 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4187 PR tree-optimization/84005
4188 * tree-data-ref.h (get_base_for_alignment): Declare.
4189 * tree-data-ref.c (get_base_for_alignment_1): New function.
4190 (get_base_for_alignment): Likewise.
4191 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4192 get_base_for_alignment to find a suitable base object, instead
4193 of always using drb->base_address.
4195 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4198 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4199 known size by default.
4201 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4204 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4205 and non BLKmode operands.
4207 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4210 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4211 Clean up attributes.
4213 2018-03-23 Richard Biener <rguenther@suse.de>
4216 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4217 we are going to emit early debug for LTO.
4219 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4222 * function.c (match_asm_constraints_1): Don't optimize if input
4223 doesn't satisfy general_operand predicate for output's mode.
4226 * alias.c (write_dependence_p): Don't require for x_canonicalized
4227 non-VOIDmode if x has VOIDmode.
4230 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4231 just don't try to optimize it rather than assert it never happens.
4233 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4235 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4236 macro expansions for definition of ST_INTERNAL_<mode> and
4237 LD_INTERNAL_<mode> builtins.
4238 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4240 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4242 (altivec_expand_st_builtin): Likewise.
4243 (altivec_expand_builtin): Remove calls to deleted functions.
4244 (rs6000_address_for_altivec): Delete this function.
4245 * config/rs6000/vector.md: Remove expands for
4246 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4248 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4251 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4252 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4253 re-computing once computed.
4254 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4255 (arm_init_machine_status): Initialize
4256 machine->static_chain_stack_bytes.
4258 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4261 * doc/extend.texi: Add four new prototypes for vec_ld.
4262 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4263 definitions for more logical presentation.
4264 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4265 entries for V1TI variants of __builtin_altivec_ld builtin.
4266 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4267 handling of V1TI variant of LVX icode pattern.
4268 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4269 (rs6000_gimple_fold_builtin): Likewise.
4270 (altivec_init_builtins): Add code to define
4271 __builtin_altivec_lvx_v1ti function.
4273 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4276 * function.c (match_asm_constraints_1): Don't do the optimization
4277 if input isn't a REG, SUBREG, MEM or constant.
4279 2018-03-22 Tom de Vries <tom@codesourcery.com>
4281 PR tree-optimization/84956
4282 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4283 bb_has_abnormal_pred.
4285 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4288 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4289 DECL_INITIAL (decl) to decl at the end.
4290 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4293 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4295 * doc/extend.texi (__builtin_tgmath): Document when complex
4296 integer types are treated as _Complex _Float64.
4298 2018-03-21 Tom de Vries <tom@codesourcery.com>
4300 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4302 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4304 PR tree-optimization/84960
4305 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4306 if it is ENTRY block, move them into single succ of ENTRY in that case.
4308 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4310 PR tree-optimization/84811
4311 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4312 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4313 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4314 (known_subrange_p): Update accordingly. Cast each value involved
4315 in the size comparison, rather than casting the result of the
4318 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4320 PR tree-optimization/84982
4321 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4322 by flipping the least significant bit rather than all bits from
4323 bitpos to bitpos + bitsize - 1.
4325 2018-03-21 Nathan Sidwell <nathan@acm.org>
4327 * doc/extend.texi (Deprecated Features): Remove mention of
4328 long-deleted deprecations.
4330 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4333 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4334 * configure: Regenerate.
4336 2018-03-21 Tom de Vries <tom@codesourcery.com>
4338 PR tree-optimization/83126
4339 * tree-parloops.c (num_phis): New function.
4340 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4342 2018-03-21 Nathan Sidwell <nathan@acm.org>
4344 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4345 mention anon-struct/union members and trailing attributes.
4347 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4349 PR tree-optimization/84969
4350 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4351 builtin memset partitions if they set different rhs values.
4353 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4355 PR rtl-optimization/84989
4356 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4357 VEC_DUPLICATE with scalar result mode.
4359 2018-03-21 Martin Liska <mliska@suse.cz>
4362 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4363 not intended return statement.
4365 2018-03-21 Martin Liska <mliska@suse.cz>
4368 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4369 (chkp_find_bound_slots_1): Limit number of iterations.
4371 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4374 * Minor grammar fixes for x86 options.
4376 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4379 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4380 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4382 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4385 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4386 (altivec_lvx_<mode>_1op): Likewise.
4387 (altivec_stvx_<mode>_2op): Likewise.
4388 (altivec_stvx_<mode>_1op): Likewise.
4389 (altivec_lvx_<VM2:mode>): New define_expand.
4390 (altivec_stvx_<VM2:mode>): Likewise.
4391 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4392 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4393 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4394 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4395 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4396 (rs6000_gen_lvx): Likewise.
4397 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4398 (altivec_expand_stv_builtin): Likewise.
4399 (altivec_expand_builtin): Likewise.
4400 * config/rs6000/vector.md: Likewise.
4402 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4405 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4408 2018-03-20 Richard Biener <rguenther@suse.de>
4411 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4412 sign-conversions as zero, fall back to standard scalar_stmt
4415 2018-03-20 Martin Liska <mliska@suse.cz>
4418 * predict.c (rebuild_frequencies): Handle case when we have
4419 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4421 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4424 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4425 flag_section_anchors.
4426 * varasm.c (use_blocks_for_decl_p): Remove hack for
4427 dw2_force_const_mem.
4430 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4432 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4433 be created, use lowpart_subreg of operands[0] rather than operands[0]
4435 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4436 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4437 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4438 and n constraint instead of aarch64_shift_imm_di and Usd.
4439 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4440 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4442 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4445 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4446 to favor GPR over NEON registers.
4447 (<shift>di3_neon): Likewise.
4449 2018-03-20 Tom de Vries <tom@codesourcery.com>
4452 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4453 (nvptx_process_pars): Emit bar.sync asap and alap.
4455 2018-03-20 Tom de Vries <tom@codesourcery.com>
4458 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4459 seen_label if seen_label is already set.
4461 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4464 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4465 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4466 Use 1U instead of 1. Formatting fixes.
4469 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4470 instead of TREE_TYPE (s1) for the return value.
4472 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4474 PR tree-optimization/84946
4475 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4476 bitsize + bitsize in poly_uint64 rather than poly_int64.
4479 * dwarf2asm.c: Include fold-const.c.
4480 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4481 of decl rather than decl itself.
4483 PR rtl-optimization/84643
4484 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4486 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4489 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4490 calling assemble_variable.
4492 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4495 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4496 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4498 2018-03-19 Jim Wilson <jimw@sifive.com>
4501 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4502 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4503 (riscv_first_stack_step): Likewise.
4504 (riscv_option_override): Use STACK_BOUNDARY instead of
4506 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4508 (BIGGEST_ALIGNMENT): Set to 128.
4509 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4510 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4513 2018-03-19 Richard Biener <rguenther@suse.de>
4515 PR tree-optimization/84933
4516 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4517 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4519 2018-03-19 Richard Biener <rguenther@suse.de>
4521 PR tree-optimization/84859
4522 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4523 (cond_if_else_store_replacement): Perform sinking operation on
4524 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4525 Generalize what a BB with a single eligible store is.
4527 2018-03-19 Richard Biener <rguenther@suse.de>
4529 PR tree-optimization/84929
4530 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4531 chrec_is_positive against non-chrec arg.
4533 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4536 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4538 2018-03-18 Martin Liska <mliska@suse.cz>
4540 PR rtl-optimization/84635
4541 * regrename.c (build_def_use): Use matches_mode only when
4544 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4546 PR tree-optimization/84913
4547 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4548 vectorize chains of COND_EXPRs.
4550 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4552 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4554 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4556 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4558 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4560 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4562 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4563 Kito Cheng <kito.cheng@gmail.com>
4565 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4566 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4567 (nds32_adjust_reg_alloc_order): New function.
4568 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4570 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4572 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4573 nds32_print_operand, nds32_print_operand_address): Use
4574 HOST_WIDE_INT_PRINT_DEC instead.
4576 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4578 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4580 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4583 * config/i386/i386.c (initial_ix86_tune_features,
4584 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4586 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4587 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4588 rather than 1u << ix86_tune. Formatting fix.
4589 (ix86_option_override_internal): Change ix86_arch_mask from
4590 unsigned int to unsigned HOST_WIDE_INT, initialize to
4591 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4592 (ix86_function_specific_restore): Likewise.
4594 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4597 * postreload.c (reload_combine_recognize_pattern): Perform
4598 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4599 truncate_int_for_mode the result for the destination's mode.
4602 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4603 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4605 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4608 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4611 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4612 regno_allocno_class_array and sorted_pseudos.
4613 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4614 insns where regno is used.
4616 2018-03-16 Martin Liska <mliska@suse.cz>
4619 * multiple_target.c (create_dispatcher_calls): Redirect
4620 reference in the symbol table.
4622 2018-03-16 Martin Liska <mliska@suse.cz>
4625 * multiple_target.c (create_dispatcher_calls): Redirect also
4628 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4632 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4633 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4634 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4636 2018-03-16 Julia Koval <julia.koval@intel.com>
4638 * doc/invoke.texi (Skylake Server): Add CLWB.
4639 Cannonlake): Remove CLWB.
4641 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4643 PR tree-optimization/84841
4644 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4646 (FLOAT_ONE_CONST_TYPE): Define.
4647 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4648 (sort_by_operand_rank): Put entries with higher constant_type last
4649 rather than first to match comments.
4651 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4653 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4656 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4659 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4660 check and use address of target always.
4662 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4665 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4666 (indirect_thunk_bnd_needed): Likewise.
4667 (indirect_thunks_used): Likewise.
4668 (indirect_thunks_bnd_used): Likewise.
4669 (indirect_return_needed): New.
4670 (indirect_return_bnd_needed): Likewise.
4671 (output_indirect_thunk_function): Add a bool argument for
4673 (output_indirect_thunk_function): Don't generate alias for
4674 function return thunk.
4675 (ix86_code_end): Call output_indirect_thunk_function to generate
4676 function return thunks.
4677 (ix86_output_function_return): Set indirect_return_bnd_needed
4678 and indirect_return_needed instead of indirect_thunk_bnd_needed
4679 and indirect_thunk_needed.
4681 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4683 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4684 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4685 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4687 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4688 Paul Hua <paul.hua.gm@gmail.com>
4691 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4693 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4695 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4696 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4699 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4702 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4703 instead of GET_MODE_SIZE when comparing Units.
4705 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4708 * varasm.c (hash_section): Return an unchangeble hash value
4709 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4710 Return !aarch64_can_use_per_function_literal_pools_p ().
4712 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4715 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4716 rather than address of cmode as last argument to prepare_cmp_insn.
4718 2018-03-15 Julia Koval <julia.koval@intel.com>
4720 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4721 F_AVX512VNNI, F_AVX512BITALG): New.
4723 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4726 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4727 insn for floating-point loads and stores.
4729 2018-03-14 Carl Love <cel@us.ibm.com>
4731 * config/rs6000/rs6000-c.c: Add macro definitions for
4732 ALTIVEC_BUILTIN_VEC_PERMXOR.
4733 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4734 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4735 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4736 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4738 * config/doc/extend.texi: Add prototypes for vec_permxor.
4740 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4743 * diagnostic-show-locus.c (class layout_point): Convert m_line
4744 from int to linenum_type.
4745 (line_span::comparator): Use linenum "compare" function when
4746 comparing line numbers.
4747 (test_line_span): New function.
4748 (layout_range::contains_point): Convert param "row" from int to
4750 (layout_range::intersects_line_p): Likewise.
4751 (layout::will_show_line_p): Likewise.
4752 (layout::print_source_line): Likewise.
4753 (layout::should_print_annotation_line_p): Likewise.
4754 (layout::print_annotation_line): Likewise.
4755 (layout::print_leading_fixits): Likewise.
4756 (layout::annotation_line_showed_range_p): Likewise.
4757 (struct line_corrections): Likewise for field m_row.
4758 (line_corrections::line_corrections): Likewise for param "row".
4759 (layout::print_trailing_fixits): Likewise.
4760 (layout::get_state_at_point): Likewise.
4761 (layout::get_x_bound_for_row): Likewise.
4762 (layout::print_line): Likewise.
4763 (diagnostic_show_locus): Likewise for locals "last_line" and
4765 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4766 * input.c (selftest::test_linenum_comparisons): New function.
4767 (selftest::input_c_tests): Call it.
4768 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4769 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4770 * selftest.h (ASSERT_GT): New macro.
4771 (ASSERT_GT_AT): New macro.
4772 (ASSERT_LT): New macro.
4773 (ASSERT_LT_AT): New macro.
4775 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4777 PR rtl-optimization/84780
4778 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4780 2018-03-14 Martin Liska <mliska@suse.cz>
4782 * tree.c (record_node_allocation_statistics): Use
4783 get_stats_node_kind.
4784 (get_stats_node_kind): New function extracted from
4785 record_node_allocation_statistics.
4786 (free_node): Use get_stats_node_kind.
4788 2018-03-14 Richard Biener <rguenther@suse.de>
4790 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4791 that the value-set of ANTIC_IN doesn't grow.
4794 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4796 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4797 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4799 2018-03-14 Julia Koval <julia.koval@intel.com>
4801 * config.gcc (icelake-client, icelake-server): New.
4803 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4804 (initial_ix86_arch_features): Ditto.
4805 (PTA_SKYLAKE): Add SGX.
4806 (PTA_ICELAKE): Remove.
4807 (PTA_ICELAKE_CLIENT): New.
4808 (PTA_ICELAKE_SERVER): New.
4809 (ix86_option_override_internal): Split up icelake on icelake client and
4811 (get_builtin_code_for_version): Ditto.
4812 (fold_builtin_cpu): Ditto.
4813 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4814 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4815 * config/i386/i386.h (processor_type): Ditto.
4816 * doc/invoke.texi: Ditto.
4818 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4821 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4822 INTEGER_CST offset, add it together with bitpos / 8 and
4823 sign extend based on POINTER_SIZE.
4827 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4830 * config/i386/constraints.md (Yc): New register constraint.
4831 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4832 Use Yc constraint for alternative 2 of operand 0. Remove
4833 preferred_for_speed attribute.
4835 2018-03-14 Richard Biener <rguenther@suse.de>
4837 PR tree-optimization/84830
4838 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4839 with the old one to avoid oscillations.
4841 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4844 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4846 (assign_by_spills): Return a flag of reload assignment failure.
4847 Do not process the reload assignment failures. Do not spill other
4848 reload pseudos if they has the same reg class. Update n if
4850 (lra_assign): Add a return arg. Set up from the result of
4851 assign_by_spills call.
4852 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4853 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4854 usage_insns if it is not NULL.
4855 (spill_hard_reg_in_range): New function.
4856 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4857 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4858 function prototypes.
4859 (lra_assign): Change prototype.
4860 * lra.c (lra): Add code to deal with fails by splitting hard reg
4863 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4865 * config/riscv/riscv.opt (mrelax): New option.
4866 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4867 "norelax" when riscv_mrelax is disabled.
4868 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4870 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4873 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4874 reassociation for int modes.
4876 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4878 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4879 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4881 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4882 * config/aarch64/aarch64-sve.md
4883 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4884 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4885 (*extend<mode><Vwide>2): Rename to...
4886 (aarch64_sve_extend<mode><Vwide>2): ...this.
4887 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4888 renaming the old pattern to...
4889 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4891 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4892 define_expand, renaming the old pattern to...
4893 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4894 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4895 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4896 account when deciding which SVE instruction the optab should use.
4897 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4899 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4901 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4902 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4903 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4904 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4905 (tlsdesc_small_<mode>): Turn a define_expand and use
4906 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4907 (tlsdesc_small_advsimd_<mode>): ...this.
4908 (tlsdesc_small_sve_<mode>): New pattern.
4910 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4912 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4913 (UNSPEC_UMUL_HIGHPART): New constants.
4914 (MUL_HIGHPART): New int iteraor.
4915 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4916 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4918 (*<su>mul<mode>3_highpart): New define_insn.
4920 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4923 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4926 2018-03-13 Martin Liska <mliska@suse.cz>
4929 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4931 (sem_item_optimizer::~sem_item_optimizer): Release it.
4932 (sem_item_optimizer::merge_classes): Register variable aliases.
4933 (sem_item_optimizer::fixup_pt_set): New function.
4934 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4935 * ipa-icf.h: Declare new variables and functions.
4937 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4940 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4941 integer_pow2p@2 and test integer_pow2p in condition.
4942 (A < 0 ? C : 0): Similarly for @1.
4945 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4946 characters starting at p contain '\0' character, don't look beyond
4950 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4951 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4954 * reg-stack.c (change_stack): Change update_end var from int to
4955 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4956 also call set_block_for_insn on the newly added insns and rescan.
4959 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4960 on the last operand.
4963 * tree.c (stabilize_reference_1): Return save_expr (e) for
4964 STATEMENT_LIST even if it doesn't have side-effects.
4966 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4968 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4970 2018-03-12 Renlin Li <renlin.li@arm.com>
4972 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4973 aarch64_output_scalar_simd_mov_immediate.
4975 2018-03-12 Martin Sebor <msebor@redhat.com>
4977 PR tree-optimization/83456
4978 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4979 for perfectly overlapping calls to memcpy.
4980 (gimple_fold_builtin_memory_chk): Same.
4981 (gimple_fold_builtin_strcpy): Handle no-warning.
4982 (gimple_fold_builtin_stxcpy_chk): Same.
4983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4985 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4987 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4988 parameter. Use it for SFmode.
4989 (rs6000_function_arg_advance_1): Adjust.
4990 (rs6000_function_arg): Adjust.
4991 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4993 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4995 PR rtl-optimization/84169
4996 PR rtl-optimization/84780
4997 * combine.c (can_combine_p): Check for a 2-insn combination whether
4998 the destination register is used between the two insns, too.
5000 2018-03-12 Richard Biener <rguenther@suse.de>
5002 PR tree-optimization/84803
5003 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5004 for refs DR analysis didn't process.
5006 2018-03-12 Richard Biener <rguenther@suse.de>
5008 PR tree-optimization/84777
5009 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5010 force-vectorize loops ignore whether we are optimizing for size.
5012 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5014 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5015 (TARGET_MD_ASM_ADJUST): Define.
5017 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5018 Kito Cheng <kito.cheng@gmail.com>
5019 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * config/nds32/nds32.c (nds32_compute_stack_frame,
5022 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5023 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5024 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5025 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5026 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5027 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5028 * config/nds32/nds32.md (prologue, epilogue): Use macro
5029 NDS32_V3PUSH_AVAILABLE_P to do checking.
5031 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5034 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5035 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5036 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5037 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5038 addition of most attributes on !orig_type_die or the attribute not
5039 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5041 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5042 Chung-Ju Wu <jasonwucj@gmail.com>
5044 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5046 * config/nds32/nds32.opt (mvh): New option.
5048 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5049 Chung-Ju Wu <jasonwucj@gmail.com>
5051 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5053 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5054 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5057 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5058 Chung-Ju Wu <jasonwucj@gmail.com>
5060 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5062 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5063 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5065 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5066 Kito Cheng <kito.cheng@gmail.com>
5067 Chung-Ju Wu <jasonwucj@gmail.com>
5069 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5070 UNSPEC_FFMISM and UNSPEC_FLMISM.
5071 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5072 for ffb, ffmism and flmism.
5073 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5074 (unspec_ffmism): Ditto.
5075 (unspec_flmism): Ditto.
5076 (nds32_expand_builtin_impl): Check if string extension is available.
5077 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5078 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5080 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5083 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5086 * lra-assigns.c (assign_by_spills): Return a flag of reload
5087 assignment failure. Do not process the reload assignment
5088 failures. Do not spill other reload pseudos if they has the same
5090 (lra_assign): Add a return arg. Set up from the result of
5091 assign_by_spills call.
5092 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5093 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5094 usage_insns if it is not NULL.
5095 (spill_hard_reg_in_range): New function.
5096 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5097 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5098 function prototypes.
5099 (lra_assign): Change prototype.
5100 * lra.c (lra): Add code to deal with fails by splitting hard reg
5103 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5106 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5108 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5111 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5112 (dw_val_node): Add val_symbolic_view.
5113 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5114 (symview_upper_bound): New.
5115 (new_line_info_table): Initialize symviews_since_reset.
5116 (dwarf2out_source_line): Count symviews_since_reset and set
5117 symview_upper_bound.
5118 (dw_val_equal_p): Handle symview.
5119 (add_AT_symview): New.
5120 (print_dw_val): Handle symview.
5121 (attr_checksum, attr_checksum_ordered): Likewise.
5122 (same_dw_val_p, size_of_die): Likewise.
5123 (value_format, output_die): Likewise.
5124 (add_high_low_attributes): Use add_AT_symview for entry_view.
5125 (dwarf2out_finish): Reset symview_upper_bound, clear
5128 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5131 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5132 Add strict argument and use it.
5133 (rs6000_split_multireg_move): Update for new strict argument.
5134 (mem_operand_gpr): Disallow all non-offsettable addresses.
5135 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5137 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5140 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5141 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5142 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5145 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5146 decl, use remap_type if we want to use the type.
5148 2018-03-09 Martin Sebor <msebor@redhat.com>
5150 PR tree-optimization/84526
5151 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5153 (builtin_access::generic_overlap): Be prepared to handle non-array
5156 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5158 PR rtl-optimization/84682
5159 * lra-constraints.c (process_address_1): Check is_address flag
5160 for address constraints.
5161 (process_alt_operands): Likewise.
5162 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5163 preprocess_constraints.
5164 * recog.h (preprocess_constraints): Add oploc parameter.
5166 * recog.c (preprocess_constraints): Test address_operand for
5167 CT_ADDRESS constraints.
5169 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5172 * lra-assigns.c (assign_by_spills): Return a flag of reload
5173 assignment failure. Do not process the reload assignment
5174 failures. Do not spill other reload pseudos if they has the same
5176 (lra_assign): Add a return arg. Set up from the result of
5177 assign_by_spills call.
5178 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5179 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5180 usage_insns if it is not NULL.
5181 (spill_hard_reg_in_range): New function.
5182 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5183 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5184 function prototypes.
5185 (lra_assign): Change prototype.
5186 * lra.c (lra): Add code to deal with fails by splitting hard reg
5189 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5192 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5193 Accept complain bool parameter. Only emit errors if it is true.
5194 (arm_parse_cpu_option_name): Likewise.
5195 (arm_target_thumb_only): Adjust callers of the above.
5196 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5197 prototype to take a default true bool parameter.
5198 (arm_parse_arch_option_name): Likewise.
5200 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5201 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5205 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5206 * configure: Regenerate.
5207 * configure.ac ("linker --version-script option"): New.
5208 ("linker soname option"): New.
5210 2018-03-09 Richard Biener <rguenther@suse.de>
5212 PR tree-optimization/84775
5213 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5214 immediate uses of predicate stmts and mark them modified.
5217 PR tree-optimization/84178
5218 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5220 (version_loop_for_if_conversion): Delay update_ssa call.
5221 (tree_if_conversion): Delay update_ssa until after predicate
5224 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5227 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5228 when the function accesses prior frames.
5230 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5233 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5234 gen_llsym, otherwise call maybe_gen_llsym.
5237 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5238 has ',' character inside of it.
5240 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5243 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5244 as clobbering CC_REGNUM.
5246 2018-03-08 Richard Biener <rguenther@suse.de>
5249 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5250 (follow_copies_to_constant): Do not follow SSA names registered
5253 2018-03-08 Richard Biener <rguenther@suse.de>
5255 PR tree-optimization/84178
5256 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5258 (version_loop_for_if_conversion): Delay update_ssa call.
5259 (tree_if_conversion): Delay update_ssa until after predicate
5262 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5264 PR tree-optimization/84178
5265 * tree-if-conv.c (release_bb_predicate): Remove the
5266 the assertion that the stmts have NULL use_ops.
5267 Discard the statements, asserting that they haven't
5268 yet been added to a BB.
5270 2018-03-08 Richard Biener <rguenther@suse.de>
5272 PR tree-optimization/84746
5273 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5274 (phi_translate): Pass in destination ANTIC_OUT set.
5275 (phi_translate_1): Likewise. For a simplified result lookup
5276 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5277 (phi_translate_set): Adjust.
5278 (do_pre_regular_insertion): Likewise.
5279 (do_pre_partial_partial_insertion): Likewise.
5281 2018-03-08 Martin Liska <mliska@suse.cz>
5283 PR gcov-profile/84735
5284 * doc/gcov.texi: Document usage of profile files.
5285 * gcov-io.h: Document changes in the format.
5287 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5291 * dwarf2out.c (struct dw_line_info_table): Update comments for
5293 (FORCE_RESET_NEXT_VIEW): New.
5294 (FORCE_RESETTING_VIEW_P): New.
5295 (RESETTING_VIEW_P): Check for -1 too.
5296 (ZERO_VIEW_P): Likewise.
5297 (new_line_info_table): Force-reset next view.
5298 (dwarf2out_begin_function): Likewise.
5299 (dwarf2out_source_line): Simplify zero_view_p initialization.
5300 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5301 view directly. Omit view when omitting .loc at line 0.
5303 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5305 PR tree-optimization/84740
5306 * tree-switch-conversion.c (process_switch): Call build_constructors
5307 only if info.phi_count is non-zero.
5309 PR tree-optimization/84739
5310 * tree-tailcall.c (find_tail_calls): Check call arguments against
5311 DECL_ARGUMENTS (current_function_decl) rather than
5312 DECL_ARGUMENTS (func) when checking for tail recursion.
5314 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5316 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5317 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5318 Volker Reichelt's entry and add entries for people that perform
5319 GCC fuzzy testing and report numerous bugs.
5321 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5324 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5325 readonly data in sdata, if that is disabled.
5326 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5327 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5328 -mreadonly-in-sdata option.
5330 2018-03-07 Martin Sebor <msebor@redhat.com>
5332 PR tree-optimization/84468
5333 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5334 basic block when looking for nul assignment.
5336 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5339 * except.h (output_function_exception_table): Adjust prototype.
5340 * except.c (output_function_exception_table): Remove FNNAME parameter
5341 and add SECTION parameter. Ouput one part of the table at a time.
5342 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5343 the first part of the exception table and emit unwind directives.
5344 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5345 (i386_pe_seh_cold_init): Likewise.
5346 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5347 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5348 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5349 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5350 * config/i386/winnt.c: Include except.h.
5351 (struct seh_frame_state): Add reg_offset, after_prologue and
5352 in_cold_section fields.
5353 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5354 (i386_pe_seh_cold_init): New function.
5355 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5356 to seh->in_cold_section.
5357 (seh_emit_push): Record the offset of the push.
5358 (seh_emit_save): Record the offet of the save.
5359 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5360 Test seh->after_prologue to disregard the epilogue.
5361 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5362 (i386_pe_end_cold_function): New function.
5364 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5367 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5368 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5371 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5373 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5374 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5377 * multiple_target.c: Include tree-inline.h and intl.h.
5378 (expand_target_clones): Diagnose and fail if node->definition and
5379 !tree_versionable_function_p (node->decl).
5381 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5383 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5385 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5386 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5387 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5389 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5392 * combine.c (try_combine): Use reg_or_subregno instead of handling
5393 just paradoxical SUBREGs and REGs.
5395 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5397 * config/arc/arc.c (arc_finalize_pic): Remove function.
5398 (arc_must_save_register): We use single base PIC register, remove
5399 checks to save/restore the PIC register.
5400 (arc_expand_prologue): Likewise.
5401 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5403 (arc_verify_short): Likewise.
5404 (arc_attr_type): Likewise.
5405 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5406 (walk_stores): Likewise.
5407 (arc_address_cost): Make it static.
5408 (arc_verify_short): Likewise.
5409 (branch_dest): Likewise.
5410 (arc_attr_type): Likewise.
5411 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5412 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5413 (arc_final_prescan_insn): Remove inserting the nops due to
5414 hardware hazards. It is done in reorg step.
5415 (insn_length_variant_t): Remove.
5416 (insn_length_parameters_t): Likewise.
5417 (arc_insn_length_parameters): Likewise.
5418 (arc_get_insn_variants): Likewise.
5419 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5421 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5424 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5427 PR tree-optimization/84687
5428 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5430 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5432 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5434 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5435 Rename to ppc_speculation_barrier.
5436 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5437 __builtin_ppc_speculation_barrier.
5439 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5442 * combine.c (combine_simplify_rtx): Don't try to simplify if
5443 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5446 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5448 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5449 to 32 bytes when compiling for POWER9.
5451 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5454 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5455 regparm >= 3 with no arg reg available also for calls with
5456 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5459 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5461 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5463 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5466 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5468 2018-03-05 Richard Biener <rguenther@suse.de>
5470 PR tree-optimization/84486
5471 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5472 When inserting a __builtin_assume_aligned call set the LHS
5473 SSA name alignment info accordingly.
5475 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5477 PR tree-optimization/84114
5478 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5479 Avoid reassociation of FLOAT_MODE addition.
5481 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5483 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5484 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5485 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5486 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5487 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5488 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5489 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5491 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5492 __builtin_ia32_wbinvd): New builtins.
5493 (SPECIAL_ARGS2): New.
5494 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5495 (SPECIAL_ARGS2): New.
5496 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5497 (ix86_valid_target_attribute_inner_p): Ditto.
5498 (ix86_init_mmx_sse_builtins): Add special_args2.
5499 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5500 TARGET_WBNOINVD_P): New.
5501 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5502 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5503 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5504 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5505 * config/i386/pconfigintrin.h: New file.
5506 * config/i386/wbnoinvdintrin.h: Ditto.
5507 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5508 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5510 2018-03-05 Richard Biener <rguenther@suse.de>
5512 PR tree-optimization/84670
5513 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5515 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5516 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5517 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5518 doesn't grow if all successors (recursively) were visited at least
5521 2018-03-05 Richard Biener <rguenther@suse.de>
5523 PR tree-optimization/84650
5524 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5525 if executed in the loop pipeline.
5527 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5529 * doc/configfiles.texi (Configuration Files): Move info about
5530 conditionalizing $target-protos.h to...
5531 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5532 differs from $target-protos.h.
5534 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5535 Chung-Ju Wu <jasonwucj@gmail.com>
5537 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5538 * config/nds32/nds32-multiple.md (setmemsi): Define.
5539 * config/nds32/nds32-memory-manipulation.c
5540 (nds32_gen_dup_4_byte_to_word_value): New.
5541 (emit_setmem_word_loop): New.
5542 (emit_setmem_byte_loop): New.
5543 (nds32_expand_setmem_loop): New.
5544 (nds32_expand_setmem_loop_v3m): New.
5545 (nds32_expand_setmem_unroll): New.
5546 (nds32_expand_setmem): New.
5548 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5549 Chung-Ju Wu <jasonwucj@gmail.com>
5551 * config/nds32/nds32-memory-manipulation.c
5552 (nds32_emit_load_store): New.
5553 (nds32_emit_post_inc_load_store): New.
5554 (nds32_emit_mem_move): New.
5555 (nds32_emit_mem_move_block): New.
5556 (nds32_expand_movmemsi_loop_unknown_size): New.
5557 (nds32_expand_movmemsi_loop_known_size): New.
5558 (nds32_expand_movmemsi_loop): New.
5559 (nds32_expand_movmemsi_unroll): New.
5560 (nds32_expand_movmemqi): Rename ...
5561 (nds32_expand_movmemsi): ... to this.
5562 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5563 (movmemsi): ... to this.
5564 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5565 (nds32_expand_movmemsi): ... to this.
5567 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5568 Monk Chiang <sh.chiang04@gmail.com>
5569 Chung-Ju Wu <jasonwucj@gmail.com>
5571 * config/nds32/nds32-protos.h
5572 (nds32_expand_load_multiple): New arguments.
5573 (nds32_expand_store_multiple): Ditto.
5574 (nds32_valid_multiple_load_store): Rename ...
5575 (nds32_valid_multiple_load_store_p): ... to this.
5576 * config/nds32/nds32-memory-manipulation.c
5577 (nds32_expand_load_multiple): Refine implementation.
5578 (nds32_expand_store_multiple): Ditto.
5579 * config/nds32/nds32-multiple.md
5580 (load_multiple): Update nds32_expand_load_multiple interface.
5581 (store_multiple): Update nds32_expand_store_multiple interface.
5582 * config/nds32/nds32-predicates.c
5583 (nds32_valid_multiple_load_store): Rename ...
5584 (nds32_valid_multiple_load_store_p): ... to this and refine
5586 * config/nds32/predicates.md
5587 (nds32_load_multiple_and_update_address_operation): New predicate.
5588 (nds32_store_multiple_and_update_address_operation): New predicate.
5590 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5591 Chung-Ju Wu <jasonwucj@gmail.com>
5593 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5594 (combo): New attribute.
5595 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5597 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5599 * config/nds32/nds32.opt: Change -mcmodel= default value.
5601 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5602 Monk Chiang <sh.chiang04@gmail.com>
5603 Chung-Ju Wu <jasonwucj@gmail.com>
5605 * config/nds32/constants.md (unspec_element): New enum.
5606 * config/nds32/constraints.md (Umw): New constraint.
5607 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5608 * config/nds32/nds32-intrinsic.md: Likewise.
5609 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5610 (nds32_valid_smw_lwm_base_p): New.
5611 (nds32_output_smw_single_word): New.
5612 (nds32_output_lmw_single_word): New.
5613 (nds32_expand_unaligned_load): New.
5614 (nds32_expand_unaligned_store): New.
5615 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5616 (nds32_output_smw_single_word): Declare.
5617 (nds32_output_lmw_single_word): Declare.
5618 (nds32_expand_unaligned_load): Declare.
5619 (nds32_expand_unaligned_store): Declare.
5620 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5621 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5622 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5623 NDS32_BUILTIN_UASTORE_DW.
5624 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5627 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5628 Kito Cheng <kito.cheng@gmail.com>
5629 Chung-Ju Wu <jasonwucj@gmail.com>
5631 * config/nds32/nds32-intrinsic.c
5632 (nds32_expand_builtin_null_ftype_reg): Delete.
5633 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5634 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5635 (nds32_read_argument): New.
5636 (nds32_legitimize_target): Ditto.
5637 (nds32_legitimize_argument): Ditto.
5638 (nds32_check_constant_argument): Ditto.
5639 (nds32_expand_unop_builtin): Ditto.
5640 (nds32_expand_unopimm_builtin): Ditto.
5641 (nds32_expand_binop_builtin): Ditto.
5642 (nds32_builtin_decl_impl): Ditto.
5643 (builtin_description): Ditto.
5644 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5645 (nds32_init_builtins_impl): Ditto.
5646 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5647 (nds32_builtin_decl): New.
5648 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5649 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5651 2018-03-02 Jeff Law <law@redhat.com>
5653 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5654 (redundant_insn, fill_simple_delay_slots): Likewise.
5655 (fill_slots_from_thread): Likewise.
5656 * resource.c (mark_referenced_resources): Likewise.
5657 (mark_set_resources, find_dead_or_set_registers): Likewise.
5659 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5661 * substring-locations.h (format_warning_va): Formatting fix for
5663 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5665 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5666 * substring-locations.c: Include intl.h.
5667 (format_warning_va): Turned into small wrapper around
5668 format_warning_n_va, renamed to ...
5669 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5670 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5672 (format_warning_at_substring_n): New function.
5673 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5674 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5675 format_warning_at_substring with just a shorter name instead of
5676 const function pointer.
5677 (fmtwarn_n): New function.
5678 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5679 appropriate, get rid of all the fmtstr temporaries, move conditionals
5680 with G_() wrapped string literals directly into fmtwarn arguments,
5681 cast dir.len to (int), formatting fixes.
5683 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5685 * doc/invoke.texi: Remove "Cilk Plus" references.
5687 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5688 Richard Biener <rguenther@suse.de>
5691 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5692 for error or warning attributes if CALL_FROM_THUNK_P is set.
5695 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5698 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5699 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5702 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5703 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5705 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5706 instead of -1U in last predictors element's probability member.
5708 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5711 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5712 arguments if they are comparable.
5714 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5716 PR tree-optimization/84634
5717 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5718 masks and masked_loop_p with a single loop_masks, making sure it's
5719 null for bb vectorization.
5721 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5723 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5724 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5725 loop->force_vectorize to check whether there is no alias.
5727 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5730 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5732 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5734 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5735 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5736 instead of a loop around prev_real_insn.
5737 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5741 * config/i386/i386.c (ix86_print_operand): Use conditional
5742 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5745 2018-03-02 Richard Biener <rguenther@suse.de>
5747 PR tree-optimization/84427
5748 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5749 (bitmap_set_subtract_values): Rewrite to handle multiple
5752 (prune_clobbered_mems): Likewise.
5753 (phi_translate): Take edge instead of pred/phiblock.
5754 (phi_translate_1): Likewise.
5755 (phi_translate_set): Likewise. Insert all translated
5756 exprs for a value into the set, keeping possibly multiple
5757 expressions per value.
5758 (compute_antic_aux): Adjust for phi_translate changes.
5759 When intersecting union the expressions and prune those
5760 not in the final value set, keeping possibly multiple
5761 expressions per value. Do not use value-insertion
5762 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5763 all expressions. Add verification that the value-sets
5764 only shrink during iteration.
5765 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5766 (do_pre_regular_insertion): Likewise.
5767 (do_pre_partial_partial_insertion): Likewise.
5769 2018-03-02 Richard Biener <rguenther@suse.de>
5772 * config/darwin.c (saved_debug_info_level): New static global.
5773 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5774 (darwin_asm_lto_end): Restore debug info generation settings.
5776 2018-03-01 Martin Liska <mliska@suse.cz>
5779 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5782 2018-03-01 Richard Biener <rguenther@suse.de>
5785 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5786 type annotation with LTO when debug was disabled at compile-time.
5788 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5790 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5792 (mips_final_postscan_insn): Likewise.
5794 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5796 PR rtl-optimization/84528
5797 * alias.c (init_alias_target): Add commentary.
5798 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5799 a unique base value if the frame pointer is not eliminated
5800 to the stack pointer.
5802 2018-03-01 Tom de Vries <tom@codesourcery.com>
5804 PR rtl-optimization/83327
5805 * lra-int.h (hard_regs_spilled_into): Declare.
5806 * lra.c (hard_regs_spilled_into): Define.
5807 (init_reg_info): Init hard_regs_spilled_into.
5808 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5809 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5810 (process_bb_lives): Handle hard_regs_spilled_into.
5811 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5812 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5814 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5816 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5817 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5818 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5819 * config/rs6000/aix72.h: New file.
5821 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5823 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5824 instead of warning_at with conditional singular and plural messages
5828 * stor-layout.c (update_alignment_for_field): For
5829 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5830 && !DECL_PACKED (field), do the alignment update, just use
5831 only desired_align instead of MAX (type_align, desired_align)
5833 (place_field): Don't do known_align < desired_align handling
5834 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5835 is non-NULL, instead do it after rli->prev_field handling and
5836 only if not within a bitfield word. For DECL_PACKED (field)
5837 use type_align of BITS_PER_UNIT.
5839 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5841 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5842 superfluous parentheses and trailing spaces.
5844 2018-02-28 Richard Biener <rguenther@suse.de>
5846 PR tree-optimization/84584
5847 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5848 SCoPs with fake exit edge.
5850 2018-02-28 Martin Liska <mliska@suse.cz>
5853 * timevar.c (timer::print): Fix format to properly print 100%
5856 2018-02-28 Richard Biener <rguenther@suse.de>
5859 * genmatch.c (capture_info::walk_match): Do not mark
5860 captured expressions without operands as expr_p given
5861 they act more like predicates and should be subject to
5862 "lost tail" side-effect preserving.
5864 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5866 PR rtl-optimization/81611
5867 * auto-inc-dec.c (attempt_change): Move dead note from
5868 mem_insn if it's the next use of regno
5869 (find_address): Take address use of reg holding
5870 non-incremented value. Add parm to limit search to the named
5872 (merge_in_block): Attempt to use a mem insn that is the next
5873 use of the original regno.
5875 2018-02-27 Martin Sebor <msebor@redhat.com>
5878 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5879 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5881 2018-02-27 Martin Sebor <msebor@redhat.com>
5883 PR translation/84207
5884 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5885 n argument to unsigned HOST_WIDE_INT.
5886 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5887 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5888 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5889 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5891 2018-02-27 Richard Biener <rguenther@suse.de>
5893 PR tree-optimization/84512
5894 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5895 Do not use the estimate returned from record_stmt_cost for
5896 the scalar iteration cost but sum properly using add_stmt_cost.
5898 2018-02-27 Richard Biener <rguenther@suse.de>
5900 PR tree-optimization/84466
5901 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5902 Adjust last change to less strictly validate use operands.
5904 2018-02-27 Martin Liska <mliska@suse.cz>
5906 PR gcov-profile/84548
5907 * gcov.c (process_file): Allow partial overlap and consider it
5908 also as group functions.
5909 (output_lines): Properly calculate range of lines for a group.
5911 2018-02-27 Martin Liska <mliska@suse.cz>
5913 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5914 'ggc' suffixes. Change first column width.
5915 (timer::print): Fix formatting of the column.
5917 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5919 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5920 preserve inline entry blocks for the sake of debug inline
5921 entry point markers alone.
5922 (remove_unused_locals): Suggest in comments a better place to
5923 force the preservation of inline entry blocks that are
5924 otherwise unused, but do not preserve them.
5926 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5928 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5930 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5933 * config/i386/constraints.md (Bs): Replace
5934 ix86_indirect_branch_register with
5935 TARGET_INDIRECT_BRANCH_REGISTER.
5937 * config/i386/i386.md (indirect_jump): Likewise.
5938 (tablejump): Likewise.
5939 (*sibcall_memory): Likewise.
5940 (*sibcall_value_memory): Likewise.
5941 Peepholes of indirect call and jump via memory: Likewise.
5942 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5943 (*sibcall_value_GOT_32): Likewise.
5944 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5945 (GOT_memory_operand): Likewise.
5946 (call_insn_operand): Likewise.
5947 (sibcall_insn_operand): Likewise.
5948 (GOT32_symbol_operand): Likewise.
5949 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5951 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5953 PR rtl-optimization/83496
5954 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5955 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5956 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5957 redundant insn, if any.
5958 (relax_delay_slots): Likewise.
5959 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5961 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5963 PR tree-optimization/83965
5964 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5965 that grouped statements are part of a reduction chain. Return
5966 true if the statement is not marked as a reduction itself but
5968 (vect_recog_dot_prod_pattern): Don't check whether the statement
5969 is part of a group here.
5970 (vect_recog_sad_pattern): Likewise.
5971 (vect_recog_widen_sum_pattern): Likewise.
5973 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5976 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5978 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5981 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5983 (ix86_output_indirect_function_return): New prototype.
5984 (ix86_split_simple_return_pop_internal): Likewise.
5985 * config/i386/i386.c (indirect_return_via_cx): New.
5986 (indirect_return_via_cx_bnd): Likewise.
5987 (indirect_thunk_name): Handle return va CX_REG.
5988 (output_indirect_thunk_function): Create alias for
5989 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5990 (ix86_output_indirect_jmp): Remove the bool argument.
5991 (ix86_output_indirect_function_return): New function.
5992 (ix86_split_simple_return_pop_internal): Likewise.
5993 * config/i386/i386.md (*indirect_jump): Don't pass false
5994 to ix86_output_indirect_jmp.
5995 (*tablejump_1): Likewise.
5996 (simple_return_pop_internal): Change it to define_insn_and_split.
5997 Call ix86_split_simple_return_pop_internal to split it for
5999 (simple_return_indirect_internal): Call
6000 ix86_output_indirect_function_return instead of
6001 ix86_output_indirect_jmp.
6003 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6006 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6007 memset and value initialization afterwards.
6009 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6011 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6013 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6016 * common/config/aarch64/aarch64-common.c
6017 (aarch_option_optimization_table[]): Switch
6018 off fomit-frame-pointer
6020 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6021 Chung-Ju Wu <jasonwucj@gmail.com>
6023 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6025 (store_multiple): Ditto.
6027 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6029 * config.gcc: Add --with-cpu support for nds32 target.
6030 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6031 * config/nds32/nds32.opt: Add -mcpu= option.
6033 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6035 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6036 isel=yes): Warn for these deprecated options.
6038 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6040 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6041 ISA_2_5_MASKS_EMBEDDED.
6043 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6045 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6046 p->max as pointers rather than using iterative_hash_expr.
6048 2018-02-23 Carl Love <cel@us.ibm.com>
6050 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6051 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6053 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6054 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6055 P8V_BUILTIN_VEC_VUNSIGNED2.
6057 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6060 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6061 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6063 * lra-constraints.c (curr_insn_transform): Set up
6064 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6066 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6067 * lra-eliminations.c (spill_pseudos): Ditto.
6068 (process_insn_for_elimination): Ditto.
6069 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6070 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6071 LRA_NON_CLOBBERED_ALT.
6073 2018-02-22 Martin Sebor <msebor@redhat.com>
6075 PR tree-optimization/84480
6076 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6077 to maybe_diag_stxncpy_trunc. Call it.
6078 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6079 from gimple_fold_builtin_strcpy. Print inlining stack.
6080 (handle_builtin_stxncpy): Print inlining stack.
6081 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6083 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6086 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6087 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6088 and -fcheck-pointer-bounds are used together.
6089 (indirect_thunk_prefix): New enum.
6090 (indirect_thunk_need_prefix): New function.
6091 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6092 "_nt" instead of "_bnd" for NOTRACK prefix.
6093 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6094 (output_indirect_thunk_function): Likewise.
6096 (ix86_code_end): Update output_indirect_thunk_function calls.
6097 (ix86_output_indirect_branch_via_reg): Replace
6098 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6099 (ix86_output_indirect_branch_via_push): Likewise.
6100 (ix86_output_function_return): Likewise.
6101 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6102 incompatible with -fcf-protection=branch and
6103 -fcheck-pointer-bounds.
6105 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6108 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6109 Change gcc_assert call to output_operand_lossage.
6111 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6113 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6115 2018-02-22 DJ Delorie <dj@redhat.com>
6116 Sebastian Perta <sebastian.perta@renesas.com>
6117 Oleg Endo <olegendo@gcc.gnu.org>
6119 * config/rx/rx.c (rx_rtx_costs): New function.
6120 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6122 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6124 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6126 2018-02-22 Martin Liska <mliska@suse.cz>
6129 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6130 Add "native" as a possible value.
6132 2018-02-22 Martin Liska <mliska@suse.cz>
6135 * config/i386/i386.c (ix86_option_override_internal):
6136 Add "native" as a possible value for -march and -mtune.
6138 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6141 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6142 to all type variants.
6144 PR tree-optimization/84503
6145 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6146 width as info->bitpos + info->bitsize - start.
6147 (merged_store_group::merge_overlapping): Simplify width computation.
6148 (check_no_overlap): New function.
6149 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6150 start + width and last_order of the group, fail if check_no_overlap
6152 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6153 to group if check_no_overlap fails.
6155 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6157 * config/rs6000/altivec.md: Delete contraint arguments to
6158 define_expand, define_split, and define_peephole2, and in
6159 define_insn_and_split if always unused.
6160 * config/rs6000/darwin.md: Ditto.
6161 * config/rs6000/dfp.md: Ditto.
6162 * config/rs6000/rs6000.md: Ditto.
6163 * config/rs6000/sync.md: Ditto.
6164 * config/rs6000/vector.md: Ditto.
6165 * config/rs6000/vsx.md: Ditto.
6167 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6169 * config/rs6000/altivec.md: Write output control strings as braced
6170 blocks instead of double-quoted strings.
6171 * config/rs6000/darwin.md: Ditto.
6172 * config/rs6000/rs6000.md: Ditto.
6173 * config/rs6000/vector.md: Ditto.
6174 * config/rs6000/vsx.md: Ditto.
6176 2018-02-21 Jason Merrill <jason@redhat.com>
6178 PR c++/84314 - ICE with templates and fastcall attribute.
6179 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6181 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6183 * ipa-cp.c (determine_versionability): Fix comment typos.
6185 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6188 * ipa-cp.c (determine_versionability): Do not version functions caling
6191 2018-02-21 Martin Liska <mliska@suse.cz>
6194 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6195 Add "native" as a possible value.
6196 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6197 the macro when native cpu detection is available.
6199 2018-02-21 Martin Liska <mliska@suse.cz>
6202 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6203 Add "native" as a possible value.
6204 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6205 when native cpu detection is available.
6207 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6208 Martin Sebor <msebor@redhat.com>
6210 PR tree-optimization/84478
6211 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6213 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6214 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6215 support which is conservatively correct, for 2 only stay conservative
6216 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6217 argument to the 2 argument get_range_strlen, adjust 6 arg
6218 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6220 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6221 (gimple_fold_builtin_strlen): Pass true as last argument to
6224 2018-02-20 Martin Sebor <msebor@redhat.com>
6227 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6228 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6229 (builtin_memref::builtin_memref): Factor out parts into
6230 set_base_and_offset and call it.
6232 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6235 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6236 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6237 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6238 search at the associated MODE_INT.
6240 2018-02-20 Jeff Law <law@redhat.com>
6243 PR tree-optimization/81592
6245 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6246 for range data rather than using global data.
6247 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6248 range data rather than using global data.
6249 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6250 pass it to children as needed.
6251 (struct directive::fmtresult): Similarly.
6252 (struct directive::set_width): Similarly.
6253 (struct directive::set_precision): Similarly.
6254 (format_integer, format_directive, parse_directive): Similarly.
6255 (format_none): Accept unnamed vr_values parameter.
6256 (format_percent, format_floating, format_character): Similarly.
6257 (format_string, format_plain): Similarly.
6258 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6259 the EVRP range analyzer for range data rather than using global data.
6260 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6261 gimple-ssa-evrp-analyze.h
6262 (class sprintf_dom_walker): Add after_dom_children member function.
6263 Add evrp_range_analyzer member.
6264 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6265 range analyzer as needed.
6266 (sprintf_dom_walker::after_dom_children): New member function.
6267 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6269 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6270 (evrp_range_analyzer::pop_to_marker): Likewise.
6272 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6274 PR tree-optimization/84419
6275 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6276 with the required type if its current type is compatible but
6279 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6282 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6283 after vectorization.
6285 2018-02-20 Martin Liska <mliska@suse.cz>
6288 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6289 possible values if we don't have a hint.
6291 2018-02-20 Martin Liska <mliska@suse.cz>
6295 * final.c (shorten_branches): Build align_tab array with one
6297 * opts.c (finish_options): Add alignment option limit check.
6298 (MAX_CODE_ALIGN): Likewise.
6299 (MAX_CODE_ALIGN_VALUE): Likewise.
6300 * doc/invoke.texi: Document maximum allowed option value for
6301 all -falign-* options.
6303 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6306 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6307 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6308 * var-tracking.c (emit_note_insn_var_location): Remove all references
6309 to NOTE_INSN_CALL_ARG_LOCATION.
6310 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6311 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6312 Use copy_rtx_if_shared.
6313 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6314 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6315 (dwarf2out_var_location): Remove handling of
6316 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6318 * final.c (final_scan_insn): Remove all references to
6319 NOTE_INSN_CALL_ARG_LOCATION.
6320 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6321 before dumping final insns.
6322 * except.c (emit_note_eh_region_end): Remove all references to
6323 NOTE_INSN_CALL_ARG_LOCATION.
6324 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6325 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6326 * config/arc/arc.c (hwloop_optimize): Likewise.
6327 * config/arm/arm.c (create_fix_barrier): Likewise.
6328 * config/s390/s390.c (s390_chunkify_start): Likewise.
6329 * config/sh/sh.c (find_barrier): Likewise.
6330 * config/i386/i386.c (rest_of_insert_endbranch,
6331 ix86_seh_fixup_eh_fallthru): Likewise.
6332 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6333 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6334 * config/frv/frv.c (frv_function_prologue): Likewise.
6335 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6337 (note_outside_basic_block_p): Remove all references to
6338 NOTE_INSN_CALL_ARG_LOCATION.
6339 * gengtype.c (adjust_field_rtx_def): Likewise.
6340 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6342 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6343 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6346 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6349 PR tree-optimization/84452
6350 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6351 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6354 2018-02-19 Martin Liska <mliska@suse.cz>
6357 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6359 2018-02-19 Martin Liska <mliska@suse.cz>
6360 Richard Sandiford <richard.sandiford@linaro.org>
6362 PR tree-optimization/82491
6363 * gimple-fold.c (get_base_constructor): Make earlier bail out
6366 2018-02-19 Carl Love <cel@us.ibm.com>
6368 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6369 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6371 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6372 P8V_BUILTIN_VEC_NEG.
6374 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6376 * config/rl78/rl78.md (movdf): New define expand.
6378 2018-02-19 Martin Liska <mliska@suse.cz>
6381 * doc/invoke.texi: Fix typo.
6382 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6384 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6386 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6387 handle rs6000_single_float and rs6000_double_float specially for
6390 2018-02-16 Jeff Law <law@redhat.com>
6392 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6395 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6398 * ipa-inline.c (inline_small_functions): Fix a typo.
6400 2018-02-16 Nathan Sidwell <nathan@acm.org>
6402 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6404 2018-02-16 Carl Love <cel@us.ibm.com>
6406 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6407 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6408 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6409 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6410 expansion to P8V_BUILTIN_VEC_FLOAT2.
6412 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6414 PR rtl-optimization/70023
6415 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6416 src_regno into account.
6418 2018-02-16 Carl Love <cel@us.ibm.com>
6420 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6421 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6422 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6423 * config/rs6000/rs6000.c: Remove case statements for
6424 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6425 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6426 and P9V_BUILTIN_VEC_VINSERT4B.
6427 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6428 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6429 * config/rs6000/vsx.md:
6430 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6433 2018-02-16 Carl Love <cel@us.ibm.com>
6435 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6437 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6439 * config/rs6000/rs6000-c.c: Add the definitions for
6440 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6441 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6442 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6443 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6444 definition for insert4b and define insn *insert3b_internal.
6445 * doc/extend.texi: Add documentation for vec_extract4b.
6447 2018-02-16 Nathan Sidwell <nathan@acm.org>
6449 * doc/extend.texi (Backwards Compatibility): Mention friend
6450 injection. Note for-scope is deprecated.
6451 * doc/invoke.texi (-ffriend-injection): Deprecate.
6453 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6455 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6456 that moved to I2, also allow destinations that are a paradoxical
6457 subreg (instead of a normal reg).
6459 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6462 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6465 2018-02-16 Richard Biener <rguenther@suse.de>
6467 PR tree-optimization/84037
6468 PR tree-optimization/84016
6470 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6471 Adjust vec_construct for the fact we need additional higher latency
6472 128bit inserts for AVX256 and AVX512 vector builds.
6473 (ix86_add_stmt_cost): Scale vector construction cost for
6476 2018-02-16 Richard Biener <rguenther@suse.de>
6478 PR tree-optimization/84417
6479 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6480 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6481 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6483 2018-02-16 Martin Liska <mliska@suse.cz>
6486 * internal-fn.def (ASAN_CHECK): Set proper flags.
6487 (ASAN_MARK): Likewise.
6489 2018-02-16 Julia Koval <julia.koval@intel.com>
6491 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6492 from PTA_CANNONLAKE.
6494 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6497 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6498 Use ++iter rather than iter++ for std::list iterators.
6499 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6500 defer deleting them until all nodes in the forest are processed. Do
6501 free even leaf nodes. Change to_process into auto_vec.
6504 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6505 * vec.h (vec_default_construct): Use memset instead of placement new
6506 if BROKEN_VALUE_INITIALIZATION is defined.
6507 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6508 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6511 PR rtl-optimization/83723
6512 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6513 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6514 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6516 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6518 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6520 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6522 PR rtl-optimization/81443
6523 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6524 from inner REGs to paradoxical SUBREGs.
6526 2018-02-16 Richard Biener <rguenther@suse.de>
6528 PR tree-optimization/84399
6529 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6530 For operands we can analyze at their definition make sure we can
6531 analyze them at each use as well.
6533 2018-02-16 Richard Biener <rguenther@suse.de>
6535 PR tree-optimization/84190
6536 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6537 volatile accesses if the decl isn't volatile.
6539 2018-02-15 Jason Merrill <jason@redhat.com>
6541 PR c++/84314 - ICE with templates and fastcall attribute.
6542 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6543 TYPE_CANONICAL on an existing type.
6545 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6547 PR tree-optimization/84383
6548 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6549 dstoff nor call operand_equal_p if dstbase is NULL.
6551 PR tree-optimization/84334
6552 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6553 also a CONSTANT_CLASS_P, punt.
6555 2018-02-14 Jim Wilson <jimw@sifive.com>
6557 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6558 first SMALL_OPERAND check. New local min_second_step. Move assert
6559 to where locals are set. Add TARGET_RVC support.
6560 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6562 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6564 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6566 2018-02-14 Martin Sebor <msebor@redhat.com>
6568 PR tree-optimization/83698
6569 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6570 arrays constrain the offset range to their bounds.
6571 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6572 (builtin_access::overlap): Avoid setting the size of overlap if it's
6574 (maybe_diag_overlap): Also consider arrays when deciding what values
6575 of offsets to include in diagnostics.
6577 2018-02-14 Martin Sebor <msebor@redhat.com>
6580 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6581 that correspond to the kind of a declaration.
6583 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6586 * config/pa/pa.md: Load address of PIC label using the linkage table
6587 if the label is nonlocal.
6589 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6591 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6592 warning message if user requests -maltivec=be.
6593 * doc/invoke.texi: Document deprecation of -maltivec=be.
6595 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6598 * config/rs6000/rs6000-c.c: Update definitions for
6599 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6600 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6602 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6605 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6606 add _get_ssp intrinsics. Remove argument from
6607 __builtin_ia32_rdssp[d|q].
6608 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6609 * config/i386/i386-builtin.def: Remove argument from
6610 __builtin_ia32_rdssp[d|q].
6611 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6612 ix86_expand_special_args_builtin for _rdssp[d|q].
6613 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6614 Clear register before usage.
6615 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6616 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6618 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6620 PR tree-optimization/84357
6621 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6622 operand 1 of an ARRAY_REF too.
6624 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6627 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6628 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6630 (set_of_reg): New struct.
6631 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6632 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6633 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6635 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6636 Split into bitclr, bitset, bitinvert patterns if appropriate.
6637 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6638 use rx_fuse_in_memory_bitop.
6639 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6640 to named insn, correct maximum insn length.
6642 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6645 * machmode.def: Define a complex mode for PARTIAL_INT.
6646 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6648 * doc/rtl.texi: Document CSPImode.
6649 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6651 (msp430_hard_regno_nregs_with_padding): Likewise.
6653 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6656 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6658 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6660 PR rtl-optimization/84169
6661 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6662 we generated a parallel as new i3 and we split that to new i2 and i3
6663 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6664 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6665 those to i2, not i1. Partially rewrite this scan code.
6667 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6670 * stor-layout.c (place_field): For variable length fields, adjust
6671 offset_align afterwards not just based on the field's alignment,
6672 but also on the size.
6675 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6676 of exps and logs in the use_exp2 case.
6678 2018-02-13 Jeff Law <law@redhat.com>
6680 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6683 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6686 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6690 * common.opt (gas-loc-support, gas-locview-support): New.
6691 (ginline-points, ginternal-reset-location-views): New.
6692 * doc/invoke.texi: Document them. Use @itemx where intended.
6693 (gvariable-location-views): Adjust.
6694 * target.def (reset_location_view): New.
6695 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6696 (TARGET_RESET_LOCATION_VIEW): New.
6697 * doc/tm.texi: Rebuilt.
6698 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6699 (dwarf2out_default_as_locview_support): New.
6700 (output_asm_line_debug_info): Use option variables.
6701 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6702 (output_loc_list): Likewise.
6703 (add_high_low_attributes): Check option variables.
6704 Don't output entry view attribute in strict mode.
6705 (gen_inlined_subroutine_die): Check option variables.
6706 (dwarf2out_inline_entry): Likewise.
6707 (init_sections_and_labels): Likewise.
6708 (dwarf2out_early_finish): Likewise.
6709 (maybe_reset_location_view): New, from...
6710 (dwarf2out_var_location): ... here. Call it.
6711 * debug.h (dwarf2out_default_as_loc_support): Declare.
6712 (dwarf2out_default_as_locview_support): Declare.
6713 * hooks.c (hook_int_rtx_insn_0): New.
6714 * hooks.h (hook_int_rtx_insn_0): Declare.
6715 * toplev.c (process_options): Take -gas-loc-support and
6716 -gas-locview-support from dwarf2out. Enable
6717 -gvariable-location-views by default only with locview
6718 assembler support. Enable -ginternal-reset-location-views by
6719 default only if the target defines the corresponding hook.
6720 Enable -ginline-points by default if location views are
6721 enabled; force it disabled if statement frontiers are
6723 * tree-inline.c (expand_call_inline): Check option variables.
6724 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6726 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6728 PR tree-optimization/84321
6729 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6730 handling. Also check whether the anti-range contains any values
6731 that satisfy the mask; switch to a VR_RANGE if not.
6733 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6736 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6738 2018-02-13 Martin Jambor <mjambor@suse.cz>
6741 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6742 of call statements, also set location of a load to a temporary.
6744 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6746 * config/rl78/rl78.c (add_vector_labels): New function.
6747 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6748 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6749 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6750 which checks that no arguments are passed.
6751 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6752 * doc/extend.texi: Documentation for the new attribute.
6754 2018-02-13 Andreas Schwab <schwab@suse.de>
6756 * config/riscv/linux.h (CPP_SPEC): Define.
6758 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6761 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6762 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6763 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6764 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6765 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6766 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6767 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6768 temporarily for AES and PCLMUL builtins.
6770 PR tree-optimization/84339
6771 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6772 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6776 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6777 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6778 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6780 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6782 * omp-simd-clone.h: New file.
6783 * omp-simd-clone.c: Include omp-simd-clone.h.
6784 (expand_simd_clones): No longer static.
6785 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6786 cgraph.h and omp-simd-clone.h.
6787 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6788 (vect_recog_widen_shift_pattern): Formatting fix.
6789 (vect_pattern_recog_1): Don't check optab for calls.
6792 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6793 operands[2] into a REG before using gen_lowpart on it.
6795 2018-02-12 Jeff Law <law@redhat.com>
6798 * config/sh/sh.c (find_barrier): Consider a sibling call
6801 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6802 successfully back substituting a reg.
6804 2018-02-12 Richard Biener <rguenther@suse.de>
6806 PR tree-optimization/84037
6807 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6808 parameter, move visited init to caller.
6809 (vect_slp_analyze_operations): Separate cost from validity
6810 check, initialize visited once for all instances.
6811 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6813 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6815 (vect_model_promotion_demotion_cost): Likewise.
6816 (vectorizable_bswap): Guard cost modeling with !slp_node
6817 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6819 (vectorizable_call): Likewise.
6820 (vectorizable_conversion): Likewise.
6821 (vectorizable_assignment): Likewise.
6822 (vectorizable_shift): Likewise.
6823 (vectorizable_operation): Likewise.
6824 (vectorizable_store): Likewise.
6825 (vectorizable_load): Likewise.
6826 (vectorizable_condition): Likewise.
6827 (vectorizable_comparison): Likewise.
6829 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6832 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6833 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6834 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6836 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6839 * params.def (inline-min-speedup): Increase from 8 to 15.
6840 (max-inline-insns-auto): Decrease from 40 to 30.
6841 * ipa-split.c (consider_split): Add some buffer for function to
6842 be considered inlining candidate.
6843 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6846 2018-02-12 Richard Biener <rguenther@suse.de>
6848 PR tree-optimization/84037
6849 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6850 matched stmts if we cannot swap the non-matched ones.
6852 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6854 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6855 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6856 _mm_maskz_scalef_round_ss): New intrinsics.
6857 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6858 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6859 __builtin_ia32_scalefss_round): Remove.
6860 (__builtin_ia32_scalefsd_mask_round,
6861 __builtin_ia32_scalefss_mask_round): New intrinsics.
6862 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6863 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6864 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6865 "<round_constraint>")): Changed to ...
6866 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6867 "<round_scalar_constraint>")): ... this.
6868 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6869 %0, %1, %2<round_op3>}"): Changed to ...
6870 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6871 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6872 %2<round_scalar_mask_op3>}"): ... this.
6873 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6875 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6877 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6878 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6879 (_mm_maskz_sqrt_round_ss): New intrinsics.
6880 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6881 (__builtin_ia32_sqrtsd_mask_round)
6882 (__builtin_ia32_sqrtss_mask_round): New builtins.
6883 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6884 (__builtin_ia32_sqrtss_round): Remove.
6885 (__builtin_ia32_sqrtsd_mask_round)
6886 (__builtin_ia32_sqrtss_mask_round): New builtins.
6887 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6888 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6889 ((match_operand:VF_128 1 "vector_operand"
6890 "xBm,<round_constraint>")): Changed to ...
6891 ((match_operand:VF_128 1 "vector_operand"
6892 "xBm,<round_scalar_constraint>")): ... this.
6893 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6894 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6895 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6896 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6897 %<iptr>1<round_scalar_mask_op3>}): ... this.
6898 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6899 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6901 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6904 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6905 Cast vec_cmpeq result to correct type.
6906 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6907 Cast vec_cmpgt result to correct type.
6909 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6911 * final.c (final_scan_insn_1): Renamed from...
6912 (final_scan_insn): ... this. New wrapper, to recover
6913 seen from the outermost call in recursive ones.
6914 * config/sparc/sparc.c (output_return): Drop seen from call.
6915 (output_sibcall): Likewise.
6916 * config/visium/visium.c (output_branch): Likewise.
6918 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6920 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6923 2018-02-10 Alan Modra <amodra@gmail.com>
6926 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6927 Specify LR as an input.
6929 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6932 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6933 remove_member_access_dummy_vars): New functions.
6934 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6935 lower_omp_1, execute_lower_omp): Use them.
6937 PR rtl-optimization/84308
6938 * shrink-wrap.c (spread_components): Release todo vector.
6940 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6942 PR rtl-optimization/57193
6943 * ira-color.c (struct allocno_color_data): Add member
6944 conflict_allocno_hard_prefs.
6945 (update_conflict_allocno_hard_prefs): New.
6946 (bucket_allocno_compare_func): Add a preference based on
6947 conflict_allocno_hard_prefs.
6948 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6949 (color_allocnos): Remove a dead code. Initiate
6950 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6952 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6955 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6956 constraint from =wa to wa. Avoid a subreg on the output operand,
6957 instead use a pseudo and subreg it in a move.
6958 (p9_xxbrd_<mode>): Changed to ...
6959 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6960 (p9_xxbrd_v2df): New expander.
6961 (p9_xxbrw_<mode>): Changed to ...
6962 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6963 (p9_xxbrw_v4sf): New expander.
6965 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6967 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6969 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6972 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6973 multiply in 32-bit mode.
6974 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6975 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6978 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6980 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6981 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6982 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6983 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6985 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6988 * dwarf2out.c (is_trivial_indirect_ref): New function.
6989 (dwarf2out_late_global_decl): Do not generate a location
6990 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6991 and that are not defined in the current unit.
6993 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6995 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6996 instead of a libcall for UNORDERED.
6998 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7001 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7002 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7004 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7007 * config/s390/s390.c (s390_set_current_function): Invoke
7008 s390_indirect_branch_settings also if fndecl didn't change.
7010 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7012 * config/rs6000/rs6000.md (blockage): Set length to zero.
7014 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7016 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7018 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7021 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7022 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7026 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7027 PARALLEL incoming that failed vt_get_decl_and_offset check.
7030 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7031 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7033 (get_variable_section): For decls in named .bss* sections pass true as
7034 second argument to bss_initializer_p.
7036 2018-02-09 Marek Polacek <polacek@redhat.com>
7037 Jakub Jelinek <jakub@redhat.com>
7040 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7041 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7042 Sync some changes from cxx_fold_indirect_ref.
7044 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7046 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7048 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7049 (BLOCK_INLINE_ENTRY_LABEL): New.
7050 (dwarf2out_var_location): Disregard inline entry markers.
7051 (inline_entry_data): New struct.
7052 (inline_entry_data_hasher): New hashtable type.
7053 (inline_entry_data_hasher::hash): New.
7054 (inline_entry_data_hasher::equal): New.
7055 (inline_entry_data_table): New variable.
7056 (add_high_low_attributes): Add DW_AT_entry_pc and
7057 DW_AT_GNU_entry_view attributes if a pending entry is found
7058 in inline_entry_data_table. Add old entry_pc attribute only
7059 if debug nonbinding markers are disabled.
7060 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7061 markers are enabled.
7062 (block_within_block_p, dwarf2out_inline_entry): New.
7063 (dwarf2out_finish): Check that no entries remained in
7064 inline_entry_data_table.
7065 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7066 (final_scan_insn, notice_source_line): Likewise.
7067 (rest_of_clean_state): Skip inline entry markers.
7068 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7070 * gimple.c (gimple_build_debug_inline_entry): New.
7071 * gimple.h (enum gimple_debug_subcode): Add
7072 GIMPLE_DEBUG_INLINE_ENTRY.
7073 (gimple_build_debug_inline_entry): Declare.
7074 (gimple_debug_inline_entry_p): New.
7075 (gimple_debug_nonbind_marker_p): Adjust.
7076 * insn-notes.def (INLINE_ENTRY): New.
7077 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7078 inline entry marker notes.
7079 (print_insn): Likewise.
7080 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7081 (INSN_DEBUG_MARKER_KIND): Likewise.
7082 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7083 * tree-inline.c (expand_call_inline): Build and insert
7084 debug_inline_entry stmt.
7085 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7086 inline entry blocks early, if nonbind markers are enabled.
7087 (dump_scope_block): Dump fragment info.
7088 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7089 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7090 (gimple_build_debug_inline_entry): New.
7091 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7092 Enable/disable inline entry points too.
7093 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7094 (DEBUG_INSN): Describe inline entry markers.
7096 * common.opt (gvariable-location-views): New.
7097 (gvariable-location-views=incompat5): New.
7098 * config.in: Rebuilt.
7099 * configure: Rebuilt.
7100 * configure.ac: Test assembler for view support.
7101 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7102 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7103 * dwarf2out.c (var_loc_view): New typedef.
7104 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7105 (dwarf2out_locviews_in_attribute): New.
7106 (dwarf2out_locviews_in_loclist): New.
7107 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7108 (enum dw_line_info_opcode): Add LI_adv_address.
7109 (struct dw_line_info_table): Add view.
7110 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7111 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7112 (zero_view_p): New variable.
7113 (ZERO_VIEW_P): New macro.
7114 (output_asm_line_debug_info): New.
7115 (struct var_loc_node): Add view.
7116 (add_AT_view_list, AT_loc_list): New.
7117 (add_var_loc_to_decl): Add view param. Test it against last.
7118 (new_loc_list): Add view params. Record them.
7119 (AT_loc_list_ptr): Handle loc and view lists.
7120 (view_list_to_loc_list_val_node): New.
7121 (print_dw_val): Handle dw_val_class_view_list.
7122 (size_of_die): Likewise.
7123 (value_format): Likewise.
7124 (loc_list_has_views): New.
7125 (gen_llsym): Set vl_symbol too.
7126 (maybe_gen_llsym, skip_loc_list_entry): New.
7127 (dwarf2out_maybe_output_loclist_view_pair): New.
7128 (output_loc_list): Output view list or entries too.
7129 (output_view_list_offset): New.
7130 (output_die): Handle dw_val_class_view_list.
7131 (output_dwarf_version): New.
7132 (output_compilation_unit_header): Use it.
7133 (output_skeleton_debug_sections): Likewise.
7134 (output_rnglists, output_line_info): Likewise.
7135 (output_pubnames, output_aranges): Update version comments.
7136 (output_one_line_info_table): Output view numbers in asm comments.
7137 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7138 Call maybe_gen_llsym.
7139 (loc_list_from_tree_1): Adjust.
7140 (add_AT_location_description): Create view list attribute if
7141 needed, check it's absent otherwise.
7142 (convert_cfa_to_fb_loc_list): Adjust.
7143 (maybe_emit_file): Call output_asm_line_debug_info for test.
7144 (dwarf2out_var_location): Reset views as needed. Precompute
7145 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7146 attribute. Set view.
7147 (new_line_info_table): Reset next view.
7148 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7149 (dwarf2out_source_line): Likewise. Output view resets and labels to
7150 the assembler, or select appropriate line info opcodes.
7151 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7152 (optimize_string_length): Catch it. Adjust.
7153 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7154 dw_val_class_view_list, and remove it if no longer needed.
7155 (hash_loc_list): Hash view numbers.
7156 (loc_list_hasher::equal): Compare them.
7157 (optimize_location_lists): Check whether a view list symbol is
7158 needed, and whether the locview attribute is present, and
7159 whether they match. Remove the locview attribute if no longer
7161 (index_location_lists): Call skip_loc_list_entry for test.
7162 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7163 Use output_dwarf_version.
7164 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7165 (struct dw_val_node): Add val_view_list.
7166 * final.c (SEEN_NEXT_VIEW): New.
7167 (set_next_view_needed): New.
7168 (clear_next_view_needed): New.
7169 (maybe_output_next_view): New.
7170 (final_start_function): Rename to...
7171 (final_start_function_1): ... this. Take pointer to FIRST,
7172 add SEEN parameter. Emit param bindings in the initial view.
7173 (final_start_function): Reintroduce SEEN-less interface.
7174 (final): Rename to...
7175 (final_1): ... this. Take SEEN parameter. Output final pending
7176 next view at the end.
7177 (final): Reintroduce seen-less interface.
7178 (final_scan_insn): Output pending next view before switching
7179 sections or ending a block. Mark the next view as needed when
7180 outputting variable locations. Notify debug backend of section
7181 changes, and of location view changes.
7182 (rest_of_handle_final): Adjust.
7183 * toplev.c (process_options): Autodetect value for debug variable
7184 location views option. Warn on incompat5 without -gdwarf-5.
7185 * doc/invoke.texi (gvariable-location-views): New.
7186 (gvariable-location-views=incompat5): New.
7187 (gno-variable-location-views): New.
7189 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7191 PR tree-optimization/84136
7192 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7193 that the result of find_edge is non-NULL.
7195 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7198 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7199 storing integer register in SImode. Fix cost of 256 and 512
7200 byte aligned SSE register store.
7202 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7204 * config/i386/i386.c (ix86_multiplication_cost): Fix
7205 multiplication cost for TARGET_AVX512DQ.
7207 2018-02-08 Marek Polacek <polacek@redhat.com>
7209 PR tree-optimization/84238
7210 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7213 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7215 PR tree-optimization/84265
7216 * tree-vect-stmts.c (vectorizable_store): Don't treat
7217 VMAT_CONTIGUOUS accesses as grouped.
7218 (vectorizable_load): Likewise.
7220 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7222 PR tree-optimization/81635
7223 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7224 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7225 (test_round_for_mask): New functions.
7226 (wide_int_cc_tests): Call test_round_for_mask.
7227 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7228 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7229 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7230 range returned by get_range_info.
7232 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7235 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7236 * symtab.c: Include builtins.h
7237 (symtab_node::output_to_lto_symbol_table_p): Move here
7238 from lto-streamer-out.c:output_symbol_p.
7239 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7240 (output_symbol_p): Move all logic to symtab.c
7241 (produce_symtab): Update.
7243 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7245 * config/s390/s390-opts.h (enum indirect_branch): Define.
7246 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7247 (s390_indirect_branch_via_thunk)
7248 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7249 (enum s390_indirect_branch_type): Define.
7250 * config/s390/s390.c (struct s390_frame_layout, struct
7251 machine_function): Remove.
7252 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7253 (indirect_branch_table_label_no, indirect_branch_table_name):
7255 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7256 (enum s390_indirect_branch_option): Define.
7257 (s390_return_addr_from_memory): New function.
7258 (s390_handle_string_attribute): New function.
7259 (s390_attribute_table): Add new attribute handler.
7260 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7261 (s390_indirect_branch_via_thunk): New function.
7262 (s390_indirect_branch_via_inline_thunk): New function.
7263 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7264 sibling call optimization for non z10 compiles.
7265 (s390_emit_call): Force indirect branch target to be a single
7266 register. Add r1 clobber for non-z10 compiles.
7267 (s390_emit_epilogue): Emit return jump via return_use expander.
7268 (s390_reorg): Handle JUMP_INSNs as execute targets.
7269 (s390_option_override_internal): Perform validity checks for the
7270 new command line options.
7271 (s390_indirect_branch_attrvalue): New function.
7272 (s390_indirect_branch_settings): New function.
7273 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7274 (s390_output_indirect_thunk_function): New function.
7275 (s390_code_end): Implement target hook.
7276 (s390_case_values_threshold): Implement target hook.
7277 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7279 * config/s390/s390.h (struct s390_frame_layout)
7280 (struct machine_function): Move here from s390.c.
7281 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7282 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7283 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7284 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7285 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7286 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7287 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7288 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7289 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7290 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7291 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7292 (mnemonic attribute): Add values which aren't recognized
7294 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7295 pattern for branch conversion. Fix mnemonic attribute.
7296 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7297 indirect branch via thunk if requested.
7298 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7299 ("*indirect_jump"): Disable for branch conversion using out of
7301 ("indirect_jump_via_thunk<mode>_z10")
7302 ("indirect_jump_via_thunk<mode>")
7303 ("indirect_jump_via_inlinethunk<mode>_z10")
7304 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7305 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7306 ("casesi_jump_via_inlinethunk<mode>_z10")
7307 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7308 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7309 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7310 ("*indirect2_jump"): Disable for branch conversion.
7311 ("casesi_jump"): Turn into expander and expand patterns for branch
7313 ("return_use"): New expander.
7314 ("*return"): Emit return via thunk and rename it to ...
7315 ("*return<mode>"): ... this one.
7316 * config/s390/s390.opt: Add new options and and enum for the
7319 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7321 * lra-constraints.c (match_reload): Unconditionally use
7322 gen_lowpart_SUBREG, rather than selecting between that
7323 and equivalent gen_rtx_SUBREG code.
7325 2018-02-08 Richard Biener <rguenther@suse.de>
7327 PR tree-optimization/84233
7328 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7329 changed flag instead of boguously re-using phi_inserted.
7331 2018-02-08 Martin Jambor <mjambor@suse.cz>
7333 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7334 static local variables.
7336 2018-02-08 Richard Biener <rguenther@suse.de>
7338 PR tree-optimization/84278
7339 * tree-vect-stmts.c (vectorizable_store): When looking for
7340 smaller vector types to perform grouped strided loads/stores
7341 make sure the mode is supported by the target.
7342 (vectorizable_load): Likewise.
7344 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7346 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7347 Increase LDP/STP opportunities by adding adjacent callee-saves.
7349 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7351 PR rtl-optimization/84068
7352 PR rtl-optimization/83459
7353 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7355 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7357 PR tree-optimization/84224
7358 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7359 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7362 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7365 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7366 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7367 count, remove one USE.
7369 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7371 * doc/install.texi (Configuration): Document the
7372 --with-long-double-format={ibm,ieee} PowerPC configuration
7376 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7377 Convert from define_expand to be define_insn_and_split. Rework
7378 float/double/_Float128 conversions to QI/HI/SImode to work with
7379 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7380 conversions to QI/HImode types did a store and then a load to
7381 truncate the value. For conversions to VSX registers, don't split
7382 the insn, instead emit the code directly. Use the code iterator
7383 any_fix to combine signed and unsigned conversions.
7384 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7385 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7386 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7387 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7388 (fix_<mode>di2_hw): Likewise.
7389 (fixuns_<mode>di2_hw): Likewise.
7390 (fix_<mode>si2_hw): Likewise.
7391 (fixuns_<mode>si2_hw): Likewise.
7392 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7393 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7394 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7395 fix<uns>_trunc<SFDF:mode>si2_p8.
7396 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7398 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7399 (fix<uns>_<mode>_mem): Likewise.
7400 (fctiw<u>z_<mode>_mem): Likewise.
7401 (fix<uns>_<mode>_mem): Likewise.
7402 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7403 the register allocator from doing a direct move to the GPRs to do
7404 a store, and instead use the ISA 3.0 store byte/half-word from
7405 vector register instruction. For IEEE 128-bit floating point,
7406 also optimize stores of 32-bit ints.
7407 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7409 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7411 * genextract.c (push_pathstr_operand): New function to support
7413 (walk_rtx): Call push_pathstr_operand.
7414 (print_path): Support [a-zA-Z].
7416 2018-02-07 Richard Biener <rguenther@suse.de>
7418 PR tree-optimization/84037
7419 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7420 (cse_and_gimplify_to_preheader): Declare.
7421 (vect_get_place_in_interleaving_chain): Likewise.
7422 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7424 (_loop_vec_info::~_loop_vec_info): Delete it.
7425 (cse_and_gimplify_to_preheader): New function.
7426 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7427 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7428 (vectorizable_load): Likewise. For grouped stores always base
7429 the IV on the first element.
7430 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7431 condition before gimplifying.
7433 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7435 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7436 *DIV_EXPR and *MOD_EXPR.
7438 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7441 * config/i386/i386.c (ix86_option_override_internal): Mask out
7442 the CF_SET bit when checking -fcf-protection.
7444 2018-02-07 Tom de Vries <tom@codesourcery.com>
7447 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7450 2018-02-07 Richard Biener <rguenther@suse.de>
7452 PR tree-optimization/84204
7453 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7456 PR tree-optimization/84205
7457 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7458 special-case isl_ast_op_zdiv_r.
7460 PR tree-optimization/84223
7461 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7462 Only add conditions from within the region.
7463 (gather_bbs::after_dom_children): Adjust.
7465 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7468 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7469 * config/avr/avr.md: Only post-reload split REG-REG moves if
7470 either register is GENERAL_REG_P.
7472 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7474 PR tree-optimization/84235
7475 * tree-ssa-scopedtables.c
7476 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7477 if the subtraction is performed in floating point type where NaNs are
7478 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7479 build 1. Formatting fix.
7481 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7484 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7485 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7486 and skip it regardless of bb boundaries. Use CALL_P macro,
7487 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7488 unnecessarily, formatting fix.
7490 2018-02-06 Michael Collison <michael.collison@arm.com>
7492 * config/arm/thumb2.md:
7493 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7494 (*thumb_mov_notscc): Ditto.
7496 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7499 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7500 unsigned_fix, not "s".
7502 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7504 * configure.ac (gcc_fn_eh_frame_ro): New function.
7505 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7506 correct .eh_frame permissions.
7507 * configure: Regenerate.
7509 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7511 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7514 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7516 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7517 Display warning message for -mno-speculate-indirect-jumps.
7519 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7521 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7523 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7525 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7527 PR tree-optimization/84225
7528 * tree-eh.c (find_trapping_overflow): Only call
7529 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7531 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7534 * config/i386/i386.c: Reimplement the check of possible options
7535 -mibt/-mshstk conbination. Change error messages.
7536 * doc/invoke.texi: Fix a typo: remove extra '='.
7538 2018-02-06 Marek Polacek <polacek@redhat.com>
7540 PR tree-optimization/84228
7541 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7543 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7546 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7547 emitted arch directives.
7548 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7549 __ARM_FEATURE_COPROC before changing architectures.
7551 2018-02-06 Richard Biener <rguenther@suse.de>
7553 * config/i386/i386.c (print_reg): Fix typo.
7554 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7556 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7558 * configure: Regenerate.
7560 2018-02-05 Martin Sebor <msebor@redhat.com>
7562 PR tree-optimization/83369
7563 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7566 2018-02-05 Martin Liska <mliska@suse.cz>
7568 * doc/invoke.texi: Cherry-pick upstream r323995.
7570 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7572 * ira.c (ira_init_register_move_cost): Adjust comment.
7574 2018-02-05 Martin Liska <mliska@suse.cz>
7576 PR gcov-profile/84137
7577 * doc/gcov.texi: Fix typo in documentation.
7579 2018-02-05 Martin Liska <mliska@suse.cz>
7581 PR gcov-profile/83879
7582 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7583 using dlopen functionality.
7585 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7587 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7588 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7589 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7590 _mm_maskz_range_round_ss): New intrinsics.
7591 (__builtin_ia32_rangesd128_round)
7592 (__builtin_ia32_rangess128_round): Remove.
7593 (__builtin_ia32_rangesd128_mask_round,
7594 __builtin_ia32_rangess128_mask_round): New builtins.
7595 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7596 __builtin_ia32_rangess128_round): Remove.
7597 (__builtin_ia32_rangesd128_mask_round,
7598 __builtin_ia32_rangess128_mask_round): New builtins.
7599 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7600 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7601 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7602 "<round_saeonly_constraint>")): Changed to ...
7603 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7604 "<round_saeonly_scalar_constraint>")): ... this.
7605 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7606 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7607 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7608 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7609 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7611 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7613 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7615 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7616 Remove all values except native, 8540 and 8548.
7618 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7620 * config/i386/i386.c (ix86_output_function_return): Pass
7621 INVALID_REGNUM, instead of -1, as invalid register number to
7622 indirect_thunk_name and output_indirect_thunk.
7624 2018-02-02 Julia Koval <julia.koval@intel.com>
7626 * config.gcc: Add -march=icelake.
7627 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7628 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7629 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7630 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7631 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7632 (processor_target_table): Add icelake.
7633 (ix86_option_override_internal): Handle new PTAs.
7634 (get_builtin_code_for_version): Handle icelake.
7635 (M_INTEL_COREI7_ICELAKE): New.
7636 (fold_builtin_cpu): Handle icelake.
7637 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7638 * doc/invoke.texi: Add -march=icelake.
7640 2018-02-02 Julia Koval <julia.koval@intel.com>
7642 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7643 to wide_int_bitmask.
7644 * wide-int-bitmask.h: New.
7646 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7649 * config/i386/i386.md: Replace Pmode with word_mode in
7650 builtin_setjmp_setup and builtin_longjmp to support x32.
7652 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7656 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7658 (rs6000_supported_cpu_names): New static variable.
7659 (linux_cpu_translation_table): Likewise.
7660 (elf_platform) <cpu>: Define new static variable and use it.
7661 Translate kernel AT_PLATFORM name to canonical name if needed.
7662 Error if platform name is unknown.
7664 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7667 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7669 2018-02-01 Jeff Law <law@redhat.com>
7672 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7673 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7674 the scratch if RELEASE_VIA_POP is false.
7675 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7676 If we have to save a temporary register, decrement SIZE appropriately.
7677 Pass new arguments to release_scratch_register_on_entry.
7678 (ix86_adjust_stack_and_probe): Likewise.
7679 (ix86_emit_probe_stack_range): Pass new arguments to
7680 release_scratch_register_on_entry.
7682 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7684 PR rtl-optimization/84157
7685 * combine.c (change_zero_ext): Use REG_P predicate in
7686 front of HARD_REGISTER_P predicate.
7688 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7690 * config/avr/avr.c (avr_option_override): Move disabling of
7691 -fdelete-null-pointer-checks to...
7692 * common/config/avr/avr-common.c (avr_option_optimization_table):
7695 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7697 PR tree-optimization/81635
7698 * tree-data-ref.c (split_constant_offset_1): For types that
7699 wrap on overflow, try to use range info to prove that wrapping
7702 2018-02-01 Renlin Li <renlin.li@arm.com>
7705 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7707 (aarch64_register_move_cost): Likewise.
7708 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7710 (REG_CLASS_NAMES): Likewise.
7711 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7712 TAILCALL_ADDR_REGS. Remove IP registers.
7713 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7715 2018-02-01 Richard Biener <rguenther@suse.de>
7717 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7718 for specifying RPO order and allow NULL for that.
7719 * domwalk.c (dom_walker::dom_walker): Likewise.
7720 (dom_walker::walk): Handle NULL RPO order.
7721 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7723 (rewrite_update_dom_walker): Likewise.
7724 (mark_def_dom_walker): Likewise.
7726 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7728 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7729 (aarch64_maybe_expand_sve_subreg_move): Declare.
7730 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7731 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7733 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7734 that are semantically a reverse operation.
7735 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7736 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7737 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7739 (aarch64_can_change_mode_class): For big-endian, forbid changes
7740 between two SVE modes if they have different element sizes.
7742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7744 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7745 the TImode handling for big-endian targets.
7747 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7749 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7750 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7752 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7753 Remove BSWAP handing for big-endian targets and use the form of
7754 LD1RQ appropariate for the mode.
7756 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7758 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7759 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7762 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7765 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7766 check for operands that need to go through aarch64_sve_reload_be.
7768 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7770 PR tree-optimization/81661
7771 PR tree-optimization/84117
7772 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7773 * tree-eh.c: Include gimplify.h.
7774 (find_trapping_overflow, replace_trapping_overflow,
7775 rewrite_to_non_trapping_overflow): New functions.
7776 * tree-vect-loop.c: Include tree-eh.h.
7777 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7778 * tree-data-ref.c: Include tree-eh.h.
7779 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7781 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7783 PR rtl-optimization/84123
7784 * combine.c (change_zero_ext): Check if hard register satisfies
7785 can_change_dest_mode before calling gen_lowpart_SUBREG.
7787 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7790 * ira.c (ira_init_register_move_cost): Remove assert.
7792 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7794 PR rtl-optimization/84071
7795 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7796 * doc/tm.texi: Regenerate.
7798 2018-01-31 Richard Biener <rguenther@suse.de>
7800 PR tree-optimization/84132
7801 * tree-data-ref.c (analyze_miv_subscript): Properly
7802 check whether evolution_function_is_affine_multivariate_p
7803 before calling gcd_of_steps_may_divide_p.
7805 2018-01-31 Julia Koval <julia.koval@intel.com>
7808 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7809 * config/i386/i386.md (rdpid_rex64) New.
7810 (rdpid): Make 32bit only.
7812 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7815 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7816 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7818 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7821 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7823 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7825 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7827 PR rtl-optimization/84071
7828 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7829 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7831 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7833 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7834 (arc_attribute_table): Add 'aux' attribute.
7835 (arc_in_small_data_p): Consider aux like variables.
7836 (arc_is_aux_reg_p): New function.
7837 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7838 (arc_get_aux_arg): New function.
7839 (prepare_move_operands): Handle aux-register access.
7840 (arc_handle_aux_attribute): New function.
7841 * doc/extend.texi (ARC Variable attributes): Add subsection.
7843 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7845 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7846 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7847 (arc_attribute_table): Add 'uncached' attribute.
7848 (arc_print_operand): Print '.di' flag for uncached memory
7850 (arc_in_small_data_p): Do not consider for small data the uncached
7852 (arc_is_uncached_mem_p): New function.
7853 * config/arc/predicates.md (compact_store_memory_operand): Check
7854 for uncached memory accesses.
7855 (nonvol_nonimm_operand): Likewise.
7856 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7858 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7861 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7862 falign-loops=): Add Optimization flag.
7864 2018-01-30 Jeff Law <law@redhat.com>
7867 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7868 INT_REGISTERS_SAVED. Check it prior to calling
7869 get_scratch_register_on_entry.
7870 (ix86_adjust_stack_and_probe): Similarly.
7871 (ix86_emit_probe_stack_range): Similarly.
7872 (ix86_expand_prologue): Corresponding changes.
7874 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7877 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7878 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7880 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7883 * lra-constraints.c (curr_insn_transform): Process AND in the
7886 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7888 PR rtl-optimization/83986
7889 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7890 dependence against last_pending_memory_flush in addition to
7893 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7895 PR tree-optimization/81611
7896 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7899 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7902 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7905 2018-01-30 Richard Biener <rguenther@suse.de>
7906 Jakub Jelinek <jakub@redhat.com>
7908 PR tree-optimization/84111
7909 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7910 inner loops added during recursion, as they don't have up-to-date
7913 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7916 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7917 (can_inline_edge_by_limits_p): ... here.
7918 (can_early_inline_edge_p, check_callers,
7919 update_caller_keys, update_callee_keys, recursive_inlining,
7920 add_new_edges_to_heap, speculation_useful_p,
7921 inline_small_functions,
7922 inline_small_functions, flatten_function,
7923 inline_to_all_callers_1): Update.
7925 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7927 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7930 2018-01-30 Richard Biener <rguenther@suse.de>
7932 PR tree-optimization/83008
7933 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7934 invariant and constant vector uses in stmts when they need
7937 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7940 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7941 * configure: Regenerate.
7943 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7945 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7947 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7948 Use gen_rtx_REG rather than gen_lowpart.
7950 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7952 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7953 rather than 0 when creating partial subregs.
7955 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7957 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7960 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7963 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7964 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7965 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7966 flags. This restores the settings used before the 2017-07-24.
7967 Turning off pre increment/decrement/modify allows IVOPTS to
7968 optimize DF/SF loops where the index is an int.
7970 2018-01-29 Richard Biener <rguenther@suse.de>
7971 Kelvin Nilsen <kelvin@gcc.gnu.org>
7974 * tree-vect-stmts.c (vectorizable_call): Don't call
7975 targetm.vectorize_builtin_md_vectorized_function if callee is
7978 2018-01-22 Carl Love <cel@us.ibm.com>
7980 * doc/extend.tex: Fix typo in second arg in
7981 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7983 2018-01-29 Richard Biener <rguenther@suse.de>
7985 PR tree-optimization/84086
7986 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7987 (flush_ssaname_freelist): When SSA names were released reset
7988 the SCEV hash table.
7990 2018-01-29 Richard Biener <rguenther@suse.de>
7992 PR tree-optimization/84057
7993 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7994 removed paths when removing edges.
7996 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7998 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7999 -mfunction-return=@var{choice}.
8001 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8004 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8006 (layout::print_source_line): Likewise.
8007 (test_get_line_width_without_trailing_whitespace): Add test cases.
8009 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8012 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8015 2018-01-26 Jim Wilson <jimw@sifive.com>
8017 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8019 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8022 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8024 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8025 and CMP + SUB-immediate -> SUBS.
8027 2018-01-26 Martin Sebor <msebor@redhat.com>
8029 PR tree-optimization/83896
8030 * tree-ssa-strlen.c (get_string_len): Rename...
8031 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8032 Avoid assuming length is constant.
8033 (handle_char_store): Use HOST_WIDE_INT for string length.
8035 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8038 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8039 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8041 2018-01-26 Richard Biener <rguenther@suse.de>
8043 PR rtl-optimization/84003
8044 * dse.c (record_store): Only record redundant stores when
8045 the earlier store aliases at least all accesses the later one does.
8047 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8049 PR rtl-optimization/83985
8050 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8051 REG_CFA_RESTORE insns.
8052 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8053 REG_CFA_RESTORE insns here.
8056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8057 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8059 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8061 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8062 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8063 (arc_init): Likewise.
8064 (arc_override_options): Likewise.
8065 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8067 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8069 * config/arc/arc.h (TARGET_DBNZ): Define.
8070 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8071 properly set the tune attribute.
8072 (dbnz): Use TARGET_DBNZ guard.
8073 * config/arc/arc.opt (mtune): Add core3 option.
8075 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8077 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8078 recognize new pic like addresses.
8079 (arc_delegitimize_address): Clean up.
8081 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8083 * config/arc/arc-arches.def: Option mrf16 valid for all
8085 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8086 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8087 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8088 * config/arc/arc-tables.opt: Regenerate.
8089 * config/arc/arc.c (arc_conditional_register_usage): Handle
8090 reduced register file case.
8091 (arc_file_start): Set must have build attributes.
8092 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8094 * config/arc/arc.opt (mrf16): Add new option.
8095 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8096 * config/arc/genmultilib.awk: Handle new mrf16 option.
8097 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8098 * config/arc/t-multilib: Regenerate.
8099 * doc/invoke.texi (ARC Options): Document mrf16 option.
8101 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8103 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8104 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8105 (arc_attribute_table): Add 'secure_call' attribute.
8106 (arc_print_operand): Print secure call operand.
8107 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8109 (arc_is_secure_call_p): New function. * config/arc/arc.md
8110 (call_i): Add support for sjli instruction.
8111 (call_value_i): Likewise.
8112 * config/arc/constraints.md (Csc): New constraint.
8114 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8115 John Eric Martin <John.Martin@emmicro-us.com>
8117 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8118 * config/arc/arc.c (_arc_jli_section): New struct.
8119 (arc_jli_section): New type.
8120 (rc_jli_sections): New static variable.
8121 (arc_handle_jli_attribute): New function.
8122 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8123 (arc_file_end): New function.
8124 (TARGET_ASM_FILE_END): Define.
8125 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8126 (arc_add_jli_section): New function.
8127 (jli_call_scan): Likewise.
8128 (arc_reorg): Call jli_call_scan.
8129 (arc_output_addsi): Remove 'S' from printing asm operand.
8130 (arc_is_jli_call_p): New function.
8131 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8133 (movhi_insn): Likewise.
8134 (movsi_insn): Likewise.
8135 (movsi_set_cc_insn): Likewise.
8136 (loadqi_update): Likewise.
8137 (load_zeroextendqisi_update): Likewise.
8138 (load_signextendqisi_update): Likewise.
8139 (loadhi_update): Likewise.
8140 (load_zeroextendhisi_update): Likewise.
8141 (load_signextendhisi_update): Likewise.
8142 (loadsi_update): Likewise.
8143 (loadsf_update): Likewise.
8144 (movsicc_insn): Likewise.
8145 (bset_insn): Likewise.
8146 (bxor_insn): Likewise.
8147 (bclr_insn): Likewise.
8148 (bmsk_insn): Likewise.
8149 (bicsi3_insn): Likewise.
8150 (cmpsi_cc_c_insn): Likewise.
8151 (movsi_ne): Likewise.
8152 (movsi_cond_exec): Likewise.
8153 (clrsbsi2): Likewise.
8164 (call_i): Remove 'S' asm letter, add jli instruction.
8165 (call_value_i): Likewise.
8166 * config/arc/arc.op (mjli-always): New option.
8167 * config/arc/constraints.md (Cji): New constraint.
8168 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8170 (subsf3_fpx): Likewise.
8171 (mulsf3_fpx): Likewise.
8172 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8174 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8176 * doc/invoke.texi (ARC): Document mjli-always option.
8178 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8180 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8181 avoid addition with 0 and use incw and decw where possible.
8183 2018-01-26 Richard Biener <rguenther@suse.de>
8185 PR tree-optimization/81082
8186 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8187 association if it requires casting to unsigned.
8188 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8189 from fold_plusminus_mult_expr to catch important cases late when
8190 range info is available.
8192 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8194 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8195 * configure.ac (hidden_linkonce): New test.
8196 * configure: Regenerate.
8197 * config.in: Regenerate.
8199 2018-01-26 Julia Koval <julia.koval@intel.com>
8201 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8202 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8203 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8204 _mm_mask_bitshuffle_epi64_mask): Fix type.
8205 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8206 USI_FTYPE_V4DI_V4DI_USI): Remove.
8207 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8208 __builtin_ia32_vpshufbitqmb256_mask,
8209 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8210 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8211 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8213 2018-01-26 Alan Modra <amodra@gmail.com>
8216 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8217 UNSPEC_VBPERMQ. Sort other unspecs.
8219 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8221 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8223 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8226 * predict.c (drop_profile): Do not push/pop cfun; update also
8228 (handle_missing_profiles): Fix logic looking for zero profiles.
8230 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8233 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8234 on functions with #pragma omp declare simd or functions with simd
8236 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8237 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8238 Remove trailing \n from warning_at calls.
8240 2018-01-25 Tom de Vries <tom@codesourcery.com>
8243 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8244 for neutered workers.
8246 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8249 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8250 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8252 2018-01-24 Jeff Law <law@redhat.com>
8255 * i386.c (get_probe_interval): Move to earlier point.
8256 (ix86_compute_frame_layout): If -fstack-clash-protection and
8257 the frame is larger than the probe interval, then use pushes
8258 to save registers rather than reg->mem moves.
8259 (ix86_expand_prologue): Remove conditional for int_registers_saved
8262 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8265 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8266 min/max for never referenced object.
8268 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8271 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8273 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8274 attributes from DECL_ATTRIBUTES (decl) without affecting
8275 DECL_ATTRIBUTES (current_function_decl).
8276 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8277 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8279 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8281 PR tree-optimization/83979
8282 * fold-const.c (fold_comparison): Use constant_boolean_node
8283 instead of boolean_{true,false}_node.
8285 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8287 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8290 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8292 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8293 Simplify the clause that sets the length attribute.
8294 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8295 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8296 clause that sets the length attribute.
8297 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8299 2018-01-24 Tom de Vries <tom@codesourcery.com>
8302 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8303 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8304 Add strict parameter.
8305 (prevent_branch_around_nothing): Insert dummy insn between branch to
8306 label and label with no ptx insn inbetween.
8307 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8309 2018-01-24 Tom de Vries <tom@codesourcery.com>
8312 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8313 for neutered threads in warp.
8314 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8316 2018-01-24 Richard Biener <rguenther@suse.de>
8318 PR tree-optimization/83176
8319 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8322 2018-01-24 Richard Biener <rguenther@suse.de>
8324 PR tree-optimization/82819
8325 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8326 code generating pluses that are no-ops in the target precision.
8328 2018-01-24 Richard Biener <rguenther@suse.de>
8331 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8333 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8335 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8336 to merge probabilities.
8337 * predict.c (probably_never_executed): Also mark as cold functions
8338 with global 0 profile and guessed local profile.
8339 * profile-count.c (profile_probability::combine_with_count): New
8341 * profile-count.h (profile_probability::operator*,
8342 profile_probability::operator*=, profile_probability::operator/,
8343 profile_probability::operator/=): Reduce precision to adjusted
8344 and set value to guessed on contradictory divisions.
8345 (profile_probability::combine_with_freq): Remove.
8346 (profile_probability::combine_wiht_count): Declare.
8347 (profile_count::force_nonzero):: Set to adjusted.
8348 (profile_count::probability_in):: Set quality to adjusted.
8349 * tree-ssa-tail-merge.c (replace_block_by): Use
8352 2018-01-23 Andrew Waterman <andrew@sifive.com>
8353 Jim Wilson <jimw@sifive.com>
8355 * config/riscv/riscv.c (riscv_stack_boundary): New.
8356 (riscv_option_override): Set riscv_stack_boundary. Handle
8357 riscv_preferred_stack_boundary_arg.
8358 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8359 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8360 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8361 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8362 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8363 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8365 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8368 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8369 of struct ix86_frame.
8370 (ix86_expand_epilogue): Likewise. Add a local variable for
8371 the reg_save_offset field in struct ix86_frame.
8373 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8375 PR tree-optimization/82604
8376 * tree-loop-distribution.c (enum partition_kind): New enum item
8377 PKIND_PARTIAL_MEMSET.
8378 (partition_builtin_p): Support above new enum item.
8379 (generate_code_for_partition): Ditto.
8380 (compute_access_range): Differentiate cases that equality can be
8381 proven at all loops, the innermost loops or no loops.
8382 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8383 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8384 (finalize_partitions, distribute_loop): Don't fuse partition of
8385 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8386 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8389 2018-01-23 Martin Liska <mliska@suse.cz>
8391 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8392 order to ignore the predictor.
8393 (PRED_POLYMORPHIC_CALL): Likewise.
8394 (PRED_RECURSIVE_CALL): Likewise.
8396 2018-01-23 Martin Liska <mliska@suse.cz>
8398 * tree-profile.c (tree_profiling): Print function header to
8399 aware reader which function we are working on.
8400 * value-prof.c (gimple_find_values_to_profile): Do not print
8401 not interesting value histograms.
8403 2018-01-23 Martin Liska <mliska@suse.cz>
8405 * profile-count.h (enum profile_quality): Add
8406 profile_uninitialized as the first value. Do not number values
8407 as they are zero based.
8408 (profile_count::verify): Update sanity check.
8409 (profile_probability::verify): Likewise.
8411 2018-01-23 Nathan Sidwell <nathan@acm.org>
8413 * doc/invoke.texi (ffor-scope): Deprecate.
8415 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8417 PR tree-optimization/83510
8418 * domwalk.c (set_all_edges_as_executable): New function.
8419 (dom_walker::dom_walker): Convert bool param
8420 "skip_unreachable_blocks" to enum reachability. Move setup of
8421 edge flags to set_all_edges_as_executable and only do it when
8422 reachability is REACHABLE_BLOCKS.
8423 * domwalk.h (enum dom_walker::reachability): New enum.
8424 (dom_walker::dom_walker): Convert bool param
8425 "skip_unreachable_blocks" to enum reachability.
8426 (set_all_edges_as_executable): New decl.
8427 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8428 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8430 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8431 but converting true to REACHABLE_BLOCKS.
8432 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8434 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8435 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8436 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8438 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8439 if check_all_array_refs will be called.
8441 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8443 * tree.c (selftest::test_location_wrappers): Add more test
8446 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8448 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8449 (selftest::test_bit_in_range): Likewise.
8451 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8454 * doc/sourcebuild.texi (vect_float): Say that the selector
8455 only describes the situation when -funsafe-math-optimizations is on.
8456 (vect_float_strict): Document.
8458 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8460 PR tree-optimization/83965
8461 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8462 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8463 instead of checking only for a reduction.
8464 (vect_recog_widen_sum_pattern): Likewise.
8466 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8468 * predict.c (probably_never_executed): Only use precise profile info.
8469 (compute_function_frequency): Skip after inlining hack since we now
8470 have quality checking.
8472 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8474 * profile-count.h (profile_probability::very_unlikely,
8475 profile_probability::unlikely, profile_probability::even): Set
8476 precision to guessed.
8478 2018-01-23 Richard Biener <rguenther@suse.de>
8480 PR tree-optimization/83963
8481 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8482 Properly terminate dominator walk when crossing the exit edge not
8483 when visiting its source block.
8485 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8488 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8489 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8491 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8493 PR tree-optimization/83957
8494 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8495 semicolon after for body surrounded by braces.
8497 PR tree-optimization/83081
8498 * profile-count.h (profile_probability::split): New method.
8499 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8500 Use profile_probability::split.
8501 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8502 when splitting a single conditional jump into 2.
8504 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8506 PR tree-optimization/69452
8507 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8510 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8512 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8513 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8514 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8516 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8518 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8519 * config/rl78/rl78.md (movdi): New define_expand.
8520 * config/rl78/rl78.c (rl78_split_movdi): New function.
8522 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8525 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8527 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8528 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8529 128-bit to produce an UNSPEC move to get the double word with the
8530 signbit and then a shift directly to do signbit.
8531 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8532 implementation with a new version that just does either a direct
8533 move or a regular move. Move memory interface to separate insns.
8534 Move insns so they are next to the expander.
8535 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8536 with signbit move. Split big and little endian case.
8537 (signbit<mode>2_dm_mem_le): Likewise.
8538 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8539 (signbit<mode>2_dm2): Likewise.
8541 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8543 * config/rl78/rl78.md (anddi3): New define_expand.
8545 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8547 * config/rl78/rl78.md (umindi3): New define_expand.
8549 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8551 * config/rl78/rl78.md (smindi3): New define_expand.
8553 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8555 * config/rl78/rl78.md (smaxdi3): New define_expand.
8557 2018-01-22 Carl Love <cel@us.ibm.com>
8559 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8560 LVX_V1TI): Add macro expansion.
8561 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8562 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8563 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8564 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8565 Change check to determine if the instruction is a byte reversing
8566 entry. Fix typo in comment.
8567 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8568 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8569 Add def_builtin calls for new builtins.
8570 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8571 Add define_insn expansion.
8573 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8575 * config/rl78/rl78.md (umaxdi3): New define_expand.
8577 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8579 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8580 for non-QImode registers.
8582 2018-01-22 Richard Biener <rguenther@suse.de>
8584 PR tree-optimization/83963
8585 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8586 including the loop exit block.
8587 (scop_detection::merge_sese): Likewise.
8588 (scop_detection::add_scop): Do it here instead.
8590 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8592 * doc/sourcebuild.texi (arm_softfloat): Document.
8594 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8597 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8598 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8599 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8601 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8602 David Edelsohn <dje.gcc@gmail.com>
8605 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8606 Change "crset eq" to "crset 2".
8607 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8608 (*call_indirect_aix<mode>_nospec): Likewise.
8609 (*call_value_indirect_aix<mode>_nospec): Likewise.
8610 (*call_indirect_elfv2<mode>_nospec): Likewise.
8611 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8612 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8613 change assembly output from . to $.
8614 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8615 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8616 (*tablejump<mode>_internal1_nospec): Likewise.
8618 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8621 * config/sh/sh_optimize_sett_clrt.cc:
8622 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8624 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8626 PR tree-optimization/83940
8627 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8628 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8629 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8630 use it to pass back the definition type.
8631 (vect_check_store_rhs): Likewise rhs_dt_out.
8632 (vect_build_gather_load_calls): Add a mask_dt argument and use
8633 it instead of a call to vect_is_simple_use.
8634 (vectorizable_store): Update calls to vect_check_load_store_mask
8635 and vect_check_store_rhs. Use the dt returned by the latter instead
8636 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8637 instead of calls to vect_is_simple_use. Pass the scalar rather
8638 than the vector operand to vect_is_simple_use when handling
8639 second and subsequent copies of an rhs value.
8640 (vectorizable_load): Update calls to vect_check_load_store_mask
8641 and vect_build_gather_load_calls. Use the cached mask_dt and
8642 gs_info.offset_dt instead of calls to vect_is_simple_use.
8644 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8647 * tree-emutls.c: Include gimplify.h.
8648 (lower_emutls_2): New function.
8649 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8650 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8651 it before further processing.
8654 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8655 UINTVAL (trueop1) instead of INTVAL (op1).
8657 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8661 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8662 INCOMING_FRAME_SP_OFFSET if not defined.
8663 (scan_trace): Add ENTRY argument. If true and
8664 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8665 emit a note to adjust the CFA offset.
8666 (create_cfi_notes): Adjust scan_trace callers.
8667 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8668 INCOMING_FRAME_SP_OFFSET in the CIE.
8669 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8670 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8672 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8673 * doc/tm.texi: Regenerated.
8675 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8677 PR rtl-optimization/83147
8678 * lra-constraints.c (remove_inheritance_pseudos): Use
8679 lra_substitute_pseudo_within_insn.
8681 2018-01-19 Tom de Vries <tom@codesourcery.com>
8682 Cesar Philippidis <cesar@codesourcery.com>
8685 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8687 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8690 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8691 spaces for function labels.
8693 2018-01-19 Martin Liska <mliska@suse.cz>
8695 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8696 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8697 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8698 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8699 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8700 (PRED_CONST_RETURN): Change from 69 to 65.
8701 (PRED_NULL_RETURN): Change from 91 to 71.
8702 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8703 (PRED_LOOP_GUARD): Change from 66 to 73.
8705 2018-01-19 Martin Liska <mliska@suse.cz>
8707 * predict.c (predict_insn_def): Add new assert.
8708 (struct branch_predictor): Change type to signed integer.
8709 (test_prediction_value_range): Amend test to cover
8711 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8712 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8713 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8714 (PRED_LOOP_IV_COMPARE): Likewise.
8715 * predict.h (PROB_UNINITIALIZED): Define new constant.
8717 2018-01-19 Martin Liska <mliska@suse.cz>
8719 * predict.c (dump_prediction): Add new format for
8720 analyze_brprob.py script which is enabled with -details
8722 * profile-count.h (precise_p): New function.
8724 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8726 PR tree-optimization/83922
8727 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8728 there are no statements that need masking.
8729 (vect_active_double_reduction_p): New function.
8730 (vect_analyze_loop_operations): Use it when handling phis that
8731 are not in the loop header.
8733 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8735 PR tree-optimization/83914
8736 * tree-vect-loop.c (vectorizable_induction): Don't convert
8737 init_expr or apply the peeling adjustment for inductions
8738 that are nested within the vectorized loop.
8740 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8742 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8745 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8749 * function.h (gimplify_parameters): Add gimple_seq * argument.
8750 * function.c: Include gimple.h and options.h.
8751 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8752 for the added local temporaries if needed.
8753 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8754 if there are any parameter cleanups, wrap whole body into a
8755 try/finally with the cleanups.
8757 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8760 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8761 Use GET_MODE_CLASS for scalar floating point.
8763 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8767 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8768 Fix call of call_cgraph_insertion_hooks.
8770 2018-01-18 Martin Sebor <msebor@redhat.com>
8772 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8774 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8777 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8780 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8783 * common.opt: (-ffile-prefix-map): New option.
8784 * opts.c (common_handle_option): Defer it.
8785 * opts-global.c (handle_common_deferred_options): Handle it.
8786 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8787 * file-prefix-map.h: New file.
8788 (remap_debug_filename, add_debug_prefix_map): ...here.
8789 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8790 * final.c (debug_prefix_map, add_debug_prefix_map
8791 remap_debug_filename): Move to...
8792 * file-prefix-map.c: New file.
8793 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8794 generalize, get rid of alloca(), use strrchr() instead of strchr().
8795 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8796 Implement in terms of add_prefix_map().
8797 (remap_macro_filename, remap_debug_filename): Implement in term of
8799 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8800 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8801 * dbxout.c: Include file-prefix-map.h.
8802 * varasm.c: Likewise.
8803 * vmsdbgout.c: Likewise.
8804 * xcoffout.c: Likewise.
8805 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8806 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8807 * doc/invoke.texi (-ffile-prefix-map): Document.
8808 (-fdebug-prefix-map): Update description.
8810 2018-01-18 Martin Liska <mliska@suse.cz>
8812 * config/i386/i386.c (indirect_thunk_name): Document that also
8814 (output_indirect_thunk): Document why both instructions
8815 (pause and lfence) are generated.
8817 2018-01-18 Richard Biener <rguenther@suse.de>
8819 PR tree-optimization/83887
8820 * graphite-scop-detection.c
8821 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8822 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8823 (scop_detection::merge_sese): Re-implement with a flood-fill
8824 algorithm that properly finds a SESE region if it exists.
8826 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8829 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8830 pointer_diff optimizations use view_convert instead of convert.
8832 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8834 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8835 Generate different code for -mno-speculate-indirect-jumps.
8836 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8837 (*call_indirect_aix<mode>): Disable for
8838 -mno-speculate-indirect-jumps.
8839 (*call_indirect_aix<mode>_nospec): New define_insn.
8840 (*call_value_indirect_aix<mode>): Disable for
8841 -mno-speculate-indirect-jumps.
8842 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8843 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8844 -mno-speculate-indirect-jumps.
8845 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8847 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8849 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8850 long double type, set the flags for noting the default long double
8851 type, even if we don't pass or return a long double type.
8853 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8856 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8859 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8861 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8862 support for merge[hl].
8863 (fold_mergehl_helper): New helper function.
8864 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8865 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8866 (altivec_vmrglw_direct): Add xxmrglw insn.
8868 2018-01-17 Andrew Waterman <andrew@sifive.com>
8870 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8871 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8873 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8876 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8877 call the lto_location_cache before reading the
8878 DECL_SOURCE_LOCATION of the types.
8880 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8881 Richard Sandiford <richard.sandiford@linaro.org>
8883 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8884 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8885 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8886 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8887 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8889 * config/aarch64/constraints.md (aarch64_movti_operand):
8891 * config/aarch64/predicates.md (Uti): Add new constraint.
8893 2018-01-17 Carl Love <cel@us.ibm.com>
8895 * config/rs6000/vsx.md (define_expand xl_len_r,
8896 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8897 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8899 (define_expand, define_insn): Move the shift left from the
8900 define_insn to the define_expand for lxvl and stxvl instructions.
8901 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8902 and XL_LEN_R definitions to PURE.
8904 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8906 * config/i386/i386.c (indirect_thunk_name): Declare regno
8907 as unsigned int. Compare regno with INVALID_REGNUM.
8908 (output_indirect_thunk): Ditto.
8909 (output_indirect_thunk_function): Ditto.
8910 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8911 in the call to output_indirect_thunk_function.
8913 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8916 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8917 rather than the size of inner_type to determine the stack slot size
8918 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8920 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8923 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8926 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8928 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8929 endian Linux systems to optionally enable multilibs for selecting
8930 the long double type if the user configured an explicit type.
8931 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8932 have no long double multilibs if not defined.
8933 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8934 warn if the user used -mabi={ieee,ibm}longdouble and we built
8935 multilibs for long double.
8936 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8937 appropriate multilib option.
8938 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8940 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8941 for building long double multilibs.
8942 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8944 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8946 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8949 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8951 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8954 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8957 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8960 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8962 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8963 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8965 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8967 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8968 different rtl trees depending on TARGET_64BIT.
8969 (rs6000_gen_lvx): Likewise.
8971 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8973 * config/visium/visium.md (nop): Tweak comment.
8974 (hazard_nop): Likewise.
8976 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8978 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8979 -mspeculate-indirect-jumps.
8980 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8981 for -mno-speculate-indirect-jumps.
8982 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8983 (*call_value_indirect_elfv2<mode>): Disable for
8984 -mno-speculate-indirect-jumps.
8985 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8986 (indirect_jump): Emit different RTL for
8987 -mno-speculate-indirect-jumps.
8988 (*indirect_jump<mode>): Disable for
8989 -mno-speculate-indirect-jumps.
8990 (*indirect_jump<mode>_nospec): New define_insn.
8991 (tablejump): Emit different RTL for
8992 -mno-speculate-indirect-jumps.
8993 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8994 (tablejumpsi_nospec): New define_expand.
8995 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8996 (tablejumpdi_nospec): New define_expand.
8997 (*tablejump<mode>_internal1): Disable for
8998 -mno-speculate-indirect-jumps.
8999 (*tablejump<mode>_internal1_nospec): New define_insn.
9000 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9003 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9005 * caller-save.c (insert_save): Drop unnecessary parameter. All
9008 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9009 Richard Biener <rguenth@suse.de>
9012 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9013 return early, inline manually is_gimple_sizepos. Make sure if we
9014 call gimplify_expr we don't end up with a gimple constant.
9015 * tree.c (variably_modified_type_p): Don't return true for
9016 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9017 * gimplify.h (is_gimple_sizepos): Remove.
9019 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9021 PR tree-optimization/83857
9022 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9023 vectorizable_live_operation for pure SLP statements.
9024 (vectorizable_live_operation): Handle PHIs.
9026 2018-01-16 Richard Biener <rguenther@suse.de>
9028 PR tree-optimization/83867
9029 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9030 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9032 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9035 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9036 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9037 If off is not INTEGER_CST, issue a may not be aligned warning
9038 rather than isn't aligned. Use isn%'t rather than isn't.
9039 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9041 <case MULT_EXPR>: Improve the case when bottom and one of the
9042 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9043 operand, in that case check if the other operand is multiple of
9044 bottom divided by the INTEGER_CST operand.
9046 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9049 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9050 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9051 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9052 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9053 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9054 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9055 (pa_function_arg_size): New function.
9057 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9059 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9060 in a separate statement.
9062 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9064 PR tree-optimization/83847
9065 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9066 group gathers and scatters.
9068 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9070 PR rtl-optimization/86620
9071 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9073 PR rtl-optimization/83213
9074 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9075 to last if both are JUMP_INSNs.
9077 PR tree-optimization/83843
9078 * gimple-ssa-store-merging.c
9079 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9080 store_immediate_info for bswap/nop orig_stores.
9082 2018-01-15 Andrew Waterman <andrew@sifive.com>
9084 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9086 <UDIV>: Increase cost if !TARGET_DIV.
9088 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9090 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9091 (define_attr "cr_logical_3op"): New.
9092 (cceq_ior_compare): Adjust.
9093 (cceq_ior_compare_complement): Adjust.
9094 (*cceq_rev_compare): Adjust.
9095 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9096 (is_cracked_insn): Adjust.
9097 (insn_must_be_first_in_group): Adjust.
9098 * config/rs6000/40x.md: Adjust.
9099 * config/rs6000/440.md: Adjust.
9100 * config/rs6000/476.md: Adjust.
9101 * config/rs6000/601.md: Adjust.
9102 * config/rs6000/603.md: Adjust.
9103 * config/rs6000/6xx.md: Adjust.
9104 * config/rs6000/7450.md: Adjust.
9105 * config/rs6000/7xx.md: Adjust.
9106 * config/rs6000/8540.md: Adjust.
9107 * config/rs6000/cell.md: Adjust.
9108 * config/rs6000/e300c2c3.md: Adjust.
9109 * config/rs6000/e500mc.md: Adjust.
9110 * config/rs6000/e500mc64.md: Adjust.
9111 * config/rs6000/e5500.md: Adjust.
9112 * config/rs6000/e6500.md: Adjust.
9113 * config/rs6000/mpc.md: Adjust.
9114 * config/rs6000/power4.md: Adjust.
9115 * config/rs6000/power5.md: Adjust.
9116 * config/rs6000/power6.md: Adjust.
9117 * config/rs6000/power7.md: Adjust.
9118 * config/rs6000/power8.md: Adjust.
9119 * config/rs6000/power9.md: Adjust.
9120 * config/rs6000/rs64.md: Adjust.
9121 * config/rs6000/titan.md: Adjust.
9123 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9125 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9126 ix86_indirect_branch_register logic.
9128 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9130 * config/i386/constraints.md (Bs): Update
9131 ix86_indirect_branch_register check. Don't check
9132 ix86_indirect_branch_register with GOT_memory_operand.
9134 * config/i386/predicates.md (GOT_memory_operand): Don't check
9135 ix86_indirect_branch_register here.
9136 (GOT32_symbol_operand): Likewise.
9138 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9140 * config/i386/predicates.md (constant_call_address_operand):
9141 Rewrite ix86_indirect_branch_register logic.
9142 (sibcall_insn_operand): Likewise.
9144 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9146 * config/i386/constraints.md (Bs): Replace
9147 ix86_indirect_branch_thunk_register with
9148 ix86_indirect_branch_register.
9150 * config/i386/i386.md (indirect_jump): Likewise.
9151 (tablejump): Likewise.
9152 (*sibcall_memory): Likewise.
9153 (*sibcall_value_memory): Likewise.
9154 Peepholes of indirect call and jump via memory: Likewise.
9155 * config/i386/i386.opt: Likewise.
9156 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9157 (GOT_memory_operand): Likewise.
9158 (call_insn_operand): Likewise.
9159 (sibcall_insn_operand): Likewise.
9160 (GOT32_symbol_operand): Likewise.
9162 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9165 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9166 type rather than type addr's type points to.
9167 (expand_omp_atomic_mutex): Likewise.
9168 (expand_omp_atomic): Likewise.
9170 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9173 * config/i386/i386.c (output_indirect_thunk_function): Use
9174 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9175 for __x86_return_thunk.
9177 2018-01-15 Richard Biener <rguenther@suse.de>
9180 * expmed.c (extract_bit_field_1): Fix typo.
9182 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9185 * config/arm/iterators.md (VF): New mode iterator.
9186 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9187 Remove integer-related logic from pattern.
9188 (neon_vabd<mode>_3): Likewise.
9190 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9193 * common.opt (fstrict-overflow): No longer an alias.
9194 (fwrapv-pointer): New option.
9195 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9196 also for pointer types based on flag_wrapv_pointer.
9197 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9198 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9199 opts->x_flag_wrapv got set.
9200 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9201 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9202 POINTER_TYPE_OVERFLOW_UNDEFINED.
9203 * match.pd: Likewise in address comparison pattern.
9204 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9206 2018-01-15 Richard Biener <rguenther@suse.de>
9209 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9210 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9211 Reset type names to their identifier if their TYPE_DECL doesn't
9212 have linkage (and thus is used for ODR and devirt).
9213 (save_debug_info_for_decl): Remove.
9214 (save_debug_info_for_type): Likewise.
9215 (add_tree_to_fld_list): Adjust.
9216 * tree-pretty-print.c (dump_generic_node): Make dumping of
9217 type names more robust.
9219 2018-01-15 Richard Biener <rguenther@suse.de>
9221 * BASE-VER: Bump to 8.0.1.
9223 2018-01-14 Martin Sebor <msebor@redhat.com>
9226 * builtins.c (check_access): Avoid warning when the no-warning bit
9229 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9231 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9232 * ira-color (allocno_hard_regs_compare): Likewise.
9234 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9237 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9238 Use .pushsection/.popsection.
9240 2018-01-14 Martin Sebor <msebor@redhat.com>
9243 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9245 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9247 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9248 entry from extra_headers.
9249 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9250 extra_headers, make the list bitwise identical to the i?86-*-* one.
9252 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9254 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9255 -mcmodel=large with -mindirect-branch=thunk,
9256 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9257 -mfunction-return=thunk-extern.
9258 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9259 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9260 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9262 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9264 * config/i386/i386.c (print_reg): Print the name of the full
9265 integer register without '%'.
9266 (ix86_print_operand): Handle 'V'.
9267 * doc/extend.texi: Document 'V' modifier.
9269 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9271 * config/i386/constraints.md (Bs): Disallow memory operand for
9272 -mindirect-branch-register.
9274 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9275 (GOT_memory_operand): Likewise.
9276 (call_insn_operand): Likewise.
9277 (sibcall_insn_operand): Likewise.
9278 (GOT32_symbol_operand): Likewise.
9279 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9280 for -mindirect-branch-register.
9281 (tablejump): Likewise.
9282 (*sibcall_memory): Likewise.
9283 (*sibcall_value_memory): Likewise.
9284 Disallow peepholes of indirect call and jump via memory for
9285 -mindirect-branch-register.
9286 (*call_pop): Replace m with Bw.
9287 (*call_value_pop): Likewise.
9288 (*sibcall_pop_memory): Replace m with Bs.
9289 * config/i386/i386.opt (mindirect-branch-register): New option.
9290 * doc/invoke.texi: Document -mindirect-branch-register option.
9292 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9294 * config/i386/i386-protos.h (ix86_output_function_return): New.
9295 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9296 set function_return_type.
9297 (indirect_thunk_name): Add ret_p to indicate thunk for function
9299 (output_indirect_thunk_function): Pass false to
9300 indirect_thunk_name.
9301 (ix86_output_indirect_branch_via_reg): Likewise.
9302 (ix86_output_indirect_branch_via_push): Likewise.
9303 (output_indirect_thunk_function): Create alias for function
9304 return thunk if regno < 0.
9305 (ix86_output_function_return): New function.
9306 (ix86_handle_fndecl_attribute): Handle function_return.
9307 (ix86_attribute_table): Add function_return.
9308 * config/i386/i386.h (machine_function): Add
9309 function_return_type.
9310 * config/i386/i386.md (simple_return_internal): Use
9311 ix86_output_function_return.
9312 (simple_return_internal_long): Likewise.
9313 * config/i386/i386.opt (mfunction-return=): New option.
9314 (indirect_branch): Mention -mfunction-return=.
9315 * doc/extend.texi: Document function_return function attribute.
9316 * doc/invoke.texi: Document -mfunction-return= option.
9318 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9320 * config/i386/i386-opts.h (indirect_branch): New.
9321 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9322 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9323 with local indirect jump when converting indirect call and jump.
9324 (ix86_set_indirect_branch_type): New.
9325 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9326 (indirectlabelno): New.
9327 (indirect_thunk_needed): Likewise.
9328 (indirect_thunk_bnd_needed): Likewise.
9329 (indirect_thunks_used): Likewise.
9330 (indirect_thunks_bnd_used): Likewise.
9331 (INDIRECT_LABEL): Likewise.
9332 (indirect_thunk_name): Likewise.
9333 (output_indirect_thunk): Likewise.
9334 (output_indirect_thunk_function): Likewise.
9335 (ix86_output_indirect_branch_via_reg): Likewise.
9336 (ix86_output_indirect_branch_via_push): Likewise.
9337 (ix86_output_indirect_branch): Likewise.
9338 (ix86_output_indirect_jmp): Likewise.
9339 (ix86_code_end): Call output_indirect_thunk_function if needed.
9340 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9342 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9343 (ix86_attribute_table): Add indirect_branch.
9344 * config/i386/i386.h (machine_function): Add indirect_branch_type
9345 and has_local_indirect_jump.
9346 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9348 (tablejump): Likewise.
9349 (*indirect_jump): Use ix86_output_indirect_jmp.
9350 (*tablejump_1): Likewise.
9351 (simple_return_indirect_internal): Likewise.
9352 * config/i386/i386.opt (mindirect-branch=): New option.
9353 (indirect_branch): New.
9356 (thunk-inline): Likewise.
9357 (thunk-extern): Likewise.
9358 * doc/extend.texi: Document indirect_branch function attribute.
9359 * doc/invoke.texi: Document -mindirect-branch= option.
9361 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9364 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9366 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9368 * ipa-inline.c (want_inline_small_function_p): Return false if
9369 inlining has already failed with CIF_FINAL_ERROR.
9370 (update_caller_keys): Call want_inline_small_function_p before
9372 (update_callee_keys): Likewise.
9374 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9376 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9378 (rs6000_quadword_masked_address_p): Likewise.
9379 (quad_aligned_load_p): Likewise.
9380 (quad_aligned_store_p): Likewise.
9381 (const_load_sequence_p): Add comment to describe the outer-most loop.
9382 (mimic_memory_attributes_and_flags): New function.
9383 (rs6000_gen_stvx): Likewise.
9384 (replace_swapped_aligned_store): Likewise.
9385 (rs6000_gen_lvx): Likewise.
9386 (replace_swapped_aligned_load): Likewise.
9387 (replace_swapped_load_constant): Capitalize argument name in
9388 comment describing this function.
9389 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9390 and stores that access quad-word aligned addresses and replace
9391 with stvx or lvx instructions when appropriate.
9392 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9393 New function prototype.
9394 (rs6000_quadword_masked_address_p): Likewise.
9395 (rs6000_gen_lvx): Likewise.
9396 (rs6000_gen_stvx): Likewise.
9397 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9398 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9399 when memory address is aligned.
9400 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9401 this split to select lvx instruction when memory address is aligned.
9402 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9403 instruction when memory address is aligned.
9404 (*vsx_le_perm_load_v16qi): Likewise.
9405 (four unnamed splitters): Modify to select the stvx instruction
9406 when memory is aligned.
9408 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9410 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9411 which appears in the queue multiple times.
9413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9414 Alan Hayward <alan.hayward@arm.com>
9415 David Sherwood <david.sherwood@arm.com>
9417 * tree-vectorizer.h (vec_lower_bound): New structure.
9418 (_loop_vec_info): Add check_nonzero and lower_bounds.
9419 (LOOP_VINFO_CHECK_NONZERO): New macro.
9420 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9421 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9422 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9423 fields. Make seg_len the distance travelled, not including the
9425 (dr_direction_indicator): Declare.
9426 (dr_zero_step_indicator): Likewise.
9427 (dr_known_forward_stride_p): Likewise.
9428 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9430 (runtime_alias_check_p): Allow runtime alias checks with
9432 (operator ==): Compare access_size and align.
9433 (prune_runtime_alias_test_list): Rework for new distinction between
9434 the access_size and seg_len.
9435 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9437 (get_segment_min_max): New function.
9438 (create_intersect_range_checks): Use it.
9439 (dr_step_indicator): New function.
9440 (dr_direction_indicator): Likewise.
9441 (dr_zero_step_indicator): Likewise.
9442 (dr_known_forward_stride_p): Likewise.
9443 * tree-loop-distribution.c (data_ref_segment_size): Return
9444 DR_STEP * (niters - 1).
9445 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9447 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9448 (vect_preserves_scalar_order_p): New function, split out from...
9449 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9450 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9451 (vect_vfa_access_size): New function.
9452 (vect_vfa_align): Likewise.
9453 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9454 (dump_lower_bound): New function.
9455 (vect_check_lower_bound): Likewise.
9456 (vect_small_gap_p): Likewise.
9457 (vectorizable_with_step_bound_p): Likewise.
9458 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9459 depencies if the vectorization factor is 1. Convert the checks
9460 for nonzero steps into checks on the bounds of DR_STEP. Try using
9461 a bunds check for variable steps if the minimum required step is
9462 relatively small. Update calls to the dr_with_seg_len
9463 constructor and to vect_compile_time_alias.
9464 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9466 (vect_loop_versioning): Call it.
9467 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9469 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9472 Alan Hayward <alan.hayward@arm.com>
9473 David Sherwood <david.sherwood@arm.com>
9475 * doc/sourcebuild.texi (vect_scatter_store): Document.
9476 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9478 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9480 * genopinit.c (main): Add supports_vec_scatter_store and
9481 supports_vec_scatter_store_cached to target_optabs.
9482 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9483 IFN_MASK_SCATTER_STORE.
9484 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9486 * internal-fn.h (internal_store_fn_p): Declare.
9487 (internal_fn_stored_value_index): Likewise.
9488 * internal-fn.c (scatter_store_direct): New macro.
9489 (expand_scatter_store_optab_fn): New function.
9490 (direct_scatter_store_optab_supported_p): New macro.
9491 (internal_store_fn_p): New function.
9492 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9493 IFN_MASK_SCATTER_STORE.
9494 (internal_fn_mask_index): Likewise.
9495 (internal_fn_stored_value_index): New function.
9496 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9498 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9499 * optabs-query.c (supports_vec_scatter_store_p): New function.
9500 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9501 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9502 true for scatter stores.
9503 (vect_gather_scatter_fn_p): Handle scatter stores too.
9504 (vect_check_gather_scatter): Consider using scatter stores if
9505 supports_vec_scatter_store_p.
9506 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9508 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9509 internal_fn_stored_value_index.
9510 (check_load_store_masking): Handle scatter stores too.
9511 (vect_get_store_rhs): Make public.
9512 (vectorizable_call): Use internal_store_fn_p.
9513 (vectorizable_store): Handle scatter store internal functions.
9514 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9515 when deciding whether the end of the group has been reached.
9516 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9517 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9518 (mask_scatter_store<mode>): New insns.
9520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9521 Alan Hayward <alan.hayward@arm.com>
9522 David Sherwood <david.sherwood@arm.com>
9524 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9525 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9526 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9528 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9529 Use vect_truncate_gather_scatter_offset if we can't treat the
9530 operation as a normal gather load or scatter store.
9531 (get_group_load_store_type): Take the gather_scatter_info
9532 as argument. Try using a gather load or scatter store for
9533 single-element groups.
9534 (get_load_store_type): Update calls to get_group_load_store_type
9535 and vect_use_strided_gather_scatters_p.
9537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9538 Alan Hayward <alan.hayward@arm.com>
9539 David Sherwood <david.sherwood@arm.com>
9541 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9542 optional tree argument.
9543 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9545 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9546 but continue to use the current value as a fallback.
9547 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9548 to compare the updates.
9549 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9550 (get_load_store_type): Use it when handling a strided access.
9551 (vect_get_strided_load_store_ops): New function.
9552 (vect_get_data_ptr_increment): Likewise.
9553 (vectorizable_load): Handle strided gather loads. Always pass
9554 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9557 Alan Hayward <alan.hayward@arm.com>
9558 David Sherwood <david.sherwood@arm.com>
9560 * doc/md.texi (gather_load@var{m}): Document.
9561 (mask_gather_load@var{m}): Likewise.
9562 * genopinit.c (main): Add supports_vec_gather_load and
9563 supports_vec_gather_load_cached to target_optabs.
9564 * optabs-tree.c (init_tree_optimization_optabs): Use
9565 ggc_cleared_alloc to allocate target_optabs.
9566 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9567 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9569 * internal-fn.h (internal_load_fn_p): Declare.
9570 (internal_gather_scatter_fn_p): Likewise.
9571 (internal_fn_mask_index): Likewise.
9572 (internal_gather_scatter_fn_supported_p): Likewise.
9573 * internal-fn.c (gather_load_direct): New macro.
9574 (expand_gather_load_optab_fn): New function.
9575 (direct_gather_load_optab_supported_p): New macro.
9576 (direct_internal_fn_optab): New function.
9577 (internal_load_fn_p): Likewise.
9578 (internal_gather_scatter_fn_p): Likewise.
9579 (internal_fn_mask_index): Likewise.
9580 (internal_gather_scatter_fn_supported_p): Likewise.
9581 * optabs-query.c (supports_at_least_one_mode_p): New function.
9582 (supports_vec_gather_load_p): Likewise.
9583 * optabs-query.h (supports_vec_gather_load_p): Declare.
9584 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9585 and memory_type field.
9586 (NUM_PATTERNS): Bump to 15.
9587 * tree-vect-data-refs.c: Include internal-fn.h.
9588 (vect_gather_scatter_fn_p): New function.
9589 (vect_describe_gather_scatter_call): Likewise.
9590 (vect_check_gather_scatter): Try using internal functions for
9591 gather loads. Recognize existing calls to a gather load function.
9592 (vect_analyze_data_refs): Consider using gather loads if
9593 supports_vec_gather_load_p.
9594 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9595 (vect_get_gather_scatter_offset_type): Likewise.
9596 (vect_convert_mask_for_vectype): Likewise.
9597 (vect_add_conversion_to_patterm): Likewise.
9598 (vect_try_gather_scatter_pattern): Likewise.
9599 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9600 (vect_vect_recog_func_ptrs): Add it.
9601 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9602 internal_fn_mask_index and internal_gather_scatter_fn_p.
9603 (check_load_store_masking): Take the gather_scatter_info as an
9604 argument and handle gather loads.
9605 (vect_get_gather_scatter_ops): New function.
9606 (vectorizable_call): Check internal_load_fn_p.
9607 (vectorizable_load): Likewise. Handle gather load internal
9609 (vectorizable_store): Update call to check_load_store_masking.
9610 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9611 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9612 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9613 (aarch64_gather_scale_operand_d): New predicates.
9614 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9615 (mask_gather_load<mode>): New insns.
9617 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9618 Alan Hayward <alan.hayward@arm.com>
9619 David Sherwood <david.sherwood@arm.com>
9621 * optabs.def (fold_left_plus_optab): New optab.
9622 * doc/md.texi (fold_left_plus_@var{m}): Document.
9623 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9624 * internal-fn.c (fold_left_direct): Define.
9625 (expand_fold_left_optab_fn): Likewise.
9626 (direct_fold_left_optab_supported_p): Likewise.
9627 * fold-const-call.c (fold_const_fold_left): New function.
9628 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9629 * tree-parloops.c (valid_reduction_p): New function.
9630 (gather_scalar_reductions): Use it.
9631 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9632 (vect_finish_replace_stmt): Declare.
9633 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9634 (needs_fold_left_reduction_p): New function, split out from...
9635 (vect_is_simple_reduction): ...here. Accept reductions that
9636 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9637 (vect_force_simple_reduction): Also store the reduction type in
9638 the assignment's STMT_VINFO_REDUC_TYPE.
9639 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9640 (merge_with_identity): New function.
9641 (vect_expand_fold_left): Likewise.
9642 (vectorize_fold_left_reduction): Likewise.
9643 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9644 scalar phi in place for it. Check for target support and reject
9645 cases that would reassociate the operation. Defer the transform
9646 phase to vectorize_fold_left_reduction.
9647 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9648 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9649 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9653 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9654 call to ifc_temp_var.
9656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9657 Alan Hayward <alan.hayward@arm.com>
9658 David Sherwood <david.sherwood@arm.com>
9660 * target.def (legitimize_address_displacement): Take the original
9661 offset as a poly_int.
9662 * targhooks.h (default_legitimize_address_displacement): Update
9664 * targhooks.c (default_legitimize_address_displacement): Likewise.
9665 * doc/tm.texi: Regenerate.
9666 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9667 as an argument, moving assert of ad->disp == ad->disp_term to...
9668 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9669 Try calling targetm.legitimize_address_displacement before expanding
9670 the address rather than afterwards, and adjust for the new interface.
9671 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9672 Match the new hook interface. Handle SVE addresses.
9673 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9678 * Makefile.in (OBJS): Add early-remat.o.
9679 * target.def (select_early_remat_modes): New hook.
9680 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9681 * doc/tm.texi: Regenerate.
9682 * targhooks.h (default_select_early_remat_modes): Declare.
9683 * targhooks.c (default_select_early_remat_modes): New function.
9684 * timevar.def (TV_EARLY_REMAT): New timevar.
9685 * passes.def (pass_early_remat): New pass.
9686 * tree-pass.h (make_pass_early_remat): Declare.
9687 * early-remat.c: New file.
9688 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9690 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9693 Alan Hayward <alan.hayward@arm.com>
9694 David Sherwood <david.sherwood@arm.com>
9696 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9697 vfm1 with a bound_epilog parameter.
9698 (vect_do_peeling): Update calls accordingly, and move the prologue
9699 call earlier in the function. Treat the base bound_epilog as 0 for
9700 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9701 this base when peeling for gaps.
9702 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9703 with fully-masked loops.
9704 (vect_estimate_min_profitable_iters): Handle the single peeled
9705 iteration in that case.
9707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9708 Alan Hayward <alan.hayward@arm.com>
9709 David Sherwood <david.sherwood@arm.com>
9711 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9712 single-element interleaving even if the size is not a power of 2.
9713 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9714 accesses for single-element interleaving if the group size is
9717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9718 Alan Hayward <alan.hayward@arm.com>
9719 David Sherwood <david.sherwood@arm.com>
9721 * doc/md.texi (fold_extract_last_@var{m}): Document.
9722 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9723 * optabs.def (fold_extract_last_optab): New optab.
9724 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9725 * internal-fn.c (fold_extract_direct): New macro.
9726 (expand_fold_extract_optab_fn): Likewise.
9727 (direct_fold_extract_optab_supported_p): Likewise.
9728 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9729 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9730 EXTRACT_LAST_REDUCTION.
9731 (get_initial_def_for_reduction): Do not create an initial vector
9732 for EXTRACT_LAST_REDUCTION reductions.
9733 (vectorizable_reduction): Leave the scalar phi in place for
9734 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9735 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9736 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9737 transform phase to vectorizable_condition.
9738 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9740 (vect_finish_stmt_generation): ...here.
9741 (vect_finish_replace_stmt): New function.
9742 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9743 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9745 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9748 Alan Hayward <alan.hayward@arm.com>
9749 David Sherwood <david.sherwood@arm.com>
9751 * doc/md.texi (extract_last_@var{m}): Document.
9752 * optabs.def (extract_last_optab): New optab.
9753 * internal-fn.def (EXTRACT_LAST): New internal function.
9754 * internal-fn.c (cond_unary_direct): New macro.
9755 (expand_cond_unary_optab_fn): Likewise.
9756 (direct_cond_unary_optab_supported_p): Likewise.
9757 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9758 loops using EXTRACT_LAST.
9759 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9760 (extract_last_<mode>): ...this optab.
9761 (vec_extract<mode><Vel>): Update accordingly.
9763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9764 Alan Hayward <alan.hayward@arm.com>
9765 David Sherwood <david.sherwood@arm.com>
9767 * target.def (empty_mask_is_expensive): New hook.
9768 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9769 * doc/tm.texi: Regenerate.
9770 * targhooks.h (default_empty_mask_is_expensive): Declare.
9771 * targhooks.c (default_empty_mask_is_expensive): New function.
9772 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9773 if the target says that empty masks are expensive.
9774 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9776 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9778 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9779 Alan Hayward <alan.hayward@arm.com>
9780 David Sherwood <david.sherwood@arm.com>
9782 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9783 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9784 (vect_use_loop_mask_for_alignment_p): New function.
9785 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9786 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9787 niters_skip argument. Make sure that the first niters_skip elements
9788 of the first iteration are inactive.
9789 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9790 Update call to vect_set_loop_masks_directly.
9791 (get_misalign_in_elems): New function, split out from...
9792 (vect_gen_prolog_loop_niters): ...here.
9793 (vect_update_init_of_dr): Take a code argument that specifies whether
9794 the adjustment should be added or subtracted.
9795 (vect_update_init_of_drs): Likewise.
9796 (vect_prepare_for_masked_peels): New function.
9797 (vect_do_peeling): Skip prologue peeling if we're using a mask
9798 instead. Update call to vect_update_inits_of_drs.
9799 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9801 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9802 alignment. Do not include the number of peeled iterations in
9803 the minimum threshold in that case.
9804 (vectorizable_induction): Adjust the start value down by
9805 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9806 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9807 Take the number of skipped iterations into account when calculating
9809 * tree-vect-stmts.c (vect_gen_while_not): New function.
9811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9812 Alan Hayward <alan.hayward@arm.com>
9813 David Sherwood <david.sherwood@arm.com>
9815 * doc/sourcebuild.texi (vect_fully_masked): Document.
9816 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9818 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9820 (vect_analyze_loop_2): ...here. Don't check the vectorization
9821 factor against the number of loop iterations if the loop is
9824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9825 Alan Hayward <alan.hayward@arm.com>
9826 David Sherwood <david.sherwood@arm.com>
9828 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9829 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9830 (dump_groups): Update accordingly.
9831 (iv_use::mem_type): New member variable.
9832 (address_p): New function.
9833 (record_use): Add a mem_type argument and initialize the new
9835 (record_group_use): Add a mem_type argument. Use address_p.
9836 Remove obsolete null checks of base_object. Update call to record_use.
9837 (find_interesting_uses_op): Update call to record_group_use.
9838 (find_interesting_uses_cond): Likewise.
9839 (find_interesting_uses_address): Likewise.
9840 (get_mem_type_for_internal_fn): New function.
9841 (find_address_like_use): Likewise.
9842 (find_interesting_uses_stmt): Try find_address_like_use before
9843 calling find_interesting_uses_op.
9844 (addr_offset_valid_p): Use the iv mem_type field as the type
9845 of the addressed memory.
9846 (add_autoinc_candidates): Likewise.
9847 (get_address_cost): Likewise.
9848 (split_small_address_groups_p): Use address_p.
9849 (split_address_groups): Likewise.
9850 (add_iv_candidate_for_use): Likewise.
9851 (autoinc_possible_for_pair): Likewise.
9852 (rewrite_groups): Likewise.
9853 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9854 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9855 (get_alias_ptr_type_for_ptr_address): New function.
9856 (rewrite_use_address): Rewrite address uses in calls that were
9857 identified by find_address_like_use.
9859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9860 Alan Hayward <alan.hayward@arm.com>
9861 David Sherwood <david.sherwood@arm.com>
9863 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9865 * gimple-expr.h (is_gimple_addressable: Likewise.
9866 * gimple-expr.c (is_gimple_address): Likewise.
9867 * internal-fn.c (expand_call_mem_ref): New function.
9868 (expand_mask_load_optab_fn): Use it.
9869 (expand_mask_store_optab_fn): Likewise.
9871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9872 Alan Hayward <alan.hayward@arm.com>
9873 David Sherwood <david.sherwood@arm.com>
9875 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9876 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9877 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9878 (cond_umax@var{mode}): Document.
9879 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9880 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9881 (cond_umin_optab, cond_umax_optab): New optabs.
9882 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9883 (COND_IOR, COND_XOR): New internal functions.
9884 * internal-fn.h (get_conditional_internal_fn): Declare.
9885 * internal-fn.c (cond_binary_direct): New macro.
9886 (expand_cond_binary_optab_fn): Likewise.
9887 (direct_cond_binary_optab_supported_p): Likewise.
9888 (get_conditional_internal_fn): New function.
9889 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9890 Cope with reduction statements that are vectorized as calls rather
9892 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9893 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9894 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9895 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9896 (UNSPEC_COND_EOR): New unspecs.
9897 (optab): Add mappings for them.
9898 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9899 (sve_int_op, sve_fp_op): New int attributes.
9901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9902 Alan Hayward <alan.hayward@arm.com>
9903 David Sherwood <david.sherwood@arm.com>
9905 * optabs.def (while_ult_optab): New optab.
9906 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9907 * internal-fn.def (WHILE_ULT): New internal function.
9908 * internal-fn.h (direct_internal_fn_supported_p): New override
9909 that takes two types as argument.
9910 * internal-fn.c (while_direct): New macro.
9911 (expand_while_optab_fn): New function.
9912 (convert_optab_supported_p): Likewise.
9913 (direct_while_optab_supported_p): New macro.
9914 * wide-int.h (wi::udiv_ceil): New function.
9915 * tree-vectorizer.h (rgroup_masks): New structure.
9916 (vec_loop_masks): New typedef.
9917 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9919 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9920 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9921 (vect_max_vf): New function.
9922 (slpeel_make_loop_iterate_ntimes): Delete.
9923 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9924 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9925 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9926 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9927 internal-fn.h, stor-layout.h and optabs-query.h.
9928 (vect_set_loop_mask): New function.
9929 (add_preheader_seq): Likewise.
9930 (add_header_seq): Likewise.
9931 (interleave_supported_p): Likewise.
9932 (vect_maybe_permute_loop_masks): Likewise.
9933 (vect_set_loop_masks_directly): Likewise.
9934 (vect_set_loop_condition_masked): Likewise.
9935 (vect_set_loop_condition_unmasked): New function, split out from
9936 slpeel_make_loop_iterate_ntimes.
9937 (slpeel_make_loop_iterate_ntimes): Rename to..
9938 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9939 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9940 (vect_do_peeling): Update call accordingly.
9941 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9943 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9944 mask_compare_type, can_fully_mask_p and fully_masked_p.
9945 (release_vec_loop_masks): New function.
9946 (_loop_vec_info): Use it to free the loop masks.
9947 (can_produce_all_loop_masks_p): New function.
9948 (vect_get_max_nscalars_per_iter): Likewise.
9949 (vect_verify_full_masking): Likewise.
9950 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9951 retries, and free the mask rgroups before retrying. Check loop-wide
9952 reasons for disallowing fully-masked loops. Make the final decision
9953 about whether use a fully-masked loop or not.
9954 (vect_estimate_min_profitable_iters): Do not assume that peeling
9955 for the number of iterations will be needed for fully-masked loops.
9956 (vectorizable_reduction): Disable fully-masked loops.
9957 (vectorizable_live_operation): Likewise.
9958 (vect_halve_mask_nunits): New function.
9959 (vect_double_mask_nunits): Likewise.
9960 (vect_record_loop_mask): Likewise.
9961 (vect_get_loop_mask): Likewise.
9962 (vect_transform_loop): Handle the case in which the final loop
9963 iteration might handle a partial vector. Call vect_set_loop_condition
9964 instead of slpeel_make_loop_iterate_ntimes.
9965 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9966 (check_load_store_masking): New function.
9967 (prepare_load_store_mask): Likewise.
9968 (vectorizable_store): Handle fully-masked loops.
9969 (vectorizable_load): Likewise.
9970 (supportable_widening_operation): Use vect_halve_mask_nunits for
9972 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9973 (vect_gen_while): New function.
9974 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9975 (aarch64_uqdec<mode>): New insn.
9977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9978 Alan Hayward <alan.hayward@arm.com>
9979 David Sherwood <david.sherwood@arm.com>
9981 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9982 (reduc_xor_scal_optab): New optabs.
9983 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9984 (reduc_xor_scal_@var{m}): Document.
9985 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9986 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9988 * fold-const-call.c (fold_const_call): Handle them.
9989 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9990 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9991 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9992 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9993 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9994 (UNSPEC_XORV): New unspecs.
9995 (optab): Add entries for them.
9996 (BITWISEV): New int iterator.
9997 (bit_reduc_op): New int attributes.
9999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10004 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10005 * optabs.def (vec_shl_insert_optab): New optab.
10006 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10007 (duplicate_and_interleave): Likewise.
10008 * tree-vect-loop.c: Include internal-fn.h.
10009 (neutral_op_for_slp_reduction): New function, split out from
10010 get_initial_defs_for_reduction.
10011 (get_initial_def_for_reduction): Handle option 2 for variable-length
10012 vectors by loading the neutral value into a vector and then shifting
10013 the initial value into element 0.
10014 (get_initial_defs_for_reduction): Replace the code argument with
10015 the neutral value calculated by neutral_op_for_slp_reduction.
10016 Use gimple_build_vector for constant-length vectors.
10017 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10018 but the first group_size elements have a neutral value.
10019 Use duplicate_and_interleave otherwise.
10020 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10021 Update call to get_initial_defs_for_reduction. Handle SLP
10022 reductions for variable-length vectors by creating one vector
10023 result for each scalar result, with the elements associated
10024 with other scalar results stubbed out with the neutral value.
10025 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10026 Require IFN_VEC_SHL_INSERT for double reductions on
10027 variable-length vectors, or SLP reductions that have
10028 a neutral value. Require can_duplicate_and_interleave_p
10029 support for variable-length unchained SLP reductions if there
10030 is no neutral value, such as for MIN/MAX reductions. Also require
10031 the number of vector elements to be a multiple of the number of
10032 SLP statements when doing variable-length unchained SLP reductions.
10033 Update call to vect_create_epilog_for_reduction.
10034 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10035 and remove initial values.
10036 (duplicate_and_interleave): Make public.
10037 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10038 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10041 Alan Hayward <alan.hayward@arm.com>
10042 David Sherwood <david.sherwood@arm.com>
10044 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10045 (can_duplicate_and_interleave_p): New function.
10046 (vect_get_and_check_slp_defs): Take the vector of statements
10047 rather than just the current one. Remove excess parentheses.
10048 Restriction rejectinon of vect_constant_def and vect_external_def
10049 for variable-length vectors to boolean types, or types for which
10050 can_duplicate_and_interleave_p is false.
10051 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10052 (duplicate_and_interleave): New function.
10053 (vect_get_constant_vectors): Use gimple_build_vector for
10054 constant-length vectors and suitable variable-length constant
10055 vectors. Use duplicate_and_interleave for other variable-length
10056 vectors. Don't defer the update when inserting new statements.
10058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10059 Alan Hayward <alan.hayward@arm.com>
10060 David Sherwood <david.sherwood@arm.com>
10062 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10063 min_profitable_iters doesn't go negative.
10065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10066 Alan Hayward <alan.hayward@arm.com>
10067 David Sherwood <david.sherwood@arm.com>
10069 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10070 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10071 * optabs.def (vec_mask_load_lanes_optab): New optab.
10072 (vec_mask_store_lanes_optab): Likewise.
10073 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10074 (MASK_STORE_LANES): Likewise.
10075 * internal-fn.c (mask_load_lanes_direct): New macro.
10076 (mask_store_lanes_direct): Likewise.
10077 (expand_mask_load_optab_fn): Handle masked operations.
10078 (expand_mask_load_lanes_optab_fn): New macro.
10079 (expand_mask_store_optab_fn): Handle masked operations.
10080 (expand_mask_store_lanes_optab_fn): New macro.
10081 (direct_mask_load_lanes_optab_supported_p): Likewise.
10082 (direct_mask_store_lanes_optab_supported_p): Likewise.
10083 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10085 (vect_load_lanes_supported): Likewise.
10086 * tree-vect-data-refs.c (strip_conversion): New function.
10087 (can_group_stmts_p): Likewise.
10088 (vect_analyze_data_ref_accesses): Use it instead of checking
10089 for a pair of assignments.
10090 (vect_store_lanes_supported): Take a masked_p parameter.
10091 (vect_load_lanes_supported): Likewise.
10092 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10093 vect_store_lanes_supported and vect_load_lanes_supported.
10094 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10095 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10096 parameter. Don't allow gaps for masked accesses.
10097 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10098 and vect_load_lanes_supported.
10099 (get_load_store_type): Take a masked_p parameter and update
10100 call to get_group_load_store_type.
10101 (vectorizable_store): Update call to get_load_store_type.
10102 Handle IFN_MASK_STORE_LANES.
10103 (vectorizable_load): Update call to get_load_store_type.
10104 Handle IFN_MASK_LOAD_LANES.
10106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10107 Alan Hayward <alan.hayward@arm.com>
10108 David Sherwood <david.sherwood@arm.com>
10110 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10112 * config/aarch64/aarch64-protos.h
10113 (aarch64_sve_struct_memory_operand_p): Declare.
10114 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10115 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10116 (VPRED, vpred): Handle SVE structure modes.
10117 * config/aarch64/constraints.md (Utx): New constraint.
10118 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10119 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10120 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10121 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10122 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10123 structure modes. Split into pieces after RA.
10124 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10125 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10127 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10128 SVE structure modes.
10129 (aarch64_classify_address): Likewise.
10130 (sizetochar): Move earlier in file.
10131 (aarch64_print_operand): Handle SVE register lists.
10132 (aarch64_array_mode): New function.
10133 (aarch64_sve_struct_memory_operand_p): Likewise.
10134 (TARGET_ARRAY_MODE): Redefine.
10136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10137 Alan Hayward <alan.hayward@arm.com>
10138 David Sherwood <david.sherwood@arm.com>
10140 * target.def (array_mode): New target hook.
10141 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10142 * doc/tm.texi: Regenerate.
10143 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10144 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10145 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10146 targetm.array_mode.
10147 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10151 Alan Hayward <alan.hayward@arm.com>
10152 David Sherwood <david.sherwood@arm.com>
10154 * fold-const.c (fold_binary_loc): Check the argument types
10155 rather than the result type when testing for a vector operation.
10157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10159 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10160 * doc/tm.texi: Regenerate.
10162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10163 Alan Hayward <alan.hayward@arm.com>
10164 David Sherwood <david.sherwood@arm.com>
10166 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10167 (sve): Document new AArch64 extension.
10168 * doc/md.texi (w): Extend the description of the AArch64
10169 constraint to include SVE vectors.
10170 (Upl, Upa): Document new AArch64 predicate constraints.
10171 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10173 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10174 (msve-vector-bits=): New option.
10175 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10176 SVE when these are disabled.
10177 (sve): New extension.
10178 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10179 modes. Adjust their number of units based on aarch64_sve_vg.
10180 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10181 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10182 aarch64_addr_query_type.
10183 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10184 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10185 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10186 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10187 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10188 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10189 (aarch64_simd_imm_zero_p): Delete.
10190 (aarch64_check_zero_based_sve_index_immediate): Declare.
10191 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10192 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10193 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10194 (aarch64_sve_float_mul_immediate_p): Likewise.
10195 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10196 rather than an rtx.
10197 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10198 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10199 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10200 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10201 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10202 (aarch64_regmode_natural_size): Likewise.
10203 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10204 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10206 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10207 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10208 for VG and the SVE predicate registers.
10209 (V_ALIASES): Add a "z"-prefixed alias.
10210 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10211 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10212 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10213 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10214 (REG_CLASS_NAMES): Add entries for them.
10215 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10216 and the predicate registers.
10217 (aarch64_sve_vg): Declare.
10218 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10219 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10220 (REGMODE_NATURAL_SIZE): Define.
10221 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10223 * config/aarch64/aarch64.c: Include cfgrtl.h.
10224 (simd_immediate_info): Add a constructor for series vectors,
10225 and an associated step field.
10226 (aarch64_sve_vg): New variable.
10227 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10228 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10229 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10230 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10231 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10232 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10233 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10234 (aarch64_get_mask_mode): New functions.
10235 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10236 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10237 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10238 predicate modes and predicate registers. Explicitly restrict
10239 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10240 to store a vector mode if it is recognized by
10241 aarch64_classify_vector_mode.
10242 (aarch64_regmode_natural_size): New function.
10243 (aarch64_hard_regno_caller_save_mode): Return the original mode
10245 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10246 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10247 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10248 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10250 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10251 does not overlap dest if the function is frame-related. Handle
10253 (aarch64_split_add_offset): New function.
10254 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10255 them aarch64_add_offset.
10256 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10257 and update call to aarch64_sub_sp.
10258 (aarch64_add_cfa_expression): New function.
10259 (aarch64_expand_prologue): Pass extra temporary registers to the
10260 functions above. Handle the case in which we need to emit new
10261 DW_CFA_expressions for registers that were originally saved
10262 relative to the stack pointer, but now have to be expressed
10263 relative to the frame pointer.
10264 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10266 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10267 IP0 and IP1 values for SVE frames.
10268 (aarch64_expand_vec_series): New function.
10269 (aarch64_expand_sve_widened_duplicate): Likewise.
10270 (aarch64_expand_sve_const_vector): Likewise.
10271 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10272 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10273 into the register, rather than emitting a SET directly.
10274 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10275 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10276 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10277 (offset_9bit_signed_scaled_p): New functions.
10278 (aarch64_replicate_bitmask_imm): New function.
10279 (aarch64_bitmask_imm): Use it.
10280 (aarch64_cannot_force_const_mem): Reject expressions involving
10281 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10282 (aarch64_classify_index): Handle SVE indices, by requiring
10283 a plain register index with a scale that matches the element size.
10284 (aarch64_classify_address): Handle SVE addresses. Assert that
10285 the mode of the address is VOIDmode or an integer mode.
10286 Update call to aarch64_classify_symbol.
10287 (aarch64_classify_symbolic_expression): Update call to
10288 aarch64_classify_symbol.
10289 (aarch64_const_vec_all_in_range_p): New function.
10290 (aarch64_print_vector_float_operand): Likewise.
10291 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10292 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10293 and the FP immediates 1.0 and 0.5.
10294 (aarch64_print_address_internal): Handle SVE addresses.
10295 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10296 (aarch64_regno_regclass): Handle predicate registers.
10297 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10299 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10300 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10301 (aarch64_convert_sve_vector_bits): New function.
10302 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10303 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10304 rather than an rtx.
10305 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10306 Handle SVE vector and predicate modes. Accept VL-based constants
10307 that need only one temporary register, and VL offsets that require
10308 no temporary registers.
10309 (aarch64_conditional_register_usage): Mark the predicate registers
10310 as fixed if SVE isn't available.
10311 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10312 Return true for SVE vector and predicate modes.
10313 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10314 rather than an unsigned int. Handle SVE modes.
10315 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10317 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10319 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10320 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10321 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10322 (aarch64_sve_float_mul_immediate_p): New functions.
10323 (aarch64_sve_valid_immediate): New function.
10324 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10325 Explicitly reject structure modes. Check for INDEX constants.
10326 Handle PTRUE and PFALSE constants.
10327 (aarch64_check_zero_based_sve_index_immediate): New function.
10328 (aarch64_simd_imm_zero_p): Delete.
10329 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10330 vector modes. Accept constants in the range of CNT[BHWD].
10331 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10332 ask for an Advanced SIMD mode.
10333 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10334 (aarch64_simd_vector_alignment): Handle SVE predicates.
10335 (aarch64_vectorize_preferred_vector_alignment): New function.
10336 (aarch64_simd_vector_alignment_reachable): Use it instead of
10338 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10339 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10341 (MAX_VECT_LEN): Delete.
10342 (expand_vec_perm_d): Add a vec_flags field.
10343 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10344 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10345 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10347 (aarch64_evpc_rev): Rename to...
10348 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10349 (aarch64_evpc_rev_global): New function.
10350 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10351 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10353 (aarch64_evpc_sve_tbl): New function.
10354 (aarch64_expand_vec_perm_const_1): Update after rename of
10355 aarch64_evpc_rev. Handle SVE permutes too, trying
10356 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10357 than aarch64_evpc_tbl.
10358 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10359 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10360 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10361 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10362 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10363 (aarch64_expand_sve_vcond): New functions.
10364 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10365 of aarch64_vector_mode_p.
10366 (aarch64_dwarf_poly_indeterminate_value): New function.
10367 (aarch64_compute_pressure_classes): Likewise.
10368 (aarch64_can_change_mode_class): Likewise.
10369 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10370 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10371 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10372 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10373 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10374 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10375 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10376 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10378 (Dn, Dl, Dr): Accept const as well as const_vector.
10379 (Dz): Likewise. Compare against CONST0_RTX.
10380 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10381 of "vector" where appropriate.
10382 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10383 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10384 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10385 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10386 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10387 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10388 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10389 (v_int_equiv): Extend to SVE modes.
10390 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10392 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10393 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10394 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10395 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10396 (SVE_COND_FP_CMP): New int iterators.
10397 (perm_hilo): Handle the new unpack unspecs.
10398 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10400 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10401 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10402 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10403 (aarch64_equality_operator, aarch64_constant_vector_operand)
10404 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10405 (aarch64_sve_nonimmediate_operand): Likewise.
10406 (aarch64_sve_general_operand): Likewise.
10407 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10408 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10409 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10410 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10411 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10412 (aarch64_sve_float_arith_immediate): Likewise.
10413 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10414 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10415 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10416 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10417 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10418 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10419 (aarch64_sve_float_arith_operand): Likewise.
10420 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10421 (aarch64_sve_float_mul_operand): Likewise.
10422 (aarch64_sve_vec_perm_operand): Likewise.
10423 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10424 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10425 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10426 as well as const_vector.
10427 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10428 in file. Use CONST0_RTX and CONSTM1_RTX.
10429 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10430 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10431 Use aarch64_simd_imm_zero.
10432 * config/aarch64/aarch64-sve.md: New file.
10433 * config/aarch64/aarch64.md: Include it.
10434 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10435 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10436 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10437 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10438 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10439 (sve): New attribute.
10440 (enabled): Disable instructions with the sve attribute unless
10442 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10443 aarch64_expand_mov_immediate.
10444 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10445 CNT[BHSD] immediates.
10446 (movti): Split CONST_POLY_INT moves into two halves.
10447 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10448 Split additions that need a temporary here if the destination
10449 is the stack pointer.
10450 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10451 (*add<mode>3_poly_1): New instruction.
10452 (set_clobber_cc): New expander.
10454 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10456 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10457 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10458 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10459 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10460 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10461 Change innermode from fixed_mode_size to machine_mode.
10462 (simplify_subreg): Update call accordingly. Handle a constant-sized
10463 subreg of a variable-length CONST_VECTOR.
10465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10466 Alan Hayward <alan.hayward@arm.com>
10467 David Sherwood <david.sherwood@arm.com>
10469 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10470 (add_offset_to_base): New function, split out from...
10471 (create_mem_ref): ...here. When handling a scale other than 1,
10472 check first whether the address is valid without the offset.
10473 Add it into the base if so, leaving the index and scale as-is.
10475 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10478 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10479 fold_for_warn before checking if arg2 is INTEGER_CST.
10481 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10483 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10484 (store_multiple_operation): Delete.
10485 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10486 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10487 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10488 guarded by TARGET_STRING.
10489 (rs6000_output_load_multiple): Delete.
10490 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10491 OPTION_MASK_STRING / TARGET_STRING handling.
10492 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10493 (const rs6000_opt_masks) <"string">: Change mask to 0.
10494 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10495 (MASK_STRING): Delete.
10496 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10498 (load_multiple): Delete.
10505 (store_multiple): Delete.
10512 (movmemsi_8reg): Delete.
10513 (corresponding unnamed define_insn): Delete.
10514 (movmemsi_6reg): Delete.
10515 (corresponding unnamed define_insn): Delete.
10516 (movmemsi_4reg): Delete.
10517 (corresponding unnamed define_insn): Delete.
10518 (movmemsi_2reg): Delete.
10519 (corresponding unnamed define_insn): Delete.
10520 (movmemsi_1reg): Delete.
10521 (corresponding unnamed define_insn): Delete.
10522 * config/rs6000/rs6000.opt (mno-string): New.
10523 (mstring): Replace by deprecation warning stub.
10524 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10526 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10528 * regrename.c (regrename_do_replace): If replacing the same
10529 reg multiple times, try to reuse last created gen_raw_REG.
10532 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10533 main to workaround a bug in GDB.
10535 2018-01-12 Tom de Vries <tom@codesourcery.com>
10538 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10540 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10542 PR rtl-optimization/80481
10543 * ira-color.c (get_cap_member): New function.
10544 (allocnos_conflict_by_live_ranges_p): Use it.
10545 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10546 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10548 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10551 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10552 (*saddl_se_1): Ditto.
10553 (*ssubsi_1): Ditto.
10554 (*ssubl_se_1): Ditto.
10556 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10558 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10559 rather than wi::to_widest for DR_INITs.
10560 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10561 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10562 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10564 (vect_analyze_group_access_1): Note that here.
10566 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10568 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10569 polynomial type sizes.
10571 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10573 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10574 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10575 (gimple_add_tmp_var): Likewise.
10577 2018-01-12 Martin Liska <mliska@suse.cz>
10579 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10580 (gimple_alloc_sizes): Likewise.
10581 (dump_gimple_statistics): Use PRIu64 in printf format.
10582 * gimple.h: Change uint64_t to int.
10584 2018-01-12 Martin Liska <mliska@suse.cz>
10586 * tree-core.h: Use uint64_t instead of int.
10587 * tree.c (tree_node_counts): Likewise.
10588 (tree_node_sizes): Likewise.
10589 (dump_tree_statistics): Use PRIu64 in printf format.
10591 2018-01-12 Martin Liska <mliska@suse.cz>
10593 * Makefile.in: As qsort_chk is implemented in vec.c, add
10594 vec.o to linkage of gencfn-macros.
10595 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10596 passing the info to record_node_allocation_statistics.
10597 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10599 * ggc-common.c (struct ggc_usage): Add operator== and use
10600 it in operator< and compare function.
10601 * mem-stats.h (struct mem_usage): Likewise.
10602 * vec.c (struct vec_usage): Remove operator< and compare
10603 function. Can be simply inherited.
10605 2018-01-12 Martin Jambor <mjambor@suse.cz>
10608 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10609 * tree-ssa-math-opts.c: Include domwalk.h.
10610 (convert_mult_to_fma_1): New function.
10611 (fma_transformation_info): New type.
10612 (fma_deferring_state): Likewise.
10613 (cancel_fma_deferring): New function.
10614 (result_of_phi): Likewise.
10615 (last_fma_candidate_feeds_initial_phi): Likewise.
10616 (convert_mult_to_fma): Added deferring logic, split actual
10617 transformation to convert_mult_to_fma_1.
10618 (math_opts_dom_walker): New type.
10619 (math_opts_dom_walker::after_dom_children): New method, body moved
10620 here from pass_optimize_widening_mul::execute, added deferring logic
10622 (pass_optimize_widening_mul::execute): Moved most of code to
10623 math_opts_dom_walker::after_dom_children.
10624 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10625 * config/i386/i386.c (ix86_option_override_internal): Added
10626 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10628 2018-01-12 Richard Biener <rguenther@suse.de>
10631 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10632 inline instance vars.
10634 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10637 * config/rx/rx.c (rx_is_restricted_memory_address):
10638 Handle SUBREG case.
10640 2018-01-12 Richard Biener <rguenther@suse.de>
10642 PR tree-optimization/80846
10643 * target.def (split_reduction): New target hook.
10644 * targhooks.c (default_split_reduction): New function.
10645 * targhooks.h (default_split_reduction): Declare.
10646 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10647 target requests first reduce vectors by combining low and high
10649 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10650 (get_vectype_for_scalar_type_and_size): Export.
10651 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10652 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10653 * doc/tm.texi: Regenerate.
10654 * config/i386/i386.c (ix86_split_reduction): Implement
10655 TARGET_VECTORIZE_SPLIT_REDUCTION.
10657 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10660 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10661 in PIC mode except for TARGET_VXWORKS_RTP.
10662 * config/sparc/sparc.c: Include cfgrtl.h.
10663 (TARGET_INIT_PIC_REG): Define.
10664 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10665 (sparc_pic_register_p): New predicate.
10666 (sparc_legitimate_address_p): Use it.
10667 (sparc_legitimize_pic_address): Likewise.
10668 (sparc_delegitimize_address): Likewise.
10669 (sparc_mode_dependent_address_p): Likewise.
10670 (gen_load_pcrel_sym): Remove 4th parameter.
10671 (load_got_register): Adjust call to above. Remove obsolete stuff.
10672 (sparc_expand_prologue): Do not call load_got_register here.
10673 (sparc_flat_expand_prologue): Likewise.
10674 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10675 (sparc_use_pseudo_pic_reg): New function.
10676 (sparc_init_pic_reg): Likewise.
10677 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10678 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10680 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10682 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10683 Add item for branch_cost.
10685 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10687 PR rtl-optimization/83565
10688 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10689 not extend the result to a larger mode for rotate operations.
10690 (num_sign_bit_copies1): Likewise.
10692 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10695 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10697 Use values-Xc.o for -pedantic.
10698 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10700 2018-01-12 Martin Liska <mliska@suse.cz>
10703 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10705 (possible_polymorphic_call_targets): Use it.
10706 (ipa_devirt): Likewise.
10708 2018-01-12 Martin Liska <mliska@suse.cz>
10710 * profile-count.h (enum profile_quality): Use 0 as invalid
10711 enum value of profile_quality.
10713 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10715 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10716 -mext-string options.
10718 2018-01-12 Richard Biener <rguenther@suse.de>
10720 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10721 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10722 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10724 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10726 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10728 * configure.ac (--with-long-double-format): Add support for the
10729 configuration option to change the default long double format on
10731 * config.gcc (powerpc*-linux*-*): Likewise.
10732 * configure: Regenerate.
10733 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10734 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10735 used without modification.
10737 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10739 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10740 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10741 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10742 MISC_BUILTIN_SPEC_BARRIER.
10743 (rs6000_init_builtins): Likewise.
10744 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10746 (speculation_barrier): New define_insn.
10747 * doc/extend.texi: Document __builtin_speculation_barrier.
10749 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10752 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10753 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10754 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10756 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10757 integral modes instead of "ss" and "sd".
10758 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10759 vectors with 32-bit and 64-bit elements.
10760 (vecdupssescalarmodesuffix): New mode attribute.
10761 (vec_dup<mode>): Use it.
10763 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10766 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10767 frame if argument is passed on stack.
10769 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10772 * ree.c (combine_reaching_defs): Optimize also
10773 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10774 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10776 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10778 PR middle-end/83189
10779 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10781 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10783 PR middle-end/83718
10784 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10785 after they are computed.
10787 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10789 PR tree-optimization/83695
10790 * gimple-loop-linterchange.cc
10791 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10792 reset cached scev information after interchange.
10793 (pass_linterchange::execute): Remove call to scev_reset_htab.
10795 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10797 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10798 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10799 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10800 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10801 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10802 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10803 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10804 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10805 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10806 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10807 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10808 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10809 (V_lane_reg): Likewise.
10810 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10812 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10813 (vfmal_lane_low<mode>_intrinsic,
10814 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10815 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10816 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10817 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10818 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10819 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10821 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10823 * config/arm/arm-cpus.in (fp16fml): New feature.
10824 (ALL_SIMD): Add fp16fml.
10825 (armv8.2-a): Add fp16fml as an option.
10826 (armv8.3-a): Likewise.
10827 (armv8.4-a): Add fp16fml as part of fp16.
10828 * config/arm/arm.h (TARGET_FP16FML): Define.
10829 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10831 * config/arm/arm-modes.def (V2HF): Define.
10832 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10833 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10834 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10835 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10836 vfmsl_low, vfmsl_high): New set of builtins.
10837 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10838 (vfml_op): New code attribute.
10839 (VFMLHALVES): New int iterator.
10840 (VFML, VFMLSEL): New mode attributes.
10841 (V_reg): Define mapping for V2HF.
10842 (V_hi, V_lo): New mode attributes.
10843 (VF_constraint): Likewise.
10844 (vfml_half, vfml_half_selector): New int attributes.
10845 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10847 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10848 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10850 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10851 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10852 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10853 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10855 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10856 Document new effective target and option set.
10858 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10860 * config/arm/arm-cpus.in (armv8_4): New feature.
10861 (ARMv8_4a): New fgroup.
10862 (armv8.4-a): New arch.
10863 * config/arm/arm-tables.opt: Regenerate.
10864 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10865 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10866 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10867 Add matching rules for -march=armv8.4-a and extensions.
10868 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10870 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10873 * config/rx/rx.md (BW): New mode attribute.
10874 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10876 2018-01-11 Richard Biener <rguenther@suse.de>
10878 PR tree-optimization/83435
10879 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10880 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10881 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10883 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10884 Alan Hayward <alan.hayward@arm.com>
10885 David Sherwood <david.sherwood@arm.com>
10887 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10889 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10890 (aarch64_print_address_internal): Use it to check for a zero offset.
10892 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10893 Alan Hayward <alan.hayward@arm.com>
10894 David Sherwood <david.sherwood@arm.com>
10896 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10897 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10898 Return a poly_int64 rather than a HOST_WIDE_INT.
10899 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10900 rather than a HOST_WIDE_INT.
10901 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10902 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10903 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10904 final_offset from HOST_WIDE_INT to poly_int64.
10905 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10906 to_constant when getting the number of units in an Advanced SIMD
10908 (aarch64_builtin_vectorized_function): Check for a constant number
10910 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10912 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10913 attribute instead of GET_MODE_NUNITS.
10914 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10915 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10916 GET_MODE_SIZE for fixed-size registers.
10917 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10918 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10919 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10920 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10921 (aarch64_print_operand, aarch64_print_address_internal)
10922 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10923 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10924 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10925 Handle polynomial GET_MODE_SIZE.
10926 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10927 wider than SImode without modification.
10928 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10929 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10930 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10931 passing and returning SVE modes.
10932 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10933 rather than GEN_INT.
10934 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10935 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10936 (aarch64_allocate_and_probe_stack_space): Likewise.
10937 (aarch64_layout_frame): Cope with polynomial offsets.
10938 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10939 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10940 polynomial offsets.
10941 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10942 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10943 poly_int64 rather than a HOST_WIDE_INT.
10944 (aarch64_get_separate_components, aarch64_process_components)
10945 (aarch64_expand_prologue, aarch64_expand_epilogue)
10946 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10947 (aarch64_anchor_offset): New function, split out from...
10948 (aarch64_legitimize_address): ...here.
10949 (aarch64_builtin_vectorization_cost): Handle polynomial
10950 TYPE_VECTOR_SUBPARTS.
10951 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10953 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10954 number of elements from the PARALLEL rather than the mode.
10955 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10956 rather than GET_MODE_BITSIZE.
10957 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10958 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10959 (aarch64_expand_vec_perm_const_1): Handle polynomial
10960 d->perm.length () and d->perm elements.
10961 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10962 Apply to_constant to d->perm elements.
10963 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10964 polynomial CONST_VECTOR_NUNITS.
10965 (aarch64_move_pointer): Take amount as a poly_int64 rather
10967 (aarch64_progress_pointer): Avoid temporary variable.
10968 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10969 the mode attribute instead of GET_MODE.
10971 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10972 Alan Hayward <alan.hayward@arm.com>
10973 David Sherwood <david.sherwood@arm.com>
10975 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10976 x exists before using it.
10977 (aarch64_add_constant_internal): Rename to...
10978 (aarch64_add_offset_1): ...this. Replace regnum with separate
10979 src and dest rtxes. Handle the case in which they're different,
10980 including when the offset is zero. Replace scratchreg with an rtx.
10981 Use 2 additions if there is no spare register into which we can
10982 move a 16-bit constant.
10983 (aarch64_add_constant): Delete.
10984 (aarch64_add_offset): Replace reg with separate src and dest
10985 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10986 Use aarch64_add_offset_1.
10987 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10988 an rtx rather than an int. Take the delta as a poly_int64
10989 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10990 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10991 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10992 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10993 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10994 and aarch64_add_sp.
10995 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10996 aarch64_add_constant.
10998 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11000 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11001 Use scalar_float_mode.
11003 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11005 * config/aarch64/aarch64-simd.md
11006 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11007 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11008 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11009 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11010 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11011 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11012 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11013 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11014 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11015 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11017 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11020 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11021 targ_options->x_arm_arch_string is non NULL.
11023 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11025 * config/aarch64/aarch64.h
11026 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11028 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11031 * expmed.c (emit_store_flag_force): Swap if const op0
11032 and change VOIDmode to mode of op0.
11034 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11036 PR rtl-optimization/83761
11037 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11038 than bytes to mode_for_size.
11040 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11042 PR middle-end/83189
11043 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11044 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11047 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11049 PR middle-end/83575
11050 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11051 when in layout mode.
11052 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11053 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11056 2018-01-10 Michael Collison <michael.collison@arm.com>
11058 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11059 * config/aarch64/aarch64-option-extension.def: Add
11060 AARCH64_OPT_EXTENSION of 'fp16fml'.
11061 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11062 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11063 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11064 * config/aarch64/constraints.md (Ui7): New constraint.
11065 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11066 (VFMLA_SEL_W): Ditto.
11069 (VFMLA16_LOW): New int iterator.
11070 (VFMLA16_HIGH): Ditto.
11071 (UNSPEC_FMLAL): New unspec.
11072 (UNSPEC_FMLSL): Ditto.
11073 (UNSPEC_FMLAL2): Ditto.
11074 (UNSPEC_FMLSL2): Ditto.
11075 (f16mac): New code attribute.
11076 * config/aarch64/aarch64-simd-builtins.def
11077 (aarch64_fmlal_lowv2sf): Ditto.
11078 (aarch64_fmlsl_lowv2sf): Ditto.
11079 (aarch64_fmlalq_lowv4sf): Ditto.
11080 (aarch64_fmlslq_lowv4sf): Ditto.
11081 (aarch64_fmlal_highv2sf): Ditto.
11082 (aarch64_fmlsl_highv2sf): Ditto.
11083 (aarch64_fmlalq_highv4sf): Ditto.
11084 (aarch64_fmlslq_highv4sf): Ditto.
11085 (aarch64_fmlal_lane_lowv2sf): Ditto.
11086 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11087 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11088 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11089 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11090 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11091 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11092 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11093 (aarch64_fmlal_lane_highv2sf): Ditto.
11094 (aarch64_fmlsl_lane_highv2sf): Ditto.
11095 (aarch64_fmlal_laneq_highv2sf): Ditto.
11096 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11097 (aarch64_fmlalq_lane_highv4sf): Ditto.
11098 (aarch64_fmlsl_lane_highv4sf): Ditto.
11099 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11100 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11101 * config/aarch64/aarch64-simd.md:
11102 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11103 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11104 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11105 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11106 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11107 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11108 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11109 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11110 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11111 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11112 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11113 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11114 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11115 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11116 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11117 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11118 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11119 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11120 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11121 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11122 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11123 (vfmlsl_low_u32): Ditto.
11124 (vfmlalq_low_u32): Ditto.
11125 (vfmlslq_low_u32): Ditto.
11126 (vfmlal_high_u32): Ditto.
11127 (vfmlsl_high_u32): Ditto.
11128 (vfmlalq_high_u32): Ditto.
11129 (vfmlslq_high_u32): Ditto.
11130 (vfmlal_lane_low_u32): Ditto.
11131 (vfmlsl_lane_low_u32): Ditto.
11132 (vfmlal_laneq_low_u32): Ditto.
11133 (vfmlsl_laneq_low_u32): Ditto.
11134 (vfmlalq_lane_low_u32): Ditto.
11135 (vfmlslq_lane_low_u32): Ditto.
11136 (vfmlalq_laneq_low_u32): Ditto.
11137 (vfmlslq_laneq_low_u32): Ditto.
11138 (vfmlal_lane_high_u32): Ditto.
11139 (vfmlsl_lane_high_u32): Ditto.
11140 (vfmlal_laneq_high_u32): Ditto.
11141 (vfmlsl_laneq_high_u32): Ditto.
11142 (vfmlalq_lane_high_u32): Ditto.
11143 (vfmlslq_lane_high_u32): Ditto.
11144 (vfmlalq_laneq_high_u32): Ditto.
11145 (vfmlslq_laneq_high_u32): Ditto.
11146 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11147 (AARCH64_FL_FOR_ARCH8_4): New.
11148 (AARCH64_ISA_F16FML): New ISA flag.
11149 (TARGET_F16FML): New feature flag for fp16fml.
11150 (doc/invoke.texi): Document new fp16fml option.
11152 2018-01-10 Michael Collison <michael.collison@arm.com>
11154 * config/aarch64/aarch64-builtins.c:
11155 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11156 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11157 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11158 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11159 (AARCH64_ISA_SHA3): New ISA flag.
11160 (TARGET_SHA3): New feature flag for sha3.
11161 * config/aarch64/iterators.md (sha512_op): New int attribute.
11162 (CRYPTO_SHA512): New int iterator.
11163 (UNSPEC_SHA512H): New unspec.
11164 (UNSPEC_SHA512H2): Ditto.
11165 (UNSPEC_SHA512SU0): Ditto.
11166 (UNSPEC_SHA512SU1): Ditto.
11167 * config/aarch64/aarch64-simd-builtins.def
11168 (aarch64_crypto_sha512hqv2di): New builtin.
11169 (aarch64_crypto_sha512h2qv2di): Ditto.
11170 (aarch64_crypto_sha512su0qv2di): Ditto.
11171 (aarch64_crypto_sha512su1qv2di): Ditto.
11172 (aarch64_eor3qv8hi): Ditto.
11173 (aarch64_rax1qv2di): Ditto.
11174 (aarch64_xarqv2di): Ditto.
11175 (aarch64_bcaxqv8hi): Ditto.
11176 * config/aarch64/aarch64-simd.md:
11177 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11178 (aarch64_crypto_sha512su0qv2di): Ditto.
11179 (aarch64_crypto_sha512su1qv2di): Ditto.
11180 (aarch64_eor3qv8hi): Ditto.
11181 (aarch64_rax1qv2di): Ditto.
11182 (aarch64_xarqv2di): Ditto.
11183 (aarch64_bcaxqv8hi): Ditto.
11184 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11185 (vsha512h2q_u64): Ditto.
11186 (vsha512su0q_u64): Ditto.
11187 (vsha512su1q_u64): Ditto.
11188 (veor3q_u16): Ditto.
11189 (vrax1q_u64): Ditto.
11190 (vxarq_u64): Ditto.
11191 (vbcaxq_u16): Ditto.
11192 * config/arm/types.md (crypto_sha512): New type attribute.
11193 (crypto_sha3): Ditto.
11194 (doc/invoke.texi): Document new sha3 option.
11196 2018-01-10 Michael Collison <michael.collison@arm.com>
11198 * config/aarch64/aarch64-builtins.c:
11199 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11200 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11201 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11202 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11203 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11204 (AARCH64_ISA_SM4): New ISA flag.
11205 (TARGET_SM4): New feature flag for sm4.
11206 * config/aarch64/aarch64-simd-builtins.def
11207 (aarch64_sm3ss1qv4si): Ditto.
11208 (aarch64_sm3tt1aq4si): Ditto.
11209 (aarch64_sm3tt1bq4si): Ditto.
11210 (aarch64_sm3tt2aq4si): Ditto.
11211 (aarch64_sm3tt2bq4si): Ditto.
11212 (aarch64_sm3partw1qv4si): Ditto.
11213 (aarch64_sm3partw2qv4si): Ditto.
11214 (aarch64_sm4eqv4si): Ditto.
11215 (aarch64_sm4ekeyqv4si): Ditto.
11216 * config/aarch64/aarch64-simd.md:
11217 (aarch64_sm3ss1qv4si): Ditto.
11218 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11219 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11220 (aarch64_sm4eqv4si): Ditto.
11221 (aarch64_sm4ekeyqv4si): Ditto.
11222 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11223 (sm3part_op): Ditto.
11224 (CRYPTO_SM3TT): Ditto.
11225 (CRYPTO_SM3PART): Ditto.
11226 (UNSPEC_SM3SS1): New unspec.
11227 (UNSPEC_SM3TT1A): Ditto.
11228 (UNSPEC_SM3TT1B): Ditto.
11229 (UNSPEC_SM3TT2A): Ditto.
11230 (UNSPEC_SM3TT2B): Ditto.
11231 (UNSPEC_SM3PARTW1): Ditto.
11232 (UNSPEC_SM3PARTW2): Ditto.
11233 (UNSPEC_SM4E): Ditto.
11234 (UNSPEC_SM4EKEY): Ditto.
11235 * config/aarch64/constraints.md (Ui2): New constraint.
11236 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11237 * config/arm/types.md (crypto_sm3): New type attribute.
11238 (crypto_sm4): Ditto.
11239 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11240 (vsm3tt1aq_u32): Ditto.
11241 (vsm3tt1bq_u32): Ditto.
11242 (vsm3tt2aq_u32): Ditto.
11243 (vsm3tt2bq_u32): Ditto.
11244 (vsm3partw1q_u32): Ditto.
11245 (vsm3partw2q_u32): Ditto.
11246 (vsm4eq_u32): Ditto.
11247 (vsm4ekeyq_u32): Ditto.
11248 (doc/invoke.texi): Document new sm4 option.
11250 2018-01-10 Michael Collison <michael.collison@arm.com>
11252 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11253 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11254 (AARCH64_FL_FOR_ARCH8_4): New.
11255 (AARCH64_FL_V8_4): New flag.
11256 (doc/invoke.texi): Document new armv8.4-a option.
11258 2018-01-10 Michael Collison <michael.collison@arm.com>
11260 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11261 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11262 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11263 * config/aarch64/aarch64-option-extension.def: Add
11264 AARCH64_OPT_EXTENSION of 'sha2'.
11265 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11266 (crypto): Disable sha2 and aes if crypto disabled.
11267 (crypto): Enable aes and sha2 if enabled.
11268 (simd): Disable sha2 and aes if simd disabled.
11269 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11271 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11272 (TARGET_SHA2): New feature flag for sha2.
11273 (TARGET_AES): New feature flag for aes.
11274 * config/aarch64/aarch64-simd.md:
11275 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11276 conditional on TARGET_AES.
11277 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11278 (aarch64_crypto_sha1hsi): Make pattern conditional
11280 (aarch64_crypto_sha1hv4si): Ditto.
11281 (aarch64_be_crypto_sha1hv4si): Ditto.
11282 (aarch64_crypto_sha1su1v4si): Ditto.
11283 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11284 (aarch64_crypto_sha1su0v4si): Ditto.
11285 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11286 (aarch64_crypto_sha256su0v4si): Ditto.
11287 (aarch64_crypto_sha256su1v4si): Ditto.
11288 (doc/invoke.texi): Document new aes and sha2 options.
11290 2018-01-10 Martin Sebor <msebor@redhat.com>
11292 PR tree-optimization/83781
11293 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11296 2018-01-11 Martin Sebor <msebor@gmail.com>
11297 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11299 PR tree-optimization/83501
11300 PR tree-optimization/81703
11302 * tree-ssa-strlen.c (get_string_cst): Rename...
11303 (get_string_len): ...to this. Handle global constants.
11304 (handle_char_store): Adjust.
11306 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11307 Jim Wilson <jimw@sifive.com>
11309 * config/riscv/riscv-protos.h (riscv_output_return): New.
11310 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11311 (riscv_attribute_table, riscv_output_return),
11312 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11313 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11314 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11316 (riscv_expand_prologue): Add early return for naked function.
11317 (riscv_expand_epilogue): Likewise.
11318 (riscv_function_ok_for_sibcall): Return false for naked function.
11319 (riscv_set_current_function): New.
11320 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11321 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11322 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11323 * doc/extend.texi (RISC-V Function Attributes): New.
11325 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11327 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11328 check for 128-bit long double before checking TCmode.
11329 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11330 128-bit long doubles before checking TFmode or TCmode.
11331 (FLOAT128_IBM_P): Likewise.
11333 2018-01-10 Martin Sebor <msebor@redhat.com>
11335 PR tree-optimization/83671
11336 * builtins.c (c_strlen): Unconditionally return zero for the empty
11338 Use -Warray-bounds for warnings.
11339 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11340 for non-constant array indices with COMPONENT_REF, arrays of
11341 arrays, and pointers to arrays.
11342 (gimple_fold_builtin_strlen): Determine and set length range for
11343 non-constant character arrays.
11345 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11347 PR middle-end/81897
11348 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11351 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11353 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11355 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11358 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11359 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11360 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11361 indexed_or_indirect_operand predicate.
11362 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11363 (*vsx_le_perm_load_v8hi): Likewise.
11364 (*vsx_le_perm_load_v16qi): Likewise.
11365 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11366 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11367 (*vsx_le_perm_store_v8hi): Likewise.
11368 (*vsx_le_perm_store_v16qi): Likewise.
11369 (eight unnamed splitters): Likewise.
11371 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11373 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11374 * config/rs6000/emmintrin.h: Likewise.
11375 * config/rs6000/mmintrin.h: Likewise.
11376 * config/rs6000/xmmintrin.h: Likewise.
11378 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11381 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11383 * tree.c (tree_nop_conversion): Return true for location wrapper
11385 (maybe_wrap_with_location): New function.
11386 (selftest::check_strip_nops): New function.
11387 (selftest::test_location_wrappers): New function.
11388 (selftest::tree_c_tests): Call it.
11389 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11390 (maybe_wrap_with_location): New decl.
11391 (EXPR_LOCATION_WRAPPER_P): New macro.
11392 (location_wrapper_p): New inline function.
11393 (tree_strip_any_location_wrapper): New inline function.
11395 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11398 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11399 stack_realign_offset for the largest alignment of stack slot
11401 (ix86_find_max_used_stack_alignment): New function.
11402 (ix86_finalize_stack_frame_flags): Use it. Set
11403 max_used_stack_alignment if we don't realign stack.
11404 * config/i386/i386.h (machine_function): Add
11405 max_used_stack_alignment.
11407 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11409 * config/arm/arm.opt (-mbranch-cost): New option.
11410 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11413 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11416 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11417 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11419 2018-01-10 Richard Biener <rguenther@suse.de>
11422 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11423 early out so it also covers the case where we have a non-NULL
11426 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11428 PR tree-optimization/83753
11429 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11430 for non-strided grouped accesses if the number of elements is 1.
11432 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11435 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11436 * i386.h (TARGET_USE_GATHER): Define.
11437 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11439 2018-01-10 Martin Liska <mliska@suse.cz>
11442 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11443 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11445 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11446 CLEANUP_NO_PARTITIONING is not set.
11448 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11450 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11451 for vectors, as a partial revert of r254296.
11452 * rtl.h (const_vec_p): Delete.
11453 (const_vec_duplicate_p): Don't test for vector CONSTs.
11454 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11455 * expmed.c (make_tree): Likewise.
11458 * common.md (E, F): Use CONSTANT_P instead of checking for
11460 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11461 checking for CONST_VECTOR.
11463 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11465 PR middle-end/83575
11466 * predict.c (force_edge_cold): Handle in more sane way edges
11467 with no prediction.
11469 2018-01-09 Carl Love <cel@us.ibm.com>
11471 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11473 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11474 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11475 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11476 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11477 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11478 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11479 * config/rs6000/rs6000-protos.h: Add extern defition for
11480 rs6000_generate_float2_double_code.
11481 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11483 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11484 (float2_v2df): Add define_expand.
11486 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11489 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11490 op_mode in the force_to_mode call.
11492 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11494 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11495 instead of checking each element individually.
11496 (aarch64_evpc_uzp): Likewise.
11497 (aarch64_evpc_zip): Likewise.
11498 (aarch64_evpc_ext): Likewise.
11499 (aarch64_evpc_rev): Likewise.
11500 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11501 instead of checking each element individually. Return true without
11503 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11504 whether all selected elements come from the same input, instead of
11505 checking each element individually. Remove calls to gen_rtx_REG,
11506 start_sequence and end_sequence and instead assert that no rtl is
11509 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11511 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11512 order of HIGH and CONST checks.
11514 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11516 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11517 if the destination isn't an SSA_NAME.
11519 2018-01-09 Richard Biener <rguenther@suse.de>
11521 PR tree-optimization/83668
11522 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11524 (canonicalize_loop_form): ... here, renamed from ...
11525 (canonicalize_loop_closed_ssa_form): ... this and amended to
11526 swap successor edges for loop exit blocks to make us use
11527 the RPO order we need for initial schedule generation.
11529 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11531 PR tree-optimization/64811
11532 * match.pd: When optimizing comparisons with Inf, avoid
11533 introducing or losing exceptions from comparisons with NaN.
11535 2018-01-09 Martin Liska <mliska@suse.cz>
11538 * asan.c (shadow_mem_size): Add gcc_assert.
11540 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11542 Don't save registers in main().
11545 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11546 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11547 * config/avr/avr.c (avr_set_current_function): Don't error if
11548 naked, OS_task or OS_main are specified at the same time.
11549 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11551 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11553 * common/config/avr/avr-common.c (avr_option_optimization_table):
11554 Switch on -mmain-is-OS_task for optimizing compilations.
11556 2018-01-09 Richard Biener <rguenther@suse.de>
11558 PR tree-optimization/83572
11559 * graphite.c: Include cfganal.h.
11560 (graphite_transform_loops): Connect infinite loops to exit
11561 and remove fake edges at the end.
11563 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11565 * ipa-inline.c (edge_badness): Revert accidental checkin.
11567 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11570 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11571 symbols; not inline clones.
11573 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11576 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11577 hard registers. Formatting fixes.
11579 PR preprocessor/83722
11580 * gcc.c (try_generate_repro): Pass
11581 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11582 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11585 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11586 Kito Cheng <kito.cheng@gmail.com>
11588 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11589 (riscv_leaf_function_p): Delete.
11590 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11592 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11594 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11596 (do_ifelse): New function.
11597 (do_isel): New function.
11598 (do_sub3): New function.
11599 (do_add3): New function.
11600 (do_load_mask_compare): New function.
11601 (do_overlap_load_compare): New function.
11602 (expand_compare_loop): New function.
11603 (expand_block_compare): Call expand_compare_loop() when appropriate.
11604 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11605 option description.
11606 (-mblock-compare-inline-loop-limit): New option.
11608 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11611 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11612 Reverse order of second and third operands in first alternative.
11613 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11614 of first and second elements in UNSPEC_VPERMR vector.
11615 (altivec_expand_vec_perm_le): Likewise.
11617 2018-01-08 Jeff Law <law@redhat.com>
11619 PR rtl-optimizatin/81308
11620 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11621 (process_switch): If group_case_labels makes a change, then set
11623 (pass_convert_switch::execute): If a switch is converted, then
11624 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11626 PR rtl-optimization/81308
11627 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11630 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11632 PR target/83663 - Revert r255946
11633 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11634 generation for cases where splatting a value is not useful.
11635 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11636 across a vec_duplicate and a paradoxical subreg forming a vector
11637 mode to a vec_concat.
11639 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11641 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11642 -march=armv8.3-a variants.
11643 * config/arm/t-multilib: Likewise.
11644 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11646 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11648 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11650 (cceq_ior_compare_complement): Give it a name so I can use it, and
11651 change boolean_or_operator predicate to boolean_operator so it can
11652 be used to generate a crand.
11653 (eqne): New code iterator.
11654 (bd/bd_neg): New code_attrs.
11655 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11656 a single define_insn.
11657 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11658 decrement (bdnzt/bdnzf/bdzt/bdzf).
11659 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11660 with the new names of the branch decrement patterns, and added the
11661 names of the branch decrement conditional patterns.
11663 2018-01-08 Richard Biener <rguenther@suse.de>
11665 PR tree-optimization/83563
11666 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11669 2018-01-08 Richard Biener <rguenther@suse.de>
11671 PR middle-end/83713
11672 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11674 2018-01-08 Richard Biener <rguenther@suse.de>
11676 PR tree-optimization/83685
11677 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11678 references to abnormals.
11680 2018-01-08 Richard Biener <rguenther@suse.de>
11683 * dwarf2out.c (output_indirect_strings): Handle empty
11684 skeleton_debug_str_hash.
11685 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11687 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11689 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11690 (emit_store_direct): Likewise.
11691 (arc_trampoline_adjust_address): Likewise.
11692 (arc_asm_trampoline_template): New function.
11693 (arc_initialize_trampoline): Use asm_trampoline_template.
11694 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11695 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11696 * config/arc/arc.md (flush_icache): Delete pattern.
11698 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11700 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11701 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11704 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11707 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11708 by not USED_FOR_TARGET.
11709 (make_pass_resolve_sw_modes): Likewise.
11711 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11713 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11716 2018-01-08 Richard Biener <rguenther@suse.de>
11718 PR middle-end/83580
11719 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11721 2018-01-08 Richard Biener <rguenther@suse.de>
11723 PR middle-end/83517
11724 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11726 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11728 PR middle-end/81897
11729 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11730 basic blocks with a small number of successors.
11731 (convert_control_dep_chain_into_preds): Improve handling of
11733 (dump_predicates): Split apart into...
11734 (dump_pred_chain): ...here...
11735 (dump_pred_info): ...and here.
11736 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11737 (can_chain_union_be_invalidated_p): Improve check for invalidation
11739 (uninit_uses_cannot_happen): Avoid unnecessary if
11740 convert_control_dep_chain_into_preds yielded nothing.
11742 2018-01-06 Martin Sebor <msebor@redhat.com>
11744 PR tree-optimization/83640
11745 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11746 subtracting negative offset from size.
11747 (builtin_access::overlap): Adjust offset bounds of the access to fall
11748 within the size of the object if possible.
11750 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11752 PR rtl-optimization/83699
11753 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11754 extract_bit_field_as_subreg to cases in which the extracted
11755 value is also a vector.
11757 * lra-constraints.c (process_alt_operands): Test for the equivalence
11758 substitutions when detecting a possible reload cycle.
11760 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11763 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11764 by default if flag_selective_schedling{,2}. Formatting fixes.
11766 PR rtl-optimization/83682
11767 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11768 if it has non-VECTOR_MODE element mode.
11769 (vec_duplicate_p): Likewise.
11771 PR middle-end/83694
11772 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11773 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11775 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11778 * config/i386/i386-builtin.def
11779 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11780 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11781 Require also OPTION_MASK_ISA_AVX512F in addition to
11782 OPTION_MASK_ISA_GFNI.
11783 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11784 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11785 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11786 to OPTION_MASK_ISA_GFNI.
11787 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11788 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11789 OPTION_MASK_ISA_AVX512BW.
11790 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11791 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11792 addition to OPTION_MASK_ISA_GFNI.
11793 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11794 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11795 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11796 to OPTION_MASK_ISA_GFNI.
11797 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11798 a requirement for all ISAs rather than any of them with a few
11800 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11802 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11803 bitmasks to be enabled with 3 exceptions, instead of requiring any
11804 enabled ISA with lots of exceptions.
11805 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11806 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11807 Change avx512bw in isa attribute to avx512f.
11808 * config/i386/sgxintrin.h: Add license boilerplate.
11809 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11810 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11811 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11812 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11814 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11815 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11816 temporarily sse2 rather than sse if not enabled already.
11819 * config/i386/sse.md (VI248_VLBW): Rename to ...
11820 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11821 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11822 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11823 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11824 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11825 mode iterator instead of VI248_VLBW.
11827 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11829 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11830 (record_modified): Skip clobbers; add debug output.
11831 (param_change_prob): Use sreal frequencies.
11833 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11835 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11836 punt for user-aligned variables.
11838 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11840 * tree-chrec.c (chrec_contains_symbols): Return true for
11843 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11846 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11847 of (x|y) == x for BICS pattern.
11849 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11851 PR tree-optimization/83605
11852 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11853 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11856 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11858 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11859 * config/epiphany/rtems.h: New file.
11861 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11862 Uros Bizjak <ubizjak@gmail.com>
11865 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11866 QIreg_operand instead of register_operand predicate.
11867 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11868 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11869 comments instead of -fmitigate[-_]rop.
11871 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11874 * cgraphunit.c (symbol_table::compile): Switch to text_section
11875 before calling assembly_start debug hook.
11876 * run-rtl-passes.c (run_rtl_passes): Likewise.
11879 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11881 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11882 range_int_cst_p rather than !symbolic_range_p before calling
11883 extract_range_from_multiplicative_op_1.
11885 2018-01-04 Jeff Law <law@redhat.com>
11887 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11888 redundant test in assertion.
11890 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11892 * doc/rtl.texi: Document machine_mode wrapper classes.
11894 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11896 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11897 using tree_to_uhwi.
11899 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11901 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11902 the VEC_PERM_EXPR fold to fail.
11904 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11907 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11908 to switched_sections.
11910 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11913 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11914 test for d.testing.
11916 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11919 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11920 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11922 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11925 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11926 is BLKmode and bitpos not zero or mode change is needed.
11928 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11931 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11934 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11937 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11938 instead of MULT rtx. Update all corresponding splitters.
11939 (*saddl_se): Ditto.
11940 (*ssub<modesuffix>): Ditto.
11941 (*ssubl_se): Ditto.
11942 (*cmp_sadd_di): Update split patterns.
11943 (*cmp_sadd_si): Ditto.
11944 (*cmp_sadd_sidi): Ditto.
11945 (*cmp_ssub_di): Ditto.
11946 (*cmp_ssub_si): Ditto.
11947 (*cmp_ssub_sidi): Ditto.
11948 * config/alpha/predicates.md (const23_operand): New predicate.
11949 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11950 Look for ASHIFT, not MULT inner operand.
11951 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11953 2018-01-04 Martin Liska <mliska@suse.cz>
11955 PR gcov-profile/83669
11956 * gcov.c (output_intermediate_file): Add version to intermediate
11958 * doc/gcov.texi: Document new field 'version' in intermediate
11959 file format. Fix location of '-k' option of gcov command.
11961 2018-01-04 Martin Liska <mliska@suse.cz>
11964 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11966 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11968 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11970 2018-01-03 Martin Sebor <msebor@redhat.com>
11972 PR tree-optimization/83655
11973 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11974 checking calls with invalid arguments.
11976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11978 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11979 (vectorizable_mask_load_store): Delete.
11980 (vectorizable_call): Return false for masked loads and stores.
11981 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11982 instead of gimple_assign_rhs1.
11983 (vectorizable_load): Handle IFN_MASK_LOAD.
11984 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11988 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11990 (vectorizable_mask_load_store): ...here.
11991 (vectorizable_load): ...and here.
11993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11995 * tree-vect-stmts.c (vect_build_all_ones_mask)
11996 (vect_build_zero_merge_argument): New functions, split out from...
11997 (vectorizable_load): ...here.
11999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12001 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12003 (vectorizable_mask_load_store): ...here.
12004 (vectorizable_store): ...and here.
12006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12008 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12010 (vectorizable_mask_load_store): ...here.
12012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12014 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12015 (vect_model_store_cost): Take a vec_load_store_type instead of a
12017 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12018 (vect_model_store_cost): Take a vec_load_store_type instead of a
12020 (vectorizable_mask_load_store): Update accordingly.
12021 (vectorizable_store): Likewise.
12022 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12027 IFN_MASK_LOAD calls here rather than...
12028 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12031 Alan Hayward <alan.hayward@arm.com>
12032 David Sherwood <david.sherwood@arm.com>
12034 * expmed.c (extract_bit_field_1): For vector extracts,
12035 fall back to extract_bit_field_as_subreg if vec_extract
12038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12039 Alan Hayward <alan.hayward@arm.com>
12040 David Sherwood <david.sherwood@arm.com>
12042 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12043 they are variable or constant sized.
12044 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12045 slots for constant-sized data.
12047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12048 Alan Hayward <alan.hayward@arm.com>
12049 David Sherwood <david.sherwood@arm.com>
12051 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12052 handling COND_EXPRs with boolean comparisons, try to find a better
12053 basis for the mask type than the boolean itself.
12055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12057 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12058 is calculated and how it can be overridden.
12059 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12060 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12062 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
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 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12070 Remove the mode argument.
12071 (aarch64_simd_valid_immediate): Remove the mode and inverse
12073 * config/aarch64/iterators.md (bitsize): New iterator.
12074 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12075 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12076 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12077 aarch64_simd_valid_immediate.
12078 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12079 (aarch64_reg_or_bic_imm): Likewise.
12080 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12081 with an insn_type enum and msl with a modifier_type enum.
12082 Replace element_width with a scalar_mode. Change the shift
12083 to unsigned int. Add constructors for scalar_float_mode and
12084 scalar_int_mode elements.
12085 (aarch64_vect_float_const_representable_p): Delete.
12086 (aarch64_can_const_movi_rtx_p)
12087 (aarch64_simd_scalar_immediate_valid_for_move)
12088 (aarch64_simd_make_constant): Update call to
12089 aarch64_simd_valid_immediate.
12090 (aarch64_advsimd_valid_immediate_hs): New function.
12091 (aarch64_advsimd_valid_immediate): Likewise.
12092 (aarch64_simd_valid_immediate): Remove mode and inverse
12093 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12094 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12095 and aarch64_float_const_representable_p on the result.
12096 (aarch64_output_simd_mov_immediate): Remove mode argument.
12097 Update call to aarch64_simd_valid_immediate and use of
12098 simd_immediate_info.
12099 (aarch64_output_scalar_simd_mov_immediate): Update call
12102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12103 Alan Hayward <alan.hayward@arm.com>
12104 David Sherwood <david.sherwood@arm.com>
12106 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12107 (mode_nunits): Likewise CONST_MODE_NUNITS.
12108 * machmode.def (ADJUST_NUNITS): Document.
12109 * genmodes.c (mode_data::need_nunits_adj): New field.
12110 (blank_mode): Update accordingly.
12111 (adj_nunits): New variable.
12112 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12114 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12115 listed in adj_nunits.
12116 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12117 listed in adj_nunits. Don't emit case statements for such modes.
12118 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12119 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12120 nothing if adj_nunits is nonnull.
12121 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12122 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12123 (emit_mode_fbit): Update use of print_maybe_const_decl.
12124 (emit_move_size): Likewise. Treat the array as non-const
12126 (emit_mode_adjustments): Handle adj_nunits.
12128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12130 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12131 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12132 (VECTOR_MODES): Use it.
12133 (make_vector_modes): Take the prefix as an argument.
12135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12136 Alan Hayward <alan.hayward@arm.com>
12137 David Sherwood <david.sherwood@arm.com>
12139 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12140 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12141 for MODE_VECTOR_BOOL.
12142 * machmode.def (VECTOR_BOOL_MODE): Document.
12143 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12144 (make_vector_bool_mode): New function.
12145 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12147 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12148 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12150 * stor-layout.c (int_mode_for_mode): Likewise.
12151 * tree.c (build_vector_type_for_mode): Likewise.
12152 * varasm.c (output_constant_pool_2): Likewise.
12153 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12154 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12155 for MODE_VECTOR_BOOL.
12156 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12157 of mode class checks.
12158 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12159 instead of a list of mode class checks.
12160 (expand_vector_scalar_condition): Likewise.
12161 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12164 Alan Hayward <alan.hayward@arm.com>
12165 David Sherwood <david.sherwood@arm.com>
12167 * machmode.h (mode_size): Change from unsigned short to
12169 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12170 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12171 or if measurement_type is not polynomial.
12172 (fixed_size_mode::includes_p): Check for constant-sized modes.
12173 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12174 return a poly_uint16 rather than an unsigned short.
12175 (emit_mode_size): Change the type of mode_size from unsigned short
12176 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12177 (emit_mode_adjustments): Cope with polynomial vector sizes.
12178 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12180 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12182 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12183 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12184 * caller-save.c (setup_save_areas): Likewise.
12185 (replace_reg_with_saved_mem): Likewise.
12186 * calls.c (emit_library_call_value_1): Likewise.
12187 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12188 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12189 (gen_lowpart_for_combine): Likewise.
12190 * convert.c (convert_to_integer_1): Likewise.
12191 * cse.c (equiv_constant, cse_insn): Likewise.
12192 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12193 (cselib_subst_to_values): Likewise.
12194 * dce.c (word_dce_process_block): Likewise.
12195 * df-problems.c (df_word_lr_mark_ref): Likewise.
12196 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12197 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12198 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12199 (rtl_for_decl_location): Likewise.
12200 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12201 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12202 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12203 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12204 (expand_expr_real_1): Likewise.
12205 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12206 (pad_below): Likewise.
12207 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12208 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12209 * ira.c (get_subreg_tracking_sizes): Likewise.
12210 * ira-build.c (ira_create_allocno_objects): Likewise.
12211 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12212 (ira_sort_regnos_for_alter_reg): Likewise.
12213 * ira-costs.c (record_operand_costs): Likewise.
12214 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12215 (resolve_simple_move): Likewise.
12216 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12217 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12218 (lra_constraints): Likewise.
12219 (CONST_POOL_OK_P): Reject variable-sized modes.
12220 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12221 (add_pseudo_to_slot, lra_spill): Likewise.
12222 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12223 * optabs-query.c (get_best_extraction_insn): Likewise.
12224 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12225 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12226 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12227 * recog.c (offsettable_address_addr_space_p): Likewise.
12228 * regcprop.c (maybe_mode_change): Likewise.
12229 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12230 * regrename.c (build_def_use): Likewise.
12231 * regstat.c (dump_reg_info): Likewise.
12232 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12233 (find_reloads, find_reloads_subreg_address): Likewise.
12234 * reload1.c (eliminate_regs_1): Likewise.
12235 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12236 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12237 (simplify_binary_operation_1, simplify_subreg): Likewise.
12238 * targhooks.c (default_function_arg_padding): Likewise.
12239 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12240 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12241 (verify_gimple_assign_ternary): Likewise.
12242 * tree-inline.c (estimate_move_cost): Likewise.
12243 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12244 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12245 (get_address_cost_ainc): Likewise.
12246 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12247 (vect_supportable_dr_alignment): Likewise.
12248 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12249 (vectorizable_reduction): Likewise.
12250 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12251 (vectorizable_operation, vectorizable_load): Likewise.
12252 * tree.c (build_same_sized_truth_vector_type): Likewise.
12253 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12254 * var-tracking.c (emit_note_insn_var_location): Likewise.
12255 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12256 (ADDR_VEC_ALIGN): Likewise.
12258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12259 Alan Hayward <alan.hayward@arm.com>
12260 David Sherwood <david.sherwood@arm.com>
12262 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12264 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12265 or if measurement_type is polynomial.
12266 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12267 * combine.c (make_extraction): Likewise.
12268 * dse.c (find_shift_sequence): Likewise.
12269 * dwarf2out.c (mem_loc_descriptor): Likewise.
12270 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12271 (extract_bit_field, extract_low_bits): Likewise.
12272 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12273 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12274 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12275 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12276 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12277 * reload.c (find_reloads): Likewise.
12278 * reload1.c (alter_reg): Likewise.
12279 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12280 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12281 * tree-if-conv.c (predicate_mem_writes): Likewise.
12282 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12283 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12284 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12285 * valtrack.c (dead_debug_insert_temp): Likewise.
12286 * varasm.c (mergeable_constant_section): Likewise.
12287 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12290 Alan Hayward <alan.hayward@arm.com>
12291 David Sherwood <david.sherwood@arm.com>
12293 * expr.c (expand_assignment): Cope with polynomial mode sizes
12294 when assigning to a CONCAT.
12296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12297 Alan Hayward <alan.hayward@arm.com>
12298 David Sherwood <david.sherwood@arm.com>
12300 * machmode.h (mode_precision): Change from unsigned short to
12302 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12304 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12305 or if measurement_type is not polynomial.
12306 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12307 in which the mode is already known to be a scalar_int_mode.
12308 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12309 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12311 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12312 for GET_MODE_PRECISION.
12313 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12314 for GET_MODE_PRECISION.
12315 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12317 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12318 (expand_field_assignment, make_extraction): Likewise.
12319 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12320 (get_last_value): Likewise.
12321 * convert.c (convert_to_integer_1): Likewise.
12322 * cse.c (cse_insn): Likewise.
12323 * expr.c (expand_expr_real_1): Likewise.
12324 * lra-constraints.c (simplify_operand_subreg): Likewise.
12325 * optabs-query.c (can_atomic_load_p): Likewise.
12326 * optabs.c (expand_atomic_load): Likewise.
12327 (expand_atomic_store): Likewise.
12328 * ree.c (combine_reaching_defs): Likewise.
12329 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12330 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12331 * tree.h (type_has_mode_precision_p): Likewise.
12332 * ubsan.c (instrument_si_overflow): Likewise.
12334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12335 Alan Hayward <alan.hayward@arm.com>
12336 David Sherwood <david.sherwood@arm.com>
12338 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12339 polynomial numbers of units.
12340 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12341 (valid_vector_subparts_p): New function.
12342 (build_vector_type): Remove temporary shim and take the number
12343 of units as a poly_uint64 rather than an int.
12344 (build_opaque_vector_type): Take the number of units as a
12345 poly_uint64 rather than an int.
12346 * tree.c (build_vector_from_ctor): Handle polynomial
12347 TYPE_VECTOR_SUBPARTS.
12348 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12349 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12350 (build_vector_from_val): If the number of units is variable,
12351 use build_vec_duplicate_cst for constant operands and
12352 VEC_DUPLICATE_EXPR otherwise.
12353 (make_vector_type): Remove temporary is_constant ().
12354 (build_vector_type, build_opaque_vector_type): Take the number of
12355 units as a poly_uint64 rather than an int.
12356 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12358 * cfgexpand.c (expand_debug_expr): Likewise.
12359 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12360 (store_constructor, expand_expr_real_1): Likewise.
12361 (const_scalar_mask_from_tree): Likewise.
12362 * fold-const-call.c (fold_const_reduction): Likewise.
12363 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12364 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12365 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12366 (fold_relational_const): Likewise.
12367 (native_interpret_vector): Likewise. Change the size from an
12368 int to an unsigned int.
12369 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12370 TYPE_VECTOR_SUBPARTS.
12371 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12372 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12373 duplicating a non-constant operand into a variable-length vector.
12374 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12375 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12376 * ipa-icf.c (sem_variable::equals): Likewise.
12377 * match.pd: Likewise.
12378 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12379 * print-tree.c (print_node): Likewise.
12380 * stor-layout.c (layout_type): Likewise.
12381 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12382 * tree-cfg.c (verify_gimple_comparison): Likewise.
12383 (verify_gimple_assign_binary): Likewise.
12384 (verify_gimple_assign_ternary): Likewise.
12385 (verify_gimple_assign_single): Likewise.
12386 * tree-pretty-print.c (dump_generic_node): Likewise.
12387 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12388 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12389 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12390 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12391 (vect_shift_permute_load_chain): Likewise.
12392 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12393 (expand_vector_condition, optimize_vector_constructor): Likewise.
12394 (lower_vec_perm, get_compute_type): Likewise.
12395 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12396 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12397 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12398 (vect_recog_mask_conversion_pattern): Likewise.
12399 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12400 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12401 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12402 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12403 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12404 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12405 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12406 (supportable_widening_operation): Likewise.
12407 (supportable_narrowing_operation): Likewise.
12408 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12410 * varasm.c (output_constant): Likewise.
12412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12413 Alan Hayward <alan.hayward@arm.com>
12414 David Sherwood <david.sherwood@arm.com>
12416 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12417 so that both the length == 3 and length != 3 cases set up their
12418 own permute vectors. Add comments explaining why we know the
12419 number of elements is constant.
12420 (vect_permute_load_chain): Likewise.
12422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12423 Alan Hayward <alan.hayward@arm.com>
12424 David Sherwood <david.sherwood@arm.com>
12426 * machmode.h (mode_nunits): Change from unsigned char to
12428 (ONLY_FIXED_SIZE_MODES): New macro.
12429 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12430 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12431 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12433 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12434 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12435 or if measurement_type is not polynomial.
12436 * genmodes.c (ZERO_COEFFS): New macro.
12437 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12439 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12440 Use ZERO_COEFFS when emitting initializers.
12441 * data-streamer.h (bp_pack_poly_value): New function.
12442 (bp_unpack_poly_value): Likewise.
12443 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12444 for GET_MODE_NUNITS.
12445 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12446 for GET_MODE_NUNITS.
12447 * tree.c (make_vector_type): Remove temporary shim and make
12448 the real function take the number of units as a poly_uint64
12449 rather than an int.
12450 (build_vector_type_for_mode): Handle polynomial nunits.
12451 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12452 * emit-rtl.c (const_vec_series_p_1): Likewise.
12453 (gen_rtx_CONST_VECTOR): Likewise.
12454 * fold-const.c (test_vec_duplicate_folding): Likewise.
12455 * genrecog.c (validate_pattern): Likewise.
12456 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12457 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12458 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12459 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12460 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12461 * rtlanal.c (subreg_get_info): Likewise.
12462 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12463 (vect_grouped_load_supported): Likewise.
12464 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12465 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12466 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12467 (simplify_const_unary_operation, simplify_binary_operation_1)
12468 (simplify_const_binary_operation, simplify_ternary_operation)
12469 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12470 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12471 instead of CONST_VECTOR_NUNITS.
12472 * varasm.c (output_constant_pool_2): Likewise.
12473 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12474 explicit-encoded elements in the XVEC for variable-length vectors.
12476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12478 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12481 Alan Hayward <alan.hayward@arm.com>
12482 David Sherwood <david.sherwood@arm.com>
12484 * coretypes.h (fixed_size_mode): Declare.
12485 (fixed_size_mode_pod): New typedef.
12486 * builtins.h (target_builtins::x_apply_args_mode)
12487 (target_builtins::x_apply_result_mode): Change type to
12488 fixed_size_mode_pod.
12489 * builtins.c (apply_args_size, apply_result_size, result_vector)
12490 (expand_builtin_apply_args_1, expand_builtin_apply)
12491 (expand_builtin_return): Update accordingly.
12493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12495 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12496 * cselib.c (cselib_hash_rtx): Likewise.
12497 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12498 CONST_VECTOR encoding.
12500 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12501 Jeff Law <law@redhat.com>
12504 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12505 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12506 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12507 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12510 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12511 explicitly probe *sp in a noreturn function if there were any callee
12512 register saves or frame pointer is needed.
12514 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12517 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12518 BLKmode for ternary, binary or unary expressions.
12521 * var-tracking.c (delete_vta_debug_insn): New inline function.
12522 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12523 insns from get_insns () to NULL instead of each bb separately.
12524 Use delete_vta_debug_insn. No longer static.
12525 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12526 delete_vta_debug_insns callers.
12527 * rtl.h (delete_vta_debug_insns): Declare.
12528 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12529 instead of variable_tracking_main.
12531 2018-01-03 Martin Sebor <msebor@redhat.com>
12533 PR tree-optimization/83603
12534 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12535 arguments past the endof the argument list in functions declared
12536 without a prototype.
12537 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12538 Avoid checking when arguments are null.
12540 2018-01-03 Martin Sebor <msebor@redhat.com>
12543 * doc/extend.texi (attribute const): Fix a typo.
12544 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12545 issuing -Wsuggest-attribute for void functions.
12547 2018-01-03 Martin Sebor <msebor@redhat.com>
12549 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12550 offset_int::from instead of wide_int::to_shwi.
12551 (maybe_diag_overlap): Remove assertion.
12552 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12553 * gimple-ssa-sprintf.c (format_directive): Same.
12554 (parse_directive): Same.
12555 (sprintf_dom_walker::compute_format_length): Same.
12556 (try_substitute_return_value): Same.
12558 2018-01-03 Jeff Law <law@redhat.com>
12560 PR middle-end/83654
12561 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12562 non-constant residual for zero at runtime and avoid probing in
12563 that case. Reorganize code for trailing problem to mirror handling
12566 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12568 PR tree-optimization/83501
12569 * tree-ssa-strlen.c (get_string_cst): New.
12570 (handle_char_store): Call get_string_cst.
12572 2018-01-03 Martin Liska <mliska@suse.cz>
12574 PR tree-optimization/83593
12575 * tree-ssa-strlen.c: Include tree-cfg.h.
12576 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12577 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12578 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12580 (strlen_dom_walker::before_dom_children): Call
12581 gimple_purge_dead_eh_edges. Dump tranformation with details
12583 (strlen_dom_walker::before_dom_children): Update call by adding
12584 new argument cleanup_eh.
12585 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12587 2018-01-03 Martin Liska <mliska@suse.cz>
12590 * cif-code.def (VARIADIC_THUNK): New enum value.
12591 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12594 2018-01-03 Jan Beulich <jbeulich@suse.com>
12596 * sse.md (mov<mode>_internal): Tighten condition for when to use
12597 vmovdqu<ssescalarsize> for TI and OI modes.
12599 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12601 Update copyright years.
12603 2018-01-03 Martin Liska <mliska@suse.cz>
12606 * ipa-visibility.c (function_and_variable_visibility): Skip
12607 functions with noipa attribure.
12609 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12611 * gcc.c (process_command): Update copyright notice dates.
12612 * gcov-dump.c (print_version): Ditto.
12613 * gcov.c (print_version): Ditto.
12614 * gcov-tool.c (print_version): Ditto.
12615 * gengtype.c (create_file): Ditto.
12616 * doc/cpp.texi: Bump @copying's copyright year.
12617 * doc/cppinternals.texi: Ditto.
12618 * doc/gcc.texi: Ditto.
12619 * doc/gccint.texi: Ditto.
12620 * doc/gcov.texi: Ditto.
12621 * doc/install.texi: Ditto.
12622 * doc/invoke.texi: Ditto.
12624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12626 * vector-builder.h (vector_builder::m_full_nelts): Change from
12627 unsigned int to poly_uint64.
12628 (vector_builder::full_nelts): Update prototype accordingly.
12629 (vector_builder::new_vector): Likewise.
12630 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12631 (vector_builder::operator ==): Likewise.
12632 (vector_builder::finalize): Likewise.
12633 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12634 Take the number of elements as a poly_uint64 rather than an
12636 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12637 from unsigned int to poly_uint64.
12638 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12639 (vec_perm_indices::new_vector): Likewise.
12640 (vec_perm_indices::length): Likewise.
12641 (vec_perm_indices::nelts_per_input): Likewise.
12642 (vec_perm_indices::input_nelts): Likewise.
12643 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12644 number of elements per input as a poly_uint64 rather than an
12645 unsigned int. Use the original encoding for variable-length
12646 vectors, rather than clamping each individual element.
12647 For the second and subsequent elements in each pattern,
12648 clamp the step and base before clamping their sum.
12649 (vec_perm_indices::series_p): Handle polynomial element counts.
12650 (vec_perm_indices::all_in_range_p): Likewise.
12651 (vec_perm_indices_to_tree): Likewise.
12652 (vec_perm_indices_to_rtx): Likewise.
12653 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12654 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12655 (tree_vector_builder::new_binary_operation): Handle polynomial
12656 element counts. Return false if we need to know the number
12657 of elements at compile time.
12658 * fold-const.c (fold_vec_perm): Punt if the number of elements
12659 isn't known at compile time.
12661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12663 * vec-perm-indices.h (vec_perm_builder): Change element type
12664 from HOST_WIDE_INT to poly_int64.
12665 (vec_perm_indices::element_type): Update accordingly.
12666 (vec_perm_indices::clamp): Handle polynomial element_types.
12667 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12668 (vec_perm_indices::all_in_range_p): Likewise.
12669 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12671 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12672 polynomial vec_perm_indices element types.
12673 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12674 * fold-const.c (fold_vec_perm): Likewise.
12675 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12676 * tree-vect-generic.c (lower_vec_perm): Likewise.
12677 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12678 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12679 element type to HOST_WIDE_INT.
12681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12682 Alan Hayward <alan.hayward@arm.com>
12683 David Sherwood <david.sherwood@arm.com>
12685 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12686 rather than an int. Use plus_constant.
12687 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12688 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12691 Alan Hayward <alan.hayward@arm.com>
12692 David Sherwood <david.sherwood@arm.com>
12694 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12695 a HOST_WIDE_INT to a poly_int64.
12697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12698 Alan Hayward <alan.hayward@arm.com>
12699 David Sherwood <david.sherwood@arm.com>
12701 * calls.c (load_register_parameters): Cope with polynomial
12702 mode sizes. Require a constant size for BLKmode parameters
12703 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12704 forces a parameter to be padded at the lsb end in order to
12705 fill a complete number of words, require the parameter size
12706 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
12713 from unsigned int to poly_uint64_pod.
12714 (alter_reg): Treat mode sizes as polynomial.
12716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12717 Alan Hayward <alan.hayward@arm.com>
12718 David Sherwood <david.sherwood@arm.com>
12720 * reload.c (complex_word_subreg_p): New function.
12721 (reload_inner_reg_of_subreg, push_reload): Use it.
12723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12724 Alan Hayward <alan.hayward@arm.com>
12725 David Sherwood <david.sherwood@arm.com>
12727 * lra-constraints.c (process_alt_operands): Reject matched
12728 operands whose sizes aren't ordered.
12729 (match_reload): Refer to this check here.
12731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12732 Alan Hayward <alan.hayward@arm.com>
12733 David Sherwood <david.sherwood@arm.com>
12735 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12736 that the mode size is in the set {1, 2, 4, 8, 16}.
12738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12739 Alan Hayward <alan.hayward@arm.com>
12740 David Sherwood <david.sherwood@arm.com>
12742 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12743 Use plus_constant instead of gen_rtx_PLUS.
12745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12746 Alan Hayward <alan.hayward@arm.com>
12747 David Sherwood <david.sherwood@arm.com>
12749 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12750 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12751 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12752 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12753 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12754 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12755 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12756 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12757 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12758 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12760 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12761 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12762 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12763 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12764 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12765 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12766 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12767 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12768 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12769 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12771 * expr.c (emit_move_resolve_push): Treat the input and result
12772 of PUSH_ROUNDING as a poly_int64.
12773 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12774 (emit_push_insn): Likewise.
12775 * lra-eliminations.c (mark_not_eliminable): Likewise.
12776 * recog.c (push_operand): Likewise.
12777 * reload1.c (elimination_effects): Likewise.
12778 * rtlanal.c (nonzero_bits1): Likewise.
12779 * calls.c (store_one_arg): Likewise. Require the padding to be
12780 known at compile time.
12782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12783 Alan Hayward <alan.hayward@arm.com>
12784 David Sherwood <david.sherwood@arm.com>
12786 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12787 Use plus_constant instead of gen_rtx_PLUS.
12789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12790 Alan Hayward <alan.hayward@arm.com>
12791 David Sherwood <david.sherwood@arm.com>
12793 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12794 rather than an int.
12796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12797 Alan Hayward <alan.hayward@arm.com>
12798 David Sherwood <david.sherwood@arm.com>
12800 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12801 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12802 via stack temporaries. Treat the mode size as polynomial too.
12804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12805 Alan Hayward <alan.hayward@arm.com>
12806 David Sherwood <david.sherwood@arm.com>
12808 * expr.c (expand_expr_real_2): When handling conversions involving
12809 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12810 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12811 as a poly_uint64 too.
12813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12814 Alan Hayward <alan.hayward@arm.com>
12815 David Sherwood <david.sherwood@arm.com>
12817 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12820 Alan Hayward <alan.hayward@arm.com>
12821 David Sherwood <david.sherwood@arm.com>
12823 * combine.c (can_change_dest_mode): Handle polynomial
12824 REGMODE_NATURAL_SIZE.
12825 * expmed.c (store_bit_field_1): Likewise.
12826 * expr.c (store_constructor): Likewise.
12827 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12828 and polynomial REGMODE_NATURAL_SIZE.
12829 (gen_lowpart_common): Likewise.
12830 * reginfo.c (record_subregs_of_mode): Likewise.
12831 * rtlanal.c (read_modify_subreg_p): Likewise.
12833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12834 Alan Hayward <alan.hayward@arm.com>
12835 David Sherwood <david.sherwood@arm.com>
12837 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12838 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12851 in a POINTER_PLUS_EXPR.
12853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12854 Alan Hayward <alan.hayward@arm.com>
12855 David Sherwood <david.sherwood@arm.com>
12857 * omp-simd-clone.c (simd_clone_subparts): New function.
12858 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12859 (ipa_simd_modify_function_body): Likewise.
12861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12862 Alan Hayward <alan.hayward@arm.com>
12863 David Sherwood <david.sherwood@arm.com>
12865 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12866 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12867 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12868 (expand_vector_condition, vector_element): Likewise.
12869 (subparts_gt): New function.
12870 (get_compute_type): Use subparts_gt.
12871 (count_type_subparts): Delete.
12872 (expand_vector_operations_1): Use subparts_gt instead of
12873 count_type_subparts.
12875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12876 Alan Hayward <alan.hayward@arm.com>
12877 David Sherwood <david.sherwood@arm.com>
12879 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12880 (vect_compile_time_alias): ...this new function. Do the calculation
12881 on poly_ints rather than trees.
12882 (vect_prune_runtime_alias_test_list): Update call accordingly.
12884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12885 Alan Hayward <alan.hayward@arm.com>
12886 David Sherwood <david.sherwood@arm.com>
12888 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12890 (vect_schedule_slp_instance): Likewise.
12892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12893 Alan Hayward <alan.hayward@arm.com>
12894 David Sherwood <david.sherwood@arm.com>
12896 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12897 constant and extern definitions for variable-length vectors.
12898 (vect_get_constant_vectors): Note that the number of units
12899 is known to be constant.
12901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12902 Alan Hayward <alan.hayward@arm.com>
12903 David Sherwood <david.sherwood@arm.com>
12905 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12906 of units as polynomial. Choose between WIDE and NARROW based
12909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12910 Alan Hayward <alan.hayward@arm.com>
12911 David Sherwood <david.sherwood@arm.com>
12913 * tree-vect-stmts.c (simd_clone_subparts): New function.
12914 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12917 Alan Hayward <alan.hayward@arm.com>
12918 David Sherwood <david.sherwood@arm.com>
12920 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12921 vectors as polynomial. Use build_index_vector for
12922 IFN_GOMP_SIMD_LANE.
12924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12925 Alan Hayward <alan.hayward@arm.com>
12926 David Sherwood <david.sherwood@arm.com>
12928 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12929 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12930 for variable-length vectors.
12931 (vectorizable_mask_load_store): Treat the number of units as
12932 polynomial, asserting that it is constant if the condition has
12933 already been enforced.
12934 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
12941 of units as polynomial. Punt if we can't tell at compile time
12942 which vector contains the final result.
12944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12945 Alan Hayward <alan.hayward@arm.com>
12946 David Sherwood <david.sherwood@arm.com>
12948 * tree-vect-loop.c (vectorizable_induction): Treat the number
12949 of units as polynomial. Punt on SLP inductions. Use an integer
12950 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12951 cast of such a series for variable-length floating-point
12954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12955 Alan Hayward <alan.hayward@arm.com>
12956 David Sherwood <david.sherwood@arm.com>
12958 * tree.h (build_index_vector): Declare.
12959 * tree.c (build_index_vector): New function.
12960 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12961 of units as polynomial, forcibly converting it to a constant if
12962 vectorizable_reduction has already enforced the condition.
12963 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12964 to create a {1,2,3,...} vector.
12965 (vectorizable_reduction): Treat the number of units as polynomial.
12966 Choose vectype_in based on the largest scalar element size rather
12967 than the smallest number of units. Enforce the restrictions
12970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12971 Alan Hayward <alan.hayward@arm.com>
12972 David Sherwood <david.sherwood@arm.com>
12974 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12975 number of units as polynomial.
12977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12978 Alan Hayward <alan.hayward@arm.com>
12979 David Sherwood <david.sherwood@arm.com>
12981 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12982 * target.def (autovectorize_vector_sizes): Return the vector sizes
12983 by pointer, using vector_sizes rather than a bitmask.
12984 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12985 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12986 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12988 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12989 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12990 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12991 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12992 * omp-general.c (omp_max_vf): Likewise.
12993 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12994 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12995 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12996 * tree-vect-slp.c (vect_slp_bb): Likewise.
12997 * doc/tm.texi: Regenerate.
12998 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13000 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13001 the vector size as a poly_uint64 rather than an unsigned int.
13002 (current_vector_size): Change from an unsigned int to a poly_uint64.
13003 (get_vectype_for_scalar_type): Update accordingly.
13004 * tree.h (build_truth_vector_type): Take the size and number of
13005 units as a poly_uint64 rather than an unsigned int.
13006 (build_vector_type): Add a temporary overload that takes
13007 the number of units as a poly_uint64 rather than an unsigned int.
13008 * tree.c (make_vector_type): Likewise.
13009 (build_truth_vector_type): Take the number of units as a poly_uint64
13010 rather than an unsigned int.
13012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13013 Alan Hayward <alan.hayward@arm.com>
13014 David Sherwood <david.sherwood@arm.com>
13016 * target.def (get_mask_mode): Take the number of units and length
13017 as poly_uint64s rather than unsigned ints.
13018 * targhooks.h (default_get_mask_mode): Update accordingly.
13019 * targhooks.c (default_get_mask_mode): Likewise.
13020 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13021 * doc/tm.texi: Regenerate.
13023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13024 Alan Hayward <alan.hayward@arm.com>
13025 David Sherwood <david.sherwood@arm.com>
13027 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13028 * omp-general.c (omp_max_vf): Likewise.
13029 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13030 (expand_omp_simd): Handle polynomial safelen.
13031 * omp-low.c (omplow_simd_context): Add a default constructor.
13032 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13033 (lower_rec_simd_input_clauses): Update accordingly.
13034 (lower_rec_input_clauses): Likewise.
13036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13037 Alan Hayward <alan.hayward@arm.com>
13038 David Sherwood <david.sherwood@arm.com>
13040 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13041 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13042 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13043 (vect_analyze_slp_cost): Likewise.
13044 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13045 (vect_model_load_cost): Likewise.
13047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13048 Alan Hayward <alan.hayward@arm.com>
13049 David Sherwood <david.sherwood@arm.com>
13051 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13052 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13053 from an unsigned int * to a poly_uint64_pod *.
13054 (calculate_unrolling_factor): New function.
13055 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13058 Alan Hayward <alan.hayward@arm.com>
13059 David Sherwood <david.sherwood@arm.com>
13061 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13062 from an unsigned int to a poly_uint64.
13063 (_loop_vec_info::slp_unrolling_factor): Likewise.
13064 (_loop_vec_info::vectorization_factor): Change from an int
13066 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13067 (vect_get_num_vectors): New function.
13068 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13069 (vect_get_num_copies): Use vect_get_num_vectors.
13070 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13071 to an unsigned int *.
13072 (vect_analyze_data_refs): Change min_vf from an int * to a
13074 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13075 than an unsigned HOST_WIDE_INT.
13076 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13077 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13078 to an unsigned int *.
13079 (vect_analyze_data_ref_dependences): Likewise.
13080 (vect_compute_data_ref_alignment): Handle polynomial vf.
13081 (vect_enhance_data_refs_alignment): Likewise.
13082 (vect_prune_runtime_alias_test_list): Likewise.
13083 (vect_shift_permute_load_chain): Likewise.
13084 (vect_supportable_dr_alignment): Likewise.
13085 (dependence_distance_ge_vf): Take the vectorization factor as a
13086 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13087 (vect_analyze_data_refs): Change min_vf from an int * to a
13089 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13090 vfm1 as a poly_uint64 rather than an int. Make the same change
13091 for the returned bound_scalar.
13092 (vect_gen_vector_loop_niters): Handle polynomial vf.
13093 (vect_do_peeling): Likewise. Update call to
13094 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13095 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13097 * tree-vect-loop.c (vect_determine_vectorization_factor)
13098 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13099 (vect_get_known_peeling_cost): Likewise.
13100 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13101 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13102 (vect_transform_loop): Likewise. Use the lowest possible VF when
13103 updating the upper bounds of the loop.
13104 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13105 rather than an int.
13106 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13107 polynomial unroll factors.
13108 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13109 (vect_make_slp_decision): Likewise.
13110 (vect_supported_load_permutation_p): Likewise, and polynomial
13112 (vect_analyze_slp_cost): Handle polynomial vf.
13113 (vect_slp_analyze_node_operations): Likewise.
13114 (vect_slp_analyze_bb_1): Likewise.
13115 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13116 than an unsigned HOST_WIDE_INT.
13117 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13118 (vectorizable_load): Handle polynomial vf.
13119 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13121 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13124 Alan Hayward <alan.hayward@arm.com>
13125 David Sherwood <david.sherwood@arm.com>
13127 * match.pd: Handle bit operations involving three constants
13128 and try to fold one pair.
13130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13132 * tree-vect-loop-manip.c: Include gimple-fold.h.
13133 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13134 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13135 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13136 Add a path that uses a step of VF instead of 1, but disable it
13138 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13139 and niters_no_overflow parameters. Update calls to
13140 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13141 Create a new SSA name if the latter choses to use a ste other
13142 than zero, and return it via niters_vector_mult_vf_var.
13143 * tree-vect-loop.c (vect_transform_loop): Update calls to
13144 vect_do_peeling, vect_gen_vector_loop_niters and
13145 slpeel_make_loop_iterate_ntimes.
13146 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13147 (vect_gen_vector_loop_niters): Update declarations after above changes.
13149 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13151 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13152 128-bit round to integer instructions.
13153 (ceil<mode>2): Likewise.
13154 (btrunc<mode>2): Likewise.
13155 (round<mode>2): Likewise.
13157 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13159 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13160 unaligned VSX load/store on P8/P9.
13161 (expand_block_clear): Allow the use of unaligned VSX
13162 load/store on P8/P9.
13164 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13166 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13168 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13169 swap associated with both a load and a store.
13171 2018-01-02 Andrew Waterman <andrew@sifive.com>
13173 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13174 * config/riscv/riscv.md (clear_cache): Use it.
13176 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13178 * web.c: Remove out-of-date comment.
13180 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13182 * expr.c (fixup_args_size_notes): Check that any existing
13183 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13184 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13185 (emit_single_push_insn): ...here.
13187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13189 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13190 (const_vector_encoded_nelts): New function.
13191 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13192 (const_vector_int_elt, const_vector_elt): Declare.
13193 * emit-rtl.c (const_vector_int_elt_1): New function.
13194 (const_vector_elt): Likewise.
13195 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13196 of CONST_VECTOR_ELT.
13198 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13200 * expr.c: Include rtx-vector-builder.h.
13201 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13202 directly on the tree encoding.
13203 (const_vector_from_tree): Likewise.
13204 * optabs.c: Include rtx-vector-builder.h.
13205 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13206 sequence of "u" values.
13207 * vec-perm-indices.c: Include rtx-vector-builder.h.
13208 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13209 directly on the vec_perm_indices encoding.
13211 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13213 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13214 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13215 * rtx-vector-builder.h: New file.
13216 * rtx-vector-builder.c: Likewise.
13217 * rtl.h (rtx_def::u2): Add a const_vector field.
13218 (CONST_VECTOR_NPATTERNS): New macro.
13219 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13220 (CONST_VECTOR_DUPLICATE_P): Likewise.
13221 (CONST_VECTOR_STEPPED_P): Likewise.
13222 (CONST_VECTOR_ENCODED_ELT): Likewise.
13223 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13224 (unwrap_const_vec_duplicate): Likewise.
13225 (const_vec_series_p): Check for a non-duplicated vector encoding.
13226 Say that the function only returns true for integer vectors.
13227 * emit-rtl.c: Include rtx-vector-builder.h.
13228 (gen_const_vec_duplicate_1): Delete.
13229 (gen_const_vector): Call gen_const_vec_duplicate instead of
13230 gen_const_vec_duplicate_1.
13231 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13232 (gen_const_vec_duplicate): Use rtx_vector_builder.
13233 (gen_const_vec_series): Likewise.
13234 (gen_rtx_CONST_VECTOR): Likewise.
13235 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13236 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13237 Build a new vector rather than modifying a CONST_VECTOR in-place.
13238 (handle_special_swappables): Update call accordingly.
13239 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13240 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13241 Build a new vector rather than modifying a CONST_VECTOR in-place.
13242 (handle_special_swappables): Update call accordingly.
13244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13246 * simplify-rtx.c (simplify_const_binary_operation): Use
13247 CONST_VECTOR_ELT instead of XVECEXP.
13249 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13251 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13252 the selector elements to be different from the data elements
13253 if the selector is a VECTOR_CST.
13254 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13255 ssizetype for the selector.
13257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13259 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13260 before testing each element individually.
13261 * tree-vect-generic.c (lower_vec_perm): Likewise.
13263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13265 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13266 * selftest-run-tests.c (selftest::run_tests): Call it.
13267 * vector-builder.h (vector_builder::operator ==): New function.
13268 (vector_builder::operator !=): Likewise.
13269 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13270 (vec_perm_indices::all_from_input_p): New function.
13271 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13272 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13273 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13274 instead of reading the VECTOR_CST directly. Detect whether both
13275 vector inputs are the same before constructing the vec_perm_indices,
13276 and update the number of inputs argument accordingly. Use the
13277 utility functions added above. Only construct sel2 if we need to.
13279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13281 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13282 the broadcast of the low byte.
13283 (expand_mult_highpart): Use an explicit encoding for the permutes.
13284 * optabs-query.c (can_mult_highpart_p): Likewise.
13285 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13286 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13287 (vectorizable_bswap): Likewise.
13288 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13289 explicit encoding for the power-of-2 permutes.
13290 (vect_permute_store_chain): Likewise.
13291 (vect_grouped_load_supported): Likewise.
13292 (vect_permute_load_chain): Likewise.
13294 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13296 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13297 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13298 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13299 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13300 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13301 (vect_gen_perm_mask_any): Likewise.
13303 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13305 * int-vector-builder.h: New file.
13306 * vec-perm-indices.h: Include int-vector-builder.h.
13307 (vec_perm_indices): Redefine as an int_vector_builder.
13308 (auto_vec_perm_indices): Delete.
13309 (vec_perm_builder): Redefine as a stand-alone class.
13310 (vec_perm_indices::vec_perm_indices): New function.
13311 (vec_perm_indices::clamp): Likewise.
13312 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13313 (vec_perm_indices::new_vector): New function.
13314 (vec_perm_indices::new_expanded_vector): Update for new
13315 vec_perm_indices class.
13316 (vec_perm_indices::rotate_inputs): New function.
13317 (vec_perm_indices::all_in_range_p): Operate directly on the
13318 encoded form, without computing elided elements.
13319 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13320 encoding. Update for new vec_perm_indices class.
13321 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13322 the given vec_perm_builder.
13323 (expand_vec_perm_var): Update vec_perm_builder constructor.
13324 (expand_mult_highpart): Use vec_perm_builder instead of
13325 auto_vec_perm_indices.
13326 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13327 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13328 or double series encoding as appropriate.
13329 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13330 vec_perm_indices instead of auto_vec_perm_indices.
13331 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13332 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13333 (vect_permute_store_chain): Likewise.
13334 (vect_grouped_load_supported): Likewise.
13335 (vect_permute_load_chain): Likewise.
13336 (vect_shift_permute_load_chain): Likewise.
13337 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13338 (vect_transform_slp_perm_load): Likewise.
13339 (vect_schedule_slp_instance): Likewise.
13340 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13341 (vectorizable_mask_load_store): Likewise.
13342 (vectorizable_bswap): Likewise.
13343 (vectorizable_store): Likewise.
13344 (vectorizable_load): Likewise.
13345 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13346 vec_perm_indices instead of auto_vec_perm_indices. Use
13347 tree_to_vec_perm_builder to read the vector from a tree.
13348 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13349 vec_perm_builder instead of a vec_perm_indices.
13350 (have_whole_vector_shift): Use vec_perm_builder and
13351 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13352 truncation to calc_vec_perm_mask_for_shift.
13353 (vect_create_epilog_for_reduction): Likewise.
13354 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13355 from auto_vec_perm_indices to vec_perm_indices.
13356 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13357 instead of changing individual elements.
13358 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13359 the vector in d.perm.
13360 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13361 from auto_vec_perm_indices to vec_perm_indices.
13362 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13363 instead of changing individual elements.
13364 (arm_vectorize_vec_perm_const): Use new_vector to install
13365 the vector in d.perm.
13366 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13367 Update vec_perm_builder constructor.
13368 (rs6000_expand_interleave): Likewise.
13369 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13370 (rs6000_expand_interleave): Likewise.
13372 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13374 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13375 to qimode could truncate the indices.
13376 * optabs.c (expand_vec_perm_var): Likewise.
13378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13380 * Makefile.in (OBJS): Add vec-perm-indices.o.
13381 * vec-perm-indices.h: New file.
13382 * vec-perm-indices.c: Likewise.
13383 * target.h (vec_perm_indices): Replace with a forward class
13385 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13386 * optabs.h: Include vec-perm-indices.h.
13387 (expand_vec_perm): Delete.
13388 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13389 (expand_vec_perm_const): Declare.
13390 * target.def (vec_perm_const_ok): Replace with...
13391 (vec_perm_const): ...this new hook.
13392 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13393 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13394 * doc/tm.texi: Regenerate.
13395 * optabs.def (vec_perm_const): Delete.
13396 * doc/md.texi (vec_perm_const): Likewise.
13397 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13398 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13399 expand_vec_perm for constant permutation vectors. Assert that
13400 the mode of variable permutation vectors is the integer equivalent
13401 of the mode that is being permuted.
13402 * optabs-query.h (selector_fits_mode_p): Declare.
13403 * optabs-query.c: Include vec-perm-indices.h.
13404 (selector_fits_mode_p): New function.
13405 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13406 is defined, instead of checking whether the vec_perm_const_optab
13407 exists. Use targetm.vectorize.vec_perm_const instead of
13408 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13409 fit in the vector mode before using a variable permute.
13410 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13411 vec_perm_indices instead of an rtx.
13412 (expand_vec_perm): Replace with...
13413 (expand_vec_perm_const): ...this new function. Take the selector
13414 as a vec_perm_indices rather than an rtx. Also take the mode of
13415 the selector. Update call to shift_amt_for_vec_perm_mask.
13416 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13417 Use vec_perm_indices::new_expanded_vector to expand the original
13418 selector into bytes. Check whether the indices fit in the vector
13419 mode before using a variable permute.
13420 (expand_vec_perm_var): Make global.
13421 (expand_mult_highpart): Use expand_vec_perm_const.
13422 * fold-const.c: Includes vec-perm-indices.h.
13423 * tree-ssa-forwprop.c: Likewise.
13424 * tree-vect-data-refs.c: Likewise.
13425 * tree-vect-generic.c: Likewise.
13426 * tree-vect-loop.c: Likewise.
13427 * tree-vect-slp.c: Likewise.
13428 * tree-vect-stmts.c: Likewise.
13429 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13431 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13432 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13433 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13434 (aarch64_vectorize_vec_perm_const): ...this new function.
13435 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13436 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13437 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13438 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13439 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13440 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13441 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13443 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13444 check for NEON modes.
13445 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13446 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13447 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13448 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13450 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13451 the old VEC_PERM_CONST conditions.
13452 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13453 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13454 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13455 (ia64_vectorize_vec_perm_const_ok): Merge into...
13456 (ia64_vectorize_vec_perm_const): ...this new function.
13457 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13458 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13459 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13460 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13461 * config/mips/mips.c (mips_expand_vec_perm_const)
13462 (mips_vectorize_vec_perm_const_ok): Merge into...
13463 (mips_vectorize_vec_perm_const): ...this new function.
13464 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13465 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13466 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13467 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13468 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13469 (rs6000_expand_vec_perm_const): Delete.
13470 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13472 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13473 (altivec_expand_vec_perm_const_le): Take each operand individually.
13474 Operate on constant selectors rather than rtxes.
13475 (altivec_expand_vec_perm_const): Likewise. Update call to
13476 altivec_expand_vec_perm_const_le.
13477 (rs6000_expand_vec_perm_const): Delete.
13478 (rs6000_vectorize_vec_perm_const_ok): Delete.
13479 (rs6000_vectorize_vec_perm_const): New function.
13480 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13481 an element count and rtx array.
13482 (rs6000_expand_extract_even): Update call accordingly.
13483 (rs6000_expand_interleave): Likewise.
13484 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13485 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13486 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13487 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13488 (rs6000_expand_vec_perm_const): Delete.
13489 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13490 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13491 (altivec_expand_vec_perm_const_le): Take each operand individually.
13492 Operate on constant selectors rather than rtxes.
13493 (altivec_expand_vec_perm_const): Likewise. Update call to
13494 altivec_expand_vec_perm_const_le.
13495 (rs6000_expand_vec_perm_const): Delete.
13496 (rs6000_vectorize_vec_perm_const_ok): Delete.
13497 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13498 reference to the SPE evmerge intructions.
13499 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13500 an element count and rtx array.
13501 (rs6000_expand_extract_even): Update call accordingly.
13502 (rs6000_expand_interleave): Likewise.
13503 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13504 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13506 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13508 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13510 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13511 vector mode and that that mode matches the mode of the data
13513 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13514 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13515 directly using expand_vec_perm_1 when forcing selectors into
13517 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13519 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13521 * optabs-query.h (can_vec_perm_p): Delete.
13522 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13523 * optabs-query.c (can_vec_perm_p): Split into...
13524 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13525 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13526 particular selector is valid.
13527 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13528 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13529 (vect_grouped_load_supported): Likewise.
13530 (vect_shift_permute_load_chain): Likewise.
13531 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13532 (vect_transform_slp_perm_load): Likewise.
13533 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13534 (vectorizable_bswap): Likewise.
13535 (vect_gen_perm_mask_checked): Likewise.
13536 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13537 implementations of variable permutation vectors into account
13538 when deciding which selector to use.
13539 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13540 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13541 with a false third argument.
13542 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13543 to test whether the constant selector is valid and can_vec_perm_var_p
13544 to test whether a variable selector is valid.
13546 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13548 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13549 * optabs-query.c (can_vec_perm_p): Likewise.
13550 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13551 instead of vec_perm_indices.
13552 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13553 (vect_gen_perm_mask_checked): Likewise,
13554 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13555 (vect_gen_perm_mask_checked): Likewise,
13557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13559 * optabs-query.h (qimode_for_vec_perm): Declare.
13560 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13561 (qimode_for_vec_perm): ...this new function.
13562 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13564 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13566 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13567 does not have a conditional at the top.
13569 2018-01-02 Richard Biener <rguenther@suse.de>
13571 * ipa-inline.c (big_speedup_p): Fix expression.
13573 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13576 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13579 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13583 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13584 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13585 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13586 cond_taken_branch_cost 3->4.
13588 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13590 PR tree-optimization/83581
13591 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13592 TODO_cleanup_cfg if any changes have been made.
13594 PR middle-end/83608
13595 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13596 convert_modes if target mode has the right side, but different mode
13599 PR middle-end/83609
13600 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13601 last argument when extracting from CONCAT. If either from_real or
13602 from_imag is NULL, use expansion through memory. If result is not
13603 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13604 the parts directly to inner mode, if even that fails, use expansion
13607 PR middle-end/83623
13608 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13609 check for bswap in mode rather than HImode and use that in expand_unop
13612 Copyright (C) 2018 Free Software Foundation, Inc.
13614 Copying and distribution of this file, with or without modification,
13615 are permitted in any medium without royalty provided the copyright
13616 notice and this notice are preserved.